This document discusses how Earth observation (EO) can provide evidence to support food security policies and decision making. It notes the global rise in food insecurity and outlines JRC's use of EO for agricultural monitoring, early warning systems, and yield forecasting to help policymakers. Machine learning methods are being used to improve predictive capacity of food insecurity indicators. The document also emphasizes strengthening local capacity to use geospatial data for monitoring agriculture in Africa and providing technical support for new regulations.
EO based information for food security policy and decision support
1. EO based information for food
security policy and decision
support
Earth Observation for Food Security International Workshop, 24 May 2023
Felix Rembold
Team Leader Food Security, D5 Unit
Joint Research Centre of the European Commission
2. • Food insecurity and malnutrition on the rise after
decades of development gains - SDG 2 out of reach
• Main drivers are: conflict, climate change, economic
impacts of COVID19 and Russia’s invasion of
Ukraine – increasing poverty
• Food, energy and fertilizer inflation remains high,
worsened locally by climate extremes – 3 billion
people have limited access to healthy diets
• Global food systems are not sustainable – need for
food systems transformation
The global food and nutrition crisis
Rapidly increasing demand for independent evidence and integrated analysis for
policy support related to food and nutrition security
3. • Agenda 2030 SDGs, international development and humanitarian aid policies
• Sustainable agri-food systems policies: Farm to Fork, food systems transformation, new regulations
like EUDR, Digitalisation
• African regional and national policies: Agenda 2063, Dakar2 Summit “Feed Africa”
Policy frameworks and need for EO based information
Food systems transformation:
How to produce more with lower
impact on the environment?
EO provides evidence for: monitoring indicators, baseline maps, inputs for
prediction models, land use change maps, GHG emissions estimates etc… at
different spatial and temporal scales, ideally open, free and transparent
4. JRC use of EO
Research and
development
Policy and decision
support
Knowledge
management
Operational
information
systems
• Evidence for policy support
• Agricultural outlook, Food sec. early warning, DRM
• Compliance with regulation (EU CAP, due diligence…)
• Impact assessments
• Reference information (e.g. land cover/use)
• Research and development
• Innovative methods for more quantitative and high
quality products
• Improved predictive capacity
Capacity building
Farm
to
Fork
Partnerships
Humanitarian
assistance
Development cooperation
5. Examples of JRC products for agricultural monitoring
Reference layers and ad
hoc mapping of crop type,
crop area etc..
Monthly agricultural hotspots of
countries vulnerable to food
insecurity and currently
experiencing below average
cereals production
Other early warning
products (floods, water
reservoirs etc…)
Monthly agricultural
production monitoring
and yield forecasting
in Europe and
neighbourhood
• EO is key to all of them, many are based on COPERNICUS Global Land products
• Coverage and frequency has improved dramatically with HR data, cloud computing, AI
6. Anomaly hotSpots of Agricultural Production (ASAP)
• Exploits the full range of latest generation global geospatial data at different scales
using innovative technologies for automated processing and visualization
1.) Hotspot analysis
2.) Warning explorer
3.) High Resolution viewer
https://mars.jrc.ec.europa.eu/asap/
Suite of 3 information platforms:
• Hotspot analysis: country level monitoring, 80 countries, monthly. For decision makers and policy analysts
• Warning explorer: global at province level, 10-daily, fully automated. For agricultural analysts
• High resolution viewer: field level monitoring, Sentinel and Landsat imagery. Analysts with RS knowledge
7. 1 - National level hotspot
South Africa classified as hotspot -February 2019 assessment
2 - Warning Explorer
Dekad: 21-28/02/2019- Free State – Warning 3
3 - High resolution viewer – Free State
JRC ASAP: Drought detection and monitoring at multiple scales
https://mars.jrc.ec.europa.eu/asap/
8. ● seasonal weather forecast from Copernicus Climate
Change Service (C3S) used as input data
● Rainfall forecast for the next 6 months from different
models
● 2 meaningful maps (terciles + skills), global
coverage, continental view
% dryer % normal % wetter
Adapting seasonal forecasts to crop
monitoring needs
https://mars.jrc.ec.europa.eu/asap/seasonal_forecast.php
9. Yield forecasting with Machine Learning
Sub-national National
● Use of agro-climatic indicators and regional
statistics in a Machine Learning workflow
● Running on JRC big data facility
● Deployed for cereals in Algeria in 2021/22
● Currently pre-operational in South Africa
● Use of agro-climatic indicators and FAOSTAT
data in a Machine Learning workflow with Univ.
Valencia
● In season forecasts for 80 countries, in test
mode for 2023
Advantages:
● Objective and reusable automatic method providing YF and accuracy, code available on GitHub
Automatic variables selection (RS, Meteo)
● Global model covering 80 ASAP countries, regional pipeline applicable to any country with reliable
official statistics
Forecasts @ 75% season
progress. Percent of
country-crop yield that can
be estimated with R2>0.3
10. What we propose
1. Review/comparison study on existing indicators, data
availability, methodologies, success and failure stories with
contributions of authors in the field (ACF, IPC , FAO, FEWSNET,
WFP, WorldBank)
2. ML-hybrid model of forecast + explainability with most
relevant variables taken from existing research and better
thematic aggregation and lagging strategies. The goal is to
have a better quantified explainability of contributing
factors for prediction of IPC3+ population share (in collaboration
with WorldBank)
3. Better characterisation of conflicts influence on Food
Insecurity using DRCM Conflicts database (JRC E1)
4. Support to PREDISAM initiative - ACF, University of Granada,
gis4tech - with data provision, data analysis, publication support
5. (Projections of Food Insecurity by 2030 - support to
WorldBank)
AI Forecasting and explainability
of Food Insecurity (FI)
Remote
Sensing
11. Examples of ad hoc support to DG INTPA:
• Cocoa is the first of a number of commodities where there is need for evidence about
traceability and sustainability (linked to EUDR)
Research to inform new policy and regulation
Implementation of new regulation like the EUDR requires major technical and scientific support
Data needed to perform due diligence:
• Cocoa plot polygons or points
• Reference forest or land-use map at the cut-off
date, including relevant classes
• Map of protected areas, including admitted
agricultural areas
• …
12. • GMES&Africa, CLIMSA, Copernicus4GEOGLAM and other projects
• Main objective: Strengthening local capacity to use geospatial and climate data for
agricultural monitoring and forecasting
• Focus on: use of latest generation free EO data, robust field data collection for calibration and
validation
• Ongoing examples:
• 10 m resolution crop type maps for parts of Kenya, Tanzania, Uganda for 2 seasons
• ICPAC Agriculture Watch and OSS GuetCrop platforms based on ASAP
• Support to mapping land use (with focus on Cocoa) in Ivory Coast
Strengthening capacity and increasing access to EO
14. • EO provides independent evidence for informing agriculture and food security
policies. Ground truth for calibration/validation is key for quality.
• Product coverage and frequency have improved dramatically with HR data
(S1, S2, L8). Access and capacity to use them remain challenging in many
parts of the world. Continuity of medium resolution data time series remains
important too (MODIS, S3, VIIRS…)
• Technology alone does not address the full information demand, the need for
expert assessments and analysis and multi-stakeholder monitoring and
reporting cannot be overstated
• The ongoing food security crises increases the need for EO based
information of high quality and improved predictive capacity
Concluding messages