Aurthy fathimaPhysical.sce

MAR-SFM Working Paper 5 / 2007
Forestry Department
Food and Agriculture Organization of the United Nations
MANUAL ON DEFORESTATION, DEGRADATION, AND
FRAGMENTATION USING REMOTE SENSING AND GIS
PREPARED BY GIRI TEJASWI
ROME, MARCH 2007
STRENGTHENING MONITORING, ASSESSMENT AND REPORTING
ON SUSTAINABLE FOREST MANAGEMENT IN ASIA (GCP/INT/988/JPN)

Strengthening Monitoring, Assessment and Reporting (MAR) on Sustainable Forest
Management (SFM) in Asia (GCP/INT/988/JPN)
FAO initiated the project “Strengthening Monitoring, Assessment and Reporting on Sustainable Forest
Management in Asia” (GCP/INT/988/JPN) in January 2006. The five-year project is funded by the Government of
Japan.
The main objective of this project is to facilitate development of harmonized forest related national monitoring,
assessment and reporting (MAR) systems in the Asia-Pacific region to contribute directly to the improvement of
sustainable forest management (SFM) regimes. An allied objective of the project is to enhance the use of the
MAR information in national decision-making, formulation of effective forest policies, and sustainable forest
management and planning.
The project accomplishes its objectives in two phases. The first two years, the Development Phase, the project
would focus on: (a) international activities like the establishment of linkages with forest-related processes; (b)
facilitating development of a globally harmonized framework, guidelines and database structure, including pilot
testing in some countries; (c) use of MAR information in forest planning and development of forest policies at the
national level; (d) establishment of a country-level network of national focal points to various forest-related
processes; and (e) initiate a set of national activities that facilitate the implementation of the harmonized MAR.
The Implementation Phase spreads over the remaining three years of the project period and focuses on the
implementation of the harmonized MAR, including facilitation in the establishment of database at the national level
in selected project countries within the Asia-Pacific region through studies, reviews, training, workshops and
expert consultations. The detailed design of this phase will be finalized on the basis of a review of the activities
and the outputs of the first phase.
All countries in the Asia-Pacific region can participate in the project, although the actual level and intensity of their
involvement may vary among them. Forestry departments in respective countries have been requested to
nominate their national focal points for this project.
The project is organized under the Forest Resources Development Service (FOMR) in the Forest Resources
Division (FOM) of FAO Forestry Department. The contact persons are:
Mr Peter Holmgren, Chief FOMR Peter.Holmgren@fao.org
Mr Kailash Govil, Senior Forestry Officer Kailash.Govil@fao.org
Monitoring Assessment and Reporting
DISCLAIMER
The MAR-SFM Working Paper Series is designed to reflect the activities and progress of the MAR on
SFM programme of FAO. Working Papers are not authoritative information sources – they do not reflect the
official position of FAO and should not be used for official purposes. Please refer to the FAO forestry website
(www.fao.org/forestry) for access to official information.
Views of participants in a workshop reported in the proceeding may be considered as their personal
views. These may be same or different from the official view of their country.
The MAR-SFM Working Paper Series provides an important forum for the rapid release of preliminary
findings needed for validation and to facilitate the final development of official quality-controlled publications.
Should users find any errors in the documents or have comments for improving their quality they should contact
Kailash.Govil@fao.org.

3
MANUAL ON DEFORESTATION, DEGRADATION, AND
FRAGMENTATION USING REMOTE SENSING AND GIS
TABLE OF CONTENTS
1. Introduction..............................................................................................................................................4
2. Basic Concepts of Deforestation, Degradation, and Fragmentation........................................................5
2.1 Deforestation..................................................................................................................................5
2.2 Forest Degradation.........................................................................................................................6
2.3 Forest Fragmentation.....................................................................................................................7
2.4 Afforestation..................................................................................................................................7
2.5 Reforestation..................................................................................................................................7
2.6 Forest improvement.......................................................................................................................8
3. Agents, causes, and consequences of Deforestation, Degradation, and Fragmentation..........................9
3.1 Agents............................................................................................................................................9
3.2 Causes..........................................................................................................................................10
3.2.1 Direct causes of deforestation.............................................................................................11
3.2.2 The underlying causes of deforestation and forest degradation .........................................11
3.3 Consequences ..............................................................................................................................13
4. Remote Sensing of Deforestation, Degradation, and Fragmentation ....................................................16
4.1 Availability, Accessibility, and Affordability of Remote Sensing Data .....................................16
4.2 Recent Advances in Remote Sensing Technology ......................................................................19
4.2.1 Methodology to Assess and Monitor Deforestation, Degradation and Fragmentation using
Remote Sensing and GIS.............................................................................................................19
4.3 Data Basis and Methodologies ....................................................................................................22
4.3.1 Deforestation.......................................................................................................................22
4.3.2 Forest Degradation..............................................................................................................25
4.3.3 Forest Fragmentation..........................................................................................................29
4.4 Data Modeling .............................................................................................................................31
5. Using ERDAS Imagine..........................................................................................................................31
5.1 Cursor Operations and Image Display.........................................................................................31
5.2 Magnification and Overlay Operations .......................................................................................33
6. Hands-On Exercise................................................................................................................................35
6.1 Opening a Raster File ..................................................................................................................35
6.2 View Menu on View Window.....................................................................................................36
6.3 Working with Spatial Modelers...................................................................................................36
6.4 Image Classification ....................................................................................................................38
6.5 Change Detection of Coastal Vegetation.....................................................................................44
BIBLIOGRAPHY......................................................................................................................................49

4
1. Introduction
Considerable uncertainty remains in our knowledge of deforestation, degradation, and forest
fragmentation. Important research and management questions such as what is the rate of
forest conversion, what is the age structure of forest re-growth, and what is the area impacted
by forest degradation, selective logging, and fragmentation, are unknown in many parts of the
world. Answers to these questions are critical to understand biogeochemical cycles (e.g.
Carbon and Nitrogen Cycles, Nutrients), hydrological cycles (Energy/Water Balance and
Climate, Sediment Transport, Erosion and Runoff), and ecological cycles (Ecosystem Health,
Biodiversity (fragmentation)).
Olympic Forest (Credit: Flickr.com)

5
2. Basic Concepts of Deforestation, Degradation, and Fragmentation
2.1 Deforestation
Deforestation is the conversion of forested areas to non-forest land use such as arable land,
urban use, logged area or wasteland. According to FAO, deforestation is the conversion of
forest to another land use or the long-term reduction of tree canopy cover below the 10%
threshold. Deforestation can result from deliberate removal of forest cover for agriculture or
urban development, or it can be an unintentional consequence of uncontrolled grazing (which
can prevent the natural regeneration of young trees). The combined effect of grazing and fires
can be a major cause of deforestation in dry areas. Deforestation implies the long-term (>10
years) or permanent loss of forest cover.
Deforestation defined broadly can include not only conversion to non-forest, but also
degradation that reduces forest quality - the density and structure of the trees, the ecological
services supplied, the biomass of plants and animals, the species diversity and the genetic
diversity. Narrow definition of deforestation is: the removal of forest cover to an extent that
allows for alternative land use. The United Nations Research Institute for Social Development
(UNRISD) uses a broad definition of deforestation, while the Food and Agriculture
Organization of the UN (FAO) uses a much narrower definition.
Definitions can also be grouped into those which refer to changes in land cover and those
which refer to changes in land use. Land cover measurements often use a percent of cover to
determine deforestation. Land use definitions measure deforestation by a change in land use.
This definition may consider areas to be forest that are not commonly considered as such. An
area can be lacking trees but still considered a forest. It may be a land designated for
afforestation or an area designated administratively as forest. Land cover based definitions can
be measured using remotely sensed data. Detailed ground survey is needed to monitor land
use type deforestation processes.

6
Meghalaya used to be one huge swath of green before deforestation began (Credit: Google Earth)
2.2 Forest Degradation
Forest degradation is a process leading to a ‘temporary or permanent deterioration in the
density or structure of vegetation cover or its species composition’. It is a change in forest
attributes that leads to a lower productive capacity caused by an increase in disturbances.
The time-scale of processes of forest degradation is in the order of a few years to a few
decades. For the purpose of having a harmonized set of forest and forest change
definitions, that also is measurable with conventional techniques, forest degradation is
assumed to be indicated by the reduction of canopy cover and/or stocking of the forest
through logging, fire, windfelling or other events, provided that the canopy cover stays
above 10%. In a more general sense, forest degradation is the long-term reduction of the
overall potential supply of benefits from the forest, which includes wood, biodiversity and
any other product or service.
Any in-depth understanding of the processes of forest degradation has to be based on an
accurate monitoring of the degradation over large areas, for at least a decade. The
detection of inter-annual changes in landscape spatial structure is more likely to reveal long
term and long lasting land cover changes, while spectral indicators are more sensitive to
fluctuations in primary productivity associated with climatic fluctuations. Different monitoring
systems may be optimal for different ecosystems. A long time series of observations is
always required with seasonal, annual and decadal changes. The monitoring of the spatio-
temporal distribution of biomass burning may also give indications of open forest
degradation.

7
2.3 Forest Fragmentation
Simply defined, forest fragmentation refers to any process those results in the conversion of
formerly continuous forest into patches of forest separated by non-forested lands. The
definitions of fragmentation are as diverse as the subject itself. For instance, a definition which
uses habitat as the qualifier is "The splitting or isolating of patches of similar habitat, typically
forest cover, but including other types of habitat...Habitat can be fragmented naturally or from
forest management activities, such as clear-cut or logging". In another example, the definition
is empirically linked to population growth, "Fragmentation is a complex phenomenon resulting
from dynamic interactions between the natural landscape and society's ever-increasing
demands on the land, creating a mosaic of natural and human-modified environments."
However, the single tie that links these two definitions together is the idea of fragmentation
referring to the process of a contiguous land base being divided into smaller pieces. In one
definition the author sums this idea, "Fragmentation has been defined as the conversion of
large areas of contiguous native forest to other types of vegetation and /or land use leaving
remnant patches of forest that vary in size and isolation".
To better understand deforestation, degradation and fragmentation, it is important to revisit
three additional terms that are often used to highlight positive changes in forest areas. These
three terms are afforestation, reforestation and forest improvement
2.4 Afforestation
Afforestation is the conversion from other land uses into forest, or the increase of the canopy
cover to above the 10% threshold. Afforestation is the reverse of deforestation and includes
areas that are actively converted from other land uses into forest through silvicultural
measures. Afforestation also includes natural transitions into forest, for example on abandoned
agricultural land or in burnt-over areas that have not been classified as forest during the barren
period. As for deforestation, the conversion should be long-term, that is areas where the
transition into forest is expected to last less than ten years, for example due to recurring fires,
should not be classified as afforestation areas. The concept “long-term” is central in this
definition and is defined as ten years. Local climatological conditions, land use contexts or the
purpose of the analysis may however justify that a longer time frame is used.
2.5 Reforestation
Reforestation is the re-establishment of forest formations after a temporary condition with less
than 10% canopy cover due to human-induced or natural perturbations. The definition of
forest clearly states that forests under regeneration are considered as forests even if the
canopy cover is temporarily below 10 per cent. Many forest management regimes include
clear-cutting followed by regeneration, and several natural processes, notably forest fires and
windfalls, may lead to a temporary situation with less than 10 percent canopy cover. In these
cases, the area is considered as forest, provided that the re-establishment (i.e. reforestation)
to above 10 percent canopy cover takes place within the relatively near future. As for
deforestation, the time frame is central. The concept "temporary" is central in this definition and
is defined as less than ten years. Local climatological or land use contexts, or the purpose of
the analysis, may however justify that a longer time frame is used.

