Contents lists available at ScienceDirectJournal of Enviro

Contents lists available at ScienceDirect
Journal of Environmental Management
journal homepage: www.elsevier.com/locate/jenvman
Research article
Climate change and the provision of biodiversity in public
temperate forests
– A mechanism design approach for the implementation of
biodiversity
conservation policies
Andrey Lessa Derci Augustynczik∗ , Rasoul Yousefpour, Marc
Hanewinkel
Chair of Forestry Economics and Forest Planning, University of
Freiburg, Tennenbacherstr. 4, 79106, Freiburg, Germany
A R T I C L E I N F O
Keywords:
Forest biodiversity
Mechanism design
Forest optimization
Conservation planning
Forest birds
A B S T R A C T
The provision of forest biodiversity remains a major challenge
in the management of forest resources.

Biodiversity is mostly considered a public good and the fact
that societal benefits from biodiversity are private
information, hinders its supply at adequate levels. Here we
investigate how the government, as a forest owner,
may increase the biodiversity supply in publicly-owned forests.
We employ a mechanism design approach to find
the biodiversity provision choices, which take into account
agents’ strategic behavior and values towards bio-
diversity. We applied our framework to a forest landscape in
Southwestern Germany, using forest birds as
biodiversity indicators and evaluating the impacts of climate
change on forest dynamics and on the costs of
biodiversity provision. Our results show that climate change has
important implications to the opportunity cost
of biodiversity and the provision levels (ranging from 10 to
12.5% increase of the bird indicator abundance). In
general, biodiversity valuations needed to surpass the
opportunity cost by more than 18% to cope with the
private information held by the agents. Moreover, higher costs
under more intense climate change (e.g.
Representative Concentration Pathway 8.5) reduced the
attainable bird abundance increase from 12.5 to 10%.
We conclude that mechanism design may provide key
information for planning conservation policies and
identify conditions for a successful implementation of
biodiversity-oriented forest management.
1. Introduction
The provision of biodiversity remains a major challenge in the
management of forest resources. Biodiversity has been
continuously
declining worldwide during the past decades, despite its
recognized
importance to human well-being, ecosystem functioning and
ecosystem

resistance and resilience under climate change (Díaz et al.,
2006; Isbell
et al., 2015; Tilman et al., 2014). A main constraint to the im-
plementation of biodiversity conservation strategies is the fact
that
biodiversity is mostly considered a public good, and in the
absence of
markets or policy mechanisms to promote its provision, there
are in-
centives for free riding and undersupply. One option to tackle
this issue,
is to enhance biodiversity goals in public forests. The
government, as a
forest owner and aiming to promote an efficient use of forest
resources,
may raise funds and apply biodiversity-oriented management
solutions
in these areas. Thereby, it is possible to mitigate the
discrepancy be-
tween current and efficient biodiversity supply, promoting
sustain-
ability and increasing social welfare (Kemkes et al., 2010).
The promotion of biodiversity in forest landscapes demands a
cost-
benefit analysis of biodiversity-oriented management strategies,
com-
patible with societal preferences for multiple forest goods and
services.
This requires that both social costs and benefits related to
biodiversity
are known. The quantification of costs is a straightforward task,
e.g.
through the computation of the opportunity costs of
biodiversity-or-

iented forest management or the value of contracts for
biodiversity
amelioration (Rosenkranz et al., 2014). Conversely, the
evaluation of
biodiversity benefits involves indirect assessments,
predominantly ap-
plying choice experiments, where participants are asked how
much
they would be willing to contribute towards an increased
biodiversity
supply or by eliciting their preferences for bundles of ecosystem
ser-
vices (e.g. Meyerhoff et al., 2012; Getzner et al., 2018; Iranah
et al.,
2018). A key issue when considering biodiversity benefits, is
the fact
that the preferences for biodiversity are private information, and
policy
makers may have at hand only prior beliefs (e.g. the probability
dis-
tribution of these preferences), in terms of the willingness-to-
pay (WTP)
for biodiversity. This means that agents may have the incentive
to
misrepresent their true preferences when asked to contri bute
towards
the cost of biodiversity provision, hindering the implementation
of
https://doi.org/10.1016/j.jenvman.2019.05.089
Received 4 February 2019; Received in revised form 13 May
2019; Accepted 21 May 2019
∗ Corresponding author.
E-mail address: [email protected] (A.L.D. Augustynczik).

Journal of Environmental Management 246 (2019) 706–716
Available online 18 June 2019
0301-4797/ © 2019 Elsevier Ltd. All rights reserved.
T
http://www.sciencedirect.com/science/journal/03014797
https://www.elsevier.com/locate/jenvman
mailto:[email protected]
http://crossmark.crossref.org/dialog/?doi=10.1016/j.jenvman.20
19.05.089&domain=pdf
biodiversity conservation programs.
Mechanism design is arguably the best tool to address problems
of
this nature. Mechanism design is a sub-field of game theory,
also known
as inverse game theory. This framework searches for the design
of
games (e.g. auctions and voting schemes) that will lead to a
desired
outcome, such as welfare maximization or other economic goals
(Nisan
and Ronen, 2001). In this sense, mechanism design can be
applied to a
variety of natural resource management problems that involve
asym-
metric information. Formally, a mechanism is composed by a
social
choice function and a payment rule ensuring agents have

incentive to
participate in the mechanism and are not better-off by
misrepresenting
their true valuations. The mechanism designer can then make
use of
these functions to decide upon the implementation of
management
solutions.
Here we use this framework to tackle the increase in
biodiversity
supply in public forests, taking into account the societal
preferences for
biodiversity and the strategic behavior of the agents. In our
setting, the
government proposes a mechanism to raise capital to cover the
costs of
an increased biodiversity supply. According to Rands et al.
(2010), to
increase the success of conservation policies it will be
necessary to
prioritize the management of biodiversity as a public good, to
integrate
biodiversity in both public and private decision-making and to
facilitate
policy implementation. These priorities may be combined under
the
mechanism design framework. Thereby, we are able to
characterize the
supply levels that can be actually realized under the private
information
held by the agents and what the minimum conditions are, in
terms of
the social benefit, that enable the implementation of
conservation po-
licies without the need of external funding. This is crucial to

create
more resilient forest landscapes in the future.
A variety of mechanisms have been studied for the private
supply of
public goods. For example, Güth and Hellwig (1986) and Csapó
and
Müller (2013) provide a framework for defining social choice
functions
and payment rules for the private supply of public goods.
Bierbrauer
and Hellwig (2016) and Grüner and Koriyama (2012) analyze
the
provision of public goods in voting mechanisms. Güth and
Hellwig
(1986) highlight that a further difficulty in the supply of public
goods
arise when a large number of participants are involved, which is
typi-
cally the case for the implementation of biodiversity
conservation po-
licies. The same authors show that in this case, the probability
that a
single participant affects the supply of the public good is small,
leading
them to reduce their willingness-to-pay and contributions
toward the
cost of the public good. Hellwig (2003) addressed this issue and
re-
ported that in the case that the supply level of the public good is
bounded, the costs are independent of the number of agents and
the
number of participants is sufficiently large, then the public
good is
eventually provided.

Traditionally, voluntary mechanisms to increase biodiversity
pro-
vision in forest landscapes have been addressed by game
theoretical
models. These include, for example, auctions to assign forest
reserves
(e.g. Hartig and Drechsler, 2010) or auctions for bundles of
ecosystem
services (Roesch-McNally et al., 2016). Despite the large body
of lit-
erature dealing with the characterization of mechanisms for the
pro-
vision of public goods, and their suitability to address a range
of natural
resource management problems involving the private supply of
public
goods, mechanism design applications are still largely missing.
Apart from the private information regarding the valuations for
biodiversity, a further challenge for the implementation of
biodiversity-
oriented management refers to the uncertainty induced by
climate
change on the dynamics of both forest and forest biota. Climate
change
is expected to modify a variety of forest processes and
interactions, e.g.
forest growth rates, species composition and disturbance
activity
(Lindner et al., 2014). These processes are closely related to
forest
profitability and are therefore predicted to cascade to the
opportunity
costs of biodiversity-oriented management. These novel
environmental
conditions will demand new management solutions to anticipate

cli-
matic impacts. Therefore, a sensible analysis of mechanisms
targeting
biodiversity provision needs to consider future forest
development and
its impacts on the implementation of conservation policies.
Still there is a major gap in the literature regarding the
definition of
adequate provision levels of forest biodiversity taking into
account
social costs and benefits. Moreover, the strategic behavior of
agents
towards contributions to implement biodiversity-oriented
management
is usually neglected. Here we tackle these issues by integrating
ecolo-
gical and economic aspects of forest biodiversity, using the
machinery
of mechanism design. We build upon the frameworks developed
by
Hellwig (2003) and Csapó and Müller (2013), and consider the
provi-
sion of biodiversity in publicly owned forests in a temperate
forest
landscape under climate change, using a coupled ecological -
economic
framework. We addressed in this study the following research
ques-
tions:
• What are the opportunity costs of biodiversity-oriented
management
in a temperate forest landscape under climate change?

• What are the minimum conditions for a feasible
implementation of
biodiversity-oriented forest management in terms of societal
bene-
fits?
• What are the impacts of climate change on the social choice
function
and what are the optimal management solutions to realize an in-
creased biodiversity supply?
To answer the research questions, we applied our mechanism
design
framework to a temperate forest landscape in southwestern
Germany.
To account for climate impacts on forest development and
opportunity
cost, we used the process-based forest growth model 4C under
three
different climate change scenarios and applied five management
stra-
tegies: 1) biodiversity provision; 2) biomass production; 3)
business-as-
usual (BAU); 4) climate adaptation and 5) no management. We
built an
optimization model to maximize forest Net Present Value (NPV)
in an
80-years planning horizon. To define the mechanism, we
considered
biodiversity valuation data from an extensive choice experiment
con-
ducted in Germany, defining the thresholds for the supply of
biodi-
versity under climate change, in terms of the social value of
biodi-
versity. We consider the case where the designer represents the

federal
state (responsible for the management of forest resources) and
agents
represent the six administrative regions the study area,
negotiating on
behalf of their population and deciding upon the contribution
towards
the biodiversity supply cost. Although we conduct our analysis
in a
temperate forest landscape in Germany, the framework proposed
is
flexible and can be easily adapted to different biomes and
conservation
practices, as long as cost and benefits related to biodiversity
supply are
available.
2. Material and methods
We conducted our analysis following three steps. Initially, we
ap-
plied a climate-sensitive growth model to evaluate forest growth
dy-
namics under climate change, assessing the opportunity cost of
biodi-
versity-oriented forest management with the help of an
optimization
model. Subsequently, we computed the social benefits of
biodiversity,
applying the results of a choice experiment conducted in our
research
area and derived a social choice function for the implementation
of
biodiversity-oriented forest management. Finally, we evaluated
climate
impacts on the social choice function and how the required

increase in
biodiversity supply can be realized through forest management.
2.1. Data
To evaluate forest development under climate change, we used
forest inventory data from 98 1-ha plots located in publicly-
owned
forests in Southwestern Germany, following two design
gradients of
forest cover (< 50%, 50–75% and > 75 in 25 km2 radius) and
forest
structure (< 5 habitat trees/ha, 5–15 habitat trees/ha and > 15
habitat
A.L.D. Augustynczik, et al. Journal of Environmental
Management 246 (2019) 706–716
707
trees/ha). The forest inventory recorded tree species identity,
the DBH
of all trees (with DBH > 7 cm), height of 7% of the trees.
Moreover,
lying and standing deadwood amounts were assessed. These
plots were
used to estimate the average forest responses for each forest age
class.
Based on these responses, we performed the optimized forest
planning
of 17503 publicly owned forest stands in the southern Black
Forest,
covering an area of 54227 ha. The forest in the region is
dominated by

Norway spruce (Picea abies (L.) H. Karst.), European beech
(Fagus syl-
vatica L.) and silver fir (Abies alba Mill.).
To assess biodiversity, we used the German Biodiversity
Strategy
indicator, given by the abundance of 10 forest bird species
(from which
seven were present in our research area), in relation to the
abundance
in the 1970's (BMUB, 2015). We computed the responses of the
bird
assemblage to different forest management alternatives in each
of our
plots applying the hierarchical Bayesian model developed in
Augustynczik et al. (2019) for the same study area.
The preferences for biodiversity were based on the results from
the
choice experiment conducted by Weller and Elsasser (2017),
assuming
homogeneous preferences. The authors computed the WTP from
an
increase in the forest biodiversity indicator used in the German
Biodi-
versity Strategy, i.e. an increase in the abundance of 10
indicator bird
species. For assessing the affected population, we retrieved the
popu-
lation statistics from the national ministry of statistics
(Statistiches
Bundesamt, 2016).
We computed forest profitability in terms of the Net Present
Value
in an 80-years planning horizon. We discounted harvesting

revenues
with a 1.5% market interest rate (Müller and Hanewinkel,
2018). This
rate represent the typical return on capital in the region and is
com-
patible with the average long term interest rates in Germany
during the
past 10 years (ECB 2019). Harvesting and planting costs were
retrieved
from Härtl et al. (2013) and for timber revenues we used the
average
market wood prices in Baden-Württemberg in 2016. Therefore,
prices
and costs were assumed to be deterministic in our analysis.
2.2. Forest simulation and biodiversity supply
To evaluate forest development under climate change, we used
the
process-based forest growth model 4C (Lasch et al., 2002). 4C
describes
forest processes at tree and stand scale under changing
environmental
conditions and is capable of simulating a variety of management
in-
terventions, including thinning, planting and final harvesting
(Lasch
et al., 2005). A detailed description of the model is available in
the
model webpage (www.pik-potsdam.de/4c/). Since our plots
spanned
over more than one forest stand, we used the growth model
Sibyla
(Fabrika, 2007) to estimate the average age of each plot and
subse-
quently computed the average responses of each forest age class

based
on the plot average age. Sibyla is an individual-tree model that
uses the
stand generator STRUGEN (Pretzsch, 1997) to generate a forest
stand
according to individual tree input data, enabling to derive the
average
age of each species in the stand.
We considered in our simulations five manage ment strategies
and
three climate change scenarios. The management strategies were
de-
fined as: 1) Biodiversity conservation: increase current rotation
age in
10 years, apply a thinning intensity of 10% of the standing
volume and
replace Norway spruce stands by European beech stands; 2)
Biomass
production: decrease current rotation age by 20 years, apply a
thinning
intensity of 25% of the standing volume and convert spruce
stands to
Douglas-fir (Pseudotsuga menziesii) stands; 3) BAU: maintain
current
rotation age, species composition and a thinning intensity of
17% of the
standing volume; 4) Climate adaptation: decrease current
rotation age
by 10 years, apply a thinning intensity of 20% of the standing
volume
and convert spruce stands to Scots pine (Pinus sylvestris) and 5)
No
management: no thinning, harvesting or conversion. Each plot
and
management alternative was simulated under three climate

scenarios,
given by the combination of the Global Climate Model (GCM)
HadGEM2-ES and the Representative Concentra tion Pathways
(RCP)
2.6, 6.0 and 8.5, bias-corrected by ISIMIP
(https://www.isimip.org/).
We refer to these scenarios in the remainder of this manuscript
as
Had2.6, Had6.0 and Had8.5.
To assess the increase in biodiversity provision through forest
management, we used the biodiversity index of the German
Biodiversity Strategy, which is computed based on the
abundance of 10
forest birds, used as indicator species. We used the outcomes of
the
forest growth model to predict the abundance of these indicator
forest
birds under different management options. Biodiversity supply
was
then evaluated as a flow of benefits along the simulation period.
The
bird abundance data was collected using a point count protocol
with
three repetitions in 2017, which was used to fit an N-mixture
Bayesian
hierarchical model (Eq. (1)). In the model, the community
process was
modelled using a Bernoulli distribution, the abundance of the
species
was described by a zero-inflation Poisson distribution and the
detect-
ability was evaluated through a Binomial observation model (for
details
see Augustynczik et al., 2019). Moreover, we set the abundance

of
microhabitats to its average value, due to limitations on the
detail of
input data to estimate this parameter. To provide more robust
projec-
tions, we also reduced one standard deviation from the mean
estimate
of the parameter related to the conifer share, due to the high
sensitivity
and uncertainty related to this parameter.
= + + + +
+ + +
λ φ b b Slope b Altitude b BA b ConiferShare
b Dvol b NDead b TMHA
exp (
)
i i i i i i i
i i i
0 1 2 3 4
5 6 7 (1)
Where: λi: abundance of species (N/ha); ϕi: zero inflation
coeffi-
cient; Slope: plot slope (°); Altitude: plot altitude (m a.s.l.);
BA: plot
basal area (m2/ha); ConiferShare: share of conifers (%); Dvol:
dead-

wood volume (m³/ha); NDead: number of snags (N/ha); TMHA:
tree
microhabitat abundance (N/ha).
2.3. Costs of biodiversity provision and optimal management
under climate
change
Based on the forest responses obtained in each climate change
scenario, in terms of wood production and biodiversity, we
quantified
the costs of biodiversity provision using an optimization
approach. We
constructed a linear programming model to maximize forest
profit-
ability while increasing biodiversity provision thresholds (for
details
see Appendix A). Thereby, it was possible to establish a Pareto
frontier
between NPV and bird abundance and to derive the total cost
related to
this increase in biodiversity supply. The total cost was defined
by the
difference in NPV between the baseline scenario (maximum
NPV and
no biodiversity requirement) and scenarios including
biodiversity re-
quirements (increase in bird abundance).
2.4. Biodiversity benefits
To identify efficient biodiversity supply levels, besides the
compu-
tation of costs, it is necessary to quantify its social benefits.
Here we
considered solely the non-use value of forest biodiversity, i.e.

existence
and bequest values. The biodiversity benefits were established
using
data from an extensive choice experiment conducted by Weller
and
Elsasser (2017). The authors computed the WTP for an increase
in the
forest biodiversity index used in the German Biodiversity
Strategy. In
this experiment, the authors asked the participants to choose
preferred
landscape structures and corresponding contributions towards a
land-
scape fund, according to the landscape selected. The authors
subse-
quently used conditional logit models assuming homogeneous
pre-
ferences to estimate the WTP for an increase in forest
biodiversity,
measured through the biodiversity index. Specifically, the WTP
for an
increase in the biodiversity index in 5 and 25 points compared
to the
current levels was estimated. This corresponds to a 5% and 25%
in-
crease in the abundance of the 10 indicator species compared to
baseline (1970's abundance), respectively. We used these data
to
Management 246 (2019) 706–716
708
http://www.pik-potsdam.de/4c/
https://www.isimip.org/

calibrate a constant elasticity of substitution (CES) utility
function (Eq.
(2)) for wood and biodiversity. This function was subsequently
used to
derive the benefits of intermediary biodiversity supply levels.
Based on the CES utility function (Eq. (2)), we calculated the
WTP
per unit (% increase in bird indicator abundance) (Eq. (3)) and
the sum
of benefits for our research area, correcting the WTP of the
affected
population by the income in the region. To establish the level of
bio-
diversity benefits, they were discounted and aggregated along
the 80-
year planning horizon, using a 1.5% pure time preference rate,
fol-
lowing the HM treasury (Treasury, 2003). This rate expresses
the pre-
ference of agents for consuming now rather than in the future.
We
weighted the biodiversity supply in public forests according to
the total
forest area in the study region and finally we defined the
benefits based
on Eq. (3) and the level of biodiversity provision.
= ⎡
⎣
+ − ⎤
⎦

− − −U αw α B(1 )
θ
θ
θ
θ
θ
θ1 1 1
(2)
=
− −− −
− −( ) ( )
MB
B w α P
α
( 1)1 1
θ
θ
θ
θ
1 1
(3)
where: U: utility; MB: WTP for biodiversity (per unit); B:
biodiversity
supply level; w: wood consumption; θ : elasticity of substitution

be-
tween wood and biodiversity; α: preference parameter for wood
over
biodiversity.
2.5. A second-best mechanism for biodiversity provision
To handle the asymmetric information on the biodiversity valua-
tions, we applied a mechanism design approach. We considered
that
biodiversity is supplied as a single indivisible unit. The
government
then defines a discrete level of biodiversity to be supplied in
public
forests and designs a mechanism to levy funds to cover the costs
of the
biodiversity-oriented forest management. We assumed that the
agents
display quasilinear utilities and are risk neutral.
A mechanism can be defined as tripleμ A q p( , , ), consisting of
a set
of agents' strategiesA, a social choice function q and a payment
rulep.
Let =i N1, ..., be agents that hold private information on their
pre-
ferences for biodiversity, which is defined by their type θi(here
the type
expresses the WTP of an agent). The types are independently
drawn
from the same probability distribution F x( ) with density
function f x( ),
which is assumed to have a monotone hazard rate (the ratio
−
f x

F x
( )
1 ( )
is
monotone increasing in x). The prior information on the
distribution of
types is common knowledge. Moreover, the agents know the
realization
of their own typeθi, but do not know the types of other agents.
The
government, as a designer, selects a social choice function and a
pay-
ment rule, so that the social choice function maps the vector of
types in
the decision of supplying biodiversity in publicly-owned
forestland
→q θ: [0,1] and the payment rule maps the vector of types i nto
the
N . Since biodi-
versity is a public good, the government can only enforce agents
to
contribute towards the cost of biodiversity supply by
threatening not to
implement biodiversity-oriented forest management, in case the
levied
agents’ contributions are not sufficient. It can be shown that
given some
conditions, the mechanism design problem can be simplified to
the
selection of the social choice function q (see details in

supplementary I).
In addition to the above mentioned assumptions, further
constraints
need to be included in the mechanism proposed. Specifically,
agents
need to derive non-negative benefits when participating in the
me-
chanism (in the interim phase), referred to as interim indi vidual
ra-
tionality constraint. Additionally, the mechanism needs to be
Bayes-
Nash incentive compatible, i.e. in expectation, reporting their
true types
is a dominant strategy for the agents (agents cannot be better -
off by
misrepresenting their type). Finally, the mechanism needs to be
ex-ante
budget balanced, meaning that in expectation the funds levied
by the
mechanism need to cover the cost of implementation of
biodiversity-
oriented management.
In a first-best mechanism, biodiversity is provided whenever the
sum of benefits is greater than the cost. This mechanism,
however,
imposes a loss on the supplier, due to the asymmetric
information on
the biodiversity valuations (Güth and Hellwig, 1986; Börgers,
2015).
This requires the application of a second-best mechanism,
which un-
dersupplies biodiversity but is implementable without external
funding.

Güth and Hellwig (1986) propose such mechanism, where the
designer
selects a choice function that maximizes social welfare, under
the
condition that no external funding is needed to cover the im-
plementation costs. This mechanism can be described by the
max-
imization of Eq. (4) (for a description of the parameters and
functions
see Table 1). In this setting, the sum of biodiversity valuations
need to
cover the cost of implementation, plus a fraction of the sum of
virtual
valuations. The virtual valuations appear in the second term of
Eq. (4)
and are the maximum surplus that the designer can extract from
the
agents in the mechanism.
∫
∫
∑
∑ ⎜ ⎟
= ⎛
⎝
⎜ −
⎞
⎠
⎟

+ ⎡
⎣
⎢
⎛
⎝
−
− ⎞
⎠
− ⎤
⎦
⎥
∈
∈
MaxZ q θ θ c dF θ
λ q θ θ
F θ
f θ
c dF θ
( ) ( )
( )
1 ( )
( )

( )
i N
i
S
i N
i
i
i (4)
When many agents are involved, valuations are reduced, due to
the
fact that each agent has a decreasing influence on the final
decision and
expects that the public good will be provided anyways. Hellwig
(2003)
shows that the valuation of an agent is corrected by the
probability that
the agent is focal to the decision of supplying the public good.
This
probability approximates N1 and the expected revenue of the
me-
chanism is proportional to Eq. (5):
∑ ⎜ ⎟= ⎛
⎝
−
− ⎞
⎠∈
R θ

N
θ
F θ
f θ
( )
1 1 ( )
( )i N
i
i
i (5)
(Hellwig, 2003).
Here we adopted a numeric optimization approach to solve the
mechanism design problem. If we replace the continuous type
space of
of 0 with
probably 0, the optimization problem of the mechanism designer
ad-
mits a Linear Programming (LP) representation (Vohra, 2012).
Using
this framework, we constructed an optimization model to find a
second-
best mechanism, as proposed by Güth and Hellwig (1986). We
used an
Integer Linear Program, based on the optimization approach
introduced
by Csapó and Müller (2013). Here we aimed to find a second-
best

mechanism that maximizes social surplus, while respecting ex-
ante
budget balance, interim individual rationality and Bayes-Nash
incentive
compatibility:
∑ ∑= ⎛
⎝
⎜ −
⎞
⎠
⎟
∈ ∈
MaxZ π q
N
θ c
1
θ Θ
θ θ
i N
i
(6)
Table 1
Description of parameters and functions used in the second-best
mechanism.
Parameter/Function Description

N Number of agents
θi Type of agent i
θ Profile of agents' types
q θ( ) Social choice function
c Cost of provision
F θ( ) Cumulative probability density of agents' types
λS Lagrange multiplier of the ex-ante budget balance
constraint
f θ( ) Probability density function of agents' types
R θ( ) Sum of virtual valuations of profile θ
Management 246 (2019) 706–716
709
∑ ∑− ≥ ∀ ∈ ∀ ′ ∈ > ′
∈ ′∈
′ ′π q π q i N θ θ Θ θ θ0 , , |
θ Θ
θ θ
θ Θ
θ θ i ii i
(7)
∑ − ≥
∈

π q R θ c( ( ) ) 0
θ Θ
θ θ
(8)
∈ ∀ ∈ q θ Θ{0,1}θ (9)
The objective function (Eq. (6)) maximizes the expected social
surplus, given by the sum of valuations minus the cost of im-
plementation, over the type space. Constraint (Eq. (7)) ensures
Bayes-
Nash incentive compatibility and interim individual rationality,
enfor-
cing monotonicity of the choice function in respect to the
agent's type.
Interim individual rationality constraints enforce that agents
receive a
nonnegative utility for participating. Bayes-Nash incentive
compat-
ibility ensures that agents have no incentives to misrepresent
their
types. Constraint (Eq. (8)) enforces ex-ante budget balance,
which re-
quires that, in expectation, the costs of biodiversity provision
are cov-
ered by the contributions of the agents and (Eq. (9)) ensures
that the
choice function takes binary values. For a description of sets,
variables
and data used in the optimization models see Table 2.
To build our optimization problem, we assumed that agents’ va-
luations were composed by 5 types =Θ {1,2,3,4,5} that were
uniformly
distributed inside the confidence interval of biodiversity

valuations
established in section 2.4. Moreover, we considered 6
agents =N {1,2,3,4,5,6}, each representing an administrative
region with
an equal share of the population (negotiating on its behalf) that
need to
agree on the implementation of biodiversity conservation
policies in
public forests. Each agent also needs to contribute towards the
cost of
the increased biodiversity supply, representing fund transfers in
a fiscal
federalism framework (Bönke et al., 2013). Finally, we analyzed
the
implementation of 6 levels of bird indicator abundance in-
crease =B {2.5%, 5%, 7.5%, 10%, 15%, 17.5%}. Besides the
uncertainty
regarding the realization of the vector of valuations of the
agents, the
government must also consider that the costs of biodiversity-
oriented
management are uncertain and contained in =C c c c{ , , }Had
Had Had2.6 6.0 8.5 .
In our analysis, we investigated the social choice on the
expected cost
= + +c c c c( )/3Had Had Had2.6 6.0 8.5 and on each climate
scenario as a sen-
sitivity analysis (Barbieri and Malueg, 2014). We disconsidered
the
trivial cases, where biodiversity should never be provided (the
costs are
lower than the sum of valuations if all agents have the lowest
type) and
never be provided (the costs are higher than the sum of
valuations if all

agents have the highest type). We solved the optimization model
using
the software Gurobi8.1
(http://www.gurobi.com/products/gurobi-
optimizer).
3. Results
3.1. Costs of biodiversity provision under climate change
We perceived that up to a 10% increase in the current bird
indicator
abundance at the end of the century, the opportunity costs
increased
almost linearly with the biodiversity requirements, whereas for
abundance increases above this threshold, it was necessary to
strongly
compromise forest profitability. This behavior is depicted in the
cost
curves (Fig. 1), where we noticed a sharp opportunity cost
increase for
high levels of bird abundance. This was a result of the limits on
the
conversion of highly profitable spruce by beech stands (with
lower
growth rates and wood value) for increasing the share of
broadleaved
forests in the region and the need to reduce the area of more
profitable
management strategies.
Climate change had important implications for the total cost of
biodiversity provision. The increase in forest growth rates under
higher
atmospheric CO2 concentration, in combination with sufficient

pre-
cipitation, led to an increase in forest growth rates and
consequently
higher forest profitability and opportunity cost. This was
determinant
for the attainable level of biodiversity supply under the
mechanism
design approach, since the total supply cost was required to be
met by
the contributions of the agents considered in our analysis.
In addition to climate impacts, the cost behavior was also
related to
the biodiversity responses to forest management in our model.
We used
the N-mixture model described in section 2.2 to estimate the
bird
abundance under novel forest structures generated by the
alternative
management regimes applied in our analysis. Three main
parameters
used to estimate bird abundance were affected by management:
the
basal area of the stand, the share of conifers and the number of
snags.
Among these parameters, the bird assemblage was most
responsive to
the share of conifers and responded marginally to the number of
snags
and the basal area of the stand. Given that the increase in the
snag
number has important economic implications due to the
reduction in
thinning revenues, the increase in the share of broadleaves was
the
most cost-effective management practice to increase

biodiversity supply
and reach the levels required by the mechanism. This
management
action, however, also reduced forest profitability due to
conversion
costs.
3.2. Second-best mechanism for biodiversity provision
The first step taken in the analysis of the mechanism was the
identification of the trivial cases, in which biodiversity is never
pro-
vided or always provided (the social choice function always
equal to 1
or 0 regardless of the profile realization), based on the lowest
and
highest possible sum of valuations. For the average cost
scenario, the
non-trivial case yielded a bird abundance increase of 12.5% at
the end
of the century (Fig. 2) and bird abundance increases below this
value
were nearly always provided.
The optimal choice function, i.e. the threshold related to the
sum of
Table 2
Sets, variables and input data used in the mechanism design
model.
Sets Description
Θ Set of profile realizations
N Set of agents
Variables

qθ Binary variable that takes value 1 if profile θ is included in
the solution
and value 0 otherwise
Data
θ Profile of agents' types
πθ Probability of observing profile θ
c Opportunity cost of biodiversity provision
R θ( ) Sum of virtual valuations of profile θ
Fig. 1. The figure shows the total opportunity cost for
increasing bird abun-
dance levels at the end of the century for each climate change
scenario.
Management 246 (2019) 706–716
710
http://www.gurobi.com/products/gurobi-optimizer
http://www.gurobi.com/products/gurobi-optimizer
valuations that would lead to the implementation of
biodiversity-or-
iented management is depicted in Fig. 2. As expected, the
second-best
mechanism undersupplied biodiversity. We perceive that the
social
choice function was only activated if the sum of valuations
surpassed 89
Million EUR, whereas the first-best mechanism would
implement bio-
diversity-oriented management for valuations above 78 Million
EUR.

