Using FME Cloud, Space Data and R Libraries to Help Assess Production Impact on Environment

Using FME Flow Hosted, space data
and R libraries to help assess
production impact on environment
Ellipso Facto / Veremes
Green Score Capital

Presenters
The
Peak
of
Data
Integration
20
23
Frédéric
Eichelbrenner
FME Addict
Veremes/Ellipso Facto
Valérie
Tiersen
CEO and founder
Green Score Capital

The
Peak
of
Data
Integration
20
23
Agenda
● Why focusing on environment ?
● Evaluating biome characteristics
● Relevant datasets
● FME process

The
Peak
of
Data
Integration
20
23
Introduction

The
Peak
of
Data
Integration
20
23
“There is now unequivocal evidence that
biodiversity loss reduces the e ciency by
which ecological communities capture
biologically essential resources, produce
biomass, decompose and recycle
biologically essential nutrients.”
Cardinale, B.J. et al. (2012) Biodiversity Loss and its
impact on humanity, Nature News.
Available at: https://www.nature.com/articles/nature11148

The
Peak
of
Data
Integration
20
23
Why focus on biodiversity here ?
Most alerts relate to climate change and carbon emissions
Biodiversity is a climate stabilizing factor
Biodiversity is central in coming regulations

The
Peak
of
Data
Integration
20
23
• Startup created in 2019 by Valérie Tiersen
• Help measure carbon and biodiversity footprint
• Supported by European Spatial Agency and CNES
• Awarded "1000+ proﬁtable solutions to protect the
environment" by the Solar Impulse Foundation in July 2023

The
Peak
of
Data
Integration
20
23
Current project objectives
• Assess impact of companies activity on environment
• Based on “connectivity scores” upon production areas

Project Team
The
Peak
of
Data
Integration
20
23
Eliette
Ascensio
Research Engineer
Morgane
Faucheux
GIS Specialist
Lucille
Marescot
Phd
Scientiﬁc Consultant

The
Peak
of
Data
Integration
20
23
Habitat connectivity
How and to what degree distinct patches of habitat are
connected, which can inﬂuence the distribution, genetic
diversity, and health of animal and plant populations

The
Peak
of
Data
Integration
20
23
Connectivity metrics
Computed using diverse mathematical
formulas depending on :
● Size
● Shape
● Spatial repartition
of spatial patches
For example
● Euclidean Nearest-Neighbor (ENN) measures the distance
between patches of the same class.
Packages available in R language

The
Peak
of
Data
Integration
20
23
Proof of Concept

The
Peak
of
Data
Integration
20
23
Input data
● Land use datasets
○ Global Land Cover (Copernicus)
■ ESA funded, 100m resolution, 21 classes, 2015 to 2019
○ World Cover (Copernicus)
■ ESA funded, 10m resolution, 11 classes, 2020 and 2021
● WWF biomes
○ Detailed forest categories
● Administrative regions

The
Peak
of
Data
Integration
20
23
Output data
● Individual metric values upon several years
● Global score by combining several metrics
○ for one speciﬁc area of interest
○ for many administrative regions (precomputed score)

The
Peak
of
Data
Integration
20
23
General algorithm
Given the area of interest
● Extract land use data
● Adapt data to computation (projection, resolution, classes, …)
● Reﬁne land use classes with external datasets (WWF)
● Run metrics computation (using R packages)
● Combine metrics to get global score
● Publish output format (excel, database, HTML, …)

The
Peak
of
Data
Integration
20
23
Focus on some key technical aspects
● Extract land use data from S3 buckets
● Adapt land use classes from Copernicus to target
● Call R functions
● Deployment on FME Flow Hosted

The
Peak
of
Data
Integration
20
23 Extract land use data from S3 buckets

The
Peak
of
Data
Integration
20
23 Adapt land use classes
Copernicus 21 classes Target 7 classes

The
Peak
of
Data
Integration
20
23 Adapt land use classes : trick with FME
● Build a fake GRAY8 palette representing the mapping
● Apply this palette to the land use image
● Replace the palette with the ﬁnal classes (RGB and names)
GRAY8
0 0
1 0
10 3
20 1
30 1
40 1
50 4
60 6
70 6
80 7
90 5
…
RGB24
0 0,0,0
1 255,255,76
2 0,204,0
3 100,140,0
4 250,0,0
5 0,150,160
6 180,180,180
7 0,50,200
8 255,255,255
…
STRING 40
0 inconnu
1 agriculture
2 foret tropicale
3 foret tempérée
4 urbain
5 zone humide
6 desert
7 eau
8 inconnu
…

The
Peak
of
Data
Integration
20
23 Call R functions
● Input as “data frame” structures
● All input features directly available
● No “group by”
● Output data as “data frame” rows
● Beneﬁts from numerous packages
Similar interface as InlineQuerier

The
Peak
of
Data
Integration
20
23 Deployment on FME Flow Hosted
● Why ?
○ CPU on-demand
○ Automations
○ Cost effective
● Tricky
○ R package manual
deployment on Linux

The
Peak
of
Data
Integration
20
23
Results
● Automate computation on
○ 200 regions of interest
○ 3 years
● Around 30 hours of Enterprise instance

The
Peak
of
Data
Integration
20
23
Conclusion

The
Peak
of
Data
Integration
20
23 Key takeaways
● Environment has a huge importance for our future
● Spatial data is very helpful to assess impact on environment
○ Copernicus data widely available in several domains
● FME as a powerful tool to build ad hoc processes
○ Specially efficient for a proof-of-concept
● (FME + R) is a winner
● FME Flow offers fast and efficient deployment capabilities
○ Automations and Server/Gallery Apps are game changers

The
Peak
of
Data
Integration
20
23
Resources
● https://greenscorecapital.com/en/
● https://solarimpulse.com/solutions-explorer/footprint-target

ThankYou!
frederic@ellipsofacto.com
frederic.eichelbrenner@veremes.com

Using FME Cloud, Space Data and R Libraries to Help Assess Production Impact on Environment

More Related Content

Similar to Using FME Cloud, Space Data and R Libraries to Help Assess Production Impact on Environment (20)

More from Safe Software (20)

Recently uploaded (20)

Using FME Cloud, Space Data and R Libraries to Help Assess Production Impact on Environment