8
2.6 Forest improvement
Forest improvement is the increase of the canopy cover or stocking within a forest. For the
purpose of having a harmonized set of forest and forest change definitions, that also is
measurable with conventional techniques, forest improvement is assumed to be indicated by
the increase of canopy cover and/or stocking of the forest through growth. In a more general
sense (cf. forest degradation) forest improvement is the long-term increase of the overall
potential supply of benefits from the forest, which includes wood, biodiversity and any other
product or service.
Palm Plantation (Credit: Flickr.com)

9
3. Agents, causes, and consequences of Deforestation, Degradation, and
Fragmentation
3.1 Agents
It is important to distinguish between the agents of deforestation and its causes. The "agents"
are those individuals, corporations, government agencies, or development projects that clear
the forests as opposed to the forces that motivate them. Much of what has been written about
deforestation fails to distinguish between "agents" and "causes". At times, this deficiency
mistakenly assigns blame to groups who are only acting in an economically rational manner
given the socioeconomic and political framework in which they find themselves.
Who is doing the deforestation? There is considerable debate about who is doing the
deforestation and why they are doing it. Small farmers? Commercial farmers? Loggers? Cattle
ranchers? The answer is all of them and more. Table 1 lists the important agents of
deforestation, degradation and fragmentation.
Overhead view of clear-cutting for slash-and-burn agriculture (Credit: Flickr.com)

10
Table 1: Important agents leading to deforestation, degradation and fragmentation
Agents Links to deforestation, degradation and fragmentation
slash-and-burn farmers - clear forest to grow subsistence and cash crops
commercial farmers - clear the forest to plant commercial cash crops, sometimes displace slash-
and-burn farmers who then move to the forest
cattle ranchers - clear the forest to plant pasture, sometimes displace slash-and-burn farmers
who then move to the forest
Livestock herders - intensification of herding activities can lead to deforestation
Loggers - remove commercial timber, logging roads provide access to other land users
commercial tree planters - clear mostly forest fallow or previously logged forests to establish plantations
to supply fibre to the pulp and paper industry
firewood collectors - intensification of firewood collection can lead to deforestation
mining and petroleum
industrialists
- roads and seismic lines provide access to other land users, localized
deforestation related to their operations
land settlement planners - relocation of people into forested areas as well as settlement projects
displacing local people who then move to the forest
infrastructure developers - new access for other land users from road and highway construction through
forested areas, flooding by hydroelectric dams
There is considerable variation region to region and country to country (see Table 2) as to
which groups are the most important agents of deforestation, degradation and fragmentation.
Table 2: Important agent groups and their regions
Region Main agents
Africa 1. slash-and-burn farmers
2. commercial farmers
3. loggers
4. livestock herders
5. refugee and civil disturbances
Asia-Oceania 1. commercial farmers
2. slash-and-burn farmers
3. loggers
4. commercial tree planters
5. infrastructure developers
Latin America and Caribbean 1. slash-and-burn farmers
2. cattle ranchers
3. commercial farmers
4. loggers
5. infrastructure developers
3.2 Causes
Deforestation has been attributed to socio-demographic factors, such as population growth
and the political economy of class structure, and specific exploitation activities like commercial
logging, forest farming, fuel wood gathering, and pasture clearance for cattle production.
Deforestation from logging operations, particularly in stands of tropical moist forest, is often
claimed to occur in a two-step sequence. First, loggers build roads into primary or old-growth
forest and remove selected trees. In many places of the world, it was observed that the logging
operation destroys 45-74% of the residual trees. The logging damage, however, is

11
compounded once the loggers have left. Then, forest farmers are likely to follow the logging
roads in search of new areas for cultivation. Famous fish-bone deforestation in Rondônia is a
classic example (Figure 1). Thus, a substantial portion of worldwide deforestation might be
explained by the logger/farmer interaction.
19 June 1975, Landsat 2 MSS bands 4 2 1 1 August 1986, Landsat 5 MSS bands 4 2 1 22 June 1992, Landsat 4 TM bands 4 3 2
Figure 1: These images show a portion of the state of Rondônia, Brazil, in which tropical deforestation has
occurred. Approximately 30% (3,562,800 sq km) of the world's tropical forests are in Brazil. The estimated
average deforestation rate from 1978 to 1988 was 15,000 sq km per year. Systematic cutting of the forest
vegetation starts along roads and then fans out to create the "feather" or "fishbone" pattern shown in the eastern
half of the 1986 image. The deforested land and urban areas appear in light blue; healthy vegetation appears red.
Source NASA
In developing countries, high population growth coupled with rapidly expanding agriculture,
and over-exploitation of forest resources is believed to be responsible for accelerated rate of
deforestation. Similarly, in many parts of the world, wood production was found to be a
significant contributor of deforestation. Causes of deforestation can be divided into two
categories: direct causes and underlying causes.
3.2.1 Direct causes of deforestation
The most important direct causes of deforestation include logging, the conversion of forested
lands for agriculture and cattle-raising, urbanization, mining and oil exploitation, acid rain and
fire. In other countries, clear-cut logging practices have been the main reason for forest loss. In
the early nineties, Canada and Malaysia were famous examples of countries where logging
companies ruthlessly cleared mile upon mile of precious primary forests. Here too, the
historical perspective should not be overlooked. Countries like Ireland and Scotland used to be
almost entirely forested, but were nearly completely cleared under British rule to provide timber
for English shipbuilders.
3.2.2 The underlying causes of deforestation and forest degradation
During the last few decades, the forest crisis has prompted many international, regional and
national preservation initiatives, yet many have had little success. There is general agreement
that this is due to the fact that these strategies were too focused on the immediate causes of
deforestation, and neglected the underlying causes which are multiple and interrelated. In
some cases they are related to major international economic phenomena, such as macro-
economic strategies which provide a strong incentive for short-term profit-making instead of
long-term sustain ability. Also important are deep-rooted social structures, which result in
inequalities in land tenure, discrimination against indigenous peoples, subsistence farmers and

12
poor people in general. In other cases they include political factors such as the lack of
participatory democracy, the influence of the military and the exploitation of rural areas by
urban elites. Over consumption by consumers in high-income countries constitutes another of
the major underlying causes of deforestation, while in some regions uncontrolled
industrialization is at the heart of forest degradation with widespread pollution resulting in acid
rain. Figure 2 diagrammatically presents the causes of declining forest area.
Conversion to agriculture by burning (Credit: Flickr.com)

13
Figure 2: These diagram shows the direct and underlying causes of forest decline Adapted from: After
Contreras-Hermosilla (2000), Underlying causes, CIFOR, p. 5
The causes of deforestation are many and varied. Here are few interesting findings.
- Deforestation results from complex socio-economic processes and in many situations it is
impossible to isolate a single cause.
- There is no clear definition of „deforestation“, neither are there reliable estimates of its
extent nor its primary causes, and – partly – as ref lection of these – there is no consensus
on the underlying causes.
- Reports of tropical deforestation indicate that it occurs in diverse circumstances which
obscure underlying patterns of deforestation.
- Deforestation has multiple causes with the particular mix of causes varying from place to
place.
- A rich crop of explanations for deforestations have appeared, none of which, however, are
definitive.
- The factors influencing deforestation are different in different continents. It may be difficult
to generalize that one or several factors are the most important.
- Findings generally support the view that processes of deforestation vary by place.
3.3 Consequences
The social consequences of deforestation are many, often with devastating long-term impacts.
For indigenous communities, the arrival of "civilization" usually means the destruction of their
traditional life-style and the breakdown of their social institutions. Individual and collective
rights to the forest resource have been frequently ignored and indigenous peoples and local
communities have typically been excluded from the decisions that directly impact upon their
Forest decline
Direct
Causes
Underlying
Causes
Natural causes
 Hurricanes
 Natural fires
 Pests
 Floods
Natural from
Human activity
 Agricultural expansion
 Cattle ranching
 Logging
 Mining and oil extraction
 Construction of dams
 Roads
Market failures
 Unpriced forest goods and services
 Monopolies and monopolistic forces
Mistaken policy interventions
 Wrong incentives
 Regulatory mechanisms
 Government investment
Governance weakness
 Concentration of land ownership
 Weak or non-existent ownership
and land tenure arrangements
 Illegal activities and corruption….
Agents
 Slash and burn farms
 Agribusiness
 Cattle ranchers
 Miners
 Oil corporations
 Loggers
 Non-timer
commercial corporations
Broader socioeconomic and
political causes
 Population growth and density
 Economic growth
 Distribution of economic
and political power
 “Excessive” consumption
 Toxification
 Global Warming
 War….

14
lives. Many of the indigenous peoples of the Brazilian states of Amazonas and Rondônia have
been encroached upon by slash-and-burn farmers, ranchers, and gold miners, often resulting
in violent confrontations. The intrusion of outsiders destroys traditional life styles, customs, and
religious beliefs.
Watersheds that once supplied communities with their drinking water and farms with irrigation
water have become subject to extreme fluctuations in water flow. The loss of safe, potable
water puts communities' health at risk for a variety of communicable diseases.
In economic terms, the tropical forests destroyed each year represent a loss in forest capital
valued at US$ 45billion. By destroying the forests, all potential future revenues and future
employment that could be derived from their sustainable management for timber and non-
timber products disappear.
Probably the most serious and most short-sighted consequence of deforestation is the loss of
biodiversity. The antiseptic phrase "loss of biodiversity" masks the fact that the annual
destruction of millions of hectares of tropical forests means the extinction of thousands of
species and varieties of plants and animals, many of which have never been catalogued
scientifically. How many species are lost each year? The exact figure is not known, a
consequence of our limited knowledge of tropical forest ecosystems and our inadequate
monitoring systems. Some estimates put the annual loss at 50,000 separate species but this is
an educated guess at best. Fragmented stands of trees left during deforestation are usually
not large enough to be self-perpetuating in terms of maintaining even an altered balance of
biodiversity. Deforestation is eroding this precious resource of biodiversity.
Although there is some debate about the rate at which the atmosphere is warming, there is
general agreement that it is warming. The currently accepted models predict a 0.3 degree
Celsius increase per decade in global temperatures over the next century. This is due to the
increase in the amount of carbon dioxide present in the atmosphere, which has risen by about
log A
log S
if deforestation results in this
much loss of habitat
then we can expect
to lose this many
species