Thus, the sum of valuations was required to exceed the cost of
im-
plementation by more than 14%, taking into account the agents
de-
scribed in our model. This was a result of the information rent
held by
the agents on their valuations. Under these conditions, the
probability
of implementation of biodiversity-oriented management was
24%, i.e.
in 24% of profile realizations the social choice function would
be ac-
tivated. Additionally, to maintain Bayes-Nash incentive
compatibility
and interim individual rationality, the payment rule would
require
agents with types 1, 2, 3, 4 and 5 to contribute with an
equivalent of
100, 90, 86, 83 and 81 % of their WTP, respectively. We
perceived that
biodiversity-oriented management was mainly implemented for
profiles
with a combination of high biodiversity valuations (e.g. types 4
and 5,
with the two highest biodiversity valuations according to the
distribu-
tion used in our analysis). This requirement yielded a reduced
prob-
ability of implementation, since all agents displayed
simultaneously
high valuations in a limited number of profile realizations. On
the other
hand, if the first-best solution was considered and external
funding was
feasible, the probability of implementation would increase to
84%, due

to the lower threshold of implementation.
In our model, we required the mechanism to be budget balanced
in
expectation. This condition, however, does not guarantee that
the funds
raised will cover the costs of implementation in all profile
realizations.
This also caused an undersupply compared to the first-best
solution,
which affected the expected surplus of the mechanism. The
first-best
case, disregarding the budget balance condition, would yield an
ex-
pected surplus of 7.5 Million EUR for the agents, whereas for
the
second-best case this figure amounted to 3.8 Million EUR. We
highlight
that, despite the lower expected surplus, this mechanism still
produces
a larger social benefit than the profit maximization mechanism
(ex-
pected surplus of 1.9 Million EUR).
Although the optimization models generated through the me-
chanism design model had high dimensionality, with 15.625
binary
variables and 165.625 non-zeros, the optimal solution could be
effi-
ciently computed, with processing time inferior to 1 s. We
emphasize
that the problem size may become computatio nally prohibitive
when
the number of types and agents is large, since the number of
profile

realizations is proportional to T N . In such cases, heuristic
solutions
may be required to compute the optimal mechanism.
3.3. Sensitivity analysis and management solutions
Climate change had a substantial effect on the choice function
due
to the varying implementation cost (Fig. 3). For the Had2.6 and
the
Had6.0 climate scenarios, the same level of bird abundance
increase
was observed (12.5%). Nevertheless, for the Had2.6 scenario,
the lower
opportunity cost led to a probability of implementation equal to
66%,
whereas for the Had6.0 it reduced to 16%. Similar to the
average cost
scenario, a higher probability of implementation was observed
for the
first-best case (> 99% for the Had2.6 and 76% for the Had6.0
sce-
nario). Considering the Had8.5 scenario, the increase in bird
abundance
amounted to 10% at the end of the century, with a probability of
im-
plementation of 76% in the second-best case. Under such
conditions,
biodiversity-oriented management would be implemented if the
sum of
the valuations surpassed 76 Million EUR, whereas in the fist-
best so-
lution the social choice function would be activated for profiles
with
valuation above 63 Million EUR.

The optimal portfolio for increasing levels of bird abundance in
each
climate scenario is shown in Fig. 4. We observed that the
increase in
bird abundance requirements caused a reduction in the area
under BAU
and biomass-oriented management, whereas the biodiversity
manage-
ment strategy largely increased. Hence, depending on the costs
of
biodiversity supply, the allocation related to the second-best
me-
chanism differed. For example, the Had2.6 scenario would
require the
biodiversity strategy in approximately 28% of the total area,
whereas
for the Had8.5 climate scenario, the biodiversity strategy would
be
reduced to 21% of the total area. Additionally, for a same level
of bird
abundance increase, climate change required tailored
management re-
gimes. The optimal portfolio under the Had2.6 scenario applied
the no
management strategy in 6% of the total area, whereas the same
figure
was reduced to 2% in the Had6.0 scenario. In general, more
intense
climate change led to a reduction in the area under no
management and
an increase in the area of the biomass production strategy.
4. Discussion
Here we analyzed how the government may implement biodi-
versity-oriented forest management in public forests, in order to

reduce
the gap between efficient and current levels of forest
biodiversity in
temperate ecosystems. We computed the costs of biodiversity
provision
and applied a mechanism design approach to account for the
strategic
behavior of agents related to the contribution towards this cost.
We
defined social choice functions for biodiversity supply in public
forests
under climate change and computed optimal management
solutions to
realize the required biodiversity indicator increase.
4.1. Costs of biodiversity provision under climate change
The total costs of biodiversity provision were moderate with up
to a
10% increase in bird abundance at the end of the century,
ranging
approximately between 892 and 1180 EUR/ha, whereas for the
max-
imum biodiversity provision within our modelling framework
increased
by up to 4346 EUR/ha. Since the conversion to broadleaved
forests is
bounded by the current area of Norway spruce, it was necessary
to
increase the abundance through less efficient management
interven-
tions and increasing the opportunity cost, e.g. applying
management
with low thinning intensity to increase mortality and snags
availability.
Rosenkranz et al. (2014) evaluated the implementation costs of

the
Habitats Directive in Germany and reported average loss of
income
Fig. 2. Social choice function value for the average cost across
the climate
scenarios. BAI stands for the level of bird abundance increase
in the non-trivial
provision level. The dotted vertical line shows the opportunity
cost for the
corresponding increase in bird abundance.
Management 246 (2019) 706–716
711
ranging from 1958 to 2496 EUR/ha, depending on the
management
applied and discounted with a 1.5% interest rate. Hily et al.
(2015)
analyzed the cost effectiveness of Natura 2000 contracts in
France, with
an average cost of contracts approaching 1900 EUR/ha.
Climate change and its implications to forest dynamics were
also
important drivers of the costs of biodiversity provision and,
thus, cas-
caded to the mechanism implementation. Climate scenarios with
in-
creased forest productivity showed higher opportunity cost,
since the
profitability of conifer stands increased. An important aspect of

forest
development under climate change not investigated here refers
to the
occurrence of forest disturbances. Disturbances may modify
forest
profitability and interact with forest biodiversity, altering
conservation
costs (Hanewinkel et al., 2013; Seidl et al., 2017), and affecting
the
probability of biodiversity supply. In this sense, a closer
investigation of
disturbances under climate change and its effects on forest
biodiversity
and profitability is encouraged.
A dynamic updating on possible climate realizations and on the
social value for forest biodiversity will help to reduce the range
of costs
and benefits, improving the efficiency of biodiversity-oriented
forest
management when new information becomes available.
Adaptive forest
management in combination with Bayesian updating provide a
natural
framework to dynamically update forest strategic plans in the
face of
new information and may be employed to tackle this issue (e.g.
Yousefpour et al., 2013). Such information may help to identify
not
only optimal conservation actions, but also identify the optimal
timing
for its implementation and avoid that thresholds related to
ecosystem
functioning are surpassed.

4.2. Second-best mechanism for biodiversity provision
The expected agents' surplus in the second-best mechanism
design
approach, as expected, was inferior to the first-best, where
biodiversity
is provided whenever the sum of benefits surpasses the costs of
im-
plementation. This occurs due to the ex-ante budget balance
require-
ment, ensuring that no external funding is needed for an
increase in
biodiversity supply. The social choice function required, in
general, that
valuations surpassed the costs by more than 18%. Yet, under
voluntary
participation, the second-best mechanism provides a powerful
frame-
work for the provision of forest biodiversity, when external
funding is
undesirable. Through this approach it is possible to derive not
only the
thresholds for biodiversity provision, but to attach probabilities
of
success, once the distribution of valuations is known.
Fig. 3. Sensitivity of the social choice function according to the
climate trajectory. BAI stands for the level of bird abundance
increase in the non-trivial provision
levels. The dotted vertical line shows the opportunity cost for
the corresponding increase in bird abundance.
Management 246 (2019) 706–716
712

The mechanism considered here refers to the case where the
gov-
ernment acts to maximize social welfare and does not have any
addi-
tional constraints apart from the budget balance. The
mechanism de-
signer, however, may have different goals. A large body of
literature is
dedicated to the supply of public goods when the designer has
the
objective of maximizing profits (e.g. Csapó and Müller, 2013),
where
the public good is only provided if the sum of virtual valuations
cover
the costs of implementation. Moreover, the government may
have ad-
ditional budget targets and minimum amounts of funds to be
raised that
would modify the mechanism. The formulation of the
mechanism de-
sign problem as an integer linear model allows to seamless
integrate
such additional requirements in the decision-making process,
e.g. by
adding extra constraints (Vohra, 2012). This may provide
valuable in-
formation when closed-form solutions for the models are not
readily
available.
The weight placed on the sum of virtual valuations decreased
when
the costs of biodiversity-oriented management approached the

upper
limit of the sum of valuations, approximating the first-best
mechanism.
This was accompanied, however, by a substantial decrease in
the
probability of implementation, since biodiversity-oriented
management
was only applied if the profile of valuations was composed by
the
highest types. Börgers (2015) shows this behavior of the
thresholds for
the second-best mechanism considering the supply of a public
good, in
which the threshold approaches the first-best criteria for costs
near
upper bound of valuations. In our analysis, when the cost was
close to
the maximum sum of valuations in the average cost and Had2.6
sce-
narios, there was an approximation to the first-best threshold.
In our study, we computed the optimal choice function for the
non-
trivial cases, considering the implementation cost in the average
case
and in each climate change scenario. One may consider the case
where
the designer wishes to guarantee the performance of the
mechanism in
the worst-case scenario. This would require that regardless of
the
climate realization, the feasibility of the mechanism is
preserved.
Hence, one may consider robustness criteria, e.g. by designing
the

mechanism based on the highest possible cost, so that in any
climate
realization the expected revenue is higher than the cost of im-
plementation (Had8.5 in our analysis). The topic of robust
mechanism
design is currently an area of active research. For example,
Bandi and
Bertsimas (2014) formulate an auction problem using robust
optimi-
zation, in which uncertainty sets are used instead of probability
dis-
tributions to characterize the agents’ valuations. Koçyiğit et al.
(2018)
developed an integer linear programming model for auction
design in
the case where the seller is ambiguity-averse. Such analyses
may im-
prove the success of mechanism under deep uncertain settings.
We investigated the application of a direct second-best
mechanism,
where agents announce their valuations and contribute towards
the
costs of biodiversity provision. There are a number of
alternative me-
chanisms dealing with the supply of public goods described in
the lit-
erature. For example, Bierbrauer and Sahm (2008) investigate
demo-
cratic mechanisms, where taxes are introduced to finance the
public
good provision and participants vote to express their
preferences for the
public good supply. Van Essen and Walker (2017) note that
theoretical
optimal mechanisms, did not always produce the desired

outcomes in
experimental studies and propose a simple market-like
mechanism that
always yield a feasible allocation. In their mechanism, the
contribution
of the participants is given by the per capita cost of provision
corrected
by the individual valuation compared to the average valuation.
Such
experimental evaluation of mechanisms for biodiversity
provision are
still scarce in the literature and deserve further investigation.
4.3. Optimal forest management
In order to provide biodiversity at a minimum cost, it was
necessary
to apply tailored management strategies to the set of forest
stands in our
Fig. 4. Optimal management portfolio for each climate scenario
under increasing levels of bird abundance at the end of the
century.
Management 246 (2019) 706–716
713
study area. A combination of management practices will be
required in
the future to balance the provision of products and ecosystem
services
in temperate forests. Gutsch et al. (2018) also show that forests

in
different regions show potential to fulfill optimally different
ecosystem
services in Germany under climate change, according to its
structure
and species composition. Naumov et al. (2018) report similar
patterns
studying forest landscapes in Northern Europe. In this context,
den
Herder et al. (2017) propose a framework for balancing the
provision of
forest goods and services, including economic, environmental
and so-
cial indicators using multi-criteria analysis. Hence, a sound
landscape
management will ask for the consideration of local forest
conditions on
the strategic planning of forest use, allowing to achieve the
desired
goals more efficiently, in terms the provision of multiple
ecosystem
goods and services.
Our solutions indicate that the conversion of spruce stands to
broadleaved forests was the most efficient practice to increase
biodi-
versity provision, measured through the abundance of bird
indicator
species. We highlight here that the indicator species had a
similar re-
sponse to the management actions considered. It is important to
con-
sider, however, that other taxa may have different requirements
re-
garding forest habitats. Particularly, saproxylic organisms
require old-

growth forest attributes, such as deadwood and habitat trees and
con-
nectivity among habitats at a finer scale (Müller et al., 2016;
Thomaes
et al., 2018). These aspects deserve to be further investigated
and both
spatial planning models and benefit assessments regarding these
taxa
are needed.
4.4. Limitations
We conducted our analysis based on the age class of each stand.
If
forest inventory data is available for each stand in the forest
area, we
may increase the accuracy of forest production forecasts and
tailor
management prescriptions to the specific stand structure.
Moreover, an
important aspect of forest dynamics under climate change refers
to the
occurrence of disturbances and how these interact with forest
pro-
ductivity and forest taxa (Hanewinkel et al., 2013; Greenville et
al.,
2018). A coupling of forest growth, disturbance and population
dy-
namics models are recommended for future studies.
We restricted our analysis to a limited set of management
options,
agents and types. Our framework, however, can be easily
extended to
encompass a larger set of management options and valuations to
pro-

vide more accurate estimates. These need to be balanced with
the
problem size generated, especially regarding the number of
types and
agents, as the possible combinations increase exponentially and
the
resulting matrix of the optimization problem has a large number
of non-
zeros.
We have included here relevant uncertainty aspects at the
strategic
planning level, with a focus on climate change and biodiversity
va-
luations. In this sense, we did not consider here all the relevant
sources
of uncertainty to the supply of biodiversity in public forests.
The un-
certainty in the biodiversity responses to management practices
may
significantly affect the operationalization of conservation
actions. This
uncertainty will have stronger influence with an increase in the
sensi-
tivity of the model and larger standard deviation of the
management-
related parameters. For example, in our analysis, the share of
conifers
showed the largest influence on the summed abundance,
compared to
other management actions. Thus, this uncertainty may affect the
total
cost of conservation (increasing the cost if the observed
response had
lower magnitude or decreasing the cost otherwise). Similarly,

un-
certainty in economic parameters (e.g. wood price and interest
rate)
may affect the opportunity costs of an increased biodiversity
supply and
deserve further investigation.
Here we considered independent valuations for forest
biodiversity.
The framework proposed by Csapó and Müller (2013) allows for
the
relaxation of this condition and is easily adaptable to our study.
The
authors accommodate dependent valuations by modifying the
virtual
valuation of agents, according to the joint probability
distribution of the
dependent random variables.
5. Conclusions
Biodiversity conservation remains a complex and important
forest
management problem. The coupling of ecological and economic
models
is key to find efficient conservation solutions and correct the
provision
of forest biodiversity, aiming to create resilient forest
landscapes.
Mechanism design offers a powerful framework to account for
the
strategic behavior of agents towards the public good provision
and
provides information on the conditions for a successful
implementation
of conservation programs. This will ultimately depend on the

re-
lationship between the social value and costs for providing
forest bio-
diversity, as well as the capacity of the government to levy
funds to
finance an increased biodiversity supply. The creation of s uch
me-
chanisms will be key to maintain the provision of multi-
functionality of
temperate forests in the face of climate change.
Acknowledgements
We acknowledge the funding of this research to the German
Research Foundation, ConFoBi project (number GRK 2123).
Supplementary data
Supplementary data to this article can be found online at
https://doi.org/10.1016/j.jenvman.2019.05.089.
Appendix A. Forest optimization model.
A description of sets, data and variables used in the
optimization model is provided in Table A1 hereafter.
=MaxZ NPV (A1)
≤ ∑ ∑ ∑ ∑
+ ∑ ∑ ∑
− ∑ ∑ ∑ − ∑ ∑ ∑
− ∑

∈ ∈ ∈ ∈ +
∈ ∈ ∈ +
∈ ∈ ∈ ∈ ∈ ∈ +
∈ +
NPV vol x price
volfin x price
volini x price planting x
fixed
i S j M t PH k T ijtk ij tk ir
i S j M k T ijk ij PHk ir
i S j M k T ijk ij k i S j M t PH ijt ij ir
t PH t ir
1
(1 )
1
(1 )
1
1
(1 )
1

(1 )
t
PH
t
t (A2)
∑ ∑ ≥
∈ ∈
bio x Biodiversity
i S j M
ijPH ij
(A3)
Management 246 (2019) 706–716
714
∑ ∑ ∑ ≥ ∀ ∈
∈ ∈ ∈
vol x b t PH0.7
i S j M k T
ijtk ij
(A4)

∑ ∑ ∑ ≤ ∀ ∈
∈ ∈ ∈
vol x b t PH1.3
i S j M k T
ijtk ij
(A5)
∑ ∑ ∑ ∑ ∑ ∑≤
∈ ∈ ∈ ∈ ∈ ∈
volini x volfin x
i S j M k T
ijk ij
i S j M k T
ijk ij
(A6)
∑ ≤ ∀ ∈
∈
x area i S
j M
ij i
(A7)
∑ ≤
∈
x totarea0.5
i S

i1
(A8)
Table A1
Sets, variables and input data used in the forest optimization
model.
Sets Description
S Set of stands
M Set of management regimes
PH Set of periods
T Set of tree species
Variables
NPV Total NPV of forest management
xij Area of stand i to be management under regime j
b Wood production bound
Data
volijtk Volume of species k produced in period t in stand i
under management j
pricetk Price of species k in period t
ir Interest rate
volfinijk Final volume of species k in stand i under management
j
voliniijk Initial volume of species k in stand i under
management j
plantingijt Planting cost of stand i under management j in
period t
fixedt Fixed cost in period t
bioijt Bird indicator abundance in stand i under management j
in period t
Biodiversity Total bird indicator abundance bound
areai Area of stand i
totarea Total forest area

The objective function (Eq. (A1)) targets the maximization of
forest NPV. Constraint (Eq. (A2)) assigns to the variable NPV
the total NPV of the
forest investment along the planning horizon, computed trough
the discounted sum of thinning revenues, the difference in
standing stock value at the
beginning and at the end of the simulation period, the planting
costs and administering costs. Constraint (Eq. (A3)) requires
that the bird indicator
abundance at the end of the period is higher than the
boundBiodiversity. Constraints (Eq. (A4)) and (Eq. (A5)) are
wood flow constraints (Bettinger
et al., 2016) and enforce that the harvested volume in every
period respects a ± 30% variation compared to the endogenously
determined volume
boundb. The bound b was a free variable in the optimization
model, enabling to achieve the highest NPV while maintaining
the wood flow stability.
Constraint (Eq. (A6)) requires that the standing volume at the
end of the simulation period to be at least equal to the standing
volume at the
beginning of the simulation period, i.e. a sustainability criteria
regarding the forest utilization rate. Constraint (Eq. (A7))
guarantees that the
managed area of each stand is bounded by the stands’ total area.
Constraint (Eq. (A8)) requires that the conversion of Norway
spruce to European
beech stands do not extend over the 50% of the total forest area,
which is the forest cover of Norway spruce in the study region.
We constructed the optimization model in Lingo 17.0 optimizer
(https://www.lindo.com), solving it multiple times, increasing
the required level
of bird abundance (Biodiversity), and establishing the efficient
frontier between NPV and biodiversity. The cost for
biodiversity provision was

subsequently calculated based on the NPV loss to attain an
increased bird indicator abundance, compared to the maximum
attainable NPV. The
solution process to the optimization problem described by Eq.
(A1) to Eq. (A8) was obtained in 7 min and 30 s for each bird
abundance level enforce
by constraint Eq. (A3).
References
Augustynczik, A.L.D., Asbeck, T., Basile, M., Bauhus, J.,
Storch, I., Mikusiński, G., et al.,
2019. Diversification of forest management regimes secures tree
microhabitats and
bird abundance under climate change. Sci. Total Environ. 650,
2717–2730.
Bandi, C., Bertsimas, D., 2014. Optimal design for multi-item
auctions: a robust optimi-
zation approach. Math. Oper. Res. 39 (4), 1012–1038.
Barbieri, S., Malueg, D.A., 2014. Increasing fundraising success
by decreasing donor
choice. J. Public Econ. Theory 16 (3), 372–400.
Bettinger, P., Boston, K., Siry, J.P., Grebner, D.L., 2016. Forest
Management and
Planning. Academic press.
Bierbrauer, F.J., Hellwig, M.F., 2016. Robustly coalition-proof
incentive mechanisms for
public good provision are voting mechanisms and vice versa.
Rev. Econ. Stud. 83 (4),
1440–1464.

Bierbrauer, Felix, Sahm, Marco, 2008. Optimal Democratic
Mechanisms for Taxation and
Public Good Provision, Preprints of the Max Planck Institute for
Research on
Collective Goods, No. 2008, vol. 9 Max Planck Inst. for
Research on Collective Goods,
Bonn.
BMUB, 2015. Indikatorenbericht 2014. Available at :
https://www.bmu.de/fileadmin/
Daten_BMU/Pools/Broschueren/indikatorenbericht_biologische
_vielfalt_2014_bf.pdf,
Accessed date: 20 September 2014.
Bönke, T., Jochimsen, B., Schroeder, C., 2013. Fiscal
Federalism and Tax Administration–
Evidence from Germany. DIW Berlin Discussion Paper No.
1307.
Börgers, T., 2015. An Introduction to the Theory of Mechanism
Design. Oxford University
Press, USA.
Management 246 (2019) 706–716
715
https://www.lindo.com
http://refhub.elsevier.com/S0301-4797(19)30715-7/sref1

https://www.bmu.de/fileadmin/Daten_BMU/Pools/Broschueren/i
ndikatorenbericht_biologische_vielfalt_2014_bf.pdf
https://www.bmu.de/fileadmin/Daten_BMU/Pools/Broschueren/i
ndikatorenbericht_biologische_vielfalt_2014_bf.pdf
Statistiches Bundesamt, 2016. Statistiken zur Einwohnerzahl in
Deutschland. Available
at : https://www.tagesschau.de/inland/einwohnerzahl-
deutschland-107.html,
Accessed date: 25 July 2018.
Csapó, G., Müller, R., 2013. Optimal mechanism design for the
private supply of a public
good. Games Econ. Behav. 80, 229–242.
den Herder, M., Kurttila, M., Leskinen, P., Lindner, M.,
Haatanen, A., Sironen, S., et al.,
2017. Is enhanced biodiversity protection conflicting w ith
ambitious bioenergy tar-
gets in eastern Finland? J. Environ. Manag. 187, 54–62.
Díaz, S., Fargione, J., Chapin III, F.S., Tilman, D., 2006.

Biodiversity loss threatens human
well-being. PLoS Biol. 4 (8), e277.
ECB (European Central Bank), 2019. Statistical data warehouse.
Available on. https://
sdw.ecb.europa.eu/browse.do?node=9691124, Accessed date: 26
April 2019.
Fabrika, M., 2007. Simulátor Biodynamiky Lesa SIBYLA,
Koncepcia, Konštrukcia a
Programové Riešenie, Habilitačná Práca. Technická univerzita
vo Zvolene, pp. 328.
Getzner, M., Meyerhoff, J., Schläpfer, F., 2018. Willingness to
pay for nature conservation
policies in state-owned forests: an Austrian case study. Forests
9 (9), 537.
Greenville, A.C., Burns, E., Dickman, C.R., Keith, D.A.,
Lindenmayer, D.B., Morgan, J.W.,
et al., 2018. Biodiversity responds to increasing climatic
extremes in a biome-specific
manner. Sci. Total Environ. 634, 382–393.
Grüner, H.P., Koriyama, Y., 2012. Public goods, participation
constraints, and democracy:
a possibility theorem. Games Econ. Behav. 75 (1), 152–167.
Güth, W., Hellwig, M., 1986. The private supply of a public
good. J. Econ. 46 (1),
121–159.
Gutsch, M., Lasch-Born, P., Kollas, C., Suckow, F., Reyer,
C.P., 2018. Balancing trade-offs
between ecosystem services in Germany’s forests under climate
change. Environ. Res.

Lett. 13 (4), 045012.
Hanewinkel, M., Cullmann, D.A., Schelhaas, M.J., Nabuurs,
G.J., Zimmermann, N.E.,
2013. Climate change may cause severe loss in the economic
value of European forest
land. Nat. Clim. Change 3 (3), 203.
Hartig, F., Drechsler, M., 2010. Stay by thy neighbor? Social
organization determines the
efficiency of biodiversity markets with spatial incentives. Ecol.
Complex. 7 (1),
91–99.
Härtl, F., Hahn, A., Knoke, T., 2013. Risk-sensitive planning
support for forest enterprises:
the YAFO model. Comput. Electron. Agric. 94, 58–70.
Hellwig, M.F., 2003. Public-good provision with many
participants. Rev. Econ. Stud. 70
(3), 589–614.
Hily, E., Garcia, S., Stenger, A., Tu, G., 2015. Assessing the
cost-effectiveness of a bio-
diversity conservation policy: a bio-econometric analysis of
Natura 2000 contracts in
forest. Ecol. Econ. 119, 197–208.
Iranah, P., Lal, P., Wolde, B.T., Burli, P., 2018. Valuing visitor
access to forested areas and
exploring willingness to pay for forest conservation and
restoration finance: the case
of small island developing state of Mauritius. J. Environ.
Manag. 223, 868–877.
Isbell, F., Craven, D., Connolly, J., Loreau, M., Schmid, B.,

Beierkuhnlein, C., et al., 2015.
Biodiversity increases the resistance of ecosystem productivity
to climate extremes.
Nature 526 (7574), 574.
Kemkes, R.J., Farley, J., Koliba, C.J., 2010. Determining when
payments are an effective
policy approach to ecosystem service provision. Ecol. Econ. 69
(11), 2069–2074.
Koçyiğit, Ç., Bayrak, H.I., Pınar, M.Ç., 2018. Robust auction
design under multiple priors
by linear and integer programming. Ann. Oper. Res. 260 (1–2),
233–253.
Lasch, P., Badeck, F.-W., Lindner, M., Suckow, F., 2002.
Sensitivity of simulated forest
growth to changes in climate and atmospheric CO2. Forstwiss.
Cent.blatt 121 (Suppl.
1), 155–171.
Lasch, P., Badeck, F.W., Suckow, F., Lindner, M., Mohr, P.,
2005. Model-based analysis of
management alternatives at stand and regional level in
Brandenburg (Germany). For.
Ecol. Manag. 207 (1–2), 59–74.
Lindner, M., Fitzgerald, J.B., Zimmermann, N.E., Reyer, C.,
Delzon, S., van der Maaten,
E.,, et al., 2014. Climate change and European forests: what do
we know, what are
the uncertainties, and what are the implications for forest
management? J. Environ.
Manag. 146, 69–83.

Meyerhoff, J., Angeli, D., Hartje, V., 2012. Valuing the benefits
of implementing a na-
tional strategy on biological diversity—the case of Germany.
Environ. Sci. Policy 23,
109–119.
Müller, F., Hanewinkel, M., 2018. Challenging the assumptions
of a standard model: how
historical triggers in terms of technical innovations, labor costs
and timber price
change the land expectation value. For. Policy Econ. 95, 46–56.
Müller, J., Simon, T., Roland, B., Khosro, S.T., Barimani, H.V.,
Sebastian, S., et al., 2016.
Protecting the forests while allowing removal of damaged trees
may imperil sa-
proxylic insect biodiversity in the Hyrcanian beech forests of
Iran. Conserv. Lett. 9
(2), 106–113.
Naumov, V., Manton, M., Elbakidze, M., Rendenieks, Z.,
Priednieks, J., Uhlianets, S.,
et al., 2018. How to reconcile wood production and biodiversity
conservation? The
Pan-European boreal forest history gradient as an “experiment”.
J. Environ. Manag.
218, 1–13.
Nisan, N., Ronen, A., 2001. Algorithmic mechanism design.
Games Econ. Behav. 35 (1–2),
166–196.
Pretzsch, H., 1997. Analysis and modeling of spatial stand
structures. Methodological
considerations based on mixed beech-larch stands in Lower
Saxony. For. Ecol. Manag.

97 (3), 237–253.
Rands, M.R., Adams, W.M., Bennun, L., Butchart, S.H.,
Clements, A., Coomes, D., et al.,
2010. Biodiversity conservation: challenges beyond 2010.
Science 329 (5997),
1298–1303.
Roesch-McNally, G.E., Rabotyagov, S., Tyndall, J.C., Ettl, G.,
Tóth, S.F., 2016. Auctioning
the forest: a qualitative approach to exploring stakeholder
responses to bidding on
forest ecosystem services. Small-scale For. 15 (3), 321–333.
Rosenkranz, L., Seintsch, B., Wippel, B., Dieter, M., 2014.
Income losses due to the im-
plementation of the Habitats Directive in forests—conclusions
from a case study in
Germany. For. Policy Econ. 38, 207–218.
Seidl, R., Thom, D., Kautz, M., Martin-Benito, D., Peltoniemi,
M., Vacchiano, G., et al.,
2017. Forest disturbances under climate change. Nat. Clim.
Change 7 (6), 395.
Thomaes, A., Dhont, P., Dekeukeleire, D., Vandekerkhove, K.,
2018. Dispersal behaviour
of female stag beetles (Lucanus cervus) in a mosaic landscape:
when should I stay and
where should I go. Insect Conserv. Divers. 11 (6), 523–533.
Tilman, D., Isbell, F., Cowles, J.M., 2014. Biodiversity and
ecosystem functioning. Annu.
Rev. Ecol. Evol. Syst. 45, 471–493.
Treasury, 2003. The Green Book. Appraisal and Evaluation in

Central Government.
Van Essen, M., Walker, M., 2017. A simple market-like
allocation mechanism for public
goods. Games Econ. Behav. 101, 6–19.
Vohra, R.V., 2012. Optimization and mechanism design. Math.
Program. 134 (1),
283–303.
Weller, P., Elsasser, P., 2017. Ökonomische Bewertung der
kulturellen
Ökosystemleistungen des Waldes. In: Fick, J., Gömann, H.
(Eds.), Wechselwirkungen
zwischen Landnutzung und Klimawandel. Springer, Berlin
Heidelberg New York (in
press).
Yousefpour, R., Temperli, C., Bugmann, H., Elkin, C.,
Hanewinkel, M., Meilby, H., et al.,
2013. Updating beliefs and combining evidence in adaptive
forest management under
climate change: a case study of Norway spruce (Picea abies L.
Karst) in the Black
Forest, Germany. J. Environ. Manag. 122, 56–64.
Management 246 (2019) 706–716
716
https://www.tagesschau.de/inland/einwohnerzahl-deutschland-
107.html

https://sdw.ecb.europa.eu/browse.do?node=9691124
https://sdw.ecb.europa.eu/browse.do?node=9691124
http://refhub.elsevier.com/S0301-4797(19)30715-7/sr0005

http://refhub.elsevier.com/S0301-4797(19)30715-
7/sref51Climate change and the provision of biodiversity in
public temperate forests – A mechanism design approach for the
implementation of biodiversity conservation
policiesIntroductionMaterial and methodsDataForest simulation
and biodiversity supplyCosts of biodiversity provision and
optimal management under climate changeBiodiversity
benefitsA second-best mechanism for biodiversity

provisionResultsCosts of biodiversity provision under climate
changeSecond-best mechanism for biodiversity
provisionSensitivity analysis and management
solutionsDiscussionCosts of biodiversity provision under
climate changeSecond-best mechanism for biodiversity
provisionOptimal forest
managementLimitationsConclusionsAcknowledgementsSupplem
entary dataAppendix A. Forest optimization model.References
NR326 Mental Health Nursing
RUA: Scholarly Article Review Guidelines
NR326 RUA Scholarly Article Review Guideline 1
Purpose
The student will review, summarize, and critique a scholarly
article related to a mental health topic.
Course outcomes: This assignment enables the student to meet
the following course outcomes.
(CO 4) Utilize critical thinking skills in clinical decision-
making and implementation of the nursing process for
psychiatric/mental health clients. (PO 4)
(CO 5) Utilize available resources to meet self-identified goals
for personal, professional, and educational
development appropriate to the mental health setting. (PO 5)
(CO 7) Examine moral, ethical, legal, and professional
standards and principles as a basis for clinical decision-making.

(PO 6)
(CO 9) Utilize research findings as a basis for the development
of a group leadership experience. (PO 8)
Due date: Your faculty member will inform you when this
assignment is due. The Late Assignment Policy applies to
this assignment.
Total points possible: 100 points
Preparing the assignment
1) Follow these guidelines when completing this assignment.
Speak with your faculty member if you have questions.
a. Select a scholarly nursing or research article, published
within the last five years, related to mental health
nursing. The content of the article must relate to evidence-based
practice.
• You may need to evaluate several articles to find one that is
appropriate.
b. Ensure that no other member of your clinical group chooses
the same article, then submit your choice for
faculty approval.
c. The submitted assignment should be 2-3 pages in length,
excluding the title and reference pages.
2) Include the following sections (detailed criteria listed below
and in the Grading Rubric must match exactly).
a. Introduction (10 points/10%)
• Establishes purpose of the paper
• Captures attention of the reader
b. Article Summary (30 points/30%)
• Statistics to support significance of the topic to mental health

care
• Key points of the article
• Key evidence presented
• Examples of how the evidence can be incorporated into your
nursing practice
c. Article Critique (30 points/30%)
• Present strengths of the article
• Present weaknesses of the article
• Discuss if you would/would not recommend this article to a
colleague
d. Conclusion (15 points/15%)
• Provides analysis or synthesis of information within the body
of the text
• Supported by ides presented in the body of the paper
• Is clearly written
e. Article Selection and Approval (5 points/5%)
• Current (published in last 5 years)
• Relevant to mental health care
• Not used by another student within the clinical group
• Submitted and approved as directed by instructor
f. APA format and Writing Mechanics (10 points/10%)
2
NR326 RUA Scholarly Article Review Guideline 2

• Correct use of standard English grammar and sentence
structure
• No spelling or typographical errors
• Document includes title and reference pages
• Citations in the text and reference page
For writing assistance (APA, formatting, or grammar) visit the
APA Citation and Writing page in the online library.
Please note that your instructor may provide you with additional
assessments in any form to determine that you fully
understand the concepts learned in the review module.
NR326 RUA Scholarly Article Review Guideline 4 3
Grading Rubric Criteria are met when the student’s application
of knowledge demonstrates achievement of the outcomes for
this assignment.
Assignment Section and
Required Criteria
(Points possible/% of total points available)
Highest Level of
Performance

High Level of
Performance
Satisfactory Level
of Performance
Unsatisfactory
Level of
Performance
Section not
present in paper
Introduction
(10 points/10%)
10 points 8 points 0 points
Required criteria
1. Establishes purpose of the paper
2. Captures attention of the reader
Includes 2 requirements for section. Includes 1
requirement for
section.
No requirements for this section presented.
Article Summary
(30 points/30%)
30 points 25 points 24 points 11 points 0 points
Required criteria

1. Statistics to support significance of the topic to
mental health care
2. Key points of the article
3. Key evidence presented
4. Examples of how the evidence can be incorporated
into your nursing practice
Includes 4
requirements for
section.
Includes 3
requirements for
section.
Includes 2
requirements for
section.
Includes 1
requirement for
section.
No requirements for
this section
presented.
Article Critique
(30 points/30%)
30 points 25 points 11 points 0 points
Required criteria
1. Present strengths of the article

2. Present weaknesses of the article
3. Discuss if you would/would not recommend this
article to a colleague
requirements for
section.
Includes 1
requirement for
section.
No requirements for
this section
presented.
Conclusion
(15 points/15%)
15 points 11 points 6 points 0 points
1. Provides analysis or synthesis of information within
the body of the text
2. Supported by ides presented in the body of the paper
3. Is clearly written
requirements for
section.
Includes 1
requirement for
section.

No requirements for
this section
presented.
Article Selection and Approval
(5 points/5%)
1. Current (published in last 5 years) Includes 4 Includes 3
Includes 2 Includes 1 No requirements for
NR326 RUA Scholarly Article Review Guideline 4 4
2. Relevant to mental health care
3. Not used by another student within the clinical group
4. Submitted and approved as directed by instructor
requirements for
section.
requirements for
section.
requirements for
section.
requirement for
section.
this section

presented.
APA Format and Writing Mechanics
(10 points/10%)
1. Correct use of standard English grammar and
sentence structure
2. No spelling or typographical errors
3. Document includes title and reference pages
4. Citations in the text and reference page
Includes 4
requirements for
section.
Includes 3
requirements for
section.
Includes 2
requirements for
section.
Includes 1
requirement for
section.
No requirements for
this section
presented.
Total Points Possible = 100 points

PurposePreparing the assignmentGrading Rubric Criteria are
met when the student’s application of knowledge demonstrates
achievement of the outcomes for this assignment.
2 7 J u l y 2 0 1 7 | V O l 5 4 7 | N A T u R E | 4 4 1
lETTER
doi:10.1038/nature23285
Global forest loss disproportionately erodes
biodiversity in intact landscapes
Matthew G. Betts1,2*, Christopher Wolf1,2*, William J.
Ripple1,2, Ben Phalan1,3, Kimberley A. Millers4, Adam
Duarte5,
Stuart H. M. Butchart3,6 & Taal levi1,4
Global biodiversity loss is a critical environmental crisis, yet
the lack
of spatial data on biodiversity threats has hindered conservation
strategies1. Theory predicts that abrupt biodiversity declines are
most likely to occur when habitat availability is reduced to very
low
levels in the landscape (10–30%)2–4. Alternatively, recent
evidence
indicates that biodiversity is best conserved by minimizing
human
intrusion into intact and relatively unfragmented landscapes5.
Here we use recently available forest loss data6 to test
deforestation
effects on International Union for Conservation of Nature Red
List
categories of extinction risk for 19,432 vertebrate species
worldwide.

As expected, deforestation substantially increased the odds of a
species being listed as threatened, undergoing recent upgrading
to a higher threat category and exhibiting declining populations.
More importantly, we show that these risks were
disproportionately
high in relatively intact landscapes; even minimal deforestation
has had severe consequences for vertebrate biodiversity. We
found
little support for the alternative hypothesis that forest loss is
most
detrimental in already fragmented landscapes. Spatial analysis
revealed high-risk hot spots in Borneo, the central Amazon and
the
Congo Basin. In these regions, our model predicts that 121–219
species will become threatened under current rates of forest loss
over the next 30 years. Given that only 17.9% of these high-risk
areas are formally protected and only 8.9% have strict
protection,
new large-scale conservation efforts to protect intact forests7,8
are
necessary to slow deforestation rates and to avert a new wave of
global extinctions.
A critical question in global efforts to reduce biodiversity loss
is
how best to allocate scarce conservation resources. To what
extent
should conservation be focused on modified and fragmented
land-
scapes where threats are potentially greatest, versus landscapes
that
are largely intact9? Although it is expected that both approaches
have
value, in some human-influenced habitats, many species seem
sur-
prisingly resilient to habitat loss and fragmentation, and can

coexist
with humans in highly modified landscapes10,11, provided that
habitat
loss does not exceed critical thresholds2. Theory predicts that
abrupt
biodiversity declines are most likely to occur when habitat
availability is
reduced to very low levels in the landscape (10–30%)3,4,12.
Alternatively,
recent evidence indicates biodiversity is best conserved by
minimizing
human intrusion into intact and relatively unfragmented
landscapes,
which implies concentrating the impacts of anthropogenic
disturbance
elsewhere5,13. This is because initial intrusion may result in
rapid deg-
radation of intact landscapes, not only via the direct effects of
habitat
loss, but also the coinciding effects of overhunting, wildfires,
selective
logging, biological invasions and other stressors5. Such
evidence has
led to recent calls to increase the protection of substantial intact
areas
of the Earth’s terrestrial ecosystems14,15. Testing the extent to
which
these alternative hypotheses explain patterns of extinction risk
globally
can improve the effectiveness of conservation efforts and
inform the
formulation of policies, affecting the future of life on Earth.
Recent advances in remote sensing have enabled the
development

of a spatially explicit, high-resolution global dataset on rates of
forest
change6, which provide the capacity to quantify the effects of
contem-
porary global forest loss on biodiversity16. We quantified the
association
between global-scale forest loss and gain within the ranges of
19,432
species and their International Union for Conservation of
Nature
(IUCN) Red List category of extinction risk, recent genuine
changes
in extinction risk, and overall population trend direction. The
species
spanned three vertebrate classes, and included 4,396 (22.6%)
listed as
threatened (Vulnerable, Endangered, or Critically Endangered)
and
15,214 (78.3%) associated with forest habitats. Under the
‘habitat
threshold’ hypothesis, we expected the effects of recent forest
loss to
be most detrimental for species that have already lost a
substantial
proportion of forest within their ranges. Under the ‘initial
intrusion’
hypothesis, we expected species with relatively intact forest
within their
ranges to show the most severe effects of deforestation.
We obtained range maps for amphibians and mammals from the
IUCN Red List17 and those for birds from BirdLife
International and
NatureServe18. We classified species as ‘non-forest’, ‘forest-
optional’,
and ‘forest-exclusive’ based on the IUCN Red List habitat

classifica-
tion data17. Within each species’ range, we used fine-resolution
forest-
change data (2000–2014)6 to calculate the amount of recent
forest
cover, loss, and gain (Fig. 1). Given that many species were
assessed
for the Red List in the early period of our recent forest-loss data
(or even before this; Methods), it would be ideal to have
contemporary
forest loss data from before 2000. The most spatially contiguous
dataset
for 1990–200019 covered > 80% of the ranges for only 58.7% of
the
species in our analyses. However, locations of forest loss were
highly
spatiotemporally correlated at the scale of species’ ranges
between
1990–2000 and 2000–2014 (Methods, Intermediate-term forest
change;
Extended Data Fig. 7).
We also expected that historical deforestation over much longer
temporal scales could influence species vulnerability, a
phenomenon
known as ‘extinction debt’20,21. We calculated historical forest
loss as
the difference between the extents of area within species’
ranges that
historically supported forest cover and the area that remained
forested
in the year 20006 (Fig. 1). We also calculated the mean ‘human
foot-
print’ value22 within each species’ range, because forest loss
could be
confounded with other broad-scale anthropogenic pressures

(Fig. 1).
Using these data, we fit a spatial autologistic regression model
to test
whether forest loss within species’ ranges is associated with the
like-
lihood that a species: (i) is listed as threatened; (ii) has
qualified for
uplisting to a higher category of extinction risk in recent
decades (see
Methods); and (iii) has a declining population trend (as
classified by
IUCN Red List assessors).
1Forest Biodiversity Research Network, Department of Forest
Ecosystems and Society, Oregon State University, Corvallis,
Oregon 97331, USA. 2Global Trophic Cascades Program,
Department of
Forest Ecosystems and Society, Oregon State University,
Corvallis, Oregon 97331, USA. 3Department of Zoology,
University of Cambridge, Downing Street, Cambridge CB2 3EJ,
UK. 4Department of
Fisheries and Wildlife, Oregon State University, Corvallis,
Oregon 97331, USA. 5Oregon Cooperative Fish and Wildlife
Research Unit, Department of Fisheries and Wildlife, Oregon
State University,
Corvallis, Oregon 97331, USA. 6BirdLife International, David
Attenborough Building, Pembroke Street, Cambridge CB2 3QZ,
UK.
* These authors contributed equally to this work.
© 2017 Macmillan Publishers Limited, part of Springer Nature.
All rights reserved.
http://www.nature.com/doifinder/10.1038/nature23285

letterreSeArCH
4 4 2 | N A T u R E | V O l 5 4 7 | 2 7 J u l y 2 0 1 7
As expected, we found a strong association between rate of
recent
forest loss and each response variable. The odds of threatened
status,
declining population trends, and uplisting increased by 5.06%
(95%
confidence interval: 1.01–9.27), 11.34% (6.45–16.45), and
8.39%
(1.53–15.70), respectively, for each 1% increase in recent forest
loss for
forest-exclusive species. This is not surprising, given that
estimated or
inferred rates of habitat loss are used to inform IUCN Red List
assess-
ments under criterion A2, particularly for species lacking direct
data
on population trends17. Nevertheless, our results confirm that
previous
categorical estimates of habitat decline (based on a mixture of
inference,
qualitative and quantitative analysis) match with our global,
systematic
analysis of quantitative data on forest loss16.
More importantly, we found strong support for the initial
intrusion
hypothesis for both forest-optional and forest-exclusive species.
Species
were more likely to be threatened, exhibit declining popul ation
trends
and have been uplisted if their ranges contained intact
landscapes

(> 90% forest cover) with high rates of recent forest loss (Fig.
2). Evidence
for this lies in the strong positive statistical interaction between
forest
loss and cover (that is, forest loss × cover, Figs 2a, 3) on all
response
variables for both forest-exclusive and forest-optional species
(maximum
false discovery rate (FDR)-adjusted P = 0.025, minimum z =
2.51,
Fig. 2a, Supplementary Table 3). For example, at high
proportions of
initial forest cover (90%), the odds of a forest-exclusive species
being
uplisted were 15.78% (95% confidence interval: 6.99–25.30)
greater
for each 1% increase in deforestation. At average proportions of
forest
cover (57%), the equivalent increase in deforestation was much
smaller,
with the odds of a forest-exclusive species being uplisted
reduced to
3.45% (95% confidence interval: − 3.91 to 11.36) (Fig. 3).
These results
were generally similar across vertebrate classes, but amphibi ans
showed
the strongest and most consistent effects across response
variables
(Extended Data Fig. 2). Predictably, forest loss and its
interaction with
forest cover had little effect on non-forest species (Figs 2, 3).
Historical forest loss also exhibited a strong negative influence
on
vertebrate biodiversity (Fig. 2), which may be evidence of an
extinc-

tion debt in which some species are capable of persisting in
landscapes
long after initial forest loss has occurred, but subsequently
decline.
0
25
50
75
100
Cell
mean (%)
a Forest cover (2000)
0
20
40
60
Cell
mean (%)
b Recent forest loss (2000−2014)
0
10

20
30
40
Cell
mean (%)
c Recent forest gain (2000−2012)
0
25
50
75
Cell
mean (%)
d Human footprint
0
10
20
30
40
50

Cell
mean (%)
e Forest loss × cover
0
25
50
75
Cell
mean (%)
f Historical forest loss
Figure 1 | Spatial distribution of the six variables used to
predict
species’ IUCN Red List response variables. a, b–d, Forest cover
in the
year 2000 (a), forest loss between 2000–2014 (b), forest gain
(2000–2012)
(c), and human footprint (d). e, The interaction term ‘forest loss
× cover’
tested alternative hypotheses that forest loss exerts the greatest
negative
influence on biodiversity at low versus high initial levels of
forest cover.
High values of this variable (shown in e) correspond to regions
of both
high forest cover and loss. f, Historical forest loss represents
long-term

forest loss in years preceding 2000. Values plotted are averages
taken over
0.4° grid cells. The maps are derived from current forest change
maps6
(a–c, e, f), an intact forest landscapes map32 (f), biomes of the
world33
(f), and human footprint22 (d).
Bene�cial effect
on biodiversity
a
c
Detrimental effect
on biodiversity
b
d
Historical forest loss Human footprint
Forest loss × cover Forest gain
−0.5 0.0 0.5 −0.5 0.0 0.5
Forest-exclusive
Forest-optional
Non-forest
Forest-exclusive
Forest-optional

Non-forest
Standardized coef�cient
Response Threatened status Declining trend
Uplisted in
threatened status
FDR adjusted P value 0 < P ≤ 0.05 0.05 < P ≤ 0.1 P > 0.1
Figure 2 | Effects of four predictors on the status of 19,432
vertebrate
species worldwide. a, Positive ‘forest loss × cover’ terms
indicate that
the negative effects of forest loss are amplified in landscapes
with greater
initial forest cover. b–d, Forest gain tended to have a positive
effect on
forest optional and exclusive species (b), whereas historical
forest loss
(c) and human footprint (d) tended to have negative effects.
‘Threatened
status’ refers to IUCN Red List categories of ‘Vulnerable’,
‘Endangered’,
or ‘Critically Endangered’. ‘Uplisted in threatened status’
means that the
most recent genuine Red List category change for a speci es has
been in the
direction of higher endangerment. Forest loss and cover
variables were
included as main effects, but coefficient estimates are not
shown here as
they are not readily interpretable in the presence of the
interaction term.
Error bars represent 95% confidence intervals. Categories for P

values
are listed as ranges (that is, 0 < P ≤ 0.05, 0.05 < P ≤ 0.1, P >
0.1), and
sample sizes (also given in Supplementary Table 1) for non-
forest/forest-
optional/forest-exclusive are 4,218/3,430/4,218,
10,457/8,827/10,457,
4,757/4,073/4,757 for Threatened status, Declining trend, and
Uplisted in
threatened status, respectively.
letter reSeArCH
2 7 J u l y 2 0 1 7 | V O l 5 4 7 | N A T u R E | 4 4 3
Predictably, increased human footprint has had a generally
negative
influence on the status of vertebrates associated with forest and
non-
forest systems (Fig. 2). We also found recent forest gain
decreased the
likelihood of threatened status (forest-exclusive and forest-
optional
species) and declining population trend (forest-optional species;
Fig. 2).
However, amphibians primarily drove these relationships; bird
and
mammal biodiversity did not show statistically significant
responses
to forest gain (Extended Data Fig. 2), indicating that young

secondary
forest does not appear to be ameliorating biodiversity declines
for
these taxa8.
Overall, the global spatial autologistic regression model
performed
remarkably well (area under the receiver operating
characteristic
curve (AUC) = 0.78; Extended Data Fig. 1), even when we
conser-
vatively excluded entire regions one at a time (Africa,
Americas,
Asia, Oceania) and evaluated models on these independent data
(AUC = 0.74). Furthermore, results remained consistent when
we
statis tically accounted for phylogenetic dependencies, latitude,
and time
since each species was initially described Extended Data Fig. 3.
We also
applied alternative approaches to account for spatial
autocorrelation
and excluded species designated as threatened due to
characteristically
small and declining or fragmented ranges (that is, under IUCN
Red List
criterion B) (Extended Data Figs 3, 6). Results were also robust
to degree
of threat; Critically Endangered, Endangered and Vulnerable
species all
showed similar patterns in response to forest loss (Extended
Data Fig. 9).
Strong support for the initial intrusion hypothesis may be
surprising,
given existing theory3,23 and that a considerable number of

conserva-
tion programs focus on areas that have already lost substantial
forest24.
However, such highly deforested landscapes may have already
passed
through a substantial local extinction filter, whereby the most
sensi-
tive species have been lost25. A recent broad-scale study
conducted
in the Brazilian Amazon revealed that landscapes still
exceeding 80%
forest cover have lost 46–60% of their conservation value5. Our
results
suggest that initial forest loss is a potential indicator of other
threats to
forest biodiversity that are more challenging to measure at large
spatial
extents. Mechanisms for intrusion effects include increased
unregulated
hunting26 (especially near new logging roads27), disease and
human
disturbance, and invasive species28, as well as the direct effects
of habitat
loss for interior forest specialists29. Indeed, many of the
species with
ranges that were characterized by high initial forest cover
(before 2000),
but intensive recent deforestation, tend to be under hunting
pressure
(for example, Sira curassow (Pauxi koepckeae)) or are habitat
specialists
(Mendolong bubble-nest frog (Philautus aurantium), Mentawi
flying
Threatened status Declining trend
Uplisted in

threatened status
N
o
n
-fo
re
st
F
o
re
st-o
p
tio
n
a
l
F
o
re
st-e
xc
lu
sive

0% 25% 50% 75% 100% 0% 25% 50% 75% 100% 0% 25% 50%
75% 100%
0%
5%
10%
15%
0%
5%
10%
15%
0%
5%
10%
15%
Forest cover (2000)
F
o
re
st
lo

ss
0.25
0.50
0.75
Probability
Figure 3 | Predicted probabilities of species status as a function
of
recent forest loss and total forest cover within a species range.
All other
covariates (forest gain, historical forest loss, and human
footprint) were
statistically held at their average values when estimating
probabilities.
For forest-optional and forest-exclusive species, the effect of
forest loss is
stronger at high levels of initial forest cover; deforestation in
intact forests
has the most negative impact, supporting the initial intrusion
hypothesis.
0
.5
×
c
u
rr
e
n

t
lo
ss
r
a
te
C
u
rr
e
n
t
lo
ss
r
a
te
1
.5
×
c
u
rr
e
n
t

lo
ss
r
a
te
IUCN
category
Ia Ib II III
IV V VI 0 20 40 60
Increase in
threatened
richness
2030–2045 2045–2075
Figure 4 | Projected increases in the number of threatened
species
under three scenarios of future forest loss. Projections are
estimated
using the global model. Increased threatened richness (blue to
red colour
scale) is relative to the fitted probabilities of a species being
threatened. For
example, a value of 20 would indicate a projected increase of 20
threatened
species in a 0.2° grid cell. Only locations with projected
increases in
threatened species are shown and only forest-exclusive species
were used
for this projection. Column labels show time spans where the

lower limit
assumes the effects of forest loss on status are entirely due to
deforestation
from 2000–2014; the upper limit assumes effects could be partly
a function
of forest loss in the decades before 2000 (global locations of
forest loss are
temporally autocorrelated; see Methods, section ‘Intermediate-
term forest
change’). IUCN protected areas (categories I–VI) are shown in
greyscale
shading. The maps are derived from the following sources:
IUCN Red List
species range maps18, recent forest change6, intact forest
landscapes32, human
footprint22, world biomes33, and the World Database of
Protected Areas34.
letterreSeArCH
4 4 4 | N A T u R E | V O l 5 4 7 | 2 7 J u l y 2 0 1 7
squirrel (Iomys sipora)) (Supplementary Table 4). If specialists’
habitat
is targeted in the initial phases of deforestation (for example,
accessible
high-economic-value forest (bottomland forest adjacent to
rivers)),
habitat will be lost at much greater rates than indicated by the
overall

rate of forest loss within a species’ range30.
As a further exploration of the habitat threshold hypothesis, we
fit
a model to test whether the strongest negative effects of recent
forest
loss occurred in landscapes with both high and low levels of
remain-
ing forest cover (a statistical interaction between forest loss and
forest
cover squared; Methods). We found no evidence for such an
effect for
either threatened status or recent uplisting (Extended Data Figs
4, 5).
Notably, the odds of a declining population trend showed
evidence for this dual effect for forest-optional and -exclusive
species;
we speculate that the increased likelihood of a declining trend
with
deforestation in landscapes with low levels of forest cover, but
no
relationship for threatened status, may constitute early signs of
an
extinction debt that remains to be fully paid. Thus, our results
do not
imply deforestation effects are benign in regions with low levels
of
remaining forest cover. Although species exposed to
deforestation in
such landscapes are less likely to be designated as threatened
than those
exposed to similar rates of deforestation in more intact areas,
their
populations will continue to decline with further habitat loss,
which
will in time inevitably lead to increased extinction risk.

The spatially explicit nature of our model enabled quantitative
pre-
dictions of global hotspots where biodiversity is at particularly
high
risk given reduced (halving current rates), continued, or
accelerated
(1.5× ) future rates of forest loss (each assumes no future forest
loss in
protected areas with IUCN categories I–VI; Fig. 4). High-risk
hot spots
emerged in southeast Asia (particularly Borneo), the central -
western
Amazon and the Congo Basin where the numbers of threatened
forest-exclusive species are predicted to increase by 82–134,
34–74,
and 5–11, respectively, over the next 30 years under current
rates of
deforestation. Together, the number of threatened species for
these
three regions is predicted to increase by 121–219. Currently,
only
17.9% of these areas are formally protected (IUCN classes I–VI;
Supplementary Table 5) and only 8.9% have strict protection
(IUCN
classes I–III). These results, alongside evidence of ongoing
erosion of
intact forest landscapes31, highlight that areas until recently
considered
to be of “low vulnerability”9 are in fact where anthropogenic
distur-
bance is increasingly putting species at most risk of extinction.
New
large-scale efforts to reduce both degradation and loss of intact
forest
landscapes7 are needed to protect against an intensified wave of

extinc-
tions in the world’s last wildernesses.
Online Content Methods, along with any additional Extended
Data display items and
Source Data, are available in the online version of the paper;
references unique to
these sections appear only in the online paper.
received 14 December 2016; accepted 13 June 2017.
Published online 19 July 2017.
1. Joppa, L. N. et al. Filling in biodiversity threat gaps. Science
352, 416–418
(2016).
2. Newbold, T. et al. Has land use pushed terrestrial
biodiversity beyond
the planetary boundary? A global assessment. Science 353, 288–
291
(2016).
3. Andrén, H. Effects of habitat fragmentation of birds and
mammals in
landscapes with different proportions of suitable habitat: a
review. Oikos 71,
355–366 (1994).
4. Betts, M. G., Forbes, G. J. & Diamond, A. W. Thresholds in
songbird
occurrence in relation to landscape structure. Conserv. Biol. 21,
1046–1058
(2007).
5. Barlow, J. et al. Anthropogenic disturbance in tropical forests

can double
biodiversity loss from deforestation. Nature 535, 144–147
(2016).
6. Hansen, M. C. et al. High-resolution global maps of 21st-
century forest cover
change. Science 342, 850–853 (2013).
7. Peres, C. A. Why we need megareserves in Amazonia.
Conserv. Biol. 19,
728–733 (2005).
8. Gibson, L. et al. Primary forests are irreplaceable for
sustaining tropical
biodiversity. Nature 478, 378–381 (2011).
9. Brooks, T. M. et al. Global biodiversity conservation
priorities. Science 313,
58–61 (2006).
10. Kareiva, P., Watts, S., McDonald, R. & Boucher, T.
Domesticated nature: shaping
landscapes and ecosystems for human welfare. Science 316,
1866–1869
(2007).
11. Mendenhall, C. D., Karp, D. S., Meyer, C. F., Hadly, E. A.
& Daily, G. C. Predicting
biodiversity change and averting collapse in agricultural
landscapes. Nature
509, 213–217 (2014).
12. Fahrig, L. When does fragmentation of breeding habitat
affect population
survival? Ecol. Modell. 105, 273–292 (1998).

13. Phalan, B., Onial, M., Balmford, A. & Green, R. E.
Reconciling food production
and biodiversity conservation: land sharing and land sparing
compared.
Science 333, 1289–1291 (2011).
14. Wilson, E. O. Half-Earth: Our Planet’s Fight For Life. (W.
W. Norton & Company,
2016).
15. International Union for Conservation of Nature World
Congress. Motion 48:
Protection of primary forests, including intact forest landscapes.
(2016).
16. Tracewski, Ł. et al. Toward quantification of the impact of
21st-century
deforestation on the extinction risk of terrestrial vertebrates.
Conserv. Biol. 30,
1070–1079 (2016).
17. International Union for Conservation of Nature. IUCN red
list of threatened
species. Version 2016.3 http://www.iucnredlist.org (2017).
18. BirdLife International and NatureServe. Bird Species
Distribution Maps of the
World Version 5.0 (BirdLife International, 2015).
19. Kim, D.-H. et al. Global, Landsat-based forest-cover change
from 1990 to 2000.
Remote Sens. Environ. 155, 178–193 (2014).
20. Tilman, D., May, R. M., Lehman, C. L. & Nowak, M. A.
Habitat destruction and
the extinction debt. Nature 371, 65–66 (1994).

21. Wearn, O. R., Reuman, D. C. & Ewers, R. M. Extinction
debt and windows of
conservation opportunity in the Brazilian Amazon. Science 337,
228–232
(2012).
22. Sanderson, E. W. et al. The human footprint and the last of
the wild. Bioscience
52, 891–904 (2002).
23. Hanski, I. Metapopulation dynamics. Nature 396, 41–49
(1998).
24. Brooks, T. M. et al. Habitat loss and extinction in the
hotspots of biodiversity.
Conserv. Biol. 16, 909–923 (2002).
25. Balmford, A. Extinction filters and current resilience: the
significance of past
selection pressures for conservation biology. Trends Ecol. Evol.
11, 193–196
(1996).
26. Ripple, W. J. et al. Bushmeat hunting and extinction risk to
the world’s
mammals. R. Soc. Open Sci. 3, 160498 (2016).
27. Benítez-López, A. et al. The impact of hunting on tropical
mammal and bird
populations. Science 356, 180–183 (2017).
28. Bellard, C., Genovesi, P. & Jeschke, J. M. Global patterns
in threats to vertebrates
by biological invasions. Proc. R. Soc. Lond. B 283, 20152454
(2016).

29. Clavel, J., Julliar, R. & Devictor, V. Worldwide decline of
specialist species:
toward a global functional homogenization. Front. Ecol.
Environ. 9, 222–228
(2011).
30. Betts, M. G. et al. A species-centered approach for
uncovering generalities in
organism responses to habitat loss and fragmentation.
Ecography 37, 517–527
(2014).
31. Potapov, P. et al. The last frontiers of wilderness: tracking
loss of intact forest
landscapes from 2000 to 2013. Sci. Adv. 3, e1600821 (2017).
32. Potapov, P. et al. Mapping the world’s intact forest
landscapes by remote
sensing. Ecol. Soc. 13, 51 (2008).
33. Olson, D. M. et al. Terrestrial ecoregions of the world: a
new map of life on
Earth: a new global map of terrestrial ecoregions provides an
innovative tool
for conserving biodiversity. Bioscience 51, 933–938 (2001).
34. IUCN and UNEP-WCMC. The World Database on Protected
Areas (WDPA)
http://www.protectedplanet.net/terms (2015).
Supplementary Information is available in the online version of
the paper.
Acknowledgements Funding from the National Science
Foundation (NSF-

DEB-1457837) and the College of Forestry IWFL Professorship
in Forest
Biodiversity Research to M.G.B. supported this research. We
are grateful for
comments from A. Hadley, U. Kormann, J. Bowman, C. Epps
and C. Mendenhall
on earlier versions of this manuscript.
Author Contributions M.G.B., C.W., S.H.M.B., W.J.R. and T.L.
conceived the
study, C.W., M.G.B. and T.L. analysed the data, and M.G.B.
and C.W. wrote
the first draft of the paper with subsequent editorial input from
C.W., B.P.,
S.H.M.B., K.A.M. and A.D.
Author Information Reprints and permissions information is
available at
www.nature.com/reprints. The authors declare no competing
financial
interests. Readers are welcome to comment on the online
version of the
paper. Publisher’s note: Springer Nature remains neutral with
regard
to jurisdictional claims in published maps and institutional
affiliations.
Correspondence and requests for materials should be addressed
to
M.G.B. ([email protected]) or C.W. ([email protected]).
reviewer Information Nature thanks J. Barlow, L. Gibson and
the other
anonymous reviewer(s) for their contribution to the peer review
of this work.

http://www.iucnredlist.org
http://www.protectedplanet.net/terms
http://www.nature.com/reprints
letter reSeArCH
MethODS
No statistical methods were used to predetermine sample size.
The experiments
were not randomized and the investigators were not blinded to
allocation during
experiments and outcome assessment.
Species data. We obtained data on three classes of terrestrial
vertebrates (mam-
mals, amphibians, and birds) from the IUCN Red List17. We
defined threatened
species as those classified as Vulnerable, Endangered, or
Critically Endangered on
the Red List. We also obtained population trends (‘Increasing’,
‘Stable’, ‘Decreasing’,
or ‘Unknown’) from the Red List. We excluded ‘Data deficient’
and ‘Extinct in the
wild’ species from our analysis with threatened status as the
response variable.
Similarly, for the decreasing population trend response, we
excluded species with
unknown population trends.