15
25 per cent in the last 150 years. Although it is less than 1/20 of one per cent of the earth's
atmosphere, carbon dioxide has a high capacity to absorb radiant heat.
The negative consequences of global warming are catastrophic -- increasing drought and
desertification, crop failures, melting of the polar ice caps, coastal flooding, and displacement
of major vegetation regimes. The amount of carbon currently in the atmosphere is estimated to
be about 800,000 million tons and is increasing at the rate of about 1 percent annually.
Deforestation is an important contributor to global warming, however, its contribution relative to
the other factors is not precisely known. The principal cause of global warming is the excessive
discharges in industrialized countries of greenhouse gases, mostly from the burning of fossil
fuels. Annual discharges from burning fossil fuels are estimated to be about 6,000 million tons
of carbon, mostly in the form of carbon dioxide. It is thought that an additional 2,000 million
tons or about 25 percent of the total carbon dioxide emissions are a consequence of
deforestation and forest fire. At the regional level, deforestation disrupts normal weather
patterns, creating hotter and drier weather. Unfortunately, efforts to find solutions to the
deforestation crisis have not been as success in capturing investment money as have
improvements to automotive exhaust emissions.
The long term impact of deforestation on the soil resource can be severe. Clearing the
vegetative cover for slash and burn farming exposes the soil to the intensity of the tropical sun
and torrential rains. This can negatively affect the soil by increasing its compaction, reducing
its organic material, leeching out its few nutrients available, increasing its aluminum toxicity of
soils, making it marginal for farming. Subsequent cropping, frequent tillage, and overgrazing by
livestock accelerate the degradation of the soil.
In the dry forest zones, land degradation has become an increasingly serious problem,
resulting in extreme cases in desertification. It affects about 3,000 to 3,500 million hectares,
about one-quarter of the world's land area, and threatens the livelihoods of 900 million people
in 100 countries of the developing world. Desertification is the consequence of extremes in
climatic variation and unsustainable land use practices including over cutting of the forest
cover. Growing populations are making ever-increasing demands on the land to produce more,
leading to an intensification of use beyond the carrying capacity of the land.
By 2050, two billion people, or 20 per cent of the world's population, will suffer from water
shortages. Most of these people will be living in developing countries. Once denuded, the
same watersheds lose their capacity to regulate stream flows and experience rapid fluctuations
in stream and river levels, often resulting in disastrous downstream flooding. Water shortage is
a major health risk in terms of inadequate sewage disposal, poor personal hygiene, and
insufficient potable water. Food security is threatened as irrigation water becomes scarcer.
Without the protection of the tree cover, soils are exposed to the rigors of severe tropical
climates and are rapidly eroded. Freshwater and coastal fisheries are devastated by the high
sedimentation loads carried by the rivers, as are wildlife-rich wetlands. Sedimentation from
degraded watersheds is also one of the principal causes of the decline of coastal coral reefs.
The economic and environmental costs are staggering.

16
4. Remote Sensing of Deforestation, Degradation, and Fragmentation
Remote Sensing provides a unique opportunity to assess and monitor deforestation,
degradation, and fragmentation for a number of reasons. First, with remote sensing we can
work at multiple scales ranging from few meters to several kilometers. This included detailed
study at local level to global forest resources assessment. Second, remotely sensed data can
be acquired repeatedly (e.g. daily, monthly) that helps us monitor forest resources in a regular
basis. Third, these measurements can be made in near real time basis which is quite useful for
monitoring events such as forest fire. Fourth, remote sensing data has synoptic coverage and
information can be acquired in places where accessibility is an issue. Fifth, we could use
wavelengths that are not visible to human eye. Thus, remote sensing is the most effective
means of assessing and monitoring forest resources. It is important to understand that remote
sensing does not replace field survey but provides complimentary information.
4.1 Availability, Accessibility, and Affordability of Remote Sensing Data
Over the past half century, a range of airborne and space-borne sensors has acquired remote
sensing data, with the number of sensors and their diversity of capability increasing over time.
Today a large number of satellite sensors observe the Earth at wavelengths ranging from
visible to microwave, at spatial resolutions ranging from sub-meter to kilometers and temporal
frequencies ranging from 30 minutes to weeks or months. In addition, archives of remotely
sensed data are increasing and these data provide a unique, but not complete, chronological
data of the Earth during the specified time period. New sensors are continually being launched
and existing sensors are often replaced to ensure continuity in the data record. Given this
enormous resource, remote sensing has proven to be an effective tool to monitor
deforestation, degradation and forest fragmentation.
Ideally, the following image characteristics are required for studying deforestation, degradation,
and fragmentation.
 Cloud free and clear atmosphere during the time of data acquisition;
 Availability of imagery for the optimum date or dates;
 Spatial resolution fine enough for accurate mapping and course enough so image size
is manageable;
 Band selection (band width, placement, and number of bands) optimized to identify
features of interest;
 Affordable;
 Study area covered on a single image
 Multi-date imagery acquired under identical conditions
o Water level (tide, river and lake level)
o Consistent and “stable” phenologic state
o Soil moisture
 Repeat interval consistent with project goals
 Same sensor and sun position when images were acquired
 Similar atmospheric conditions

17
Pragmatically, it is rather difficult to acquire the data with the above characteristics. Instead,
the following problems are common in data acquisition process:
 unavailability of data for specific time period;
 Persistent cloud coverage throughout the year and for many years;
 Cost is too high specially for commercial satellite data;
 Availability of data in usable format (digital or hard copy);
 Licensing concerns in sharing the data;
 Classified, secure or sensitive data;
 Cost of processing, in producing value added product,
 Lack of expertise,
 Calibration/ground truth problems,
 Lack of knowledge within senior management as to how data can be used,
 Lack of equipment/software for analysis,
 Volume of data (storage/archiving),
 Resistance within the agency to change.
There are a number sources where you can get variety of data, often times, free of cost. See
Annex II for selected and popular data sources. However, the data available may not
necessarily be accessible to many developing countries because of slow and often times
non-connectivity internet problems.
Optical sensors providing systematic observations at the regional/global level and at coarse
(≥1 km) spatial resolution include the NOAA advanced very high resolution radiometer
(AVHRR) and SPOT VEGETATION. At finer (10–30 m) spatial resolution, Landsat sensor
(currently the Enhanced Thematic Mapper Plus or ETM+) and SPOT sensor (currently the high
resolution visible infrared or HRVIR) data can be combined to provide regional and even
continental level observations. The Terra-1 Moderate Resolution Imaging Spectroradiometer
(MODIS) and ADEOS II Global Land Imager (GLI) represent a new generation of medium
(250–500 m) spatial resolution sensors and an important bridge between those observing at
fine and coarse spatial resolutions.
The acquisition of optical data is hampered by problems associated with cloud cover, haze and
smoke that often limit the number of usable scenes. These limitations can be overcome using
fine (10–30 m) space borne synthetic aperture radar (SAR) sensors that observe under all
weather conditions and also during day and night. Key sensors include the European
Resources Satellite (ERS-1/2), RADARSAT-1/2, the Japanese Earth Resources Satellite
(JERS-1) SAR and the ENVISAT Advanced SAR (ASAR). These sensors have the capacity to
provide more consistent coverage at continental scales and over short timeframes—if
systematic acquisition strategies are implemented–and their use for regional-scale vegetation
mapping has been well-proven both for the tropics and the boreal zone.
LIDAR technology has only recently been explored for vegetation characterization and
mapping. The LIDAR transmits a vertical near infrared (NIR) laser pulse towards the land
surface which then reflects from objects and returns to the sensor. The delay in time between
the transmission and receipt of the pulse is related directly to distance and hence the height of
objects (e.g. tree crowns) and the ground surface. LIDAR footprint sizes may vary from 0.25 to

18
5m with achievable accuracies of <25 cm in elevation. Using small footprint laser, strong
relationships with variables such as tree height, stem volume, biomass and canopy closure
have been obtained. Large footprint LIDAR are able to provide a complete waveform of
reflected light allowing attributes such as stem diameter, basal area, stem volume and above
ground biomass to be retrieved. Such sensors are currently available on airborne platforms.
VCL is the first multi-beam, waveform-recording LIDAR to fly in space. The VCL is not an
imaging instrument but has been designed to collect a series of point samples along the flight
path. Thematic products can only be foreseen if these data are used in combination with other
spatially extensive data (e.g. optical or SAR).
Table 3: Major satellite remote sensing data available
Name Spatioal
Resolution
(m)
Spectral
Resolution
Temporal
resolution
(days)
launched
(year)
Web Site
Landsat 15-80 V/NIR,
SWIR, TIR
16 1972 www.landsat.org
SPOT 2.5-20 V/NIR,
SWIR,
26 1986 www.spotimage.fr
IRS 6-188 V/NIR,
SWIR,
24 1995 www.isro.org
IKONOS 1-4 V/NIR,
SWIR
3 1999 www.spaceimaging.com
QuickBird 0.61-2.44 V/NIR,
SWIR,
2002 www.digitalglobe.com/
MODIS 250-1000 V/NIR,
SWIR, TIR
1 1999 modis.gsfc.nasa.gov/
VEGETATION 1000 V/NIR,
SWIR, TIR
1 1998 www.spot-
vegetation.com/
AVHRR 1000 V/NIR,
SWIR, TIR
1 1978 www.noaa.gov
MERIS 300 V/NIR,
SWIR
3 2002 www.envisat.esa.int
ASTER 15-90 V/NIR,
SWIR, TIR
4-16 1999 asterweb.jpl.nasa.gov/
Hyperion 30 V/NIR,
SWIR,
16 2000 eo1.usgs.gov
ALI 10-30 V/NIR,
SWIR,
16 2000 eo1.usgs.gov
CBRES 20-260 V/NIR,
SWIR,
3-26 2003 www.dgi.inpe.br/
JERS/SAR 18 L-band 44 1992 www.eorc.jaxa.jp
Radarsat 8-100 C-band 24 1995 www.rsi.ca

19
The cost of Landsat data has decreased significantly in the last 5-10 years. Similar resolution
data e.g. SPOT is still expensive. The next generation of the Landsat data may be available
free of charge. At the same, file sharing is likely to be one of the popular methods for data
sharing.
4.2 Recent Advances in Remote Sensing Technology
Remote Sensing technology is evolving through time. Significant improvements have been
made in terms of spectral, spatial, temporal and radiometric resolutions. For example, with the
improvement in spectral and radiometric characteristics quality and quantity of data acquired
and archived has improved. More specifically, improvements have been observed in
1) visibility and clarity that includes more detailed image of a smaller piece of land;
2) clear definition involving more precisely the specific colors or light responses
reflecting off of the field; and
3) frequent data acquisition on a regular interval of every other day or every 5-7 days.
The images below visually demonstrate the difference in the quality of data.
14 March 1986, Landsat 5 TM bands 4 3 2 13 April 2000, Landsat 7 ETM+ bands 4 3 2
Source: http://earthshots.usgs.gov/Angangueo/Angangueo
4.2.1 Methodology to Assess and Monitor Deforestation, Degradation and
Fragmentation using Remote Sensing and GIS
Remote Sensing can assess and monitor two broad categories of forest cover change:
categorical vs. continuous. Both types of changes can be used to assess and monitor
deforestation, degradation and fragmentation. Some changes such as deforestation are
inherently categorical in which areas previously classified as forests are no longer. Examples
of continuous change include the change in LAI or cover of vegetation canopy over time. While
the distinction between categorical and continuously measured change is simple in the
abstract, it worth noting that in practice the differences are often blurred. For example, the
magnitude of change in a surface property (such as burn severity) is often reduced to ordinal
categories (high, medium, and low).