For analyses in which we examined change in Red List
category, it is necessary
to compare time points in which all species in the taxonomic
group were assessed,
and to consider only those Red List category changes between
such assessments
that resulted from genuine improvement or deterioration in
status (that is, exclud-
ing changes owing to improved knowledge or revised
taxonomy). These genuine
changes underpin the Red List Index35,36. We considered
species to have been
uplisted if their most recent genuine Red List category change
was in the direction
of increasing endangerment (Least Concern < Near Threatened
< Vulnerable
< Endangered < Critically Endangered). These data were
obtained from Hoffmann
et al.37, and were updated to match the taxonomy on the 2016
IUCN Red List; the set
of genuine changes for birds was also updated using data in
BirdLife International38.
The relevant periods of our primary uplisting dataset are 1980–
2004 for amphibi-
ans, 1996–2008 for mammals, and 1988–1994, 1994–2000,
2000–2004, 2004–2008,
2008–2012, and 2012–2016 for birds. Additionally, we used all
available genuine
category change data from 2008–2016 for mammals and 2006–
2016 for amphibi-
ans. Although these more recent category change data
(approximately 100 category
changes) are not yet comprehensive (that is, not all species in
these taxa have been
checked for genuine category changes over these times), they

cover a wide range of
species and are likely to be reflective of recent changes in
forest cover for these species.
Genuine category change data are currently unavailable for
other time periods.
We classified non-avian species according to habitat usage
(forest-exclusive,
forest-optional, and non-forest) using the IUCN Red List data
coding species
against the IUCN habitats classification scheme
(http://www.iucnredlist.org/
technical-documents/classification-schemes/habitats-
classification-scheme-ver3).
We treated species using only forest habitat as forest-exclusive,
those using forest
habitat and at least one other habitat type as forest-optional, and
those not using
forest at all as non-forest. To categorize bird species, we used
higher-quality data on
forest dependency from BirdLife International38, treating
species with high forest
dependency as forest-exclusive, medium and low forest
dependency as forest-
optional, and not normally using forest as non-forest.
The species range maps used in the analysis were derived from
the IUCN Red List
for mammals and amphibians, and from BirdLife International
and NatureServe18
for birds. For each species, we used only range polygons where
presence was classi-
fied as ‘Extant’ or ‘Probably extant’. Vertebrates without range
maps available were
omitted from the analyses (108 mammals, 39 amphibians, and
30 birds). Reptiles

were excluded from the analysis as IUCN reptile data are
relatively limited39.
After screening for data availability using the steps above, the
dataset consisted
of 19,432 species (19,615 including data-deficient species),
4,396 (22.6%) of which
are listed as threatened. The entire dataset represents 58.2% of
the terrestrial
vertebrate species globally (98.9% birds, 84.9% mammals,
63.1% amphibians)
(based on described species totals from IUCN Red List
summary table 1).
Predictor variables. We used six predictor variables in our
primary analysis
(Fig. 1). Here, we describe these variables in detail.
We used the forest change maps (version 1.2) given in Hansen
et al.6 for our
analyses. The forest cover map indicates the percentage forest
cover in each 30 m
pixel in the year 2000. The forest loss and gain maps are both
binary and indicate
whether forest loss or gain occurred in each pixel. Following
Hansen et al., we
considered forest to have been ‘lost’ if a stand-replacing
disturbance (that is, complete
removal of tree cover canopy at the Landsat pixel scale) had
occurred between 2000
and 2014, and ‘gained’ if establishment of tree canopy from a
non-forest state had
occurred between 2000 and 2012. In addition, we included a
forest loss × forest cover
interaction term to test the hypothesis that the effects of forest
loss are dependent
upon the total amount of forest within a species’ range. A

positive coefficient for such
a term would indicate that the effect of recent forest loss on our
response variables
was amplified at when initial forest cover was high (support for
the initial intrusion
hypothesis). Conversely, a negative coefficient for this
interaction term would indi-
cate that the effect of recent forest loss on our response
variables was greatest at low
forest cover (support for the habitat threshold hypothesis; see
main text).
The human footprint map that we used (Global Human Footprint
v.2,
1995–2004) measures the extent of human impacts on the
environment and is
created from nine global data layers covering biome type and
biogeographic realm,
human population density, human land use and infrastructure
(that is, built-up
areas, night-time lights, land use/land cover), and human access
(coastlines, roads,
railroads, navigable rivers)40. Among land cover types, built-up
environments
increase the human influence index the most, followed by
agricultural land cover,
and mixed-use land cover (other types do not contribute to the
index)22. Thus,
loss of forest to these land cover types could cause human
footprint to be partially
confounded with our forest loss variable, potentially causing
our analysis to under-
estimate the effects of forest loss. A more recent version of this
map (1993–2009)
was recently released41,42 but the original and updated human

footprint maps are
highly correlated (r = 0.935 at 2° resolution), so our choice of
human footprint map
is unlikely to have influenced the results.
In our analysis, ‘historical forest loss’ is an estimate of long-
term patterns in
forest loss that is not captured by contemporary forest change.
To construct this
variable, we took the following steps. First, we used a random
forest regression
model to develop a historical (or potential) forest cover map.
We modelled the
continuous variable ‘percentage forest cover’ in the year 2000
(from Hansen et al.7)
as a function of x and y coordinates, 19 bioclimatic variables
(derived from monthly
temperature/precipitation) from the WorldClim database13
along with a categor-
ical variable representing forest biomes33. Importantly, to
exclude the effects of
contemporary anthropogenic disturbance on percentage forest
cover we only used
data from within ‘intact forest landscapes’ (IFLs) in the
regression model. An IFL
is defined as “an unbroken expanse of natural ecosystems within
areas of current
forest extent, without signs of significant human activity, and
having an area of at
least 500 km2”32. We assumed that forest cover in intact forest
landscapes (IFLs) is
representative of the degree of canopy cover that could be
historically supported in
across the globe. We then extrapolated the fitted values of this
model to the areas
for a map of potential or historical forest cover (Extended Data

Fig. 10a). Second,
we subtracted recent forest cover from historical cover to
estimate historical loss
(Extended Data Fig. 10b) to yield a map of historical forest loss
(Extended Data
Fig. 10c). We restricted our modelling to within forest biomes,
excluding non-forest
biomes and the boreal forest/taiga. Although some forest cover
may be present out-
side forest biomes (for example, in savannahs), limitations in
available IFL data for
these cover types and the taiga make reconstructing historic
forest cover in these
biomes impractical. Moreover, forest obligate species—our
primary focus—seldom
occur outside forest biomes. Modelling was conducted at 5-km
resolution using
rasters in Behrmann cylindrical equal-area projection. We used
ArcGIS 10.1 and
R for the geospatial analyses43,44. The random forest model
was fit using the Rborist
R package with the default settings45. We acknowledge that the
period of time since
historical deforestation can vary widely across locations
globally. Nevertheless, in
the absence of globally available forest loss data before 2000,
this variable is the best
available test of whether long-term reductions in forest cover
within a species range
affects Red List category and overall direction of population
trend.
Statistical analysis. We used a 2-decimal degree equivalent
equal-area grid
(constructed using the Behrmann cylindrical equal-area
projection). This resolu-
tion is considered appropriate for macroecological analyses that

involve species’
range maps46. We rescaled covariates to this resoluti on by
taking their average
values across each grid cell (ignoring regions over water). We
rescaled species’
ranges to the grid by treating a species as present in a grid cell
if any part of its
range overlapped that cell. We then averaged covariates acr oss
species ranges
using the averages of their cell values weighted by the
proportion of land in
each grid cell.
We modelled the probability of species being threatened, having
a declining
population trend, or having been uplisted (three separate binar y
responses) using
autologistic regression to account for potential spatial
autocorrelation47. The
spatial autocovariate was calculated for each species using a
symmetric spatial
weights matrix as:
∑=
∈
A w yi
j k
ij j
i
where i is the ith species, ki is the set of its neighbours, yj is
the response for the jth
species, and wij = 1 corresponding to the (i, j) entry of the
binary spatial weights

matrix48. Geographic distance was calculated using species’
range centroids. The
spatial weights matrix and spatial autocovariate were calculated
using the spdep
package for R44,49.
We used the generalized linear model (GLM) function glm in R
to fit the logistic
regression model, including the covariates described above, the
spatial autocovariate,
and taxonomic class (as a fixed effect). We estimated
standardized coefficients
and 95% confidence intervals for all predictor variables (each
was standardized
(z-transformed) before analysis). Our hypothesis tests were
conducted across all three
vertebrate classes with six predictor variables, which risks
inflating Type I error rate.
Sequential Bonferroni-type multiple comparisons are sometimes
used to account
for such error inflation, but are highly conservative50.
Therefore, we used a FDR
procedure (the ‘graphically sharpened method’50) which does
not suffer from the
http://www.iucnredlist.org/technical-documents/classification-
schemes/habitats-classification-scheme-ver3
http://www.iucnredlist.org/technical-documents/classification-
schemes/habitats-classification-scheme-ver3
letterreSeArCH

same loss of power but corrects for multiple comparisons. FDR-
adjusted P values
were calculated with p.adjust in R44,51.
Projecting future status changes. We used our model for
threatened status of for-
est-exclusive species to map the predicted increase in
threatened species richness at
multiple forest loss rates over time. We did this by simulating
continued forest loss
at rates of 50%, 100% and 150% the current rates for the time
spans 2030–2045 and
2045–2070. For example, at the current loss rate the area of
forest lost would double
in 15 years (by 2030). We modified these forest loss projections
by setting predicted
future loss to zero within IUCN category I–VI protected areas
using the polygon
type protected area maps in the World Database of Protected
Areas (WDPA)34.
Assuming that there are no substantial time lags between forest
loss and species
being listed as threatened, the resulting predictions
(probabilities of species being
threatened) correspond to 2030. In the event that intermediate
term (approximately
1950–2000) forest loss is also closely linked to threatened
status (that is, there are
time lags between forest loss and status decisions), we included
conservative upper
time limits corresponding to half the stated forest loss rates. In
all cases, predicted
current probabilities of being threatened (from the fitted model)
were subtracted
from the estimated future probabilities; we then mapped the
result by summing
probabilities for all species in each raster grid cell. As the maps

show qualitatively
similar patterns, they can conservatively be interpreted as
showing ‘relative hot
spots’—an interpretation that is valid even if the true
intermediate-term forest rate
of loss is substantially higher than in our scenario.
To assess overlap between existing protected areas and hot
spots (at high risk
of increases to the Red List), we used the ‘predicted increase in
threatened spe-
cies richness’ map for 2030–2045 at the current loss rate.
Within each regional
panel of this map set in Fig. 4, we considered hot spot areas to
be those with at
least one quarter of the maximum predicted increase in
threatened richness for
that region. We estimated the percentage of these areas that is
protected using the
World Database of Protected Areas (WDPA)34. For this
analysis, we report both
strictly protected areas: IUCN categories Ia (Strict Nature
Reserve), Ib (Wilderness
Area), II (National Park), and III (Natural Monument or
Feature) and all other
IUCN categories (IV, Habiat/Species Management Area; V,
Protected Landscape;
VI, Protected area with sustainable use of natural resources). In
addition, we only
consider protected areas with polygon data in the WDPA, which
results in a con-
servative estimate of the percentage of high-risk area that is
protected.
Assessing model performance. We used the area under the
receiver operating
characteristic curve (AUC) to assess model performance for our

primary model
(predicting threatened status for forest exclusive species). The
AUC reflects the true
versus false positive rates for a binary classifier with
continuous output as a function
of the threshold used to determine which outputs correspond to
which categories52.
We calculated AUC both for the ‘All species’ model (with
‘class’ as a fixed effect)
and separately for each class using models fit to individual
classes. We did this
with and without the spatial autocovariate term. In each case,
we also quanti-
fied model performance using fourfold cross-validation by
regions of the world
(Supplementary Table 2). We used a regional grouping (Africa,
Americas, Asia,
Oceania) based on the United Nations Statistics Division
classification system53.
Using entire regions as hold-out test datasets further reduces the
positive effects
of dependency (spatial, taxonomic, and so on) on model
performance metrics54.
The raw and cross-validated AUCs (0.784 without cross-
validation, 0.743 with
cross-validation) for ‘All species’ together (with the
autocovariate) indicate that our
models perform well (Extended Data Fig. 1). For each model,
we also calculated
P values from a Wilcoxon rank-sum test55 to quantify whether
the AUCs were
significantly greater than 0.5 (a baseline at which the model is
performing no better
than random chance). All P values except the one for mammals
with cross validation

and no auto-covariate term were highly significant (< 0.001)
(Extended Data Fig. 1).
Alternative statistical methods to account for spatial
autocorrelation. We tested
the residuals of our global autologistic regression model for
spatial autocorrelation;
for all response variables, Moran’s I was < 0.15 across all
distance classes, indicating
that the autocovariate had removed spatial autocorrelation.
Further more, to ensure
that our results were robust to the sort of spatial model applied,
we fit other spatial
logistic regression models (that is, Moran eigenvector filtering,
simultaneous
spatial autoregressive models (SAR), and Bayesian conditional
autoregressive models
(CAR)) to assess sensitivity to the procedure used for modelling
or accounting
for spatial autocorrelation. We also fit a non-spatial generalized
linear model for
reference along with our primary spatial autologistic regression
model using the
50 nearest neighbours of each species (instead of 5). In each
case, the models were
fit using forest-exclusive species with threatened status as the
response. We fit
models for each taxonomic class separately, as not of all of the
procedures could
readily incorporate the hierarchical structure of the data. Our
results were robust
to the spatial autocorrelation modelling method (Extended Data
Fig. 6). Details
on other spatial models applied are given below.
We fit a Moran eigenvector GLM filtering model by adding
covariates to the

generalized linear model that were computed using the ME
function in the spdep R
package49,56. This spatial filtering model involves augmenting
the predictor matrix
with eigenvectors computed from the spatial weights matrix so
as to reduce the
spatial autocorrelation of the residuals (as estimated using the
Moran’s I statistic).
The smallest subset of eigenvectors that causes the permutation-
based Moran’s I
test P value to exceed a threshold α is chosen for inclusion (we
used α = 0.2, which
is a common default value).
We fit CAR and SAR models using the binary spatial weights
matrix described
above. The conditional autoregressive model was fit using the
CARBayes R
package57. Markov chain Monte Carlo sampling errors were
encountered when
fitting a few of the CAR models. In such cases, the model
results are not available.
The simultaneous autoregressive model was fit using the splogit
function in the
MCSpatial package58. It is based on an approximation
(linearization), which allows
the model to be fit to large datasets59.
Estimates within taxonomic classes. While the primary results
presented in the
main text (Fig. 2) are for all classes together (with class
included as a fixed effect),
we also fit models using data from each class separately
We did this to assess the extent to which our results,
particularly for the forest

loss × cover interaction, are consistent between classes.
Accounting for the effects of latitude. We fit models including
latitude as a main
effect (Extended Data Fig. 3a). We did this to test whether our
results were robust
to this potential confounding variable, which is correlated with
numerous variables
that may be linked to endangerment such as net primary
productivity (NPP) and
per capita gross domestic product (GDP). The estimated forest
loss × cover inter-
action term did not change substantially when accounting for
(absolute) latitude
(Extended Data Fig. 3a).
Quadratic models (loss × cover squared interaction). We fit
models with quad-
ratic interaction terms corresponding to forest loss × cover2 to
test whether the
models with only the linear forest loss × cover terms were
adequate for forest
exclusive and optional species. Support for a quadratic
interaction term would
provide evidence for both the initial intrusion hypothesis and
the threshold
hypothesis; in other words, the effects of forest loss on species
status and trends
are most substantial both at very high and very low initial forest
amounts (see main
text). These quadratic terms were generally non-significant
(Extended Data Fig. 4)
supporting the hypothesis that the effect of forest loss on the
odds of species being
threatened, declining, or uplisted varies linearly with forest
cover. However, in the
overall (all species) models, we found strong evidence that the
forest loss × cover2

term was positive when declining trend was the response
variable (Extended Data
Fig. 4). This suggests that the effect of loss on population
trends may be most
negative at both low and high levels of forest cover, and
smallest (near zero) at
intermediate levels of forest cover (Extended Data Fig. 5).
Tropical forest species. As the ecology of tropical forests often
responds differently
to non-tropical forests, we also examined model results for
species found exclusively
in tropical forests (Extended Data Fig. 3b). We did this by
restricting the species set
to those with ranges containing only grid cells that overlap
tropical forests. We deter-
mined tropical forest regions using a map of biomes33 and
treating the following
biomes as tropical forest: ‘Tropical & Subtropical Moist
Broadleaf Forests’, ‘Tropical
& Subtropical Dry Broadleaf Forests’, ‘Tropical & Subtropical
Coniferous Forests’
and ‘Mangroves’. The restriction of our dataset to tropical
forest species did not
substantially alter our primary results, although it did weaken
the forest loss × cover
effect on the likelihood of declining population trends
(Extended Data Fig. 3b).
Range area. Species’ geographic range area is a key predictor of
extinction risk,
and extent of occurrence and area of occupancy are two
parameters used to assess
species under criterion B of the IUCN Red List. This can pose a
circularity issue
for comparative extinction risk analyses, particularly those that
attempt to assess
the effect of geographic range area relative to the effects of

other predictors on
species endangerment60. A common remedy is to run the
analysis on species
classified as Least Concern and those that are listed as Near
Threatened or threat-
ened for reasons not directly linked to small geographic range
area (that is, not
under criterion B)60. We followed this procedure as part of our
sensitivity analysis.
Specifically, we excluded species listed as threatened under
criterion B. Such
species made up 2,529 (approximately 58%) of the 4,396
threatened species in our
full dataset. The results (Extended Data Fig. 3c) show that our
overall conclusions
are robust to the exclusion of these species.
Forest loss and cover threshold. In our primary analysis, we
used the forest loss
and cover variables directly as given in Hansen et al.6. Forest
cover is a continuous
variable ranging from 0% to 100% cover within each pixel and
forest loss is a binary
variable indicating whether or not tree cover canopy had been
completely removed
between 2000 and 2014. Since the effects of forest loss and
cover on endangerment
(status/trends/uplisting) probably vary depending on the initial
amount of forest
cover, we replicated our analyses, but truncated forest loss and
cover at the 75%
threshold (Extended Data Fig. 3d). That is, we treated cover and
loss as zero in
pixels that had less than 75% initial forest cover. This change
did not influence our
results substantially (Extended Data Fig. 3d).

letter reSeArCH
Forest loss and gain standardization. The forest loss and gain
variables in our
analysis can be thought of in terms of percentages of species’
ranges since they are
averages of spatial variables across species’ ranges. An
alternative way to compute
the forest loss and gain variables is as percentages of forest
cover within species’
ranges. We used these standardized loss and gain variables (that
is, loss divided by
cover and gain divided by cover) as part of our sensitivity
analysis (we similarly
standardized historical loss by dividing by potential cover), and
found that their
use had little effect on our results (Extended Data Fig. 3e). This
provides another
way of quantifying forest loss and gain, which may be
particularly appropriate for
species that have little forest cover within their ranges. This
was uncommon in our
core dataset as we focused on forest-optional and -exclusive
species, that tend to
have high forest cover across their ranges.
Accounting for phylogeny. The models that we fit assume that
the dependence
structure of the observations is purely spatial. However, this
may not be valid as
species that are phylogenetically similar may be more likely to
have the same status,

trend, or uplisting variable values, even after accounting for the
covariates in the
models. To explore this issue of potential phylogenetic
dependence and its effect
on our results, we fit generalized linear mixed models using
glmer in the lme4
R package61, including random effects by taxonomic order
(Extended Data
Fig. 3f ). We were unable to fit more complex phylogenetic
models that use full trees
(for example, phylogenetic logistic regression) because detailed
phylogenetic data
are not available for many of the species in our analysis62.
However, the addition of
taxonomic-based random effects did not substantially alter our
results, suggesting
that the effects of phylogenetic dependence are weak after
accounting for spatial
autocorrelation and the other predictors (Extended Data Fig. 3f
).
Assessing sensitivity to resolution. We tested the sensitivity of
the results to the
spatial resolution used in our analysis (2 decimal degree
equivalent equal-area) by
re-computing the covariates (averages across species’ ranges) at
a finer resolution
of approximately 5 km. In this analysis, we refined the species’
ranges by clipping
them using the species’ altitude limits coded on the IUCN Red
List, when available
(6,047 of 19,615 species). We also excluded forest loss, gain,
and cover inside of
known tree plantations using a map of plantations for seven
tropical countries63.
Covariate averages at high resolution were calculated using
Google Earth Engine.

Coefficient estimates show relatively low sensitivity to our
choice of resolution,
clipping ranges by altitudinal limits, and masking out forest
variables within known
plantations (Extended Data Fig. 3g).
Intermediate-term forest change. Our primary forest change
variables are from
2000 to 2014. We also included a derived ‘historical forest
cover’ variable to account
for long-term forest change. However, given that many species
were listed in the
early period of our recent forest-loss data (or even before this),
it would be ideal to
have contemporary forest loss data from before 2000.
Unfortunately, no spatially
contiguous datasets exist for this period. Nevertheless, to
extend the time span
for the more recent forest change variables, we added 1990–
2000 forest loss and
gain estimates to the 2000–2014 estimates, producing estimates
of loss and gain
for the period 1990–201419. This summed dataset covered >
80% of the ranges for
only 58.7% of the species in our analyses. Using these data, the
forest loss × cover
interaction term was weaker. However, consistent with our
primary analyses, esti-
mates still tended to be positive for forest-optional and -
exclusive species (Extended
Data Fig. 3h). It is likely that the smaller effect size estimates
are related to uncer-
tainty in the 1990–2000 dataset caused by missing data
Importantly, we found a high correlation between 1990–2000
and 2000–2014
forest loss at low levels of missing data, which suggests that

locations of interme-
diate-term and recent forest loss are correlated at the scale of
species’ ranges (there
is temporal autocorrelation in forest loss; Extended Data Fig.
7). This correlation
is further supported by the country-level correlations between
1990–2000 and
2000–2015 net forest loss (that is, change in percentage cover)
obtained using the
Food and Agriculture Organization’s (FAO) Global Forest
Resources Assessment
country-level data64 (unweighted correlation 0.705, country
land-area-weighted
correlation 0.805; Extended Data Fig. 8). This explains strong
effects of forest loss
during the 2000–2014 period even though some species may not
yet have fully
felt the effects of this most recent loss (or had their status
updated accordingly).
Year of discovery. Newly described species are often from
remote areas (that is,
with initial high forest cover) where development is starting to
take place (dis-
covery was facilitated by access); such species are highly likely
to be classed as
threatened65. To explore how time since initial species
description influenced our
results, we conducted a sensitivity analysis including ‘year of
species description’ as
a predictor. We gleaned year of description from the taxonomic
authority sections
of Red List fact sheet accounts. For 18 of the species in our
analysis, two adjacent
years were reported (for example, “Highton, 1971 (1972)”). In
these cases, we used
the average of the two years. In addition to a main effect for

year, we included
the three-way forest loss × forest cover × year interaction.
This directly tests the
hypothesis that the initial intrusion effect (the statistical
interaction between forest
loss and cover) is mediated by the time when a species was
initially described,
with the expectation that most recently described species are
more likely to show
such effects. However, there was little statistical support for
this hypothesis; the
strength of the forest loss × forest cover interaction (our
primary focus) was largely
unchanged (Extended Data Fig. 3i).
Threshold for threatened species. It is possible that species in
different threat cat-
egories could respond in contrasting ways to forest loss. For
instance, we expected
species listed as Endangered and Critically Endangered to be
more likely to support
the habitat threshold hypothesis; these species only become
extremely threatened
when forest continues to be lost at high rates after most original
habitat has been lost.
Therefore, we tested effects of forest loss, forest amount and
their interaction on suc-
cessive levels of IUCN threat categories (Extended Data Fig. 9).
We compared model
results to those obtained when threatened species were taken to
be Endangered or
Critically Endangered species and Critically Endangered species
alone. Our overall
conclusions were consistent across threat categories (Extended
Data Fig. 9).
Data availability. Data that support the findings of this study

have been deposited
with figshare at:
https://doi.org/10.6084/m9.figshare.4955465.v4.
35. Butchart, S. H. et al. Improvements to the red list index.
PLoS ONE 2, e140 (2007).
36. Butchart, S. H. et al. Measuring global trends in the status
of biodiversity: red
list indices for birds. PLoS Biol. 2, e383 (2004).
37. Hoffmann, M. et al. The changing fates of the world’s
mammals. Phil. Trans. R.
Soc. Lond. B 366, 2598–2610 (2011).
38. BirdLife International. IUCN Red List for birds.
http://www.birdlife.org (2017).
39. Böhm, M. et al. The conservation status of the world’s
reptiles. Biol. Conserv.
157, 372–385 (2013).
40. Wildlife Conservation Society and Center for International
Earth Science
Information Network, Columbia University. Last of the Wild,
v2: Global Human
Footprint Dataset (Geographic).
http://dx.doi.org/10.7927/H4M61H5F (2005).
41. Venter, O. et al. Sixteen years of change in the global
terrestrial human
footprint and implications for biodiversity conservation. Nat.
Commun. 7,
12558 (2016).
42. Venter, O. et al. Data from: Global terrestrial Human
Footprint maps for 1993

and 2009. http://dx.doi.org/10.5061/dryad.052q5 (2016).
43. ESRI. ArcGIS Desktop: Release 10.1 (Environmental
Systems Research
Institute, 2012).
44. R Core Team. R: A Language and Environment for
Statistical Computing.
http://www.R-project.org/ (2013).
45. Seligman, M. Rborist: extensible, parallelizable
implementation of the random
forest algorithm. https://cran.r-
project.org/web/packages/Rborist/index.html
(2015).
46. Hurlbert, A. H. & Jetz, W. Species richness, hotspots, and
the scale dependence
of range maps in ecology and conservation. Proc. Natl Acad.
Sci. USA 104,
13384–13389 (2007).
47. Augustin, N. H., Mugglestone, M. A. & Buckland, S. T. An
autologistic model for
the spatial distribution of wildlife. J. Appl. Ecol. 33, 339–347
(1996).
48. Dormann, C. F. Assessing the validity of autologistic
regression. Ecol. Modell.
207, 234–242 (2007).
49. Bivand, R. et al. spdep: spatial dependence: weighting
schemes, statistics and
models. https://cran.r-
project.org/web/packages/spdep/index.html (2017).

50. Pike, N. Using false discovery rates for multiple
comparisons in ecology and
evolution. Methods Ecol. Evol. 2, 278–282 (2011).
51. Benjamini, Y. & Hochberg, Y. Controlling the false
discovery rate: a practical and
powerful approach to multiple testing. J. R. Stat. Soc. B 57,
289–300 (1995).
52. Swets, J. A. Signal Detection Theory and ROC Analysis in
Psychology and
Diagnostics: Collected Papers (Psychology Press, 2014).
53. United Nations. United Nations Statistics Division. Standard
Country and Area
Codes Classifications (M49).
http://unstats.un.org/unsd/methods/m49/
m49regin.htm (2013).
54. Bahn, V. & McGill, B. J. Testing the predictive performance
of distribution
models. Oikos 122, 321–331 (2013).
55. Gilleland, M. E. verification: weather forecast verification
utilities. https://
cran.r-project.org/web/packages/verification/index.html (2015).
56. Dray, S., Legendre, P. & Peres-Neto, P. R. Spatial
modelling: a comprehensive
framework for principal coordinate analysis of neighbour
matrices (PCNM).
Ecol. Modell. 196, 483–493 (2006).
57. Lee, D. CARBayes: an R package for Bayesian spatial
modelling with
conditional autoregressive priors. J. Stat. Softw. 55, 13 (2013).