20
A. Deforestation
There are two main approaches for assessing and monitoring deforestation: wall-to-wall
mapping and sampling strategy. Wall-to-wall mapping could be a very expensive proposition if
the study is conducted in a very large area spanning different time periods. For example, wall-
to-wall coverage of the entire tropical region would require thousands of Landsat scenes.
Depending on the time period of interest, these scenes can cost anywhere from $400 to $600
each, adding up to a considerable investment, especially if one wishes to examine change by
conducting analyses at two different dates.
The alternative to wall-to-wall coverage is the use of a sampling strategy. The sampling
strategy specifies the portion of the entire area of interest that must be analyzed, and results
are then extrapolated from the sample to the entire region. The sampling strategy also
specifies how the location of the samples is determined. For example, in the commonly used
stratified random sampling schemes, the sampling is equally partitioned out by some
independent stratification of the region, such as ecological biome, rainfall, elevation, or some
landcover classification. Thus, if the 'lowland evergreen forest' strata of a given classification
covers 35% of the total area, then 35% of the samples will be constrained to this strata, and
changes for that strata can be independently computed.
Unfortunately, for the issue of deforestation, these types of random sampling schemes often
do not work well. Deforestation is rarely a randomly located occurrence in a region. Rather,
most deforestation occurs next to or near previous deforestation, or near the roads which
provide access. Industrial logging, on the one hand, may be centered on company's
concessions. Deforestation due to clearance for pasture or farming will likely be centered on
previous farms or pasturelands. In either case, it is not randomly distributed around the
landscape. Thus, schemes using random sampling of a non-random process will not provide
very accurate results. Some authors describe stratification and sampling schemes might be a
better option to evaluate tropical deforestation than simple random sampling. Their
methodology relies upon expert knowledge of deforestation hot spots. Tucker and Townshend
(2000) used Monte Carlo simulations to investigate this issue with Landsat data for Bolivia,
Colombia, and Peru, and found that, on average, approximately 88% of the area needed to be
sampled in order to estimate deforestation to within 20% of the actual amount 90% of the time.
Thus, results from anything much less than wall-to-wall coverage must be closely scrutinized
to ensure the adequacy of the sampling strategy.
A variety of methodologies exist for deriving deforestation estimates from satellite data. The
simplest methods use manual interpretation of printed images, similar to methods long used
for interpretation of aerial photographs. The interpreter simply draws lines around areas of
different forest types, which are then manually summed. This requires that the interpreter know
the region, or at least how to differentiate its forest types from one another and from various
forms of agriculture and plantations. Change can be assessed by either comparing the results
from two separately interpreted images, or by performing a multi-date analysis; in this case,
image transparencies are overlaid and areas of change between the images are delimited.
However, several problems arise from this method. First, there is the uncertainty of the human
element; although human knowledge of the region may make differentiating forest areas
straightforward, it will also make the results dependent on the specific interpreter involved and
their experience in this type of analysis. Second, the quality and scale of the printed images
will also affect the results. The scale will determine the minimum mappable unit, which may

21
affect results in areas with highly fragmented landcover or deforestation. Despite these
uncertainties, on-screen digitizing using remote sensing and GIS software is quite popular in
many countries.
More sophisticated methodologies use computers to distinguish the various forest types,
based on all the spectral information available in the images. Conceptually, this is a much
more appealing process, since the computer can analyze changes in all of the bands of the
image, while humans are generally limited to viewing a falsecolor combination of three bands.
Nevertheless, computer classification is also dependent on human interaction; in one form,
supervised classification, the human analyst delineates sample regions (training classes) of
the different landcover types. The computer then compares the spectral information from the
remainder of the image to that from the sample regions, and thereby indicates which pixels are
most similar to which of the sample regions. Thus, this method depends on the specific sample
regions chosen by the analyst, which, again, is dependent on the professional knowledge and
experience of the analyst. This method is particularly attractive if sufficient field data is
available. The alternate method, unsupervised classification, lets the computer determine the
best classes based on the entire image. However, this method then requires the human
analyst to aggregate the resulting computer defined classes into standard landcover and
forestry classes. Again, this requires knowledge of the area or at least of the types of land
cover in the area, and experience with the methods. As with the manual methods, the
computer can also perform multi-image analyses, in which areas of change between two
images are classified. Several other change detection methods are described in section 4.3.1.
These computer methods also greatly facilitate the registration of one image to another; due to
slight changes in orbit or sensor pointing, images of the same region, by the same satellite,
generally will not perfectly align with one another. For coarser resolution imagery, this is
generally less of a problem, but for the high resolution images needed for forest cover
monitoring, such misregistration can cause major problems for multi-image change detection
procedures.

22
B. Forest Degradation
Forest degradation is a process leading to a ‘temporary or permanent deterioration in the
density or structure of vegetation cover or its species composition’. It is a change in forest
attributes that leads to a lower productive capacity. It is caused by an increase in disturbances.
The time-scale of processes of forest degradation is in the order of a few years to a new few
decades. Thus, it is more difficult to identify and monitor using remotely sensed data. For
example, rapid regeneration tends to obscure the spectral/spatial signature of the recently
clear-felled or degraded forest.
Any in-depth understanding of the processes of forest degradation has to be based on an
accurate monitoring of the degradation over large areas, for at least a decade. The detection of
inter-annual changes in landscape spatial structure is more likely to reveal long term and long
lasting land cover changes, while spectral indicators are more sensitive to fluctuations in
primary productivity associated with climatic fluctuations. Different monitoring systems may be
optimal for different ecosystems. A long time series of observations is always required with
seasonal, annual and decadal changes. The monitoring of the spatio-temporal distribution of
biomass burning may also give indications of open forest degradation.
Degradation is generally grouped into classes based on the degrading activity (selective
logging, burning) and intensity. Quantification of the impacts is measured in terms of forest
structure and composition. For example, forest degradation classes can be grouped into intact
forest, moderately logged, heavily logged, lightly burned, and heavily burned. They can be
further described in terms of physical characteristics, biomass and canopy cover.
C. Forest Fragmentation
Forest fragmentation is generally defined as the process of subdividing continuous forests into
smaller patches, which results in the loss of original habitat, reduction in patch size, and
increasing isolation of patches. Forest fragmentation can be measured and monitored in a
powerful new way by combining Remote Sensing, GIS, and analytical software (e.g.
FRAGSTAT). For example, the length of forest clearance edges and patches of fragmented
forest can be directly measured in satellite images.
Forest fragmentation can be enumerated using high-altitude aerial photography and satellite
data. In the past both high resolution and coarse resolution data have been used for this
purpose depending on the objective of the study. AVHRR and MODIS data remain useful at
continental and global scales, however, finer resolution satellite data (e.g. Landsat, SPOT,
IRS) is recommended for studying smaller geographic areas. Existing land cover data can also
be used to build forest fragmentation database. Other GIS data layers particularly roads data
play a prominent role in the fragmentation assessment.
4.3 Data Basis and Methodologies
4.3.1 Deforestation
Digital image processing in conjunction with spatial analysis in a Geographic Information
System are effective means for quantifying deforestation. When implementing a deforestation
monitoring project, three major steps are involved:

23
(1) selection of appropriate satellite data
(2) image preprocessing including geometrical rectification and image registration,
radiometric and atmospheric correction, and topographic correction if the study
area is in mountainous regions;
(3) selection of suitable techniques to implement change detection analyses; and
(4) accuracy assessment.
Results may depend on several factors including the following.
(1) precise geometric registration between multi-temporal images,
(2) calibration or normalization between multi-temporal images,
(3) availability of quality ground truth data,
(4) the complexity of landscape and environments of the study area,
(5) change detection methods or algorithms used,
(6) classification and change detection schemes,
(7) analyst’s skills and experience,
(8) knowledge and familiarity of the study area, and
(9) time and cost restrictions.
Several change detection methods and algorithms are available. Because of the impacts of
complex factors, different authors often arrived at different and sometimes controversial
conclusions about which change detection techniques are most effective. In practice, it is not
easy to recommend a suitable algorithm for a specific change detection project. Hence, a
review of change detection techniques used in previous research and applications is useful to
understand how these techniques can be best used to help address specific problems.
Identification of a suitable change detection technique play an important role in producing good
quality change detection results.
Change detection methods are grouped into seven categories: (1) algebra, (2) transformation,
(3) classification, (4) advanced models, (5) Geographical Information System (GIS)
approaches, (6) visual analysis, and (7) other approaches.
The algebra category includes image differencing, image regression, image ratioing,
vegetation index differencing, change vector analysis (CVA) and background subtraction.
These algorithms have a common characteristic, i.e. selecting thresholds to determine the
changed areas. These methods (excluding CVA) are relatively simple, straightforward, easy to
implement and interpret, but these cannot provide complete matrices of change information.
CVA is a conceptual extension of image differencing. This approach can detect all changes
greater than the identified thresholds and can provide detailed change information. One
disadvantage of the algebra category is the difficulty in selecting suitable thresholds to identify
the changed areas. In this category, two aspects are critical for the change detection results:
selecting suitable image bands or vegetation indices and selecting suitable thresholds to
identify the changed areas.
The transformation category includes Principal Component Analysis (PCA), Gramm–Schmidt
(GS), and Chi-square transformations. One advantage of these methods is in reducing data
redundancy between bands and emphasizing different information in derived components.
However, they cannot provide detailed change matrices and require selection of thresholds to

24
identify changed areas. Another disadvantage is the difficulty in interpreting and labeling the
change information on the transformed images.
The classification category includes post-classification comparison, spectral–temporal
combined analysis, expectation–maximization algorithm (EM), unsupervised change detection,
hybrid change detection, and Artificial Neural Network (ANN). These methods are based on
the classified images, in which the quality and quantity of training sample data are crucial to
produce good quality classification results. The major advantage of these methods is the
capability of providing a matrix of change information and reducing external impact from
atmospheric and environmental differences between the multi-temporal images. However,
selecting high-quality and sufficiently numerous training sample sets for image classification is
often difficult, in particular for historical image data classification. The time consuming and
difficult task of producing highly accurate classifications often leads to unsatisfactory change
detection results, especially when high-quality training sample data are not available.
The advanced model-based change detection category includes the Li–Strahler reflectance
model, spectral mixture models, and biophysical parameter estimation models. In these
methods, the image reflectance values are often converted to physically based parameters or
fractions through linear or nonlinear models. The transformed parameters are more intuitive to
interpret and better to extract vegetation information than are spectral signatures. The
disadvantage of these methods is the time-consuming and difficult process of developing
suitable models for conversion of image reflectance values to biophysical parameters. The
GIS-based change detection category includes the integrated GIS and remote sensing method
and the pure GIS method. The advantage of using GIS is the ability to incorporate different
source data into change detection applications. However, different source data associated with
different data accuracies and formats often affect the change detection results.
The visual analysis category includes visual interpretation of multi-temporal image composite
and on-screen digitizing of changed areas. This method can make full use of an analyst’s
experience and knowledge. Texture, shape, size and patterns of the images are key elements
useful for identification of LULC change through visual interpretation. These elements are not
often used in the digital change detection analysis because of the difficulty in extraction of
these elements. However, in visual interpretation, a skilled analyst can incorporate all these
elements in helping make decisions about LULC change. The disadvantage of this method is
the time consumed for a large-area change detection application and it is difficult to timely
update the change detection results. It is also difficult to provide detailed change trajectories.
In addition to the six categories of change detection techniques discussed above, there are
also some methods that cannot be attributed to one of the categories indicated above and that
have not yet frequently been used in practice. They include a knowledge-based vision system,
area production model, and use of indices etc.
In general, change detection techniques can be grouped into two types: (1) those detecting
binary change/non-change information, for example, using image differencing, image ratioing,
vegetation index differencing and PCA; and (2) those detecting detailed ‘from–to’ change, for
example, using post-classification comparison, CVA and hybrid change detection methods.
One critical step in using the methods for change/non-change detection is to select appropriate
threshold values in the tails of the histogram representing change information. In conclusion,
some generalizations are possible:

25
To summarize,
 There is not a single method of change detection in remote sensing that is best for all
kinds of imagery, deforestation, and locations; and
 There are clearly sets of methods more appropriate for some applications than others.
4.3.2 Forest Degradation
Forest degradation is more difficult to measure using remotely sensed data compared to
deforestation. Moderate spatial resolution images often do not provide detailed information
about degraded forests. For example, forest structure and composition (e.g. crown size, liana
density, tree species, etc) cannot be retrieved at these moderate spatial resolutions (i.e. 20–
30mpixel size). In addition, old degraded forests (i.e. .2 years) are difficult to distinguish from
intact forests with both Landsat and SPOT images. However, there are a number of
techniques available and used to assess forest degradation. In addition, high resolution
satellite data such as IKONOS and QuickBird seems promising (Figure 3).
Figure 3: Ikonos image of a degraded forest showing log landings, logging roads, canopy damaged areas, burned
forest, regeneration and forest (cover).