58. McMillen, D. McSpatial: nonparametric spatial data
analysis. https://
cran.r-project.org/web/packages/McSpatial/index.html (2013).
59. Klier, T. & McMillen, D. P. Clustering of auto supplier
plants in the United States.
J. Bus. Econ. Stat. 26, 460–471 (2012).
60. Cardillo, M. et al. Multiple causes of high extinction risk in
large mammal
species. Science 309, 1239–1241 (2005).
61. Bates, D. M. lme4: Mixed-effects modeling with R.
http://lme4.r-forge.r-project.
org/book (2010).
62. Ives, A. R. & Garland, T. Jr. Phylogenetic logistic
regression for binary
dependent variables. Syst. Biol. 59, 9–26 (2010).
63. Petersen, R. et al. Mapping Tree Plantations with
Multispectral Imagery: Preliminary
Results for Seven Tropical Countries. (World Resources
Institute, 2016).
64. MacDicken, K. G. Global Forest Resources Assessment
2015: what, why and
how? For. Ecol. Manage. 352, 3–8 (2015).
65. Giam, X. et al. Reservoirs of richness: least disturbed
tropical forests are centres
of undescribed species diversity. Proc. R. Soc. Lond. B 279,
67–76 (2012).
66. Hijmans, R. J., Cameron, S. E., Parra, J. L., Jones, P. G. &

Jarvis, A. Very high
resolution interpolated climate surfaces for global land areas.
Int. J. Climatol.
25, 1965–1978 (2005).
https://doi.org/10.6084/m9.figshare.4955465.v4
http://www.birdlife.org
http://dx.doi.org/10.7927/H4M61H5F
http://dx.doi.org/10.5061/dryad.052q5
http://www.R-project.org/
https://cran.r-project.org/web/packages/Rborist/index.html
https://cran.r-project.org/web/packages/spdep/index.html
http://unstats.un.org/unsd/methods/m49/m49regin.htm
http://unstats.un.org/unsd/methods/m49/m49regin.htm
https://cran.r-project.org/web/packages/verification/index.html
https://cran.r-project.org/web/packages/verification/index.html
https://cran.r-project.org/web/packages/McSpatial/index.html
https://cran.r-project.org/web/packages/McSpatial/index.html
http://lme4.r-forge.r-project.org/book
http://lme4.r-forge.r-project.org/book
letterreSeArCH
Extended Data Figure 1 | Receiver operating characteristic
(ROC)
curves for the models predicting status of forest exclusive
species. Class
was included as a fixed effect (as in our main results) for the
‘All species’
group. The other results (by class) are based on models fit to
each class
separately. The left column is based on results where the model

was fit to
the entire dataset. The right column shows ROC curves for
predictions
using a fourfold cross-validation scheme where the probability
of species
being threatened was predicted for each of four regions with the
model
fit using data from all other regions. P values are based on the
Mann–
Whitney U statistic and test whether the population AUC is
greater than
0.5 (that is, better than random predictions). Results are
presented both
with (bottom row) and without (top row) the spatial
autocovariate.
letter reSeArCH
Extended Data Figure 2 | Model results for models fit by class
(mammals, amphibians, birds) and for all classes together (All).
Each row shows
standardized coefficient estimates and 95% confidence intervals
(as error bars) for each single model. All covariates are shown
in this figure.

letterreSeArCH
Extended Data Figure 3 | Sensitivity analysis results. The
plotted
variable is the estimated standardized coefficient for the forest
loss × cover
term with 95% confidence interval (as error bars). Each column
corresponds to a different sensitivity analysis (other covariates
are not
shown). a–i, In general, we found that our primary results were
robust to
the inclusion of absolute latitude as a predictor variable (a), the
restriction
of the dataset to tropical species only (b), the exclusion of
species listed as
threatened based on small geographic range (c), using a 75%
pixel-scale
threshold for the forest loss and forest cover variables (d),
standardizing
forest loss and gain by forest cover (that is, dividing forest loss
and gain
by forest cover so that these variables can be interpreted as
approximate
percentages of species’ forested range) (e), accounting for
potential
phylogenetic dependence using generalized linear mixed models
with
random intercepts by taxonomic order (and by class for the ‘all
species’
model) (f), using high-resolution species’ range maps and
covariate
maps (approximately 5 km), clipping species ranges based on
altitudinal
limits, and setting forest loss and cover to zero in regions of
known tree

plantations (g), including forest loss and gain from 1990–2000
by adding
1990–2000 and 2000–2014 forest change variables (h), and the
inclusion
of year of initial species description as a main effect and in a
three-way
interaction term with forest loss × cover (i).
letter reSeArCH
Extended Data Figure 4 | Estimated standardized coefficients
for
each model term (with 95% confidence intervals as error bars)
when
a quadratic forest loss × cover2 interaction (forest loss ×
cover2)
is included in the model. This allows for the effect of loss to
vary
quadratically with cover. A significant and positive forest loss ×
cover2
interaction term would suggest that the (negative) effects of
forest loss
are greatest in areas with both high and low proportions of
forest cover.
However, this term was non-significant for most taxa and
response
variables, indicating that the linear model for the interaction is
more
parsimonious.

letterreSeArCH
Extended Data Figure 5 | The effect of forest loss (for 2%
additional
loss) in relation to total forest cover using quadratic models.
These
models allow the effect of forest loss to vary nonlinearly as a
function
of forest cover, allowing us to test the hypothesis that forest
loss is
detrimental to species at both high and low levels of forest
cover. However,
the quadratic model reveals very similar results to the linear
model.
The exception is when ‘declining trend’ is used as the response;
species’
populations were more likely to be in decline when forest
amount is very
low (the habitat threshold hypothesis), and upon initial
intrusion into
intact forests (the initial intrusion hypothesis). For statistical
significance
of the quadratic models, see confidence intervals in Extended
Data
Fig. 4, far right panel. For context, the histograms (grey bars)
show the
(normalized to maximum 100%) distributions of forest cover
across
species. For example, if one bar in a panel is twice as high as
another, then

twice as many species have average forest cover of this
percentage in their
ranges. The black lines show the cumulative percentages of
species with
at most x per cent forest cover. For example, approximately half
of forest-
optional species have 50% forest cover or less.
letter reSeArCH
Extended Data Figure 6 | Results of multiple spatial models
(estimates
and 95% confidence intervals as error bars) for forest exclusive
species
when status (that is, whether or not a species is threatened) is
used as
the response. Coefficients across multiple models that account
for spatial
autocorrelation were very similar. ‘Method’ indicates the
procedure
(if any) used to account for spatial autocorrelation: non-spatial
ordinary
GLM (non_spatial), autologistic model with spatial
autocovariate (AL_b),
autologistic model using 50 nearest neighbours in the spatial
weights
matrix (AL_b_50), Moran eigenvector filtering (filtering),
spatial
autoregressive model (SAR_approx), or Bayesian condition
autoregressive

model (CAR_Bayes). Details on each method are given in the
sensitivity
analyses section of the Methods.
letterreSeArCH
Extended Data Figure 7 | Relationship between forest loss
1990–2000
(from ref. 34) and 2000–2014 (from ref. 7). Overall, rates of
forest loss
are temporally autocorrelated; species ranges with high forest
loss in
the 1990s also show high forest loss in 2000s. However, this
relationship
is strongly affected by data availability; approximately 12.1%
of forest
loss data are missing across the globe and as we expected, the
more data
missing from a species range, the weaker the relationship
between 1990s
and 2000s rates of forest loss. The plots show correlations (in
red; top right
of each panel) between forest loss across the two time periods
for various
levels of missing data. Each point corresponds to a single
species and the x
and y axis values indicate average values of each variable
across its range.
Panel titles show the proportion of missing 1990–2000 forest
loss data in

species ranges. For example, the top left panel contains results
for species
with between 0% and 4% of their ranges missing 1990 forest
data (owing
to clouds, lack of satellite coverage, and so on). The correlation
between
1990–2000 and 2000–2014 forest loss is highest for species
with the least
missing data.
letter reSeArCH
Extended Data Figure 8 | Country-level forest net loss (that is,
change
in percentage forest cover) for the 1990–2000 and 2000–2015
periods
according to the Food and Agriculture Organization’s (FAO)
Global
Forest Resources Assessment. Based on these data, the
correlation between
1990–2000 and 2000–2015 forest loss is 0.705. Weighting by
country area
increases the correlation to 0.805. The relatively high
correlation suggests
that the spatially explicit recent (2000–2014) forest loss data
that we used is
closely related to less recent (1990–2000) forest loss.

letterreSeArCH
Extended Data Figure 9 | Sensitivity of our results to alternative
categories of threat. In the main text we considered a species to
be
‘threatened’ if it fell into the IUCN Red List category
Vulnerable,
Endangered or Critically Endangered. We conducted further
analysis
considering as threatened only species that are Endangered and
Critically
Endangered, and again for only species that are Critically
Endangered.
Dots show estimated standardized coefficients for each model
term (with
95% confidence intervals as error bars) for all main effects and
the forest
loss × cover interaction term. Our overall conclusions were
consistent
across these different definitions of threat.
letter reSeArCH
Extended Data Figure 10 | Maps showing the methods used to
quantify
historical forest loss. First, we used random forests (a machine -
learning

method) to estimate potential forest cover globally (within
forest
biomes33). a–c, This model was fit using current forest cover
within intact
forest landscapes36 and bioclimatic and other predictor
variables66 (a; see
Methods). We then subtracted current forest cover (b; Hansen et
al.6) from
this map to obtain estimated historical forest loss (c). The map
of land is
taken from
http://thematicmapping.org/downloads/world_borders.php.
http://thematicmapping.org/downloads/world_borders.php
Reproduced with permission of copyright owner.
Further reproduction prohibited without permission.
Global forest loss disproportionately erodes biodiversity in
intact landscapes
AuthorsAbstractReferencesAcknowledgementsAuthor
ContributionsFigure 1 Spatial distribution of the six variables
used to predict species’ IUCN Red List response
variables.Figure 2 Effects of four predictors on the status of
19,432 vertebrate species worldwide.Figure 3 Predicted
probabilities of species status as a function of recent forest loss
and total forest cover within a species range.Figure 4 Projected
increases in the number of threatened species under three
scenarios of future forest loss.Extended Data Figure 1 Receiver
operating characteristic (ROC) curves for the models predicting
status of forest exclusive species.Extended Data Figure 2 Model
results for models fit by class (mammals, amphibians, birds)

and for all classes together (All).Extended Data Figure 3
Sensitivity analysis results.Extended Data Figure 4 Estimated
standardized coefficients for each model term (with 95%
confidence intervals as error bars) when a quadratic forest loss
× cover2 interaction (forest loss × cover2) is included in the
model.Extended Data Figure 5 The effect of forest loss (for 2%
additional loss) in relation to total forest cover using quadratic
models.Extended Data Figure 6 Results of multiple spatial
models (estimates and 95% confidence intervals as error bars)
for forest exclusive species when status (that is, whether or not
a species is threatened) is used as the response.Extended Data
Figure 7 Relationship between forest loss 1990–2000 (from
ref.Extended Data Figure 8 Country-level forest net loss (that
is, change in percentage forest cover) for the 1990–2000 and
2000–2015 periods according to the Food and Agriculture
Organization’s (FAO) Global Forest Resources
Assessment.Extended Data Figure 9 Sensitivity of our results to
alternative categories of threat.Extended Data Figure 10 Maps
showing the methods used to quantify historical forest loss.
Contents lists available at ScienceDirect
Journal of Environmental Management
journal homepage: www.elsevier.com/locate/jenvman
Research article
Modelling the washoff of pollutants in various forms from an
urban
catchment
Jarrod Gauta,∗ , Lloyd HC. Chuaa, Kim N. Irvineb, Song Ha

Lec
a School of Engineering, Faculty of Science Engineering &
Built Environment, Deakin University, 75 Pigdons Road, Waurn
Ponds, VIC, 3220, Australia
b National Institute of Education and Nanyang Environment and
Water Research Institute Nanyang Technological University, 1
Nanyang Walk, 637616, Singapore
c Surbana Jurong Pte Ltd, Coastal Engineering, Infrastructure
and Land Survey Department, #01-01 Connection One, 168
Jalan Bukit Merah, 150168, Singapore
A B S T R A C T
The exponential washoff model was originally developed based
on observations of particulate pollutants, however, its
applicability when applied to different forms of
pollutants is not well understood. Data from a previous study of
6 stormwater pollutants from 126 events at 12 sites in Singapore
was used for event based model
parameter calibration using a Monte Carlo technique. The
accuracy of the calibrated exponential washoff model was
clearly best for particulate pollutant total
suspended solids (TSS), and worst for dissolved pollutants
Ortho-Phosphate (PO4), nitrate (NO3) and ammonium-nitrogen
(NH4). Model accuracy for mixed forms of
pollutants total Phosphorus (TP) and total Nitrogen (TN) were
in between these two extremes. Relationships between model
parameters with rainfall and flow
characteristics were also investigated. Statistically significant
relationships could only be found for TSS, where the total
rainfall depth was identified as being the
most significant variable to explain model parameter behaviour.
Antecedent dry period (ADP) was shown to have little or no
importance across all land uses and
pollutant forms. The results showed that the model parameter

behaviour could be explained only for particulate pol lutants and
small (≤10 ha) sub-catchments, and
that replicating washoff of mixed or dissolved forms of
pollutants as a fraction of solids is likely to lead to misleading
results.
1. Introduction
Stormwater runoff from urban areas has been identified as the
lar-
gest nonpoint source of pollutants entering waterbodies.
Stormwater
runoff models for flow prediction have been well researched
and are
supported by much field data, resulting in a high benchmark for
ac-
curacy in urban runoff prediction (Imteaz et al., 2012). The
modelling
of stormwater quality is still less understood with a “vast
number of
complex, interrelated processes influencing urban stormwater
quality”
(May and Sivakumar, 2008).
Although different models have been developed for prediction
of
pollutant washoff and associated changes to stormwater quality,
it re-
mains the case that high levels of confidence have not been
demon-
strated in their predictive ability. A proper understanding of the
model
parameters, and the relationship of these parameters with
different
rainfall conditions and physical catchment characteristics is still
lacking, with some existing descriptive views of the processes

described
as being inadequate (Duncan, 1995a, 1995b; Shaw et al., 2010).
Novotny (1994) describes that the prediction of total event load
from
urban runoff may be more important than the accurate
quantification of
event flux regarding the assessment of the impact on receiving
waters.
However, the accurate quantification and an understanding of
the
within-event behaviour remains an important prerequisite to
effectively
design treatment interventions to manage urban stormwater
(Francey
et al., 2010).
Washoff behaviour is commonly modelled as an exponential
decay
function, based on earlier experimental observations in studies
by
Metcalf & Eddy (1971) and Sartor and Boyd (1972). This
empirically
calibrated washoff model has been widely applied and verified
for
particulate pollutants, with a large number of studies focusing
on this
application (Bonhomme and Petrucci, 2017; Charbeneau and
Barrett,
1998; Gaume et al., 1998; Le et al., 2017; Wicke et al., 2012).
Although
the original focus of the exponential model was on particulates,
in
practice its use has also been extended to model dissolved
pollutants by
applying a potency factor (simple ratio) to the washoff model

predic-
tions of particulate pollutants (Rossman, 2015). Consequently,
there
remains “inadequate knowledge as to whether or not they are
suitable
for dissolved pollutants” (Xiao et al., 2017). This is because the
practice
of adopting the washoff of particulates as a surrogate for
pollutants in
the dissolved or mixed (particulate and dissolved) form is
tenuous since
the washoff behaviour of dissolved and particulate substances
can be
expected to be different (Miguntanna et al., 2013; Xiao et al.,
2017).
Most recently, the suitability of the exponential washoff model
for
dissolved pollutants has been investigated in research by Xiao
et al.
(2017), including working toward a new modelling approach
based on
their laboratory observations, but with upscaling to the
catchment scale
identified as a major challenge.
An added complexity in the use of the exponential model is
related
Received 15 January 2019; Received in revised form 24 May
2019; Accepted 25 May 2019
∗ Corresponding author.
E-mail addresses: [email protected] (J. Gaut), [email protected]
(L.H. Chua).

Journal of Environmental Management 246 (2019) 374–383
Available online 10 June 2019
0301-4797/ © 2019 Elsevier Ltd. All rights reserved.
T
http://www.sciencedirect.com/science/journal/03014797
https://www.elsevier.com/locate/jenvman
http://crossmark.crossref.org/dialog/?doi=10.1016/j.jenvman.20
19.05.118&domain=pdf
to the fact the model is empirical. The original washoff
observations
and model development were based on experimental conditions
and
particulate pollutants. That original work by the U.S.
Environmental
Protection Agency (1983) and consequent research studies such
as
Goonetilleke et al. (2005) have shown that the particle size of
the
pollutant may indeed affect model parameters and accuracy.
Energy
from rainfall and runoff have traditionally been thought to best
re-
present the total energy available for both detachment and
transport,
but it is unknown to what extent the washoff parameters are
related to

the rainfall and runoff conditions, and whether any dependency
varies
on the basis of the form of pollutant. An understanding of the
behaviour
of washoff models that are currently in use for pollutants in
different
forms, as a function of rainfall and flow characteristics, can
allow these
models to be used with increased confidence.
The objectives of this study were: (i) establish a range of
calibrated
parameters for the washoff model as a function of the form of
pollutant,
defined as either particulate, dissolved or mixed; (ii) deduce the
sig-
nificance of rainfall and flow variables on washoff behaviour
for these
forms of pollutants; and (iii) assess the appropriateness of the
washoff
model for pollutants in their different forms. The results of this
study
provide insights on the strength and limitations of washoff
modelling,
and shed light on the variation of model parameters as a
function of
rainfall and flow variables and thus help to explore the
significant
processes governing washoff.
2. Data used
Data from Le (2014) where a total of 1424 discrete stormwater
samples, collected from 126 rain events between November
2006 and
October 2012 across 12 sites in the north-west of Singapore

were
analysed for this study. In that study, a total of six (four at
fixed loca-
tions and two mobile) RIMCO 8020 tipping bucket rain gauges
were
used to collect rainfall data. Storm samples were collected from
au-
tonomous sampling stations and analysed for concentrations of
6 dif-
ferent stormwater pollutants: total suspended solids (TSS); total
Phos-
phorus (TP); total Nitrogen (TN); Ortho-Phosphate (PO4);
Nitrate
(NO3); and ammonium-nitrogen (NH4). ISCO 2150 flow-area-
velocity
sensors were used to obtain flow data at 2 min intervals, to cope
with
the rapid runoff response from urban catchments in the tropics.
The
sampling intervals were set to ensure that samples were
collected
during the rising limb and peak of the hydrograph. For a
sampled event
to be deemed representative and therefore suitable for data
analysis,
the sampled volume needed to have covered at least 70% of the
total
runoff volume with a minimum of 8 samples collected. Pertinent
in-
formation on the sampling sites, and the rainfall events for each
site are
provided in Table 1. Further information on the sampling study
can be
found in Le et al. (2017). The authors note the exclusion of the
catch-
ment area from Table 1 at the request of the data owners.

3. Methodology
3.1. Washoff modelling
Washoff is described as the process by which accumulated
pollu-
tants on a surface are removed by rainfall and runoff (Duncan,
1995b).
Washoff is triggered by rain impact as well as the runoff it
causes. In
this sense, washoff of surface pollutants is caused by either of,
or a
combination of two processes being (i) kinetic energy from
rainfall
impact and/or (ii) shear stress applied to accumulated surface
pollu-
tants by the surface runoff. James and Irvine (1992) reviewed
erosion
processes in urban areas and noted that previous research (e.g.
Moss,
1988) showed that overland transport may be more important
when the
interaction between rainfall and shallow overland flow creates
greater
erosive and transport capacity than either rainfall acting on a
surface in
the absence of flow or overland flow acting alone. This
phenomenon
becomes less important with deeper overland flow depths.
Earlier studies by Metcalf & Eddy (1971) and Sartor and Boyd
(1972) assumed that the mass of pollutant washed off a surface
in any
time interval is proportional to the mass remaining on the
surface. A

first-order decay model was therefore proposed. A major
drawback of
the first-order decay model is that the model will always predict
a de-
creasing concentration, which is sometimes at odds with field
ob-
servations (Duncan, 1995b; Yaziz et al., 1989). A variation of
the ori-
ginal exponential formula that is able to predict the first flush
has since
been proposed. Version 4 of the Stormwater Management Model
or
SWMM by the U.S. Environmental Protection Agency (2017)
adopts the
following empirical equation for the washoff rate, W, expressed
as mass
per unit time in mg/h or similar (Rossman, 2015):
=W c q Bc3 4 (1)
where c3 is the washoff coefficient, c4 is the washoff exponent,
q is the
runoff rate per unit area (in/h or mm/h), and B is the mass
remaining
on the surface. Note that units for c3 are also dependent on the
value
adopted for c4, which contributes toward making the exact
physical
processes these coefficients represent difficult to understand.
Table 1
Summary of monitored catchments and rainfall event statistics.
Site Dominant land use No of events sampled Ave rainfall depth
± 1 std dev.(mm) Ave intensity, ± 1 std dev.(mm/h)

RES1 Residential 13 17.8 ± 16.0 10.7 ± 7.8
RES2 Residential 10 21.5 ± 19.4 13.8 ± 11.5
RES3 Residential 12 14.3 ± 14.7 9.1 ± 8.9
RES4 Residential 10 28.2 ± 30.3 23.1 ± 16.9
MIX1 Residential/Commercial 6 19.4 ± 20.7 16.2 ± 15.0
MIX2 Forest/Agricultural 6 47.2 ± 74.3 17.5 ± 17.3
MIX3 Residential/Commercial/Forest 13 32.3 ± 20.2 14.7 ±
11.0
FOR1 Forest 11 34.7 ± 17.2 21.3 ± 18.4
AG1 Agricultural 12 18.1 ± 21.3 13.2 ± 6.4
AG2 Agricultural 9 33.7 ± 17.9 15.2 ± 10.5
AG3 Agricultural 13 24.7 ± 21.5 10.8 ± 7.0
AG4 Agricultural 11 23.2 ± 26.7 8.5 ± 7.3
Table 2
Exponential washoff model parameters from previous studies.
Suggested Parameter
Ranges
Comments Reference
c3, 0.001–1.0 Modelling TSS, using same
data set as this study
Le (2014)
c4, 0.9–2.0
c3, 0.001–0.5 (0.014) Modelling SS, best fitting
values shown in bold
Gaume et al. (1998)
c4, 0.5–2.5 (1.6)
c3, 0.2–0.34 (0.27) Modelling TSS and heavy
metals, best fitting value for

TSS shown in bold
Wicke et al. (2012)
c4, 1.0
c3, 0.01–10 Modelling TSS Bonhomme and
Petrucci (2017)c4, 0.2–3
J. Gaut, et al. Journal of Environmental Management 246 (2019)
374–383
375
Fig. 1. Box plots of best fitting values for model parameters c3,
c4, and Bini for particulates (TSS, left), mixed (TN, middle)
and dissolved pollutants (NH4, right).
374–383
376
The washoff model is typically coupled with a buildup function
to
predict the initial mass of pollutant Bini on the surface at the
start of a
rain event generally as some function of the antecedent dry
period
(ADP), forming a build-up washoff (BUWO) model.
Interestingly, stu-
dies such as those by Egodawatta et al. (2007) have shown that
the total

amount of pollutant washed off may often be only a fraction of
the
pollutant mass available on the surface before the rain event,
indicating
that washoff, not buildup, may most often be the limiting
process in
pollutant runoff. Washoff behaviour in the tropics is also
expected to be
different from temperate areas due to the differences in rainfall
char-
acteristics (Le et al., 2017). There have however to date been
relatively
limited detailed studies on washoff completed in tropical
catchments
and of dissolved pollutants.
3.2. Monte Carlo Analysis
The model parameters (c3, c4 and Bini) were calibrated using
the
field measured values of W for TSS, TP, TN, PO4, NO3 and
NH4 on an
individual event basis using a Monte Carlo approach. Values of
q
measured in the field were used as this removes uncertainties
in-
troduced when relying on a runoff model as part of the washoff
model
calibration process. Secondly, although it is common practice to
relate
Bini at the start of an event to the antecedent dry period (ADP),
previous
studies such as that by Shaw et al. (2010) have indicated that
Bini may
have little or no relationship with the antecedent dry period
(ADP) as

the values tend to be higher for larger rain events. Any washoff
de-
pendence on the ADP has also been shown to be weak in other
studies
such as that by Egodawatta et al. (2007); Pitt et al. (2005),
particularly
under tropical conditions (Le et al., 2017). Thus, although Bini
is usually
estimated based on buildup models, it was decided to include
para-
meter Bini in the Monte Carlo calibration of Eqn. (1), to ensure
any
factors affecting the best values can be studied independently,
and
avoid uncertainties introduced when a value for this parameter
is es-
timated using a separate buildup model.
This method of event based washoff model parameter
calibration
differs to that used in some previous studies of this model, for
example
by Chow et al. (2011), where the authors equated Bini to the
measured
washed off mass for each event, and not via a calibration
process as
adopted in this study. By equating Bini to the observed mass,
the un-
derlying assumption is that all available material represented by
Bini is
removed under any rainfall conditions (i.e. for all events). This
as-
sumption may be plausible for particulates, however, the
justification
for pollutants in other forms is less strong. Therefore, any

dependence
of Bini on rainfall and flow characteristics, and the equality of
Bini with
observations, will result from analysis arising from the current
treat-
ment of Bini as a calibration parameter.
In accordance with Eqn. (1), if the initial washoff rate is:
=W c q Bini c ini3 4 (2)
then, the washoff rate Wi at t = ti is:
= −− −W c q B W t. ( . Δ )i i
c
i i3 1 1
4 (3)
where Δt is the model time step, and Bi-1 is the mass remaining
at t = ti-
1. The entire loadograph for any given event can be constructed
by
successive application of Eqn. (3) for a given set of model
parameters
(c3, c4 and Bini). The parameter set that produced the best
prediction of
the W time series for each event was estimated using the Monte
Carlo
approach, in a process similar to that used by Le et al. (2017)
which
involved the creation of random ‘trial’ parameter sets from
within a
predefined, uniformly distributed parameter space. Ranges
(upper and
lower bound) for parameters c3 and c4 of 0.001–1.0 and 0.1–

2.0, re-
spectively, were adopted as suggested in previous studies listed
in
Table 2. Although based on particulates, these values were
considered
as reasonable starting points for the mixed and dissolved forms.
Given
that parameter B is normalised by the catchment area, initial
upper
limits for Bini of 100 kg/ha for TSS and 10 kg/ha for all other
pollutants
were adopted based on published literature of stormwater
pollutant
concentrations (Duncan, 1999; Fletcher et al., 2004; Miguntanna
et al.,
2013; U.S. Environmental Protection Agency, 1983).
For a given event, 10,000 parameter sets were generated
randomly
and applied in the model, and the parameter set that provided
the
smallest Standard-Squares Error (SSE) was identified. This
process was
repeated 10 times, and the 10 best parameter sets recorded.
Next, the
initial parameter ranges were then adjusted to extend 20%
beyond the
range of these 10 best fitting parameter sets, and the process
was re-
peated. Where any of the best parameter sets were seen to be
limited by
the upper or lower bounds, the range was increased accordingly.
The
average of the parameter values from this second list of 10
parameters
sets was adopted as the model parameters, calibrated for that

event. In
this way, the process ensures 200,000 parameter sets
(acceptance level
of 0.005%) were reviewed for each event with the latter 100,000
fo-
cusing in the region identified as having a higher likelihood of a
good
model fit in a method similar to the Shuffled Complex
Evolution
Metropolis Algorithm described by Dotto et al. (2012). The
Nash-
Fig. 2. Box plots of ME and NS scores (all events and all sites
combined).
374–383
377
Sutcliffe efficiency (NS) and Mass-Error (ME) were then
calculated for
each event, based on the event specific set of model parameters
ob-
tained. NS provides a measure of how well the predicted time
series fits
with the observations whereas ME provides a measure of the
bulk mass
balance. Having independent error measures calculated after and
out-
side the calibration process (which was instead based on the
SSE)
provides a second reference and assessment of model
performance.

4. Results and discussion
4.1. Monte Carlo Analysis
The analysis carried out in this study included calibration of the
model for each of the six pollutants across the 126 rainfall
events. This
resulted in a total of 711 independently calibrated parameters
sets for
c3, c4 and Bini. The results for the best fitting parameters for
each event
at the 12 sites for selected particulate (TSS), mixed (TN) and
dissolved
pollutants (NH4) are presented as box plots in Fig. 1. On each
box, the
dotted circle shows the median value, and the left and right
edges of the
box indicate the 25th and 75th percentiles, respectively. The
whiskers
extend to the most extreme data points not considered outliers,
and
these outliers are denoted using the circle symbol. The results
reflected
findings from earlier studies of washoff models by Jewell and
Adrian
(1982), and showed that the event based values for c3 and Bini
varied
greatly, sometimes over three orders of magnitude for the same
pollu-
tant at the same site. In contrast, c4 typically centred around a
median
of 1.0 for all pollutants, with no clear difference between
particulate
and dissolved forms. The inter quartile ranges for all parameters

are
also noticeably larger for dissolved pollutants than particulates,
in-
dicating more spread or variance in the values for best fitting
model
parameters for different events at the same site.
The boxplots of the event based NS and ME scores are
combined
across all sites in Fig. 2 with the event producing the median
results
indicated by the dotted circle. The ME results are similar across
all
pollutants, with the inter quartile range slightly larger for
dissolved
pollutants. The median values were close to zero in all cases,
showing
good mass conservation. However, the NS scores show a
distinct trend
from being the best for particulate (TSS) and worst for
dissolved pol-
lutants (NO3, PO4 and NH4) indicating that although the
overall mass is
preserved reasonably well, there was a tendency for better
model
Fig. 3. Observed and simulated pollutographs for the different
pollutant forms (event selection based on median NS scores).
374–383
378

representation of within-event washoff behaviour for
particulates. TN
and TP which contain both particulate and dissolved fractions
has NS
scores between those of particulate and dissolved pollutants.
The ac-
ceptance limits adopted for all consequent analysis ar e also
shown with
dashed red lines, defined as within |30%| for ME and at least 0.7
for NS.
Events with ME and NS scores outside these ranges were not
considered
for subsequent analysis. An additional criterion of the total
rainfall
depth of the event being at least 5 mm was also applied, based
on re-
commendations in the earlier study of this dataset (Le et al.,
2017).
Given there was only one forest site, this is also excluded from
sub-
sequent analysis.
The model results for W are compared with the observed data
for
events with the median NS score, in Fig. 3. The 5% and 95%
quantiles
were found from the 100,000 model predictions (based on the
initial
100,000 trial parameter sets), and the 90% prediction bounds
are
shown as the shaded areas in Fig. 3. The darker grey bands
indicate the
90% prediction bounds for the best 10-parameter sets, with the
solid
black line indicating the simulated washoff rate using the
calibrated

parameter set (average of the 10 best parameter sets) adopted
for the
event. The plots show a closer fit between observed and
modelled va-
lues of W for particulate and mixed pollutants TSS, TP and TN,
than for
dissolved pollutants PO4, NO3 and NH4. This corresponds with
the
median NS scores for these six pollutants of 0.92, 0.92, 0.93,
0.74, 0.68
and 0.78 respectively. The 95% confidence intervals of the 10
best
parameter sets for each pollutant are also observed to
incorporate more
of the observed values of W when modelling particulate and
mixed
pollutants, than when modelling those that are dissolved.
4.2. Sensitivity analysis
Fig. 4 compares the SSE results for the different model
parameter
values within the Monte Carlo Analysis for selected particulate
(TSS),
mixed (TP) and dissolved (PO4) forms of pollutants. The results
shown
in Fig. 4 are based on data obtained on 19 Jan 2012 at RES1,
where the
total rainfall depth recorded was 18 mm, very close to the
median of
17.8 mm at this site. The ME score for each parameter value is
also
shown based on the colorbar provided. The stronger parabolic
shape
and more localised SSE minimum for TSS than PO4 indicates

that the
Monte Carlo calibration was able to identify a more definite
combina-
tion of c3, c4 and Bini values for particulates than dissolved
pollutants.
TP which has both particulate and dissolved portions falls in
between
these two extremes.
4.3. Dependence of model parameters with rainfall and flow
variables
The study aimed to understand how the washoff parameters are
related to the rainfall and runoff conditions, and whether these
re-
lationships are a function of the form of the pollutant. A series
of sta-
tistical analyses was thus conducted. Firstly, the Kruskal -Wallis
test was
used to test if the model parameters for the same pollutants and
sites
with the same land use could be combined for further analysis.
The
Kruskal-Wallis test looks at whether the means of the parameter
values
for each pollutant at each site are significantly different from
each other
(William and Wallis, 1952). This test is preferred as it is non-
para-
metric, without requiring an understanding of the sample
distributions.
The results indicate that sample means for the same pollutant at
dif-
ferent sites with the same land use were similar, such that only
a small
number (less than 5% of the 198 sets) of samples were found to

be
significantly different at p = .05. Therefore, the sample sets of
cali-
brated parameter values for the same pollutant and land use
could be
combined for further analysis.
The Lilliefors test was then used on the combined parameter
sets to
determine the representative distributions of c3, c4 and Bini
(Hubert,
1967). The normal, lognormal and exponential distributions
were
tested. The representative distribution for each parameter was
identi-
fied as the one with the highest mean p value (averaged over all
pol-
lutants and land uses) which was log-normal for c3 and Bini,
and normal
for c4. Next, a one-way analysis of variance (ANOVA) was
conducted to
check if the combined parameter sample sets were significantly
dif-
ferent to each other. This was performed with regard to both
pollutant
Fig. 4. SSE results for c3, c4 and Bini for particulates (TSS,
left), mixed (TP, middle) and dissolved (PO4, right) forms of
pollutants. The colorbar shows corresponding
ME score for each trial parameter value. Event was on 19 Jan
2012 at CCK Ave where total rainfall depth was 18 mm.
374–383
379

type and land use. Parameters c3 and Bini were log-transformed
to suit
the distributions indicated by the Lilliefors tests. The key
results from
this analysis are as follows: (i) For parameter c3 the only
significant
difference between ranges was for TSS, between land uses:
mixed and
residential (p value = .018), and mixed and agricultural (p value
<
.001). (ii) The ranges of c4 values were not significantly
different be-
tween any pollutants. Nor did the values show any change
associated
with land uses which agrees with earlier studies for TSS in Le
et al.
(2017). (iii) For ranges of Bini, all pollutants showed
significantly
different ranges between residential and agricultural land uses,
with the
ranges for mixed land use generally falling between these
extremes.
This observation is consistent with the differences between the
amount
of pervious surfaces and material available for washoff for these
types
of land use.
The Lilliefors test was also used to determine the closest re -
presentative distribution of rainfall variables including the total
event
rainfall depth (d) in mm, average event rainfall intensity (iave)

in mm/h,
maximum 5 min event intensity (imax) in mm/h, the antecedent
dry
Fig. 5. Correlation matrices of c3, c4 and Bini with rainfall and
flow variables for catchments with residential land use. Log
transforms were used for model parameter
and rainfall variable values where appropriate based on
representative distributions.
374–383
380
period (ADP) in days, maximum runoff rate (Qmax) in m
3/s and total
flow (V) in m3. These were determined as being lognormal for
all
variables with the exception of d, which was exponentially
distributed,
and imax which was normally distributed. Correlation analysis
between
the three model parameters with rainfall and flow variables was
per-
formed using the CORRPLOT tool within MATLAB, including
any log
transforms of parameters and rainfall variables as appropriate
based on
their distributions. This was completed for each pollutant and
land use,
creating a total of 18 correlation matrices. The correlation
matrices and