26
The estimation by remote sensing of biophysical attributes that are indicative of forest
degradation has been conducted using either an empirical approach or a physically-based
approach. In the empirical approach, many authors have characterized the surface using
vegetation indices that are arithmetic combinations of spectral bands. Most of these indices
are variants of the normalized difference vegetation index (NDVI), computed as the ratio
between the difference and the sum of the radiances measured in the near infra-red and red
parts of the electromagnetic spectrum. These variants display different levels of sensitivity to
perturbing factors such as soil colour changes or atmospheric effects.
Empirical studies and simulations with radiative transfer models support the
interpretation of vegetation indices in terms of the fraction of photosynthetically active
radiation absorbed by the vegetation canopy, canopy attributes (e.g. green biomass or
green leaf area index), state of the vegetation (i.e. vegetation vigour or stress) and
instantaneous rates associated with the activity of the vegetation (Myneni et at., 1995).
Inter-annual variations in primary production in the Sahel have been monitored using
annual integration of vegetation index data at a coarse spatial resolution. Similar data
have also been used to detect large-scale tropical deforestation. The methods used in
these studies are not necessarily suitable for the detection of changes associated with
slow rates of forest degradation due to spatial and temporal sampling issues. First, the
level of spatial aggregation of the data used in these studies (i.e. 8 by 8 km in the case
of pre-processed time series of Global Area Coverage data from the Advanced Very-
High Resolution Radiometer on the NOAA series of orbiting platforms) is too coarse to
allow for the detection of subtle changes in the density or structure of vegetation cover.
Second, integrating vegetation index data on an annual basis hides variations in the
phenological cycle of the vegetation cover. Third, the time consistency of the data over
several years is not always ensured due to, in the case of AVHRR, a lack of on-board
calibration of radiances, instrument change-over, and a drift in equator crossing-time of
the satellite's orbit over the years. Time series of high spatial resolution data can
overcome several of these limitations.
Land surface temperature (Ts), derived from the thermal channels of satellite sensors
has also been used as a biophysical indicator of the surface in empirical studies. Ts is
related, through the surface energy balance equation, to surface moisture availability
and evapotranspiration, as a function of latent heat flux. A few empirical studies have
demonstrated the potential of these data for forest-cover change analysis, e.g. to detect
forest disturbances, to locate the boundary between shrub and woodland savannas in
West Africa, or to map the boundary between tropical rain forest and savanna
vegetation. A few authors have combined a vegetation index such as NDVI with Ts data
to describe change studies. The slope of the Ts-NDVI relationship is interpreted in terms
of surface resistance to evapotranspiration.
Physically-based approaches, relying on model inversion, have also been developed. In
these methods, a scene model is constructed to describe the form and nature of the
energy and matter within the scene and their spatial and temporal order. This scene
model is coupled with anatmospheric and a sensor model. This coupled remote sensing
model is then inverted to infer, from the remotely sensed measurements, some of the
properties or parameters of the scene that were unknown. Concerning tropical forest

27
ecosystems, these methods are not yet sufficiently robust to be applied routinely for a
large-scale monitoring of forest-cover changes, due to the difficulty for representing
adequately in a model the natural variability in landscape structure and composition.
For example, Franklin & Strahler (1988) applied a model inversion approach to generate
regional estimates of tree size and density in West Africa, using Landsat TM data. They
used the Li-Strahler canopy reflectance model. A woodland stand is modeled
geometrically as a group of objects casting shadow on the background. The reflectance
of a pixel is modeled as a linear combination of the signatures of the major scene
components (i.e. illuminated tree crown, illuminated background, shadowed tree and
shadowed background), weighted by their relative areas. Inversion results over a sample
of observations were satisfactory for the estimation of tree density and cover but poor for
tree size. The application of this type of approach for monitoring woodland degradation
would probably require further developments for a more robust approach.
Remote sensing offers the possibility to analyze changes in spatial structure at the scale
of landscapes. Indicators of the degradation of the vegetation cover can be derived from
such measures. For example, Pickup & Foran (1987) developed a method to monitor
and landscapes used for pastoralism based on the spatial variability of the vegetation.
The spatial autocorrelation function and mean-variance plots of a spectral indicator were
found to be successful in discriminating between the cover responses typical of good
and poor rainfall years. For drought conditions, the decrease in spatial autocorrelation
with increasing spatial lag was rapid since the ground surface is bare and most of the
vegetation signal comes from scattered areas of trees and shrubs. A low decay rate of
the autocorrelation function indicated a greater spatial uniformity of the landscape, e.g.
during wet periods, when more ground cover is present so reducing the contrast
between the bare soil signal and that produced by trees and shrubs. Similar
observations were made by Lambin (1996) over the seasonal and inter-annual cycle of
three West African landscapes. Vogt (1992) also analyzed the seasonal changes in
spatial structure of a West African landscape, showing that there is a marked seasonal
cycle in the spatial structure of a vegetation index (NDVI), surface temperature and
albedo, and that zones of ecological transition have an identifiable seasonal dynamic in
spatial structure. However, the monitoring of these spatial variability measures only
provides for a qualitative description of the cover state. Lambin & Strahler (1994)
showed that the detection by remote sensing of land-cover change processes is Index
and suface temp) and spatial indicators of surface condition. Their study suggested that
the detection of inter-annual changes in landscape spatial structure is more likely to
reveal long term and long lasting land-cover changes while spectral indicators are more
sensitive to fluctuations in primary productivity associated with the inter-annual variability
in climatic conditions. The long term monitoring of landscape spatial pattern in addition
to other biophysical variables can thus lead to the detection of a greater range of
processes of landscape modification.
The spatial pattern of a changing landscape also contains some information on the
processes of land-cover change. Certain categories of changes in human land use tend
to fragment the landscape (e.g. expansion of extensive agricultural exploitation, forest
degradation driven by small-scale logging, overgrazing or desertification around deep
wells). Other land use changes increase landscape homogeneity (e.g. large-scale

28
intensive cultivation or ranching). Spatial processes of gap expansion in a forest cover
have been modeled to predict the total gap area and gap size distribution. One can
hypothesize that landscapes with a very low or very high level of disturbance are
characterized by a low spatial heterogeneity, while landscapes with a medium level of
disturbance are very heterogeneous.
Land-cover changes take place at a variety of temporal scales, i.e. short events with
detectable effects only for a few months, modifications in seasonal trajectories of
ecosystem attributes, processes that affect the land cover through several seasonal
cycles, and long term, permanent changes. Open forest degradation may affect, and
therefore be indicated by, the phenology of the vegetation cover. The analysis of the
temporal trajectories of vegetation indices based on high temporal frequency remote
sensing data allows us to monitor vegetation phenology and biome seasonality (Justice
et al., 1985).Processes such as a shortening of the growing season, a dephasing of the
phenology of different vegetation layers, or modifications of the cover due to
disturbances such as fires, can only be detected if inter-annual changes in the
seasonal trajectories of vegetation covers are analyzed. For any landscape with a
strong seasonal signal, the detection of inter-annual changes needs to take into account
explicitly the fine-scale temporal variations. If data from only one or a few dates a year
are used to measure inter-annual changes, the undersampling of the temporal series
hinders the change detection accuracy and might lead to the detection of spurious
changes.
A quantitative evaluation of differences in seasonal development curves of remotely
sensed data was applied for land-cover change analysis by Lambin & Ehrlich (1997). In
this study, a measure of the deviation in seasonal trajectories was computed for 10
years of continental-scale remote sensing data. Subtle processes of land-cover change
could be detected this way. Results suggested that land-cover changes in Africa mostly
involve erratic variations in land-cover conditions due to droughts, temporary
modifications in seasonality, Ikonos images, which can be acquired with resolution as fine as
one meter, offer a new capability to map forest degradation in the Amazon region.
In summary, some important lessons may be learned from this review.
- Degradation of forest-cover is often a complex process, with some degree of
reversibility as the productivity of forests is partially controlled y climatic
fluctuations;
- Regular monitoring of several biophysical variables can lead to an accurate
detection of forest degradation;
- Repetitive measurements of spectral, spatial and temporal indicators of the land
surface have to be performed;
- long time series of observations is required to be able to detect trends in forest
degradation that depart is a significant way from short-term, climate-driven
fluctuations in forest conditions; and
- monitoring system can combine indicators of the degradation itself, its proximate
causes, and other surface processes linked with the vegetation cover.

29
4.3.3 Forest Fragmentation
Fragmentation is generally defined as the process of subdividing a continuous habitat type into
smaller patches, which results in the loss of original habitat, reduction in patch size, and
increasing isolation of patches. Forest fragmentation can be measured and monitored in a
powerful new way.
Land cover or Forest/Non-forest classification can be used together with other GIS data layers
to compute fragmentation indices. GIS layers such as road maps, urban/peri-urban maps,
ecoregions etc. can be used together with fragmentation model to infer important information
on fragmentation. One of the commonly used software is Fragstats, which provides a powerful
and comprehensive set of descriptors of spatial pattern at the landscape, class and patch
levels. Several of these metrics are partially or wholly redundant, so, one can select a set of -
indicators at the patch and class level as being of the most interest to them in assessing
fragmentation.
At the patch level, the three indicators namely patch size (area in hectares), patch shape
(using the shape index, which computes the complexity of patch shape by comparing it with
the equivalent most compact patch of identical area, starting from 1 when most compact and
increasing without limit as the patch becomes more irregular) and nearest neighbor index (the
nearest edge–edge distance in meters between a patch and its nearest neighbor of the same
category) is considered to quantify distinct aspects of patch structure. The one-tailed Mann–
Whitney U-tests can be used assess whether patch size, shape and nearest neighbor distance
differed significantly (p=0.05) from year to year and across different landscape zones.
At the class level, the interest may be to compare descriptive metrics of land-cover pattern
between forest and non-forest classes, and across categories of land-cover change. These
indices can be grouped into categories of area, shape, core, diversity, and
contagion/interspersion. To simplify interpretation, the following indices can be considered,
assuming that they quantify different aspects of structure
(a) Number of patches: the total number of patches in this category.
(b) Edge density: total edge per unit area for the class, in meters per hectare.
(c) Mean area: average patch area for the class, in hectares.
(d) Mean shape index: measures the average complexity for a category, of patch shape,
compared to a square patch of identical area. For a single patch, the shape index is 1 when
square, and increases without limit as the patch becomes more irregular.
(e) Mean nearest neighbor distance: average nearest neighbor distance in meters for all
patches belonging to the class.
(f) Clumpy: Measures the extent to which the habitat is aggregated, by comparing the values
observed with those expected under a spatially random distribution, and takes values from -1
to 1.

30
(g) Interspersion-juxtaposition index: measures the degree of interspersion of patches of this
class, with all other categories. This index takes values from 0 to 100, decreasing as the
distribution of patch adjacencies among types becomes increasingly uneven.
The indices of number of patches, mean area and edge density correspond to
area/density/edge metrics, and provide indications of the degree of fragmentation for different
forest/land-cover types and forest/land-cover change trajectories. The mean shape index is a
metric of shape, the mean nearest neighbor distance an indicator of proximity or isolation, and
the clumpiness and interspersion-juxtaposition index measure the contagion and interspersion
of pixels in the landscape.
Techniques for Forest Fragmentation Analysis
There are a number of commercial as well as well free software and programs available to
perform fragmentation analysis. Fragstats, ArcView/ArcInfo, and APACK are few examples.
Fragstats is a shareware, new version of which is based available in Graphical User Interface
(GUI) interface. The software can be downloaded from this website:
http://www.umass.edu/landeco/research/fragstats/fragstats.html. Path Analyst
(http://flash.lakeheadu.ca/~rrempel/patch/) is also a shareware graphical interface for Fragstats
that runs as an extension to ArcView GIS. One of the benefits of Patch Analyst is its ability to
use ARC grid format raster files directly rather than the images that are required by the native
Fragstats implementation, Processing time may be longer compared to Fragstats.
The ARC software packages do not provide for analysis of landscape fragmentation analysis
directly, but with a limited amount of knowledge of the command syntax, users can create
many custom analyses that perform some of the same analyses as other bundled
fragmentation analysis software. For this purpose GRID program within ArcInfo is necessary.
Some of the functions that can be used for the analysis include: gridclip, imagegrid, region
group, statistics, frequency, reclass, zonalgeometry, and core areas.
APACK (landscape.forest.wisc.edu/Projects/APACK/apack.html) is a shareware tool for
processing large raster data sets consisting of landscape information (e.g. classified satellite
imagery). It calculates most of the statistics available from Fragstats. APACK has a command
line parameter entry.

31
4.4 Data Modeling
Generally speaking, deforestation models may be grouped into (1) Malthusian, (2)
economic/technologic, (3) dependency, and (4) political economic (or ecologic) categories. In
the Malthusian models, population growth is held accountable for deforestation. In the
economic/technologic framework, property relations, externalities, and inappropriate
technology are considered responsible. Political economic explanations point to social
structures such as class relations, and dependency theory suggests that the adoption of crops
ill-suited to their environments (by domestic capitalists or foreign interests) for sale and world
markets is at the basis of ecological degradation. We could also relate this with respect to the
time scale.
1. Short time scales: individual & social responses to new opportunities & constraints
created by markets & policies.
2. Long time scales: demographic factors: population increase & decrease, breakdown of
extended families, migration.
Extreme biophysical events trigger further change.
5. Using ERDAS Imagine
Objective
√ To introduce basic ERDAS IMAGINE display and screen cursor control
procedures.
5.1 Cursor Operations and Image Display
We will be analyzing two images in this exercise. Both files are copied in your computer
under/training directory.
After you have successfully logged onto the system, launch IMAGINE by going to the Start
Menu - Remote Sensing - ERDAS IMAGINE 8.7. (or later version) Wait a few seconds for all
menus to appear. Examine the icon panel along the top of the screen. These icons represent
the various components and add-on modules available with the current license. You have the
option of displaying the icon panel horizontally across the top of the screen or vertically down
the left side of the screen using the [Session | Flip Icons] menu item.

32
Cursor Operations: Familiarize yourself with the five menu items located along the top of the
icon panel in the left corner: [Session], [Main], [Tools], [Utilities], & [Help]. The Session menu
controls many of the session settings such as user preferences and configuration options. The
[Main] menu allows access to all the modules located along the icon panel. The [Tools] menu
allows you to display and edit annotation, image, and vector information, access surface
draping capabilities, manage post script and true type fonts, convert coordinates, and view
Erdas Macro Language (EML) script files. The [Utilities] menu allows access to a variety of
compression and conversion algorithms including JPEG, ASCII, image to annotation, and
annotation to raster. The [Help] menu brings up the IMAGINE On-Line Documentation as well
as icon panel and version information. An index of keywords helps you to quickly locate a help
topic by title. A text search function also helps you find topics in which a word or phrase
appears. The menu you will probably use the most under the [Session] menu is the Preference
Editor. The Preference Editor is accessed under [Session | Preferences]. It allows you to
customize and control many individual or global IMAGINE parameters and default settings.
Use the left mouse button (lmb) on the scroll arrows on the side of this menu to examine the
available categories. With the User Interface & Session category open, change the Default
Data and Output Directories to whatever you like (for example):
c:TempEx02
Save these changes using the [User Save] or [Global Save] buttons and close the editor.
Proper use of these preferences can reduce the time it takes you to perform image processing
tasks.
Image Display: Now, you are ready to display the first image. Move the cursor back to the
IMAGINE Viewer and select the File dropdown menu with the lmb. Select [File | Open | Raster
Layer] to get to the Select Layer to Add: dialog. You can also type [Ctrl R] or click on the
Viewer icon that looks like a manila folder that is half open to accomplish the same task.
Additional Viewers may be opened by clicking the [Viewer] icon on the IMAGINE icon panel.
On the left side of the menu you should see a list of files in the current folder. Navigate to the
desired folder if necessary. Position the cursor over the file you want to display (e.g.
mangrove00.img) and click the lmb once (do not double-click). The file name should appear in
the file name window in the Viewer. If you do not see the correct files then you are either not
looking in the correct folder or you do not have the [Files of type] specified as IMAGINE Image
(*.img).

33
Before clicking [OK], you need to assign the spectral bands of the image to the color planes
red, green, blue (RGB). Click on the [Raster Options] folder tab and assign band 3 (NIR) to
red, band 2 (Red) to green, and band 1 (Green) to blue. Make sure that the [Display] option is
set to [True Color] . You also have the option of making the image fit the Viewer frame by
depressing the small box next to Fit to Frame. Now you are ready to click [OK]. If the SPOT
image is requiring less space in the IMAGINE Viewer (there are large black borders on the
sides) then you can resize the IMAGINE Viewer to use your screen desktop area more
efficiently. This will become important in future exercises when many IMAGINE Viewers will
need to be open at once. To remove an image displayed in the IMAGINE Viewer move to the
[File] dropdown menu in that Viewer and select it with the lmb, then find the [Clear] option and
select it. You can also click on the Close Top Layer or Eraser tool icons in the Viewer.
To find out additional information about this image, go to the [Utility] drop down menu in the
open Viewer. Choose [Layer Info] and wait for the ImageInfo dialog box to appear. You can
also access ImageInfo by clicking on the "info" icon in the Viewer icon menu (third one from
the left). Now answer the following questions:
1a) What is the pixel size in the X and Y direction?
1b) What are the units of measurement?
1c) What map projection is the image georeferenced to?
5.2 Magnification and Overlay Operations
The next image we will browse is a Landsat Thematic Mapper (TM) scene of Burma,
mangrove90.img.
Magnification: To magnify (or reduce) an image, the easiest option is to use the Interactive
Zoom In (or Interactive Zoom Out) tools that are located immediately above the image in the
gray Viewer area. The area over which you place the cursor will be the general center for the
area that is magnified. However, you may at some times wish to magnify the image by a
certain factor, such as 2X or 4X. To do so you can select [Zoom] under the [View] menu or
[Quick View] menu and then select the appropriate choice. When you choose Zoom in by X or
Zoom out by X you can also choose the interpolation method. These methods will be
evaluated in class at a future time, but you might wish to try each method for the sake of
understanding them. When you have completed your selection click [OK] and the magnified
image will appear. Another method of explicitly specifying the zoom factor is under the [Raster
Options] feature when you open a file. When [Fit to Frame] is not highlighted, you can enter in

34
the [Zoom by] factor in the lower left hand corner. Finally, you also have the ability to change
the frame scale of the image. The process can be implemented using the [View | Scale] option.
The icon with the hand also gives you panning capabilities within the Viewer.
You can also create a magnifying window by either choosing [View | Create Magnifier] or
accessing the [Quick View] menu and selecting [Zoom | Create Magnifier]. This brings up an
additional window that corresponds to the AOI Box in your Viewer. The AOI Box can be
resized by dragging on the corners. To close the magnifier, place your cursor inside it and
select the [Close Window] option in the [QuickView] menu.
Sometimes it is necessary to determine the coordinates and brightness values of specific
pixels in the displayed image. The inquire cursor allows you to do this. Go into the [Quick View]
menu of the IMAGINE Viewer and select [Inquire Cursor]. This will open a pixel information
menu that allows you to move a crosshair cursor on the Viewer. You can use the black arrows
to move the crosshair cursor in any pixel increment you set. For now leave the increment at
1.00 and note that the increment is variable between the file and map coordinate system. You
can move the crosshair cursor using the black arrows or by pressing and holding the lmb while
the mouse cursor touches the crosshair cursor. For "fine tuning" use the keyboard arrows to
move the cursor. The black circle will move the crosshair cursor back to the center of the
Viewer.
Reference system values for the image can be obtained in either Map, Paper, File, or Lat/Lon,
or MGRS coordinates. Notice that the coordinate system is defined for you. The image
projection is also shown but if you have not selected the Map option then you may not
necessarily be viewing the x, y coordinates of that projection system. The table shows the
R,G,B pixel brightness values for both the image file (File Pixel) and the color lookup table
(LUT Value). Move the Viewer Cursor and notice how the values change. To move the Inquire
Cursor using the mouse you must initially place the arrow cursor at the center of the crosshairs
and click on the lmb. Keep the lmb depressed to move the Inquire Cursor.
Overlay: Now close the Inquire Cursor dialog and open another image in Viewer #1 without
closing the TM scene. You can use IMAGINE to overlay imagery that is georeferenced to the
same coordinate system. To do this, be sure to uncheck the [Clear Display] option under
Raster in the [Select Layer to Add] dialog box. Now overlay the file mangrove00.img on top of
the TM scene.

35
Ottawa, Canada (Credit: Digital Globe)
6. Hands-On Exercise
6.1 Opening a Raster File
 Choosing a file: Click on icon go to File/Open/Raster Layer in menu
o Type in file name
o Browse
o Click on recent files
 Raster options
o Assigning bands
 True color (bands 3,2,1)
 False color composite (other combinations)
o Color, black & white, relief
Utility Menu (on View Window):
 Cursor Inquiry (via Utility or + icon)
o Geographic coordinate information
o Pixel value information
o Changing color/shape of cursor (use Utility menu)
 Measurement Tool (via Utility menu or Ruler icon)
o Coordinate data
o Polyline data

36
o Polygon
o Changing units (Edit tab on Measurement Tool window)
 Layer Information (via Utility Menu)
 Mage merges
o First open second image
o Swipe
o Blend
o Flicker
6.2 View Menu on View Window
 Arrange Layers
o Make sure second image is open
o Open arrange layers dialog
 Zooming (via right click on window, icons, or view menu)
o In/out icons
o Magnifying glass icons
o Right click in view window
o Magnifying box (Create Magnifier in Viewer menu)
 Display two images that are linked geographically
o Have both images open, one on top of the other
o Go to View/Split/Split Horizontal on the View Menu
o Open the file name in the Select Layer to Add on the new viewer
o Go to View/Link-Unlink Viewers/Geographical
 Rotate/flip etc. –on the View Menu
Raster Menu on View Window:
 Band Combinations on Raster Menu
o Play with variations
o Think on what meaning of image is when all IR bands are shown
6.3 Working with Spatial Modelers
Objectives:
√ Introduce the ERDAS Imagine Spatial Modeler.
√ Compute various image statistics using Spatial Modeler.
Introduction to Spatial Modeler:
Begin by opening the Spatial Modeler menu by selecting the Modeler icon in the Imagine icon
panel. Review the function of each of the Model Maker's tools before going on.