Pearson's r value for each pollutant for catchments with
predominantly
residential land use is shown in Fig. 5. The Pearson's r values
are in-
dicated in red if the p < .05. A summary of the correlations that
were
significant at the p < .05 level, and that had Pearson's r values
of at
least 0.5 is listed in Table 3.
The number of significant (p < .05 and r ≥ 0.5) correlations be -
tween model parameters and rainfall and flow variables was
strongest
for particulates (TSS) and seemingly decreased based on the
particulate
fraction, with fully dissolved pollutants being the least
correlated. The
largest occurrence of correlations between rainfall variables and
model
parameters c3 and Bini for particulate pollutant TSS was with
rainfall
depth (d), which occurred for all three types of land use (giving
the
total of six shown in Table 3). This aligns with the results of the
pre-
vious study of this dataset by (Le et al., 2017). There were no
corre-
lations between ADP and any model parameters for any
pollutant,
casting further doubt on how important the build-up process is
with
regard to predicting pollutant washoff, particularly in tropical
catch-
ments. Parameter c4 had the least number of correlations with
the

rainfall variables, and can be seen as further evidence that it
was the
parameter the model was least sensitive to. Correlations were
also
found between model parameters and flow variables qmax and
V,
however rainfall data may represent the best opportunity for
model
calibration and application in practice, given the comparative
ease of
collecting the data.
4.4. Total mass at the start of an event
The results show that there was a strong correlation between
Bini
with rainfall and flow variables, but no significant correlations
of Bini
with ADP. The main reason for this could be the relatively short
dry
periods between rain events in the tropics, which raises
questions on
the use of time dependent buildup models in tropical
catchments. Since
Bini can generally interpreted as the “unit mass available for
washoff”,
see Le (2014), this would imply that M’ (= Mobs/MB) can be
defined as
the ratio of the observed washoff mass, Mobs to the “mass
available for
washoff”, MB where:
∑=M W t( . Δ )obs obs i i, (4)
=M B A.B ini (5)

and A is the area of the catchment. An M′ value of 1.0 indicates
that all
the mass available on the surface at the start of the event
(represented
by Bini) was removed. In contrast, where values of M’ < 1, the
total
observed washoff mass was less than the amount available, as
re-
presented by Bini. Based on this analysis, it was found that the
median
value of Μ′ was close to 1.0 for particulates, and lowest for
dissolved
pollutants, with TP and TN falling somewhere in between these
ex-
tremes, with no observed significant differences in this ratio
based on
land use.
The relationship of M′ with catchment size is presented in Fig.
6
where scatterplots of M′ are shown for particulates (TSS),
mixed (TP
and TN) and dissolved pollutants (PO4, NO3 and NH4) for all
events.
The figure indicates that for particulate pollutant TSS, there is a
clear
trend of W′ decreasing as the catchment size increases (r2 of
0.36 and
p < .001). The results also show that for catchments less than 10
ha,
the W′ value for TSS centred around a mean of 0.98, with
narrow 95%
Table 3
Total number of significant correlations (defined as where p <
.05 and

r ≥ 0.5) of model parameter values (c3,c4,Bini) with rainfall and
flow variables
for all land uses.
rainfall and flow variable TSS TP TN PO4 NO3 NH4
d (mm)1 6 2 0 0 0 0
iave (mm/h)
1 3 2 0 0 0 0
imax (mm/h) 3 2 0 0 0 0
ADP (days)1 0 0 0 0 0 0
Qmax (m
3/s)1 3 2 2 0 0 1
V (m3)1 3 3 2 0 0 1
Total number of correlations 18 11 4 0 0 2
Note1: log transform used for correlation analysis.
Fig. 6. Scatterplot of M′ with catchment size for TSS (115
events, left), mixed pollutants TP and TN (220 events, middle)
and dissolved pollutants PO4, NO3 and NH4
(306 events, right). Correlation trends shown in red. (For
interpretation of the references to color in this figure legend,
the reader is referred to the Web version of this
article.)
374–383
381
confidence interval limits of 0.92–1.04. This indicates that for

smaller
catchments, the model is returning results that reflect a closer
physical
interpretation of Bini being “the unit mass available for
washoff”, and
that this mass is indeed fully washed off or very close to fully
washed off
for all events. In contrast, this trend of M′ with catchment size
was not
observed in a meaningful way for mixed and dissolved
pollutants. Al-
though correlations in the same direction were found to be
significant
at a p < .05, the r2 values of 0.06 and 0.03 respectively indicate
that
the change in catchment size is explaining only a small part of
the
variance in W’. This makes any physical representation of Bini
for these
pollutants more difficult to understand, and reflects the model's
diffi-
culty in accurately predicting the behaviour of these forms of
pollu-
tants.
The results show that the calibrated exponential washoff model
parameters may indeed be linked to underlying physical and
rainfall
conditions, however, the strength of any dependence appears to
be
limited to particulate pollutants being modelled within small
sub-
catchments. Larger catchments present difficulty in use of this
model,
given that the explanatory nature of the model's time series
input

parameter q (mm/h) begins to breakdown as this is a normalised
value
of the flow rate taken at a specific outlet location, and is
assumed to be
simultaneously distributed over the entire catchment area.
Adding to
this affect is that the testing is based on a mixed sample at a
specific
outlet location. The mixed sample may represent a sum of
pollutant
mass that was washed off different surfaces of origin within the
catchment at slightly different times and for large catchments
the model
is unable to discriminate between the number of complex,
interrelated
processes that affect washoff. These results reflect conclusions
from
Miguntanna et al. (2013) that replicating washoff of nutrients
based on
solids is likely to lead to misleading results. More studies with
high
quality data at small catchment scales and different climatic
conditions
are required to garner a comprehensive understanding of these
differ-
ences and work toward confident application of representative
models
for different forms of pollutants.
5. Conclusions
Calibrated values for exponential washoff model parameters c3
and
Bini were found to be log-normally distributed and varied by up
to 3
orders of magnitude for the same pollutant at the same site.

Parameter
c4 however was found to be normally distributed with an
average value
of close to 1.0 for all pollutants. Quartile ranges or the spread
of
parameter values were typically greater for dissolved pollutants
PO4,
NO3, NH4 than particulates (TSS).
The accuracy of the washoff model was clearly best for
particulate
pollutant TSS, and worst for dissolved pollutants, with mixed
pollutants
TP and TN falling in between these two extremes. While the
model was
able to capture both the within event variation and mass balance
well
for particulates, the model was unable to accurately predict the
within
event behaviour for dissolved pollutants, although mass seems
to be
reasonably conserved.
Total rainfall depth, d, was identified as having strong potential
as
an explanatory variable for modelling washoff behaviour of TSS
with
statistically significant correlations between c3 and d, and Bini
and d for
all three types of land use. While this observation is true in the
tropics,
independent studies should be carried out for other climatic
conditions.
As the dissolved proportion of the pollutant being modelled
increased,
there appeared to be less ability to identify predictors of model

para-
meters, with almost no correlations found between model
parameters
and rainfall variables for PO4, NO3, and NH4. ADP was
identified as
having little or no importance across residential, mixed and
agricultural
land uses for any pollutant, which continues to question the use
of
build-up models when predicting washoff, particularly for
tropical re-
gions.
Finally, the analysis of the ratio of observed washoff mass and
the
“mass available for washoff” for different pollutants and
different
catchment sizes showed that the model was only able to return
values
close to the observed washoff mass for particulates and for
catchments
≤10 ha in size. Any equating of Bini with the washoff mass
should
therefore be undertaken with caution.
Acknowledgements
The authors would like to thank the Public Utilities Board -
Singapore, Singapore’s National Water Agency for use of the
data and
the City of Greater Geelong for their collaboration with Deakin
University on this project.
References

Bonhomme, C., Petrucci, G., 2017. Should we trust build-
up/wash-off water quality
models at the scale of urban catchments? Water Res. 108, 422–
443. https://doi.org/
10.1016/j.watres.2016.11.027.
Charbeneau, R.J., Barrett, M.E., 1998. Evaluation of methods
for estimating stormwater
pollutant loads. Water Environ. Res. 70, 1295–1302.
https://doi.org/10.2175/
106143098X123679.
Chow, M.F., Yusop, Z., Mohamed, M., 2011. Quality and first
flush analysis of stormwater
runoff from a tropical commercial catchment. Water Sci.
Technol. 63, 1211–1216.
https://doi.org/10.2166/wst.2011.360.
Dotto, C.B.S., Mannina, G., Kleidorfer, M., Vezzaro, L.,
Henrichs, M., McCarthy, D.T.,
Freni, G., Rauch, W., Deletic, A., 2012. Comparison of
different uncertainty techni-
ques in urban stormwater quantity and quality modelling. Water
Res. 46, 2545–2558.
https://doi.org/10.1016/j.watres.2012.02.009.
Duncan, H., 1995a. A Bibliography of Urban Stormwater
Quality, Report 95/8.
Cooperative Research Centre for Catchment Hydrology,
Clayton, VIC.
Duncan, H., 1995b. A Review of Urban Stormwater Quality
Processes. Report 95/9.
Clayton, VIC.

Duncan, H., 1999. Urban Stormwater Quality: a Statistical
Overview, Report 99/3.
Clayton, VIC.
Egodawatta, P., Thomas, E., Goonetilleke, A., 2007.
Mathematical interpretation of pol-
lutant wash-off from urban road surfaces using simulated
rainfall. Water Res. 41,
3025–3303. https://doi.org/10.1016/j.watres.2007.03.037.
Fletcher, T.D., Duncan, H., Poelsma, P.J., Lloyd, S., 2004.
Stormwater Flow and Quality,
and the Effectiveness of Non-proprietary Stormwater Treatment
Measures - A Review
and Gap Analysis. Report 04/8. Cooperative Research Centre
for Catchment
Hydrology, Clayton, VIC.
Francey, M., Fletcher, T.D., Deletic, A., Duncan, H., 2010. New
insights into the quality of
urban storm water in south eastern Australia. J. Environ. Eng.
136, 381. https://doi.
org/10.1061/(ASCE)EE.1943-7870.0000038.
Gaume, E., Villeneuve, J., Desbordes, M., 1998. Uncertainty
assessment and analysis of
the calibrated parameter values of an urban storm water quality
model. J. Hydrol.
210, 38–50. https://doi.org/10.1016/S0022-1694(98)00171-1.
Goonetilleke, A., Thomas, E., Ginn, S., Gilbert, D., 2005.
Understanding the role of land
use in urban stormwater quality management. J. Environ.
Manag. 74, 31.

Hubert, W.L., 1967. On the Kolmogorov-smirnov test for
normality with mean and var-
iance unknown. J. Amer. Stat. Assoc. 62, 399.
https://doi.org/10.2307/2283970.
Imteaz, M.A., Ahsan, A., Rahman, A., Mekanik, F., 2012.
Modelling stormwater treatment
systems using MUSIC: Accuracy. Resour. Conserv. Recycl. 71,
15–21. https://doi.org/
10.1016/j.resconrec.2012.11.007.
James, W., Irvine, K.N., 1992. Urban Surface Water Pollution.
Lewis Publishers, Boca
Raton, FL.
Jewell, T.K., Adrian, D.D., 1982. Statistical analysis to derive
improved stormwater
quality models. J. Water Pollut. Control Fed. 54, 489–499.
Le, S.H., 2014. Characterization and Modeling of Storm Runoff
Quality from a Tropical
Catchment with Diverse Land Use (Ph.D. Thesis). Nanyang
Technological University,
Singapore.
Le, S.H., Chua, L.H.C., Irvine, K.N., Eikaas, H.S., 2017.
Modeling washoff of total sus-
pended solids in the tropics. J. Environ. Manag. 200, 263–274.
https://doi.org/10.
1016/j.jenvman.2017.05.091.
May, D.B., Sivakumar, M., 2008. Prediction of urban
stormwater quality using artificial
neural networks. Environ. Model. Softw 24 296-230.
https://doi.org/10.1016/j.
envsoft.2008.07.004.

Metcalf & Eddy, 1971. Storm Water Management Model. U.S.
Government Printing
Office.
Miguntanna, N.P., Liu, A., Egodawatta, P., Goonetilleke, A.,
2013. Characterising nu-
trients wash-off for effective urban stormwater treatment
design. J. Environ. Manag.
120, 61–97.
Moss, A., 1988. Effects of flow velocity variation on rain
driven transportation and the
role of rain impact in the movement of solids. Soil Res. 26,
443–450. https://doi.org/
10.1071/SR9880443.
Novotny, V., 1994. Water Quality : Prevention, Identification,
and Management of Diffuse
Pollution/Vladimir Novotny. Harvey Olem. Van Nostrand
Reinhold, New York.
Pitt, R.E., Williamson, D., Voorhees, J., Clark, S., 2005.
Review of historical street dust
and dirt accumulation and washoff data. Journal of Water
Management Modeling. .
https://doi.org/10.14796/JWMM.R223-12.
Rossman, L., 2015. Storm Water Management Model User's
Manual Version 5.1. Office of
Research and Development, Water Supply and Water Resources
Division,
374–383

382
https://doi.org/10.1016/j.watres.2016.11.027
https://doi.org/10.2175/106143098X123679
https://doi.org/10.2175/106143098X123679
https://doi.org/10.2166/wst.2011.360
https://doi.org/10.1061/(ASCE)EE.1943-7870.0000038
https://doi.org/10.1016/S0022-1694(98)00171-1
https://doi.org/10.2307/2283970
https://doi.org/10.1016/j.resconrec.2012.11.007
https://doi.org/10.1016/j.resconrec.2012.11.007

https://doi.org/10.1016/j.envsoft.2008.07.004
https://doi.org/10.1016/j.envsoft.2008.07.004
https://doi.org/10.1071/SR9880443
https://doi.org/10.1071/SR9880443
https://doi.org/10.14796/JWMM.R223-12
Cincinnati, OH.
Sartor, J.D., Boyd, G.B., 1972. Water Pollution Aspects of
Street Surface Contaminants.
Environmental Protection Agency Report EPA-R2-72-081. U.S.
Government Printing
Office, Washington D.C.
Shaw, S.B., Stedinger, J.R., Walter, M.T., 2010. Evaluating
urban pollutant buildup/wash-
off models using a madison, Wisconsin catchment. J. Environ.
Eng. 136, 194. https://
doi.org/10.1061/(ASCE)EE.1943-7870.0000142.
U.S. Environmental Protection Agency, 1983. Results of the
Nationwide Urban Runoff
Program. Final Rep. No NTIS PB84-185552, Washington DC.
U.S. Environmental Protection Agency, 2017. Storm Water
Management Model (SWMM)

v5.1. (Cincinatti, OH).
Wicke, D., Coochrane, T., O'Sullivan, A., 2012. Build-up
dynamics of heavy metals de-
posited on impermeable urban surfaces. J. Environ. Manag. 113,
347–354. https://
doi.org/10.1016/j.jenvman.2012.09.005.
William, H.K., Wallis, W.A., 1952. Use of ranks in one-
criterion variance analysis. J.
Amer. Stat. Assoc. 47, 583. https://doi.org/10.2307/2280779.
Xiao, Y., Zhang, T., Wang, L., Liang, D., Xu, X., 2017.
Analytical and experimental study
on dissolved pollutant wash-off over impervious surfaces.
Hydrol. Process. 2017,
1–10. https://doi.org/10.1002/hyp.11372.
Yaziz, M.I., Gunting, H., Sapari, N., Ghazali, A.W., 1989.
Variations in rainwater quality
from roof catchments. Water Res. 23, 761–765.
https://doi.org/10.1016/0043-
1354(89)90211-X.
374–383
383
http://refhub.elsevier.com/ S0301-4797(19)30746-7/sref26

https://doi.org/10.2307/2280779
https://doi.org/10.1002/hyp.11372
https://doi.org/10.1016/0043-1354(89)90211-X
https://doi.org/10.1016/0043-1354(89)90211-XModelling the
washoff of pollutants in various forms from an urban
catchmentIntroductionData usedMethodologyWashoff
modellingMonte Carlo AnalysisResults and discussionMonte
Carlo AnalysisSensitivity analysisDependence of model
parameters with rainfall and flow variablesTotal mass at the
start of an eventConcl usionsAcknowledgementsReferences
O R I G I N A L P A P E R
Why conserve biodiversity? A multi-national exploration
of stakeholders’ views on the arguments for biodiversity
conservation
Pam M. Berry1 • Veronika Fabók2,3 • Malgorzata Blicharska4,5
•
Yennie K. Bredin6 • Marina Garcı́a Llorente7,8 •
Eszter Kovács
2,3
• Nicoleta Geamana
9

• Adina Stanciu
9
•
Mette Termansen
10
• Tiina Jääskeläinen
11
• John R. Haslett
12
•
Paula A. Harrison1
Received: 17 February 2016 / Revised: 14 June 2016 /
Accepted: 21 June 2016 /
Published online: 4 July 2016
� Springer Science+Business Media Dordrecht 2016
Abstract Given the concern about biodiversity loss, there are a
number of arguments used
for biodiversity conservation ranging from those emphasisi ng
the intrinsic value of bio-
diversity to those on the direct use value of ecosystems. Yet
arguing the case for biodi-
versity conservation effectively requires an understanding of
why people value
biodiversity. We used Q methodology to explore and understand
how different conser-

vation practitioners (social and natural science researchers,
environmental non-Govern-
mental organisations and decision-makers) in nine European
countries argue for
conservation. We found that there was a plurality of vi ews
about biodiversity and its
Communicated by Rob Bugter, Paula Harrison, John Haslett and
Rob Tinch.
This is part of the special issue on ‘BESAFE’.
& Pam M. Berry
[email protected]
1
Environmental Change Institute, University of Oxford, South
Parks Road, Oxford OX1 3QY, UK
2
Institute of Nature Conservation and Landscape Management,
Szent István University, Páter
Károly u. 1, Gödöll}o H-2100, Hungary
3
Environmental Social Science Research Group (ESSRG), Rómer
Flóris u. 38, Budapest H-1024,
Hungary
4
Department of Aquatic Sciences and Assessment, Swedish
University of Agricultural Sciences,
Box 7050, 750 07 Uppsala, Sweden

5
Swedish Biodiversity Centre, 7016, 750 07 Uppsala, Sweden
6
Norwegian Institute for Nature Research, Fakkelgården, NO-
2624, Lillehammer, Norway
7
Department of Applied Research and Agricultural Extension,
Madrid Institute for Rural,
Agricultural and Food Research and Development (IMIDRA),
Ctra. Madrid-Barcelona (N-II), KM.
38.200, 28802 Alcalá De Henares, Madrid, Spain
8
Social-Ecological Systems Laboratory, Department of Ecology,
Universidad Autónoma de Madrid,
c. Darwin 2, Biology, 28049 Madrid, Spain
123
Biodivers Conserv (2018) 27:1741–1762
https://doi.org/10.1007/s10531-016-1173-z
http://orcid.org/0000-0002-1201-072X
http://crossmark.crossref.org/dialog/?doi=10.1007/s10531-016-
1173-z&domain=pdf
http://crossmark.crossref.org/dialog/?doi=10.1007/s10531-016-
1173-z&domain=pdf
https://doi.org/10.1007/s10531-016-1173-z
conservation. A moral argument and some arguments around the
intrinsic and ecological
value of biodiversity were held by all stakeholder groups. They

also shared the view that
species valuation does not justify the destruction of nature.
However, there were also some
differences within and between the groups, which primarily
reflected the espousal of either
ecocentric or anthropocentric viewpoints. Our findings suggest
that moral arguments and
those around biodiversity’s intrinsic and ecological value could
potentially serve as a
starting point for building consensus among conservation
practitioners.
Keywords Intrinsic value � Ecological value � Utilitarian value
� Ecosystem services �
Q-methodology � Conservation practitioners
Introduction
While the loss of biodiversity continues to be of global concern
(Secretariat of the Con-
vention on Biological Diversity 2014; McCallum 2015), the
debate over its importance
remains a hot topic amongst conservation practitioners,
researchers and policy makers. A
wide variety of arguments for the conservation of biodiversity
have been proposed, ranging
from those based on its intrinsic value to more utilitarian

perspectives (Ehrlich and Ehrlich
1992; Nunes and van der Bergh 2001; Montgomery 2002;
Raffaelli et al. 2009). This
spectrum of arguments is not new, as already in the late 19th
century there were tensions
between these two views as expressed by John Muir’s
‘‘preservationism’’ and Gifford
Pinochet’s ‘‘conservationism’’ (Meyer 1997), with Muir
emphasising the need to protect
wilderness and Pinochet the sustainable use of natural
resources. Armsworth et al. (2007)
suggested that pluralism between the two schools of thought is
the norm in conservation
practice. However, the rise of the ecosystem services concept as
a means of making human
dependence on ecosystems explicit (Norgaard 2010) and
attempts to mainstream the
commodification of nature (Gómez-Baggethun and Ruiz-Pére
2011), have led to renewed
debate about the different arguments for biodiversity and the
reasons for its conservation
(e.g. McCauley 2006; Ridder 2008; Peterson et al. 2009).
In a study of the motivations behind conservation in urban
areas, Dearborn and Kark

(2010) suggested that there was a spectrum, ranging from
motivations associated with
benefits to nature, to those associated with benefits to humans.
Moreover, they showed that
often there were multiple motivations, arising from cultural and
value differences, amongst
the different groups of people involved in any given situation.
Thus, those involved in
biodiversity conservation can have different legitimate
motivations and operate as different
actors (Hermelingmeier 2014). It has been suggested that in
order to increase the effec-
tiveness of arguments supporting conservation, this diversity of
opinions in biodiversity
conservation science and practice should be embraced
(Sandbrook et al. 2010).
9
Research Center in Systems Ecology and Sustainability,
University of Bucharest, Splaiul
Independentei 91-95, Sector 5, 050095 Bucharest, Romania
10
Department of Environmental Science, Aarhus University,
Aarhus, Denmark
11
Finnish Environment Institute, P.O. Box 140, 00251 Helsinki,

Finland
12
Division of Animal Structure and Function, Department of Cell
Biology, University of Salzburg,
Hellbrunnerstrasse 34, A-5020 Salzburg, Austria
1742 Biodivers Conserv (2018) 27:1741–1762
123
The ways in which we understand and interact with nature
determine our human-nature
relationship (Russell et al. 2013) and in practice, this personal
perception of nature may
affect the motivation of biodiversity conservation actors and
how they seek to deliver
conservation. Understanding the value systems of those working
with biodiversity con-
servation is, therefore, important if we want to avoid
unnecessary conflicts and design
conservation solutions that take into account different, often
diverging perspectives. Couix
and Hazard (2013), for example, emphasise the importance of
understanding the beliefs
and values of different researchers and other stakeholders for
effective cooperation.

The aim of this paper is to compare the personal views of
individuals from different
conservation practitioner stakeholder groups in different
European countries concerning
arguments for biodiversity conservation. Such an investigatio n
is important because dif-
ferences in argumentation may affect decisions concerning
biodiversity conservation and
thus the delivery of essential ecosystem services and their
impact on human wellbeing.
Also, understanding the interests and views of stakeholders is
important for the successful
implementation of natural resources and biodiversity policy
(Grimble and Wellard 1997;
Bagnoli et al. 2008). This explains why stakeholder analyses are
often connected to
stakeholder participation and conflict management (Grimble and
Wellard 1997; Mushove
and Vogel 2005; Reed 2008). Most papers applying stakeholder
theory and stakeholder
analysis in the conservation context, concentrate on a specific
local or national conser-
vation context with the purpose of revealing stakeholders’
perceptions, interests or rela-

tionships, or making a complex analysis in a particular policy
situation (e.g. De Lopez
2001; Mushove and Vogel 2005; Suškevičs et al. 2013). There
are also a number of
qualitative studies investigating certain stakeholder groups’
perceptions about biodiversity
or nature in general, sometimes also including their attitudes
toward conservation measures
(e.g. Fischer and Young, 2006; Buijs et al. 2008; Kelemen et al.
2013). However, there is a
lack of analysis of the broad spectrum of values of biodiversity
that might be held by the
stakeholders working in biodiversity conservation. Particularly,
there is a lack of studies
that investigate the views of different stakeholders in a
European context. Hence in this
paper, we focus on three stakeholder groups particularly
involved in informing, formu-
lating, influencing and implementing conservation policy:
researchers, non-governmental
organisations (NGOs) and decision-makers, and we explore
their views through a Q
analysis.

Methodology
Q methodology combines quantitative and qualitative
information to explore social per-
spectives on a particular issue. The Q methodology is
particularly suitable for this study as
it enables elicitation of the personal views of stakeholders
involved in conservation on
arguments associated with biodiversity conservation, and the
identification of common-
alities and differences in their perspectives in a quantitative
manner. While, the qualitative
information obtained during the Q-interviews allows for deeper
investigation of the reasons
underlying personal views.
It has been broadly applied to investigate a range of
environmental issues, for example:
potential for sustainable forestry (Swedeen 2006) and small
scale forestry and market
reform in the Ukraine (Ninjik et al. 2009); motivations for
urban biodiversity conservation
(Dearborn and Kark, 2010); or portrayal of climate change
(O’Neill et al. 2013) and values
and attitudes of locals living along the Tisza River, Hungary, to
water issues and adaptation

Biodivers Conserv (2018) 27:1741–1762 1743
123
(Marjainé Szerényi et al. 2011). It is especially suitable for
studying contested issues, such
as those concerning the environment (Barry and Proops 1999),
as it seeks to capture a
range of perspectives. It also provides the opportunity to
understand how different
stakeholders, characterised by a variety of points of view,
perceive an issue (e.g. attitudes
to conservation on private land in Poland; Kamal and
Grodzı́nska-Jurczak Kamal and
Grodzinska-Jurczak 2014; landscape preferences of locals in
southern Transylvania,
Romania (Milcu et al. 2014); identifying which are the greater
points of conflict; but also,
uncovering the common ground of agreement and shared
understanding between different
actors (e.g. Chamberlain et al. 2012). While Q-methodology has
some potential drawbacks,
such as a lack of possibility to generalise the findings to a
larger population, it has proved

useful in revealing a range of perspectives existing on a
particular topic. It is an exploratory
tool that gives qualitative data some quantitative support
(Kamal et al. 2014).
The Q method involves six steps: (1) identification of a
discourse area of interest; (2)
collection of a full range of statements about the discourse; (3)
selection of a representative
set of statements from the full range (the concourse); (4)
selection of participants and
execution of Q sorts (i.e. sorting of the statements by the
participants according to their
level of agreement with the statements); (5) statistical analysis
of the Q sorts; and (6)
interpretation of the identified perspectives using both the
results of the statistical analysis
and qualitative information from the discussion of the
statements during the sorting (based
on Swedeen, 2006). In this study, the discourse of interest
concerned views related to
conservation, as reflected in arguments about the value of
biodiversity and why it is worth
investing effort in biodiversity conservation.

A literature review of biodiversity value arguments identified
549 relevant articles from
which 180 statements representing the range of views on the
importance of biodiversity
were selected (Howard et al. 2013). These were sorted into the
following broad categories:
direct economic use (TEEB 2010); biophilia (Wilson 1984);
non-use values (i.e., not
associated with actual use of a good or service, such as the
moral satisfaction obtained
from biodiversity conservation known as its existence value
(Kahneman and Knetsch 1992;
TEEB 2010) or the satisfaction gained from preserving a natural
environment for future
generations known as bequest value (TEEB 2010); aesthetic
value (e.g. Montgomery
2002); intrinsic value (i.e. biodiversity has a value in itself
independent of its usefulness for
humans; Brondizio et al. 2010); ecological value/importance of
ecological functions
(Cardinale et al. 2012); and ecosystem service reasoning (e.g.
Daily et al. 2000; Ulyshen
2013). These broad categories were chosen to ensure that the
final list of statements

represented the diversity of arguments in the literature (Brown
1980). A number of
statements between 40 and 80 has been recommended in Q-
literature (e.g. Watts and
Stenner 2005), so from the initial 180 statements, 42 were
selected (Appendix 1), covering
the broad categories (the Q concourse). We selected statements
that were salient (i.e. ones
that people were likely to have opinion about and could be
interpreted in slightly different
ways by different people), and understandable (i.e. meaningful
to people doing the sorts).
In addition, as far as possible, both positively and negatively
worded statements were
selected from each category. Some editing of the statements
(but without altering their
meaning) was undertaken, so that they: (i) were understandable
when taken out of context;
(ii) understandable in different countries and (iii) were easily
translated into other lan-
guages for application in non-English speaking countries. To
avoid misunderstandings due
to translation to national languages, we provided the
participants with both national lan-

guage and English versions of the statements. Overall piloting
was not carried out as the Q
methodology had already been used by a number of the
researchers for other studies in
their country (e.g. Denmark, Poland and Spain).
123
While Q methodology can be carried out with relatively small
numbers of participants,
between 40 and 60 individuals is thought to be good (Stainton
Rogers 1995). The Q
participants selected included 53 researchers (both natural and
social scientists), 25 rep-
resentatives from NGOs and 43 decision-makers from different
governance levels from
nine European countries: Denmark, Finland, Hungary, Poland,
Norway, Romania, Spain,
UK and Austria/Salzburg Province (Table 1). The Q interviews
were conducted between
April 2013 and April 2014.
The sorting of the Q statements can be according to a pre-
defined (forced) quasi-normal

distribution or can be freely placed relative to each other
(Brown 1980), in both cases
ranging from most like respondents think to most unlike they
think. In this study, the
participants were given the 42 statements, each on a separate
card and asked to place them
onto a sort chart with a quasi-normal distribution (Fig. 1).
Given that the centre (0) may not
represent the inflexion point between an individual’s agreement
and disagreement with the
statements, they were asked at the end of the sort to draw a line
on the chart to represent
this point, as suggested by Webler et al. (2009). The interviews
were, if possible, audio
recorded to ensure a complete capture of their expressed
thoughts. While placing the
statements the participants were encouraged to reflect aloud on
the reasons for their
positioning of them. This information was used to help
understand their thinking and to
facilitate the interpretation of the results.
We used the free software PQmethod 2.35
1

to analyse the Q sorts. Firstly, we undertook
principal component analysis on the statement response matrix
and then rotated the
resulting factors using the varimax rotation method, where a
factor represents a cluster of
respondents with similar views (Brown 1993). The rotation
helps to reduce noise from
sorts which load significantly on more than one factor (Wolf
2006). The decision making
process of factor extraction is a complex process, and there is
not one single mathemati-
cally best solution, as besides statistical considerations
(eigenvalue of factors, total vari-
ance, number of significantly loading sorts, correlation between
factors), it is important to
take into account the meaning and significance of the factors
(Watts and Stenner 2012).
Extracting a set of factors with a relatively high eigenvalue, a
reasonable proportion of the
total variance (above 40 %) and two (or more) significantly
loaded Q sorts is important in
the decision making process (Watts and Stenner 2012). We
considered variance above
40 % acceptable and chose factor solutions with more than three

significantly loading Q
Table 1 Number of respondents
from each country and stake-
holder group
a
Salzburg Province was
considered as equal to the
countries in our study, as
biodiversity conservation
decisions are the responsibility of
Austrian Provincial Governments
Country Researchers Decision-makers NGOs Total
Denmark 9 1 2 12
Finland 7 4 0 11
Hungary 6 4 3 13
Norway 6 3 4 13
Poland 6 4 5 15
Romania 4 11 2 17
Salzburg, Austria
a
6 2 3 11
Spain 5 5 4 14

UK 4 9 2 15
Total 53 43 25 121
1
http://schmolck.org.qmethod/pqmanual.html.
123
http://schmolck.org.qmethod/pqmanual.html
sorts per factor. We also tried to choose factors with the lowe st
factor correlation, (ideally
below 0.4) as significant correlations could mean that the
factors cannot be easily distin-
guished (Watts and Stenner 2012), but in some cases other
considerations were more
important. Before choosing the final factor solutions we
checked other possible solutions (3
and 4 factor solutions for researchers and NGOs and 2 and 4
factor solutions for decision-
makers). None of these had a correlation lower than 0.4 and
although there were factor
solutions that might have performed better on correlation than
the chosen ones, they were
more difficult to interpret or there were only a few significantly

loading sorts on one of the
factors.
We undertook the factor analysis and extracted the factors
separately for the three
stakeholder groups. We conducted a qualitative analysis of the
different factors of the
stakeholder groups using the data analysis software NVivo to
compare the final perspec-
tives of the different groups. Those individuals whose sorts
correlate with a specific factor
are called loaders and a sort loading of [ ±0.39 for a given
factor was considered sig-
nificant at the P 0.01 level, based on Brown (1980, p.283). The
idealised sort pattern (i.e.
-4 to ?4) for the factor was constructed from the weighted
averages of the loaders. The
perspectives for each factor were primarily defined by the
statements with the highest/
lowest z-scores, as z-scores provide a measure of how far the
statement is from the centre
of the distribution of all statements typical for that factor. Thus,
they are useful in iden-
tifying those statements most important for describing that
particular factor (Webler et al.
2009). Qualitative data from the reflections made during the

sort by the significantly loaded
respondents was also included in the description in order to
understand better the
respondents’ interpretations of the different statements. Thus,
the perspectives were
interpreted in a narrative style (Watts and Stenner 2012).
Kampen and Tamás (2014) identified several potential
limitations of the Q methodol-
ogy. One of them is the concern as to whether the concourse
selected represents the full
range of views on the particular topic. To address this, we
conducted an extensive literature
review, to cover as many issues as possible. Nevertheless, we
acknowledge, that some
extreme views may have been omitted from the concourse.
Other potential drawbacks are
the potential to affect participants’ opinion by forcing them to
sort the statements into the
normal distribution, which may not fully represent their views
and biased interpretation of
the factors by the researcher. To avoid these problems, the
sorting included interviews
where the participants could fully express their views about
particular statements. Then the

recorded interviews were used in the interpretation of sorting
and thus contributed to un-
biased understanding of the respondents’ views.
Least like Most like
I think I think
-4 -3 -2 -1 0 1 2 3 4
Fig. 1 Distribution used for
sorting the 42 statements
123
Results
The factor analysis on the researchers’ dataset (n = 53) resulted
in two factors representing
two main perspectives (Table 2). All of the respondents loaded
onto a factor, 34 respon-
dents loaded onto the first and 19 onto the second factor. These
factors explained 41 % of
the variance among all the researchers Q-sorts. When analysing
the data we also consid-
ered an alternative with three factors. However, in the three
factor solution the correlation

between the first and third factors was relatively high (0.58). In
addition, in this case the
second factor, that explained 9 % of the variance, did not
identify any meaningful per-
spective. Thus, we decided in favour of the two factor solution.
Analysing the dataset of the NGO respondents (n = 25), we
distinguished two factors,
demonstrating two main perspectives; 17 of the respondents
loaded onto the first and seven
onto the second factor. The correlation between the tw o factors
was 0.53, while the
explained variance was 43 %. Although the highest correlation
between the factors was
lower in the three factor solution (0.48) and the explained
variance was higher (52 %), we
decided against this solution as the number of significantly
loaded Q sorts were higher in
the two factor solution.
The decision-makers (n = 43) loaded onto three main factors,
which together explained
47 % of the variance; sixteen respondents loaded onto the first,
14 onto the second, and
eight onto the third factor. The highest correlation was 0.56

between the first and the third
factors. The highest correlation was slightly lower in the two
factor solution (0.52), but the
explained variance was lower in this case (42 %). Moreover, the
perspectives represented
by the factors were more meaningful for the three factor
solution, so we opted for this.
In the following sections, based on the seven factors derived
from the scoring of
particular Q-statements, and combined with respondents’
reflections on these statements,
we present the perspectives that best describe the researchers,
NGOs, and decision-makers
views on the arguments for biodiversity conservation. Numbers
in brackets refer to specific
Q statements (see Table 3).
Researchers’ perspectives on biodiversity conservation:
intrinsic values
and ecosystem services
The first researchers’ perspective (R1, Table 2) highlights some
of the intrinsic and eco-
logical values of biodiversity—that species are not superfluous
and each species is
important (17) and that species have a right to exist even if they

do not benefit humans
Table 2 Results for two and three factor solutions for the groups
of respondents
Stakeholder
group
No. of respondents
loaded on the factor
% of
variance
explained
Correlation between
factors (F)
Factor 1 Factor 2 Factor 3 F1 and F2 F1 and F3 F2 and F3
Researchers (two
factor solution)
34 19 Not relevant 41 0.44 Not relevant Not relevant
NGOs (two factor
solution)
17 7 Not relevant 43 0.53 Not relevant Not relevant
Decision-makers
(three factor

solution)
16 14 8 47 0.46 0.56 0.41
123
T
a
b
le
3
T
h
e
4
2
Q
st
a
te
m
e
n
ts

u
se
d
fo
r
th
e
in
te
rv
ie
w
s
a
n
d
th
e
sc
o
re
s
fo
r

o
n
ly
d
is
ti
n
g
u
is
h
in
g
st
a
te
m
e
n
ts
fo
r
e
a
c
h

o
f
th
e
p
e
rs
p
e
c
ti
v
e
s
S
ta
te
m
e
n
ts
R
e
se
a

rc
h
e
rs
N
G
O
s
D
e
c
is
io
n
-m
a
k
e
rs
S
ta
te
m
e
n

t
p
o
si
ti
o
n
s
P
e
rs
p
e
c
ti
v
e
R
1
in
tr
in
si
c
v

a
lu
e
P
e
rs
p
e
c
ti
v
e
R
2
u
ti
li
ta
ri
a
n
v
a
lu
e
(f

o
c
u
s
o
n
E
S
)
P
e
rs
p
e
c
ti
v
e
N
G
O
1
in
tr
in

si
c
v
a
lu
e
P
e
rs
p
e
c
ti
v
e
N
G
O
2
M
ix
e
d
(i
n

tr
in
si
c
a
n
d
u
ti
li
ta
ri
a
n
v
a
lu
e
)
P
e
rs
p
e
c

ti
v
e
D
1
in
tr
in
si
c
v
a
lu
e
P
e
rs
p
e
c
ti
v
e
D
2

u
ti
li
ta
ri
a
n
v
a
lu
e
(f
o
c
u
s
o
n
E
S
)
P
e
rs
p

e
c
ti
v
e
D
3
m
ix
e
d
(i
n
tr
in
si
c
a
n
d
sp
ir
it
u
a

l
v
a
lu
e
)
1
.
W
e
d
o
n
o
t
k
n
o
w
h
o
w
e
c
o
sy

st
e
m
s
w
il
l
b
e
a
ff
e
c
te
d
b
y
th
e
lo
ss
o
f
sp
e

c
ie
s,
th
e
re
fo
re
w
e
b
e
tt
e
r
p
re
se
rv
e
th
e
m
?
3

?
3
?
4
2
.
P
ro
te
c
ti
n
g
e
c
o
sy
st
e
m
se
rv
ic
e

p
ro
v
id
e
rs
is
im
p
o
rt
a
n
t
b
e
c
a
u
se
th
e
y
a
re

a
so
u
rc
e
o
f
e
c
o
n
o
m
ic
v
a
lu
e
?
4
?
4
3
.
T

h
e
e
c
o
sy
st
e
m
se
rv
ic
e
a
p
p
ro
a
c
h
h
a
s
p
o

te
n
ti
a
l
to
im
p
ro
v
e
sp
e
c
ie
s
c
o
n
se
rv
a
ti
o
n
in

E
u
ro
p
e
?
4
?
2
?
3
4
.
B
io
d
iv
e
rs
it
y
c
o
n

se
rv
a
ti
o
n
is
n
o
t
a
m
o
ra
l
m
a
tt
e
r
-
3
-
4

-
4
-
4
-
4
5
.
S
o
m
e
sp
e
c
ie
s
a
re
im
p
o
rt
a
n

t
sy
m
b
o
ls
o
f
h
u
m
a
n
v
a
lu
e
s,
su
c
h
a
s
fr
e
e

d
o
m
-
3
6
.
S
p
e
c
ie
s
a
re
p
ri
c
e
le
ss
?
4
?
4
?

4
7
.
T
h
e
re
a
so
n
b
io
d
iv
e
rs
it
y
m
a
tt
e
rs
is

b
e
c
a
u
se
it
c
o
n
fe
rs
o
n
u
s
a
n
im
p
re
c
is
e
,
im

m
e
a
su
ra
b
le
w
e
ll
-
b
e
in
g
th
a
t
is
lo
c
a
te
d
in

th
e
sp
ir
it
ra
th
e
r
th
a
n
in
th
e
w
a
ll
e
t
8
.
T
h
e

e
x
ti
n
c
ti
o
n
o
f
a
sp
e
c
ie
s
is
li
k
e
th
e
d
e
st

ru
c
ti
o
n
o
f
a
g
re
a
t
w
o
rk
o
f
a
rt
-
3
-
3
-
3

9
.
It
is
n
o
t
c
le
a
r
w
h
y
a
ll
sp
e
c
ie
s
th
a
t
e

n
v
ir
o
n
m
e
n
ta
li
st
s
c
a
m
p
a
ig
n
to
c
o
n
se
rv
e

o
u
g
h
t
to
b
e
sa
v
e
d
-
3
-
3
-
4
1
0
.
P
ro
te
c

ti
n
g
b
io
d
iv
e
rs
it
y
a
n
d
e
c
o
sy
st
e
m
se
rv
ic
e
s

is
p
a
rt
ic
u
la
rl
y
im
p
o
rt
a
n
t
fo
r
p
o
v
e
rt
y
a
ll
e

v
ia
ti
o
n
in
d
e
v
e
lo
p
in
g
c
o
u
n
tr
ie
s
?
4
1
1

.
C
o
n
se
rv
in
g
g
e
n
e
ti
c
d
iv
e
rs
it
y
is
im
p
o

rt
a
n
t
to
fe
e
d
fu
tu
re
h
u
m
a
n
p
o
p
u
la
ti
o
n
s

?
3
?
4
123
T
a
b
le
3
c
o
n
ti
n
u
e
d
S
ta
te
m

e
n
ts
R
e
se
a
rc
h
e
rs
N
G
O
s
D
e
c
is
io
n
-m
a
k
e
rs

S
ta
te
m
e
n
t
p
o
si
ti
o
n
s
P
e
rs
p
e
c
ti
v
e
R
1

in
tr
in
si
c
v
a
lu
e
P
e
rs
p
e
c
ti
v
e
R
2
u
ti
li
ta

ri
a
n
v
a
lu
e
(f
o
c
u
s
o
n
E
S
)
P
e
rs
p
e
c
ti
v
e

N
G
O
1
in
tr
in
si
c
v
a
lu
e
P
e
rs
p
e
c
ti
v
e
N
G
O

2
M
ix
e
d
(i
n
tr
in
si
c
a
n
d
u
ti
li
ta
ri
a
n
v
a
lu

e
)
P
e
rs
p
e
c
ti
v
e
D
1
in
tr
in
si
c
v
a
lu
e
P
e
rs

p
e
c
ti
v
e
D
2
u
ti
li
ta
ri
a
n
v
a
lu
e
(f
o
c
u
s
o
n

E
S
)
P
e
rs
p
e
c
ti
v
e
D
3
m
ix
e
d
(i
n
tr
in
si
c
a

n
d
sp
ir
it
u
a
l
v
a
lu
e
)
1
2
.
C
o
u
n
tr
ie
s
c
a

n
b
e
n
e
fi
t
fr
o
m
th
e
ir
c
o
n
se
rv
a
ti
o
n
e
ff
o

rt
s
th
ro
u
g
h
to
u
ri
sm
?
2
?
3
1
3
.
N
a
tu
re
p
ro

v
id
e
s
u
s
w
it
h
m
a
n
y
v
a
lu
a
b
le
e
x
p
e
ri
e

n
c
e
s.
W
e
h
u
n
t,
fi
sh
,
h
ik
e
,
m
o
u
n
ta
in
c
li
m

b
,
a
n
d
e
n
g
a
g
e
in
n
u
m
e
ro
u
s
a
c
ti
v
it
ie
s

in
w
h
ic
h
w
e
in
te
ra
c
t
w
it
h
n
a
tu
re
?
3
1
4

.
L
o
si
n
g
it
s
b
io
lo
g
ic
a
l
ri
c
h
n
e
ss
a
n
d
d

iv
e
rs
it
y
,
th
e
w
o
rl
d
lo
se
s
it
s
m
a
g
ic
-
3
?
3

1
5
.
It
is
im
p
o
rt
a
n
t
to
c
o
n
se
rv
e
th
e
g
e
n

e
ti
c
re
se
rv
o
ir
in
a
re
g
io
n
,
in
c
a
se
w
e
n
e
e
d
to

b
re
e
d
d
is
e
a
se
-r
e
si
st
a
n
t
p
la
n
ts
o
r
p
ro
d

u
c
e
fo
o
d
a
d
a
p
te
d
to
lo
c
a
l
c
o
n
d
it
io
n
s

?
4
1
6
.
W
e
w
a
n
t
to
e
x
p
e
ri
e
n
c
e
a
re
a
s

w
h
e
re
h
u
m
a
n
s
a
re
m
e
re
ly
v
is
it
o
rs
a
n
d
n

o
t
in
h
a
b
it
a
n
ts
1
7
.
M
o
st
sp
e
c
ie
s
a
re
su
p

e
rfl
u
o
u
s
-
4
-
4
-
4
-
4
-
4
-
4
-
4
1
8
.
W

e
v
a
lu
e
so
m
e
sp
e
c
ie
s
fo
r
th
e
ir
b
e
a
u
ty
,
b

u
t
th
is
is
o
n
ly
re
le
v
a
n
t
fo
r
a
v
e
ry
sm
a
ll
n
u
m

b
e
r
o
f
sp
e
c
ie
s.
T
h
e
re
fo
re
,
b
e
a
u
ty
is
n
o
t

a
p
a
rt
ic
u
la
rl
y
im
p
o
rt
a
n
t
b
a
si
s
fo
r
c
o
n

se
rv
a
ti
o
n
1
9
.
W
e
d
o
n
o
t
n
e
e
d
to
re
c
o
g

n
iz
e
o
th
e
r
b
e
in
g
s
a
s
o
u
r
m
o
ra
l
e
q
u
a

ls
to
re
a
li
z
e
th
a
t
w
e
sh
o
u
ld
n
o
t
k
il
l
th
a
t

w
h
ic
h
is
n
o
t
a
th
re
a
t
2
0
.
A
ll
sp
e
c
ie
s
h

a
v
e
a
ri
g
h
t
to
e
x
is
t,
re
g
a
rd
le
ss
o
f
th
e
ir
a
b

il
it
y
to
b
e
n
e
fi
t
h
u
m
a
n
s
?
4
?
4
?
4
?
4

?
3
123
T
a
b
le
3
c
o
n
ti
n
u
e
d
S
ta
te
m
e
n

ts
R
e
se
a
rc
h
e
rs
N
G
O
s
D
e
c
is
io
n
-m
a
k
e
rs
S

ta
te
m
e
n
t
p
o
si
ti
o
n
s
P
e
rs
p
e
c
ti
v
e
R
1
in

tr
in
si
c
v
a
lu
e
P
e
rs
p
e
c
ti
v
e
R
2
u
ti
li
ta
ri
a

n
v
a
lu
e
(f
o
c
u
s
o
n
E
S
)
P
e
rs
p
e
c
ti
v
e
N
G

O
1
in
tr
in
si
c
v
a
lu
e
P
e
rs
p
e
c
ti
v
e
N
G
O
2

M
ix
e
d
(i
n
tr
in
si
c
a
n
d
u
ti
li
ta
ri
a
n
v
a
lu
e
)

P
e
rs
p
e
c
ti
v
e
D
1
in
tr
in
si
c
v
a
lu
e
P
e
rs
p
e

c
ti
v
e
D
2
u
ti
li
ta
ri
a
n
v
a
lu
e
(f
o
c
u
s
o
n
E

S
)
P
e
rs
p
e
c
ti
v
e
D
3
m
ix
e
d
(i
n
tr
in
si
c
a
n
d

sp
ir
it
u
a
l
v
a
lu
e
)
2
1
.
N
a
tu
re
is
a
la
b
o
ra

to
ry
fo
r
th
e
p
u
rs
u
it
o
f
sc
ie
n
c
e
th
ro
u
g
h
w

h
ic
h
so
c
ie
ty
g
a
in
s
k
n
o
w
le
d
g
e
,
a
n
d
u

n
d
e
rs
ta
n
d
in
g
o
f
th
e
w
o
rl
d
?
3
2
2
.
T
h

e
d
iv
e
rs
it
y
o
f
li
fe
is
so
m
e
th
in
g
li
k
e
th
e
ri

v
e
ts
o
n
a
n
a
ir
p
la
n
e
,
w
it
h
e
a
c
h
sp
e
c
ie
s
p

la
y
in
g
a
sm
a
ll
b
u
t
si
g
n
ifi
c
a
n
t
ro
le
in
th
e
w

o
rk
in
g
o
f
th
e
w
h
o
le
.
T
h
e
lo
ss
o
f
e
a
c
h
ri

v
e
t
w
e
a
k
e
n
s
th
e
p
la
n
e
b
y
a
sm
a
ll
b
u
t

n
o
ti
c
e
a
b
le
a
m
o
u
n
t—
u
n
ti
l
it
lo
se
s
a
ir
w

o
rt
h
in
e
ss
a
n
d
c
ra
sh
e
s
?
3
?
3
?
3
?
4
2
3
.

N
a
tu
re
p
ro
v
id
e
s
a
p
la
c
e
to
ta
k
e
c
a
lc
u
la
te

d
ri
sk
s,
to
le
a
rn
th
e
lu
c
k
o
f
th
e
w
e
a
th
e
r,
to
lo

se
a
n
d
fi
n
d
o
n
e
’s
w
a
y
,
to
re
fl
e
c
t
o
n
su
c
c
e

ss
a
n
d
fa
il
u
re
2
4
.
E
v
e
n
if
o
n
ly
a
fe
w
sp
e
c

ie
s
a
re
n
e
e
d
e
d
fo
r
o
u
r
w
o
rl
d
to
b
e
p
ro
d

u
c
ti
v
e
w
e
h
a
v
e
to
c
o
n
se
rv
e
m
o
re
sp
e
c
ie

s
a
s
a
b
a
c
k
-u
p
.
O
th
e
rw
is
e
a
p
e
st
o
r
c
li

m
a
te
c
h
a
n
g
e
c
o
u
ld
w
ip
e
o
u
t
th
e
fe
w
sp
e

c
ie
s
w
e
h
a
v
e
sa
v
e
d
,
a
n
d
w
e
w
o
u
ld
h
a

v
e
n
o
th
in
g
in
re
se
rv
e
?
2
-
4
?
4
2
5
.
P
ri
st

in
e
n
a
tu
re
is
v
a
lu
a
b
le
in
it
se
lf
?
4
?
4
?
3
2

6
.
E
c
o
sy
st
e
m
s
h
a
v
e
c
o
-e
v
o
lv
e
d
w
it

h
h
u
m
a
n
s
c
re
a
ti
n
g
la
n
d
sc
a
p
e
s
o
f
im
p

o
rt
a
n
t
c
u
lt
u
ra
l
v
a
lu
e
?
3
2
7
.
A
n
y
e

ff
o
rt
to
c
o
n
se
rv
e
b
io
d
iv
e
rs
it
y
m
u
st
b
e
li

m
it
e
d
b
y
c
o
n
si
d
e
ra
ti
o
n
s
o
f
o
th
e
r
v
a

lu
e
s
su
c
h
a
s
fr
e
e
d
o
m
,
e
q
u
a
li
ty
,
h
e
a
lt
h
,

a
n
d
ju
st
ic
e
-
3
2
8
.
D
e
st
ro
y
in
g
n
a
tu
re
is

li
k
e
b
u
rn
in
g
u
n
re
a
d
b
o
o
k
s
-
3
-
3
123

T
a
b
le
3
c
o
n
ti
n
u
e
d
S
ta
te
m
e
n
ts
R
e
se
a

l
v
a
lu
e
)
2
9
.
V
a
lu
in
g
sp
e
c
ie
s
in
e
c
o
n
o
m

ic
te
rm
s
im
p
li
e
s
a
ju
st
ifi
c
a
ti
o
n
fo
r
th
e
d
e
st

ru
c
ti
o
n
o
f
th
e
b
io
sp
h
e
re
-
3
-
4
-
3
-
4
-

3
-
4
-
3
3
0
.
N
a
tu
re
p
ro
d
u
c
e
s
w
o
rk
s
o
f

g
ra
c
e
w
h
ic
h
p
le
a
se
th
e
e
y
e
3
1
.
S
p
e
c

ie
s
su
rv
iv
a
l
u
lt
im
a
te
ly
d
e
p
e
n
d
s
o
n
la
rg

e
n
u
m
b
e
rs
o
f
o
th
e
r
sp
e
c
ie
s
?
2
3
2
.
N
a

tu
re
p
ro
v
id
e
s
th
e
p
ro
fo
u
n
d
e
st
h
is
to
ri
c
a
l

m
u
se
u
m
o
f
a
ll
-
3
?
4
3
3
.
S
p
e
c
ie
s
e
x
ti

n
c
ti
o
n
re
d
u
c
e
s
p
o
ss
ib
il
it
ie
s
fo
r
fu
tu
re
g

e
n
e
ra
ti
o
n
s
?
4
?
3
?
3
?
3
3
4
.
T
h
e
k
n
o
w

le
d
g
e
o
f
th
e
m
e
re
e
x
is
te
n
c
e
o
f
sp
e
c
ie
s

is
v
a
lu
a
b
le
,
e
v
e
n
if
it
is
c
e
rt
a
in
th
a
t
I
w

il
l
n
e
v
e
r
e
x
p
e
ri
e
n
c
e
th
e
m
in
si
tu
?
3
3
5

.
G
e
n
e
ti
c
d
iv
e
rs
it
y
is
g
o
o
d
b
e
c
a
u
se
e

a
c
h
p
a
rt
ic
u
la
r
sp
e
c
ie
s
re
p
re
se
n
ts
th
e
su
c
c

e
ss
o
f
g
e
n
e
ra
ti
o
n
s
o
f
e
v
o
lu
ti
o
n
a
ry
tr

ia
l
a
n
d
e
rr
o
r
?
3
3
6
.
B
io
d
iv
e
rs
it
y
is
a
n

u
n
q
u
a
li
fi
e
d
g
o
o
d
,
i.
e
.
b
io
d
iv
e
rs
it
y

is
g
o
o
d
n
o
m
a
tt
e
r
w
h
a
t
-
3
3
7
.
H
u
m

a
n
s
a
re
m
o
ra
ll
y
p
e
rm
it
te
d
to
e
x
ti
n
g
u
is
h

a
n
y
sp
e
c
ie
s
h
a
rm
fu
l
to
h
u
m
a
n
su
rv
iv
a
l
-

4
-
3
-
4
-
4
-
4
-
3
-
4
3
8
.
W
e
c
a
n
’t
a
im
to

c
o
n
se
rv
e
b
io
d
iv
e
rs
it
y
in
a
ll
it
s
a
sp
e
c
ts
.

In
st
e
a
d
,
w
e
h
a
v
e
to
m
a
k
e
c
h
o
ic
e
s
a
b
o
u
t

in
c
re
a
si
n
g
,
m
a
in
ta
in
in
g
,
o
r
e
v
e
n
d
im

in
is
h
in
g
b
io
d
iv
e
rs
it
y
in
p
a
rt
ic
u
la
r
c
ir
c
u
m

st
a
n
c
e
s
-
3
-
3
123
T
a
b
le
3
c
o
n
ti
n
u
e
d

S
ta
te
m
e
n
ts
R
e
se
a
rc
h
e
rs
N
G
O
s
D
e
c
is
io
n

-m
a
k
e
rs
S
ta
te
m
e
n
t
p
o
si
ti
o
n
s
P
e
rs
p
e
c

ti
v
e
R
1
in
tr
in
si
c
v
a
lu
e
P
e
rs
p
e
c
ti
v
e
R
2

e
c
ti
v
e
N
G
O
1
in
tr
in
si
c
v
a
lu
e
P
e
rs
p
e
c
ti

v
e
N
G
O
2
M
ix
e
d
(i
n
tr
in
si
c
a
n
d
u
ti
li
ta
ri

a
n
v
a
lu
e
)
P
e
rs
p
e
c
ti
v
e
D
1
in
tr
in
si
c
v
a
lu

e
P
e
rs
p
e
c
ti
v
e
D
2
u
ti
li
ta
ri
a
n
v
a
lu
e
(f
o

c
u
s
o
n
E
S
)
P
e
rs
p
e
c
ti
v
e
D
3
m
ix
e
d
(i
n

tr
in
si
c
a
n
d
sp
ir
it
u
a
l
v
a
lu
e
)
3
9
.
A
s
n
a

tu
re
is
a
lw
a
y
s
c
h
a
n
g
in
g
th
e
re
is
n
o
p
o
in
t

in
c
o
n
se
rv
in
g
a
fi
x
e
d
e
c
o
sy
st
e
m
st
a
te
4
0

.
S
p
e
c
ie
s
e
x
ti
n
c
ti
o
n
s
a
re
n
o
t
n
e
c
e
ss

a
ri
ly
b
a
d
-
4
-
3
-
3
-
3
-
3
4
1
.
N
a
tu
re
a

n
d
it
s
d
iv
e
rs
it
y
m
a
k
e
o
u
r
li
v
e
s
m
e
a
n
in

g
fu
l
?
4
4
2
.
T
h
e
e
a
rt
h
’s
b
io
d
iv
e
rs
it
y

sh
o
u
ld
b
e
c
o
n
se
rv
e
d
b
e
c
a
u
se
g
e
n
e
ti
c
d

iv
e
rs
it
y
m
a
y
b
e
v
a
lu
a
b
le
in
th
e
d
e
v
e
lo
p

m
e
n
t
o
f
n
e
w
d
ru
g
s
a
g
a
in
st
d
is
e
a
se
?
4

N
e
g
a
ti
v
e
sc
o
re
s
in
d
ic
a
te
d
is
a
g
re
e
m
e
n

t
(‘
‘m
o
st
u
n
li
k
e
I
th
in
k
’’
)
a
n
d
p
o
si
ti
v

e
sc
o
re
s
in
d
ic
a
te
a
g
re
e
m
e
n
t
(‘
‘m
o
st
li
k
e

I
th
in
k
’’
)
123
(20). This perspective also embraces a moral argument, as
humans have no right ‘‘to decide
about the lives of other creatures’’ (37) and do not have a
‘‘superior moral management role
over nature’’, although there are specific cases when humans
can kill some species in order
to protect themselves (e.g. from pathogens). On the contrary,
humans have responsibility to
protect the planet, ‘‘it is a moral duty’’, as species extinction is
considered to be bad,
although there are natural extinctions (40). Species are seen as
priceless (6). According to
the respondents some species cannot be valued in monetary
terms, and the lack of eco-

nomic value of a particular species does not make it superfluous
since ‘‘most species are
necessary for one or another’’ (17). Humans are seen as part of
nature ‘‘not detached from
everything’’. Moreover, biodiversity is seen as important for
future options (33). This
reflects the precautionary principle, because of uncertainty and
gaps in our knowledge
about ecosystem functioning (1) and the need to try and
conserve all aspects of biodiversity
(38). In fact, respondents behind this perspective reflect on the
moral relevance of biodi-
versity conservation ‘‘it is foremost a moral question’’ (4, 37)
and express disagreement
with the claim that valuing species in economic terms is
harmful (29).
The second perspective represented by the researchers (R2) is
more utilitarian, as, while
it contains elements of the intrinsic value of nature, it also
emphasizes the role of biodi-
versity in fulfilling human needs. Biodiversity is seen as
fundamental in providing food
security (11, 15), producing new drugs in the future (42) and
offering recreational

opportunities (13). However, respondents strongly disagree with
the statement that bio-
diversity is good no matter what (36), as ‘‘in some cases the
costs of conserving biodi-
versity may exceed the benefits’’. The precautionary principle
is also mentioned in this
perspective (1, 24) and biodiversity is seen as useful because of
its economic value and the
ecosystem services it delivers (2). Thus, applying the ecosystem
services approach in
conservation is seen as potentially useful as an effective
argument about the benefits of
nature ‘‘the anthropocentric framing can be very effective
politically’’ and it is a ‘‘very
important new tool if properly applied’’ (3). However, the
respondents representing this
perspective do not agree that economic valuation can be a
justification for destroying
nature (29).
Respondents representing this perspective also believe that
biodiversity is a moral
matter (4) and do not support the idea that humans have the
right to kill any species that is

dangerous (37) or that most species are superfluous (17). The
respondents do not agree
with the arguments that compare the destruction of nature to
destroying works of art or
books (8, 28). Some found the metaphors not relevant or useful
as arguments for biodi-
versity conservation, while one respondent saw the art metaphor
as a ‘‘creationist view-
point’’, as art is created by humans and the species were created
through evolution.
According to another respondent, extinction can be a part of
nature, while ‘‘a work of art is
irreplaceable’’. In the same manner, ‘‘destroying books is
irreversible, with nature it is
different, not all kind of nature is destroyed irreversibly, it can
be restored’’.
NGOs’ perspectives on biodiversity conservation: intrinsic
and anthropocentric values
Similar to the researchers, the first perspective of the NGOs
(NGO1) focuses on the
intrinsic value of biodiversity. According to the respondents
behind this perspective,
species have a right to exist and they are valuable regardless of
their economic value and