37
Create a NDVI image using Spatial Modeler:
Select the Model Maker button in the Spatial Modeler menu. Now, create your own NDVI
spatial model with your knowledge of NDVI (NDVI = (NIR -Red)/(NIR+Red)). Run your model
and open your output image. If you didn't get the proper image, please make it sure that you
declare the raster as float, not integer, in the input/output raster object windows.
Save and open the model. Explore other menu and tool options.
2a. Check your output image. Does it show the similar pattern with your
knowledge of NDVI?
There are built-in models of NDVI for several sensors (e.g., TM, MSS, and AVHRR) in ERDAS
IMAGINE. Select the Indices button in the Spectral Enhancement in the Image Interpreter
menu. Set Select Function as NDVI in the Indices window and click View button in the
bottom, and you get the spatial model like this.
Description of the Model Maker Tools
Use this tool to select items on the Model Maker page. Once selected, these graphics (or text) can be moved or deleted.
Click and drag a selection box to select multiple elements. Multiple selected elements can be dragged to a new location
as a unit. You can also use the arrow to double click on any of the graphics below to further define their contents.
Creates a raster object, which is a single or layer-set of raster data typically used to contain or manipulate data from
image files.
Places a vector object, which is usually an Arc/Info coverage or an Annotation layer.
Creates a matrix object, which is a set of numbers arranged in a fixed number of rows and columns in a two-
dimensional array. Matrices may be used to store numbers such as convolution kernels or neighborhood definitions.
Creates a table object, which is a series of numeric values or character strings. A table has one column and a fixed
number of rows. Tables are typically used to store columns from an attribute table, or a list of values which pertain to
individual layers of a raster layer-set.
Creates a scalar object, which is simply a single numeric value.
Creates a function definition, which are written and used in the Model Maker to operate on the objects. The function
definition is an expression (like "a + b + c") that defines your input. You can use a variety of mathematical, statistical,
Boolean, neighborhood, and other functions, plus the input objects that you set up, to write function definitions.
Use this tool to connect objects and functions together. Click and drag from one graphic to another to connect them in
the order they are to be processed in the model. To delete a connection, simply click and drag in the opposite direction
(from the output to the input).
Creates descriptive text to make your models readable. The Text String dialog is opened when you click on this tool.

38
6.4 Image Classification
Objectives:
√ Define and evaluate a signature.
√ Perform supervised classification.
√ Generate clusters using an unsupervised classification approach (ISODATA).
√ Evaluate the clusters in image feature space.
Part I Training Site Selection
Signature Extraction:
To begin, open mangrove90.img in a viewer and fit to frame. The ERDAS Imagine Signature
Editor allows you to create, manage, evaluate, edit, and classify signatures (.sig extension).
Both parametric (statistical) and non-parametric (feature space) signatures can be defined. In
this exercise, we will be defining signatures by collecting them from the image to be classified
using the Signature Editor and Area of Interest (AOI) tools. The Signature Editor can be
accessed through the Classifier icon in the Imagine icon panel. This device will enable you to
select and save training sites and make them available for future use in a supervised
classification. You may launch the Signature Editor without having obtained any previous
signatures or you can retrieve a .sig file using Load under the File menu within the Signature
Editor. The Signature Editor has many interesting and useful tools. The tools you should
concern yourself with are the buttons directly beneath the menu bar, especially the three that

39
have pluses and minuses on them. These will be used in conjunction with the AOI editor to
enter training sites into a .sig file. The first button, it looks like an L with a plus next to it, is
used to add a currently selected AOI site to the file. The next one to the right will replace the
highlighted field with the current AOI site. The third button is used to merge training sites
(signatures) once you feel they have similar spectral characteristics.
Create New Signature(s) from AOI
Replace Current Signature(s) with AOI
Merge Selected Signatures
To gather the spectral signature of the sites you would like to place in the signature editor as
training sites, you will need to use the AOI (Area Of Interest) tools. The AOI menu can be
accessed through the current viewer's menu bar. In the AOI pull down menu you will be
presented with many choices (AOI Styles changes the way the cursor styles look). The Tools
and Commands options are important because they allow you to select the type of polygon,
modify the polygon, etc. with which you want to encompass your AOI. The Seed Properties
option is also important because it allows you to modify the limits of seed area growth by area
and/or distance in addition to letting you select the Neighborhood selection criteria. We will be
using the Neighborhood default setting which specifies that four pixels are to be searched,
then only those pixels above, below, to the left, and to the right of the seed or any accepted
pixels are considered contiguous. Under Geographic Constraints, the Area check box should
be turned on to constrain the region area in pixels. Enter 500 into the Area number field and
press Return. This will be the maximum number of pixels that will be in the AOI. Enter 10.00 in
the Spectral Euclidean Distance number field. The pixels that are accepted in the AOI will be
within this spectral distance from the mean of the seed pixel. Before closing the Seed
Properties window, click on Options and make sure that the Include Island Polygons box is
turned on in order to include nonadjacent polygons within the logical growth region.
To begin the process, you must select an area on the image using one of the AOI tools, such
as the polygon or rectangle tool, or you can place a seed and grow a region using the Region
Grow tool (looks like a magnifying glass in the AOI menu). Use whatever you need in that
particular instance; just make sure you think you know what the area represents in terms of
ground cover. In the viewer, zoom into an area where you want to select an AOI using the
viewer's magnifier tool and then select the AOI polygon tool and draw a polygon around your
chosen area (or you may plant a seed to grow). After the AOI is created, a bounding box
surrounds the polygon or region, indicating that it is currently selected. While the area is
selected, use the Create New Signature button to add the selected area into the Signature

40
Editor. Now click inside the Signature Name column for the signature you just added and give
it a name (use names like urban1, urban2, etc. to define your individual AOIs). You may also
want to change the color in the Color column. You can use the Image Alarm tool under View
in the Signature Editor to get a preview of the extent that the classes you have chosen
represent the rest of the image. If you select the Image Alarm option a pop-up box titled
Signature Alarm will open. In this box you can choose to indicate classes that overlap and the
color that represents overlap. This can be useful if you are considering merging classes. The
signature alarm will also, as mentioned, let you see the extent of each of the classes (you will
need to do this anyway before you can see the overlap). Do this by selecting (highlighting) a
class or a set of classes in the signature editor using the cursor. You can select the color you
would like to represent the class as by clicking on the colored square with the right mouse
button. Once you have made your selections click on the OK button in the signature alarm and
let Imagine do its work. Using this tool, you can see what areas are covered and which are not
using the classes you have selected (according to the rules of parallelepiped classification
logic). For this exercise, take at least three relatively distinct training sites for each of the
following classes found in the scene:
1. Mangrove
2. Agriculture
3. Water bodies
4. Barren lands
When you are done generating the training sites for these 4 classes and you feel they are
representative of the whole scene based on your use of the signature alarm, save the
signature editor file using the Save As menu item under file in the signature editor menu.
Feature Selection:
The Signature Mean Plot button to the left of the histogram in the signature editor, allows you
to view the mean plots of your training data on the screen and thus estimate which of the TM
bands best discriminates between the different training sites that you have selected. Select this
option as well as the histogram (if you feel this is more helpful, use the all selected signatures
and all bands option and make sure you have all of the classes highlighted when you do this).
The most precise way to accomplish the task of determining which bands to use is to through
the use of the separability option under the Evaluate menu in the Signature Editor. Select 3
for the Layers Per Combination choice. Use the Best Average listing method and click OK.
The results of this operation will appear in a pop-up box titled Separability CellArray. Note
which 3 bands seem to do the best job of spectrally separating your classes. Also note which
classes overlap and which are spectrally separable. You will need this information for the next
part of the exercise.
Display Signature Mean Plot Window

41
Display Signature Histogram Window
Supervised Classification:
Now that you have specified your training sites, you are ready to proceed with the supervised
classification. Under the Classify menu in the Signature Editor, choose the Supervised
Classification option. Because you have already selected a signature file it will not ask for
one. If you were to close the signature editor and access the supervised classification through
the Imagine Classifier menu, you would be able to open a .sig file. In the Supervised
Classification pop-up box that appears give a name for your output file. The Parametric Rule
setting should be set to Minimum Distance (note: the textbook gives a good description of the
differences, advantages and disadvantages of the various classification logic schemes) and
everything else should be left as you find them. Select OK when everything is in place. Open a
new viewer and display the results, then answer the following questions concerning signature
extraction and supervised classification:
1) What function does the Image Alarm present?
2) Explain why you think the mean plot and the separability cellarray seem to
differ on the three most important bands in certain instances.
3) Which Distance Measure method did you choose and why? Which three bands
appear to best separate the classes? (Use the results from the Signature
Separability measures on this one)
4) Which land cover classes seem to be confused the most and why do think this
is the case?
5) If you could combine or throw out certain signatures (training sites) to create
sufficient separability between classes, which would you manipulate and how?
Unsupervised Classification (Clustering):
Unsupervised classification differs from a supervised classification in that the computer
develops the signatures that will be used to classify the scene rather than the user. The
classification process results in a number of spectral classes which the analyst must then
assign (a posteriori) to information classes of interest. This requires knowledge of the terrain
present in the scene as well as its spectral characteristics.
The Unsupervised Classification option is selected in the Classification menu under the
Imagine Classifier icon. You will notice that the Unsupervised Classification dialog box
states that it is an ISODATA unsupervised classification. The Iterative Self-Organizing Data
Analysis Technique (ISODATA) is a widely used clustering algorithm and is different from the
formerly used chain method because it makes a large number of passes through the remote
sensing dataset, not just two passes. It uses the minimum spectral distance formula to form
clusters. It begins with either arbitrary cluster means or means of an existing signature set, and
each time the clustering repeats, the means of these clusters are shifted. The new cluster
means are used for the next iteration.

42
The ISODATA utility repeats the clustering of the image until either a maximum number of
iterations has been performed, or a maximum percentage of unchanged pixels has been
reached between two iterations. Performing an unsupervised classification is simpler than a
supervised classification, because the signatures are automatically generated by the ISODATA
algorithm. However, as stated before, the analyst must have ground reference information and
knowledge of the terrain, or useful ancillary data if this approach is to be successful.
To begin the unsupervised classification, click on the Classification icon and then select
Unsupervised Classification... Fill in the input and output information in the Unsupervised
Classification dialog box. Give both the Output Cluster Layer and Output Signature Set a
similar name. Make sure that under Clustering Options the Initialize from Statistics box is
on and set Number of Classes to 30. Set Maximum Iterations to 20 and leave the
Convergence Threshold set to 0.950. Maximum Iterations is the number of times that the
ISODATA utility will recluster the data. It prevents the utility from running too long, or from
getting stuck in a cycle without reaching the convergence threshold. The convergence
threshold is the maximum percentage of pixels whose cluster assignments can go unchanged
between iterations. This prevents the ISODATA utility from running indefinitely. Leave
everything else in its default state. When you have entered all the relevant information click OK
to begin the process.
Cluster Identification:
To aid in evaluation we will need to view the results of the clustering so that we may see how
the clusters are arranged in feature space and thereby make informed decisions about the
nature of the cluster. The first step that will allow us to do so is the creation of feature space
images. The Feature Space Image button can be found on the Classification menu. When it
has been selected a dialog box will appear saying Create Feature Space Images at the top.
Select the original image (not the clustered one) as the Input Raster Layer and make sure the
Output Root Name is similar to the raster layer and the directory path is correct. Leave the rest
of the selections at their default settings and click OK. When the processing is complete open
a new viewer and view the output images. Note that the 2_5 (and other options) represents the
layers that are being shown in the image. In this case layer 2 (band 2) will be displayed on the
x-axis and layer 5 (band 5) on the y-axis. Pay close attention when you look in the book for
help in determining what clusters represent which ground elements.
The next step is to open the Signature Editor (under the Classification menu) with the *.sig file
you created in the unsupervised classification. Select all the clusters (they should all be

43
highlighted in yellow). In the Signature Editor main menu select Feature and then in that pull-
down menu select Objects. This will display a Signature Objects dialog box that allows you to
tell Imagine which viewer you want to receive the signature editor information about the
clusters. In this case we want the viewer in which you have displayed your chosen feature
space graphic. Select that viewer # in the Signature Objects space provided that represents
this viewer. Select Plot Ellipses and Plot Means (or you can try the others if you like). Leave
everything else in it's default state and click OK. Only selected clusters in the Signature Editor
window will be drawn. More than likely your ellipses and means are multi-tonal in nature. If you
would like them all to be white, red, green, etc... select all the classes in the Signature Editor
dialog box using the mouse and change the color to the one you desire. Save the information
as an Annotation Layer.
To analyze the content of the clusters, you should use a combination of techniques. The first
should be to use the mean scatter plot to make some educated guess about the information in
each cluster. You might want to label each of the 30 clusters on the scatter plot using the
Label option in the Signature Objects dialog box so you know what cluster is containing
which class. You will more than likely have to zoom in to get a better look at some of the
clusters given the close proximity of clusters to each other. You should also have a viewer
open with the original scene displayed. This will further help you identify the land cover class. If
you are feeling adventurous, you can overlay your classified image on the original image and
set all the clustered image's colors to transparent using the Raster Attribute Editor found
under the Viewer menu (Raster - Attributes). Once you have set all classes to transparent then
you can individually color (by making them opaque) particular classes and see where they are
on the image. Another method may be to use the Utility - Swipe or the Utility - Fade tools in
the Viewer by opening the classified image on top of the raw data (do not Clear Display after
opening the first raw image). Regardless of how you choose to proceed you should not rely on
any one particular method but a combination of methods and some common sense to arrive at
a sound classification.
When you have decided upon the class breakdowns, use the Raster Attribute editor to assign
class names and colors to the classification image. Create the same five classes you used in
the supervised classification and place each of the 30 clusters into one of the classes by giving
it the same color and class name as every other cluster in that class. When you satisfied with
your unsupervised classification, finish the lab.

44
6.5 Change Detection of Coastal Vegetation
Objectives:
√ Compare and contrast two different change detection methods to analyze
changes in land use/land cover between two dates.
Introduction:
Change detection is the process of identifying differences in the state of a feature or
phenomenon by observing it at different times. In remote sensing it is useful in land use/land
cover change analysis such as monitoring deforestation or vegetation phenology. However,
there are many remote sensor system and environmental parameters that must be considered
whenever performing change detection. Failure to understand the impact of the various
parameters on the change detection process can lead to inaccurate results. Ideally, the
remotely sensed data used should be acquired by a remote sensor system that holds
temporal, spatial, spectral, and radiometric resolutions constant. For example, changes in
radiance values between images may be caused by a number of different factors such as a
field which may have different soil moisture content and therefore appear different in two
individual images.
Bring up two viewers and display the 1990 and 2000 images and compare them side-by-side.
These images depict mangrove forests in Burma. Visually examine the differences as an initial
familiarization technique. It is important to have an idea of where you might expect to see
changes.
Change Detection Methods:
Of the many change detection methods available, we will be examining three. Ideally, it is
important that the analyst has knowledge of the cultural and biophysical characteristics of the
area and preferably some ancillary data. The analyst should also be aware of the different
techniques available, including the limitations and advantages of their respective algorithms.
Method 1: Image Differencing in Spatial Modeler
This involves the subtraction of two images and the addition of a constant value to the result.
These results can be demonstrated in a differenced distribution for each band.
The Spatial Modeler function in ERDAS Imagine allows the user to graphically create a spatial
model and execute it. In this simple example, we will create a change detection model which
uses both mangrove90.img and mangrove00.img as inputs, develops an image differencing
algorithm as the function, and creates a change detection image as an output.

45
Begin by opening the Spatial Modeler menu by selecting the Modeler icon in the Imagine icon
panel. Review the function of each of the Model Maker's tools before going on.
Now select the Model Maker button in the Spatial Modeler menu. Wait for the Model Maker
dialog box and the model tools to appear. Select the raster object tool and place a raster
object in the model window (towards the top left of the window). It will have a question mark as
a title for now, but you will assign the input raster file later. Repeat the process and place a
second and third raster icon in the window (one on the top right and one near the bottom
center. If you make a mistake, use the Edit menu to cut the selected mistake out of the model.
Now select the function tool and place a function symbol near the center of the model
window. Use the connect tool to connect the raster objects on top to the function symbol by
selecting a point inside the top left raster icon and dragging a line to the center of the function
symbol. Release the mouse and a connection arrow should appear. Now connect the upper
right raster icon to the function symbol, and then the function symbol to the lower raster object.
The resulting function should look somewhat like the model depicted below:
Now double click on the top left raster object. The Raster Object dialog box will open. Select
mangrove90.img as the input and leave all other options in their default state. When this is
completed, select OK. The name of the image should now be present below the raster object.
Complete the same process for the upper right raster object with mangrove00.img as the input.

46
Next, double click on the function symbol. In the Function Definition window that appears,
you will create the image differencing algorithm to be used in this model. In the list showing the
available inputs, the number in parentheses corresponds to the individual raster layer. We will
be using the full scene image for the calculations and NOT the individual layers. Use the dialog
box calculator to create the following algorithm in the blank space in the bottom of the dialog
box.
$n1_mangrove90 - $n2_mangrove00 + 128
Finally, double click on the bottom raster object, which is your raster model output, and give it
an output file name. Again, leave the rest of the selections in their default state. When all
objects are labeled and the function definition complete, look at the top of the model window
and find the Process option. Run the model by selecting Run. When the model is processed,
select OK and exit the Model Maker without saving any changes. In a new viewer, display the
model output image using the same RGB layers you used in the first part of the exercise.
1) Why add a constant of 128 to the image difference equation?
2) After running an improved image difference equation of your choice, compare
the output model image with the two original images. What areas appear to have
the most land cover change? What do the different colors represent?
Method 2: Multi-Date Visual Change Detection Using Write-Function Memory Insertion
This method involves the use of one band from each date of imagery. Each band is put in an
image plane to create a layer stack and the composite is displayed. The corresponding colors
represent change in either direction or no change.
Select the Interpreter icon in the Image icon panel and then select the Utilities option. In the
menu that appears, select the Layer Stack. We will be creating a layer stack using only the
NIR bands in each of the 1994 and 1996 images. When the Layer Stack dialog box appears,
only select layer 3 of mangrove90.img as the first input file and click Add. This should add the
first input image name and path into the blank space above the Add button. Now add the 1996
NIR band to the image by selecting layer 3 of the mangrove00.img in the input file space. After
you have specified an output file name, leave the rest of the information in its default state and
click OK. Wait until the processing is complete and then display the output image in true color
mode. Assign layer 1 to red, layer 2 to green, and blue to either layer 1 or 2 (Note: 1:2:2 seems
to be provide the most clearly defined change areas). After the image is displayed, go to
Raster - Band Combinations and turn off the blue gun by clicking on the button next to the

47
word Blue. This combination leaves you with just the red (1994) and green (1996) layers in the
viewer window. The resulting image should have only red, green and yellow shades.
Method 3: Change Detection using Specific Threshold
Select this option to compute the differences between two images and to highlight changes
that exceed a user-specified threshold. You will be creating two new output image files, which
you can save to a directory. Open a viewer to see the files. To open this dialog, select Image
Interpreter | Utilities... | Change Detection.... Before Image: The Before Image
(mangrove90.img) is the earlier of the two images. If you are comparing bands of one image,
you do not have to click to select from the Before Image popup list. Layer: If you are working
with a multiband image, you may choose any band for analysis in conjunction with the Before
Image. The Before Image should match the After Image when you compare the two images.
After Image: The After Image (mangrove00.img) is subtracted from the Before Image to
provide the image difference and Highlight Change image. The After Image is the more recent
of your two themes and reflects change over time. Layer: If you are working with a multiband
image, you may choose any band for analysis in conjunction with the After Image. The After
Image should match the Before Image when you compare the two images. Image Difference
File: A grayscale image composed of single band continuous data. This image is the direct
result of subtraction of the Before Image from the After Image. Since Change Detection
calculates change in brightness values over time, the Image Difference File reflects that
change using the grayscale image. Type in a name for this file here and save it to a directory.
Highlight Change File: The Highlight Change Image is a five-class thematic image, typically
divided into the five categories of Background, Decreased, Some Decreased, Unchanged,
Some Increase, and Increased. Type in a name for this file here and save it to a directory.
Good Luck!

48
Surabaya Coast from Space (Credit: Digital Globe)

49
BIBLIOGRAPHY
1. FAO, (2004). FAO Global forest resources assessment update 2005 Terms and definitions
(Final version), Food and Agriculture Organization of the United Nations, Rome. Available at:
http://www.fao.org/docrep/007/ae156e/AE156E00.htm#TopOfPage.
2. Lambin E. (1996). Monitoring Forest Degradation in Tropical Regions by Remote Sensing:
Some Methodological Issues. Global Ecology and Biogeography, Vol. 8, No. 3/4, Tropical
Open Woodlands Special Issue (May - Jul., 1999), pp. 191-198
3. Riitters, K., J. Wickham, R. O'Neill, B. Jones, and E. Smith. 2000. Global-scale patterns of
forest fragmentation. Conservation Ecology 4(2): 3. [online] URL:
http://www.consecol.org/vol4/iss2/art3/
4. Remote Sensing Tutorial by NASA, available at: http://rst.gsfc.nasa.gov/
5. SOUZA C. M. ROBERTS D. (2005). Mapping forest degradation in the Amazon region with
Ikonos images. International journal of remote sensing, vol. 26, no
3, pp. 425-429.

Aurthy fathimaPhysical.sce

More Related Content

Similar to Aurthy fathimaPhysical.sce (20)

Recently uploaded (20)

Aurthy fathimaPhysical.sce