their ability to serve human needs (6, 20). The knowledge of
their mere existence is also of
great value, no matter whether we will have the possibility to
see them in our lifetime or
not (34), yet biodiversity conservation should not be limited by
considerations of other
123
values such as freedom, equality, health, and justice (27).
Although little untouched nature
is left, pristine nature has a value in itself (25). Therefore,
species are seen as invaluable
and ‘‘standing above economic valuation’’ (6). Nevertheless,
economic valuation would
not necessarily lead to the destruction of nature (29). NGO1
also believe that humans are
not ‘‘the lords of nature’’, and thus cannot simply kill species
that are harmful (with the
exception of extreme cases, such as ‘‘smallpox virus’’) (37).
However, this is a complex
issue that depends on the particular context, e.g. local people
suffer more from large

carnivores protected at the national level. Moreover, in this
perspective, the extinction of
species is seen as bad (40). Ecological values are reflected in it
not being right to say that
most species are superfluous, as ‘‘everything complements each
other’’, ‘‘species’ survival
depend on a large number of other species’’ and ‘‘every species
is a part of a whole’’ (17,
22). Therefore, one needs to be careful, particularly when
considering the needs of future
generations as the extinction of species will reduce their options
(33).
The second NGOs’ perspective (NGO2) includes very diverse
statements. While there
are intrinsic and ecological value arguments, concerning
species’ inherent right to exist
(20) no matter if they benefit humans, and seeing every species
as having some role (17), a
more anthropocentric thinking about nature also is present.
Respondents underline the role
of humans in shaping (cultural) landscapes (26). Biodiversity
conservation is not seen as a
moral matter (4), but species extinction is bad and humans are
not permitted to kill any

species even if they are harmful to human survival (40, 37).
Unlike any other group, NGO2
see nature as a laboratory for learning (21), but not like a
museum (32). Within this
perspective, economic valuation is seen as potentially useful in
communicating the value
of biodiversity, and the ecosystem services concept can serve as
a justification for species
protection (3, 29). Yet, according to one respondent the
economic valuation can also be
problematic, as some ecologically important species, such as
soil fauna, do not have a
direct economic value in conventional markets. Respondents
acknowledge the importance
of different species for future generations (33), and they see the
need to apply the pre-
cautionary principle (1) due to the imperfections in our
knowledge of how ecosystems
might respond to future changes and thus the need to conserve
species as back-ups (24).
Decision-makers: intrinsic values, spiritual values and
ecosystem services
As in the other two stakeholder groups, the main focus of the
first perspective represented

by decision-makers (D1) is the intrinsic and ecological value of
biodiversity and pristine
nature (25). Intrinsic value is fundamental; it is like a ‘‘religion
or a belief’’. All species
have a right to exist, ‘‘they cannot be valued as they are
invaluable’’, they are all important,
but usually not irreplaceable (20, 6, 17). This perspective also
highlights that species
depend on each other, even though there can be replacements,
and that ‘‘nature is flexible’’,
although sometimes the system can break down (22). D1
respondents claim that humans
are not allowed to extinguish any species, but there can be
exceptions (e.g. ‘‘mosquitos’’)
(37). Extinctions that are caused by humans are seen as wrong,
and referring to extinctions
as a natural process can be used as ‘‘an excuse from our
responsibility to conserve’’ (40).
The respondents also believe that we should aim to conserve
biodiversity in all its aspects
(38) because we have no right to diminish it, although some
admitted that, in reality,
sometimes we have other goals (as in the case of dealing with
invasive species). Although

the main focus of this perspective is on the intrinsic and
ecological values of nature, it also
mentions conserving possibilities for future generations (33).
The ecosystem services
approach and economic valuation are seen as good tools in
communication, although both
123
can be ‘‘dangerous as they can be abused’’ (3, 29).
Nevertheless, economic valuation of
nature itself is not seen as justifying the destruction of the
biosphere (29).
The most prominent message of the second perspective of the
decision-makers (D2) is
the role of biodiversity in providing certain goods and services
for humans. The usefulness
of biodiversity in poverty alleviation and food provision are
mentioned (10, 11). Main-
taining genetic diversity for food security is very important for
reducing vulnerability,
especially to changes, like climate change and in developing
countries. The insurance

value of biodiversity is also seen as fundamental for
maintaining functioning ecosystems
under drivers of change, as particular species can replace each
other (11, 24). This per-
spective also highlights the possible co-benefits of conservation
and tourism (12). The
economic value of biodiversity is considered important (2),
although biodiversity should
not be reduced to its economic benefits only. The ecosystem
services concept and eco-
nomic valuation, however, is described as a ‘‘promising
approach’’ in convincing society of
the importance of conservation. According to one respondent,
ecosystem services provide a
link between nature conservation and instrumental values.
Economic valuation is not
meant to destroy nature, and such thinking is seen as a
‘‘misconception’’ (29). Besides all
these utilitarian arguments, the decision-makers behind the
second perspective mostly see
biodiversity as a moral matter (4). Thus, humans are not
permitted to kill any harmful
species, although in some cases there is a ‘‘room for discussion
(e.g. pathogens)’’ (37). As

we do not know enough to say that most species are
superfluous, ‘‘it cannot be a reason to
not conserve species’’ (17).
The third perspective of the decision-makers (D3), besides
intrinsic value, includes
ecological, spiritual and aesthetic value elements. The
respondents behind this perspective
believe that nature makes our lives meaningful, and that without
it ‘‘human existence has
no sense’’ as we are ‘‘part of nature’’ (41). The world would
lose its magic without
biodiversity, and it would be a ‘‘poorer place’’ (14). D3
respondents also argue for the
intrinsic value of biodiversity and pristine nature. According to
them, the existence of
species and their conservation should not depend on their
capacity to provide services for
humans (20, 25). Biodiversity conservation is seen as a moral
matter and we have a moral
responsibility towards nature (4). Species extinction is bad ‘‘no
matter the motivation’’
(40). We do not have a right to kill any harmful species, except
those with which it is
impossible to co-exist (e.g. ‘‘smallpox virus’’) (37). Those

representing this perspective
also believe that ‘‘all species are important’’ (17), and thus, it
is essential to maintain the
integrity of ecosystems. ‘‘If we want a healthy environment we
need to stress the inter-
connectedness of it all (the system), we need to think about it as
a whole, rather than
(focusing on) individual species’’ (22). This perspective also
underlines the possibilities
that can be lost for future generations due to species
extinctions. According to one
respondent, we will never know what we will lose with species
extinctions, from a
medicine to the experience of seeing a tiger (33).
Discussion
There are various approaches to eliciting different views on
aspects of biodiversity and
conservation, with each serving different purposes. For
example, questionnaires have been
used to explore the thoughts of professional and public on
nature and landscape (Buijs and
Eland Buijs and Elands 2013); focus groups in France, Hungary
and Italy to compare

organic and conventional farmer’s perceptions of biodiversity
(Kelemen et al. 2013) and
123
semi-structured interviews followed by questionnaires to
capture views on protected area
expansion in Poland (Grodzinska-Jurczak and Cent 2011).
While Q methodology has its
limitations (see Methods), one obvious advantage is the
relatively small sample size
(number of respondents) needed for the exploration of different
perspectives within a
population, especially in comparison with traditional survey
methods, such as question-
naires, although these serve different purposes. Furthermore,
the Q methodology allows for
fairly straightforward and easy analyses compared to other
social science methods, e.g.
discourse analyses, designed to exploring people’s attitudes,
and embraces the subjective
nature of attitudes by asking for people’s subjective views on
specific topics. Another

strength of the Q methodology is that it allows a combination of
quantitative and quali-
tative information in the interpretation of results. Thus, the Q
methodology allows for a
relatively robust statistical analysis of people’s subjectivities,
while at the same time
offering enough flexibility in the interpretation of results to
allow for accurate represen-
tations of respondents’ views.
Comparison of perspectives across the stakeholder groups
Our analysis has revealed seven different perspectives among
the conservation practi-
tioners interviewed. Given that they focus on different aspects
of biodiversity conservation
in their work, one could expect considerable variation in
perspectives amongst the groups.
However, our study identified a wide range of perspectives in
all of these groups, from
intrinsic to utilitarian, with certain statements (e.g. 17 and 37
around intrinsic ecological
value and a moral argument) sorted similarly by the respondents
from all the perspectives
across all the stakeholder groups. These are the values which
traditionally have been seen

as central to motivations for conservation (McCauley 2006).
Moreover, no respondents
thought that economic valuation led to a justification for
destruction of the biosphere (29),
and most thought that human-induced species extinctions were
bad (40). There were,
however, within particular perspectives, certain views that were
stressed and statements
that were not shared by any other group. For example, for the
potential of biodiversity for
delivering ecosystem services, R2 emphasised the importance of
cultural experiences and
genetic resources, especially for dealing with future change (13,
15, 42), while D2 focused
on poverty alleviation and the benefits of (eco)tourism (10, 12).
It is also interesting to note
that the decision-makers have a mix of perspectives and that we
identified more per-
spectives in this group than in the others.
The stakeholder groups we investigated are all directly involved
in and influencing
biodiversity policy making and working for biodiversity
conservation, sharing a common

goal. We acknowledge that our sample of respondents included
more researchers, while
two other groups were less represented (Table 1). This is partly
because we sought to
include both natural and social scientists, so this was more
heterogeneous group, which
needed to be represented. Also, it may both reflect the fact that
biodiversity governance in
Europe is dominated by researchers, and the availability of
respondents that have agreed to
take part in the study. However, we believe that the greater
number of researchers in our
investigation do not undermine the importance of our findings,
as each group was analysed
separately and our findings show that the main differences in
perspectives on biodiversity
conservation are not between the particular stakeholder groups,
but rather between eco-
centric and anthropocentric views. These two views could be
clearly distinguished in the
NGO group, while the researchers and decision-makers focused
more on the ecosystem
services (perspectives R2 and D2).

123
A number of statements did not reflect the thinking of any
participants and thus they
may represent arguments which are less likely to be relevant in
discussions about biodi-
versity conservation in Europe, or that these arguments were
considered less critical or not
as strong as the others. These statements included some about
the spiritual and aesthetic
aspects of biodiversity (e.g. 7, 30), the desire to experience
pristine or untouched areas (16)
and beauty as a potential basis for conservation (18). For some
of these there are related
statements, for example statements 16 and 25 express similar
sentiments, and, from
observations of the sorting, in such cases respondents preferred
one statement over another,
and only placed one of them in the ‘‘more like I think’’ part of
the sort. In fact, the well-
known argumentation line followed by McCauley (2006)
referred to ‘‘The reason biodi-
versity matters is because it confers on us an imprecise,

immeasurable well-being that is
located in the spirit rather than in the wallet’’ (7) was not
supported or opposed in any
perspective. Also, the respondents did not think in line with
statement 19, ‘‘We do not need
to recognize other beings as our moral equals to realize that we
should not kill that which is
not a threat’’, although they reacted strongly to other, closely
linked statements, such as all
species having a right to exist (20) and humans (not) being
morally permitted to extinguish
species which are harmful to our survival (37). These less
salient statements could offer a
good starting point for discussions to reach an understanding or
agreement among con-
flicting interest groups, as they could represent less
controversial statements, which are not
so critical to their standpoints. Then, once they are already
talking, it could be easier to
move on to discussing issues, which are actually important to
the parties.
Perspectives on ecosystem services, economic valuation and
biodiversity
conservation

The ecosystem service approach was represented explicitly in
four statements (2, 3, 10,
29), which refer to the concept itself and to the economic
valuation process in particular.
They were important for certain perspectives and this may be a
result of the mainstreaming
of the ecosystem service concept, which created a forum of
debate where different
stakeholders can express their various opinions (Primmer et al.
2015). In our study,
decision-makers were often pragmatic and remarked on the key
potential contributions of
ecosystem services (e.g. to poverty alleviation).
The respondents also all agreed that economic value or the lack
of it should not be used
as a justification to destroy nature or species, but otherwise
opinions about economic value
and valuation seemed to be rather mixed. This is in line with the
long ongoing debate about
the opportunities, but also the limits, drawbacks, and problems
of economic valuation of
biodiversity (e.g. Spash 1997; Daily et al. 2000; Norgaard 2010;
TEEB 2010). Many
authors have previously pointed out the risk of economic

valuation, which may lead to
commodification of nature (Gómez-Baggethun and Ruiz-Pére
2011; Salles 2011; Gómez-
Baggethun et al. 2016). However, others point to the potential
benefits of economic val-
uation that does not necessarily lead to such commodification
(Costanza et al. 2014).
As presented by Kallis et al. (2013), to value or not to value
could be a false dilemma,
and the decision will depend on many other contextual factors
that should be analysed.
Following our findings, it seems that all respondents were
aware of the potential downsides
of economic valuation. The range of perspectives about
biodiversity identified in our study
supports the suggestion that a more integrated and pluralistic
approach to biodiversity and
its valuation is needed Intergovernmental Science-Policy
Platform on Biodiversity and
Ecosystem Services (IPBES ) 2014; Gómez-Baggethun et al.
2016). This should consider
123

different values (anthropocentric and non-anthropocentric) and
valuation techniques
including ecological, economic and socio-cultural approaches.
Why conserve biodiversity?
Little published research is available comparing the views of
conservation stakeholders
across Europe, as opposed to studies in individual countries
(e.g. Sandbrook et al. 2010;
Buijs and Elands 2013; Couix and Hazard 2013). This study is
the first to identify these
differing conservation perspectives between stakeholder groups
and revealing them can
help communication and cooperation. Our study has shown that,
while the stakeholders do
have a variety of reasons for why nature should be conserved,
ranging from moral to more
anthropocentric and utilitarian arguments, there is an overall
appreciation of certain aspects
of the intrinsic value of nature. This, and the existence of
elements of utilitarian or
anthropocentric perspectives in all groups provides, some
common ground with opportu-

nities for building an inclusive discourse around biodiversity
conservation. Also, realising
and acknowledging the differences in stakeholders’ arguments
for biodiversity when
engaging in discussions about conservation is likely to lead to
better communication and
thus more effective delivery of conservation solutions
(Gustafsson 2013).
The ecosystem services concept seems to lie between the
intrinsic value and the utili-
tarian perspectives and may form an important conceptual and
communication bridge. In
our analysis, it appeared in the utilitarian perspective of
researchers (R2), in the mixed
perspective of NGOs (NGO2) and in the intrinsic value and
utilitarian perspectives of
decision-makers (D1 and D2). It might be explained by the main
feature of ecosystem
services, which is their linking of social and ecological systems
(e.g. the cascade model of
Haines-Young and Potschin 2010). Despite being inherently
anthropocentric, concentrat-
ing on the benefits people obtain from ecosystems (Millennium
Ecosystem Assessment

2005), the ecosystem services approach has potential for raising
awareness about the
importance of nature and nature conservation across various
stakeholder groups.
In conclusion, this study has provided insights into the reasons
underlying European
conservation practitioners’ value of biodiversity. However,
while we found an overall
appreciation of certain aspects of the intrinsic value of nature,
we also revealed a broad
spectrum of perspectives on biodiversity conservation from
intrinsic to utilitarian ones. The
main differences appeared to result from the espousal of
ecocentric or anthropocentric
viewpoints, rather than from differences between the various
stakeholder groups. Under-
standing of these different and sometimes diverse perspectives
represented by the con-
servation practitioners can provide the basis for better
cooperation and more effective
argumentation for maintaining biodiversity. Thus,
understanding how arguments for
conservation are considered by different stakeholders is of
crucial importance for the

planning of effective biodiversity conservation and the use of a
variety of arguments based
on the plurality of views may enhance the acceptability and
success of conservation action.
Acknowledgments We would like to thank all the stakeholders
who took part in this study and gave freely
of their time and thoughts. This work was also supported by the
European Union, under FP7 project
BESAFE (FP7-ENV.2011.282743). MGL was also funded by a
postdoctoral grant from the Spanish
National Institute for Agriculture and Food Research and
Technology (INIA), which is co-funded by the
European Social Fund. Authors from the Szent István University
were also supported by the Research
Centre of Excellence (9878/2015/FEKUT, 9878-
3/2016/FEKUT). We would also like to thank two
anonymous reviewers for their comments, which helped to
strengthen this paper.
123
Appendix: the 42 Q statements
1. We do not know how ecosystems will be affected by the loss
of species, therefore we
better preserve them.
2. Protecting ecosystem service providers is important because
they are a source of

economic value.
3. The ecosystem service approach has potential to improve
species conservation in
Europe.
4. Biodiversity conservation is not a moral matter.
5. Some species are important symbols of human values, such as
freedom.
6. Species are priceless.
7. The reason biodiversity matters is because it confers on us an
imprecise,
immeasurable well-being that is located in the spirit rather than
in the wallet.
8. The extinction of a species is like the destruction of a great
work of art.
9. It is not clear why all species that environmentalists
campaign to conserve ought to be
saved.
10. Protecting biodiversity and ecosystem services is
particularly important for poverty
alleviation in developing countries.
11. Conserving genetic diversity is important to feed future
human populations.

12. Countries can benefit from their conservation efforts
through tourism.
13. Nature provides us with many valuable experiences. We
hunt, fish, hike, mountain
climb, and engage in numerous activities in which we interact
with nature.
14. Losing its biological richness and diversity, the world loses
its magic.
15. It is important to conserve the genetic reservoir in a region,
in case we need to breed
disease-resistant plants or produce food adapted to local
conditions.
16. We want to experience areas where humans are merely
visitors and not inhabitants.
17. Most species are superfluous.
18. We value some species for their beauty, but this is only
relevant for a very small
number of species. Therefore, beauty is not a particularly
important basis for
conservation.
19. We do not need to recognize other beings as our moral
equals to realize that we
should not kill that which is not a threat.

20. All species have a right to exist, regardless of their ability
to benefit humans.
21. Nature is a laboratory for the pursuit of science through
which society gains
knowledge, and understanding of the world.
22. The diversity of life is something like the rivets on an
airplane, with each species
playing a small but significant role in the working of the whole.
The loss of each rivet
weakens the plane by a small but noticeable amount—until it
loses airworthiness and
crashes.
23. Nature provides a place to take calculated risks, to learn the
luck of the weather, to
lose and find one’s way, to reflect on success and failure.
24. Even if only a few species are needed for our world to be
productive we have to
conserve more species as a back-up. Otherwise a pest or climate
change could wipe
out the few species we have saved, and we would have nothing
in reserve.
25. Pristine nature is valuable in itself.

26. Ecosystems have co-evolved with humans creating
landscapes of important cultural
value.
123
27. Any effort to conserve biodiversity must be limited by
considerations of other values
such as freedom, equality, health, and justice.
28. Destroying nature is like burning unread books.
29. Valuing species in economic terms implies a justification
for the destruction of the
biosphere.
30. Nature produces works of grace which please the eye.
31. Species survival ultimately depends on large numbers of
other species.
32. Nature provides the profoundest historical museum of all.
33. Species extinction reduces possibilities for future
generations.
34. The knowledge of the mere existence of species is valuable,
even if it is certain that I

will never experience them in situ.
35. Genetic diversity is good because each particular species
represents the success of
generations of evolutionary trial and error.
36. Biodiversity is an unqualified good, i.e. biodiversity is good
no matter what.
37. Humans are morally permitted to extinguish any species
harmful to human survival.
38. We can’t aim to conserve biodiversity in all its aspects.
Instead, we have to make
choices about increasing, maintaining, or even diminishing
biodiversity in particular
circumstances.
39. As nature is always changing there is no point in conserving
a fixed ecosystem state.
40. Species extinctions are not necessarily bad.
41. Nature and its diversity make our lives meaningful.
42. The earth’s biodiversity should be conserved because
genetic diversity may be
valuable in the development of new drugs against disease.
References
Armsworth PR, Chan KMA, Daily GC, Ehrlich PR, Kremen C,

Ricketts TH, Sanjayan MA (2007)
Ecosystem-service science and the way forward for
conservation. Conserv Biol 21:1383–1384
Bagnoli P, Goeschl T, Kovács E (2008) People and biodiversity
policies, impacts, issues and strategies for
Policy Action. OECD, Paris
Barry J, Proops J (1999) Seeking sustainability discourses with
Q methodology. Ecol Econ 28:337–345
Brondı́zio E, Gatzweiler FW, Zografos C, Kumar M (2010) The
socio-cultural context of ecosystem and
biodiversity valuation. In: Kumar P (ed) The economics of
ecosystems and biodiversity: ecological and
economic foundations. Earthscan, London, pp 149–181
Brown SR (1980) Political subjectivity: applications of Q
methodology in political science. Yale University
Press, New Haven, CT
Brown SR (1993) A primer on Q methodology. Operant Subj
16:91–138
Buijs AE, Elands BHM (2013) Does expertise matter? An in-
depth understanding of people’s structure of
thoughts on nature and its management implications. Biol
Conserv 168:184–191
Buijs AE, Fischer A, Rink D, Young JC (2008) Looking beyond
superficial knowledge gaps: understanding
public representations of biodiversity. Int J of Biodi ver Sci and
Manag 4:65–80
Cardinale BJ, Duffy JE, Gonzalez A et al (2012) Biodiversity
loss and its impact on humanity. Nature

486(7401):59–67
Chamberlain EC, Rutherford MB, Gibeau ML (2012) Human
perspectives and conservation of grizzly bears
in Banff National Park, Canada. Conserv Biol 26:420–431
Costanza R, de Groot R, Sutton P, van der Ploeg S, Anderson
SL, Kubiszewski I, Farber S, Turner RK
(2014) Changes in the global value of ecosystem services.
Global Environ Change 26:152–158
Couix N, Hazard L (2013) When the future of biodiversity
depends on researchers’ and stakeholders’
thought-styles. Futures 53:13–21
Daily GC, Soderquist T, Aniyar S, Arrow K, Dasgupta P,
Ehrlich PR, Folke C, Jansson AM, Jansson BO,
Kautsky N, Levin S et al (2000) The value of nature and the
nature of value. Science 289:395–396
123
De Lopez TT (2001) Stakeholder management for conservation
projects: a case study of Ream National
Park, Cambodia. Environ Manage 28:47–60
Dearborn DC, Kark S (2010) Motivations for conserving urban
biodiversity. Conserv Biol 4:432–440
Ehrlich PR, Ehrlich AH (1992) The value of biodiversity.
Ambio 21:219–226
Fischer A, Young JC (2006) Understanding mental constructs of
biodiversity: implications for biodiversity

management and conservation. Biol Conserv 136:271–282
Gómez-Baggethun E, Ruiz-Pére M (2011) Economic valuation
and the commodification of ecosystem
services. Prog in Phys Geog 35:613–628
Gómez-Baggethun, E. Barton DN, Berry P, Dunford R, Harrison
PA (2016) Concepts and methods in
Ecosystem Service Valuation. In: Haines-Young R, Fish R,
Turner RK (eds) Potschin M. Routledge,
Handbook of Ecosystem Services
Grimble R, Wellard K (1997) Stakeholder methodologies in
natural resource management: a review of
principles, context, experiences and opportunities. Agr Sys
55:173–193
Grodzinska-Jurczak M, Cent J (2011) Expansion of nature
conservation areas: problems with Natura 2000
implementation in Poland? Environ Manag 47:11–27
Gustafsson KM (2013) Environmental discourses and
biodiversity: the construction of a storyline in
understanding and managing an environmental issue. J Integr
Environ Sci 10:39–54
Haines-Young R, Potschin M (2010) The links between
biodiversity, ecosystem services and huma n well-
being. In: Raffaelli DG, Frid CLJ (eds) Ecosystem ecology: a
new synthesis. Cambridge University
Press, New York, pp 110–139
Hermelingmeier V (2014) Harmonizing OPERAs voices. An
investigation of different perspectives on the
ecosystem services concept and their implications for research

and practice. Masters Thesis, University
of Lund
Howard BM, Braat L, Bugter R, Hails RH (2013) Deliverable
1.1: Report on the classification of arguments
and the provisional framework http://www.besafe-
project.net/files/DOWNLOAD2/D1%201%
20Report%20on%20classification%20of%20arguments%20-
%20FINAL%20.pdf
IPBES (Intergovernmental Science-Policy Platform on
Biodiversity and Ecosystem Services) (2014) Pre-
liminary guide regarding diverse conceptualization of multiple
values of nature and its benefits,
including biodiversity and ecosystem functions and services,
IPBES/3/INF/7)
Kahneman D, Knetsch JL (1992) Valuing public goods: the
purchase of moral satisfaction. J of Environ
Econ and Manag 22:57–70
Kallis G, Gómez-Baggethun E, Zogrofos C (2013) To value or
not to value? That is not the question. Ecol
Econ 94:97–105
Kamal S, Grodzińska-Jurczak M (2014) Should conservation of
biodiversity involve private land? A Q
methodological study in Poland to assess stakeholders’ attitude.
Biodiver and Conserv 23:2689–2704
Kamal S, Kocór M, Grodzińska-Jurczak M (2014) Quantifying
human subjectivity using Q method: when
quality meets quantity. Qual Sociol Rev. 10:60–79
Kampen JK, Tamás P (2014) Overly ambitious: contributions
and current status of Q methodology. Qual

and Quant 48:3109–3126
Kelemen E, Nguyen G, Gomiero T, Kovács E, Choisis J-P,
Choisis N, Paoletti MG, Podmaniczky L,
Ryschawy J, Sarthou J-P, Herzog F, Dennis P, Balázs K (2013)
Farmers’ perceptions of biodiversity:
lessons from a discourse-based deliberative valuation study.
Land Use Policy 35:318–328
Marjainé Szerényi Z, Zsóka Á, Ásványi K, Flachner Z (2011)
The Role of Adaptation to Climate Change in
Rural Development. Reg and Bus Stud 3(Suppl 1):189–198
McCallum M (2015) Vertebrate biodiversity losses point to a
sixth mass extinction. Biodivers Conserv
24:2497–2519
McCauley DJ (2006) Selling out on nature. Nature 443:27–28
Meyer JM (1997) Gifford Pinchot, John Muir, and the
Boundaries of Politics in American Thought. Polity
30:267–284
Milcu AI, Sherren K, Hanspach J, Absom D, Fischer J (2014)
Navigating conflicting landscape aspirations:
application of a photo-based Q-method in Transylvania (Central
Romania). Land Use Policy
41:408–422
Millennium Ecosystem Assessment (2005) Ecosystems and
human well-being: synthesis. World Resource
Institute, Washington DC
Montgomery C (2002) Ranking the benefits of biodiversity: an
exploration of relative values. J of Environ
Manag 65:313–326

Mushove P, Vogel C (2005) Heads or tails? Stakeholder
analysis as a tool for conservation area manage-
ment. Glob Environ Change 15:184–198
Ninjik M, Ninjik A, Bizikova L (2009) Analysing the
development of small-scale forestry in central and
eastern Europe. Small-scale For 8:159–174
123
http://www.besafe-
project.net/files/DOWNLOAD2/D1%25201%2520Report%2520
on%2520classification%2520of%2520arguments%2520-
%2520FINAL%2520.pdf
http://www.besafe-
project.net/files/DOWNLOAD2/D1%25201%2520Report%2520
on%2520classification%2520of%2520arguments%2520-
%2520FINAL%2520.pdf
Norgaard R (2010) Ecosystem services: from eye-opening
metaphor to complexity blinder. Ecol Econ
69:1219–1227
Nunes PALD, van der Bergh JCJM (2001) Economic valuation
of biodiversity: sense or nonsense? Ecol
Econ 39:203–222
O’Neill SJ, Boykoff M, Niemeyer S, Day SA (2013) On the use
of imagery for climate change engagement.
Glob Environ Change 23(2):413–421
Peterson MJ, Hall DM, Feldpausch-Parker AM, Peterson TR

(2009) Obscuring ecosystem function with
application of the ecosystem services concept. Conserv Biol
24:113–119
Primmer E, Jokinen P, Blicharska M, Barton DN, Bugter R,
Potschin M (2015) Governance of ecosystem
services: a framework for empirical analysis. Ecosyst Serv
16:159–166
Raffaelli D, Smart J, Austen M, Mangi S, Hattam C, Termansen
M, Fraser E, Abson D (2009) Valuation of
biodiversity? A NERC scoping study. Natural Environment
Research Council, Swindon
Reed MS (2008) Stakeholder participation for environme ntal
management: a literature review. Biol Conserv
141:2417–2431
Ridder B (2008) Questioning the ecosystem services argument
for biodiversity conservation. Biodivers
Conserv 17:781–790
Russell R, Guerry AD, Balvanera P, Gould RK, Basurto X, Chan
KMA, Klain S, Levine J, Tam J (2013)
Humans and nature: how knowing and experiencing nature
affect well-being. Annu Rev Environ
Resour 38:473–502
Salles JM (2011) Valuing biodiversity and ecosystem services:
why put economic values on nature? CR
Biol. doi:10.1016/j.crvi.2011.03.008
Sandbrook C, Scales I, Vira B, Adams WM (2010) Value
plurality among conservation professionals.
Conserv Biol 25:285–294

Secretariat of the Convention on Biological Diversity (2014)
Global Biodiversity Outlook 4. Montréal
Spash CL (1997) Ethics and environmental attitudes with
implications for economic valuation. J of Environ
Manag 50:403–416
Stainton Rogers R (1995) Q methodology. In: Smith JA, Harre
R, Van Langenhove L (eds) Rethinking
methods in psychology. Sage, Thousand Oaks, CA, pp 178–192
Suškevičs M, Tillemann K, Külvik M (2013) Assessing the
relevance of stakeholder analysis for national
ecological network governance: the case of the Green Network
in Estonia. J for Nat Conserv
21:206–213
Swedeen P (2006) Post-normal science in practice: a Q study of
the potential for sustainable forestry in
Washington State, USA. Ecol Econ 57:190–208
TEEB (2010) In: Kumar P (ed) The Economics of Ecosystems
and Biodiversity: Ecological and Economic
Foundations. Earthscan, London and Washington
Ulyshen MD (2013) Strengthening the case for saproxylic
arthropod conservation: a call for ecosystem
services research. Insect Conserv Divers 6:393–395
Watts S, Stenner P (2005) The subjective experience of
partnership love: a Q methodological study. Brit J
Soc Psychol 44:85–107
Watts S, Stenner P (2012) Doing Q methodological research
theory, method and interpretation. Sage,
London

Webler T, Danielson S, Tuler S (2009) Using Q method to
reveal social perspectives in environmental
research. Greenfield (MA): Social and Environmental Research
Institute. Available from http://www.
seri-us.org/content/primer-q-methodology-available-free-
download
Wilson EO (1984) Biophilia. Harvard University Press,
Cambridge
Wolf J (2006) Climate change and citizenship: a case study of
responses in Canadian coastal communities.
Department of Development Studies, University of East Anglia,
Norwich
123
http://dx.doi.org/10.1016/j.crvi.2011.03.008
http://www.seri-us.org/content/primer-q-methodology-available-
free-download
http://www.seri-us.org/content/primer-q-methodology-available-
free-download
Reproduced with permission of copyright owner.
Further reproduction prohibited without permission.
Why conserve biodiversity? A multi-national exploration of
stakeholders’ views on the arguments for biodiversity
conservationAbstractIntroductionMethodologyResultsResearche
rs’ perspectives on biodiversity conservation: intrinsic values
and ecosystem servicesNGOs’ perspectives on biodiversity
conservation: intrinsic and anthropocentric valuesDecision-
makers: intrinsic values, spiritual values and ecosystem

servicesDiscussionComparison of perspectives across the
stakeholder groupsPerspectives on ecosystem services,
economic valuation and biodiversity conservationWhy conserve
biodiversity?AcknowledgmentsAppendix: the 42 Q
statementsReferences

Contents lists available at ScienceDirectJournal of Enviro

More Related Content

More from AlleneMcclendon878 (20)

Recently uploaded (20)

Contents lists available at ScienceDirectJournal of Enviro