ESA African Research Fellowship: Celebrating Achievements and Looking Ahead

Integrating Machine Learning with Hyper-Resolution Land Surface Modelling: Enhancing Climate Resilience and Water Management for Zimbabwean Smallholder Farmers: 

During her fellowship at ESA ESRIN, Ms. Moreblessing Muzava (assistant researcher, Zimbabwe Power Company, Zimbabwe) developed an innovative hybrid modelling framework that combines climate reanalysis data, satellite EO, land surface modelling, and machine learning techniques to improve flood prediction in Zimbabwe’s Save and Runde catchments. The framework produced high-resolution flood maps and achieved an accuracy of up to 89% when compared with satellite-derived flood observations. Beyond mapping flood extents, the methodology generates probabilistic flood-risk information that can support early-warning systems, disaster preparedness activities, water resources management, and climate-resilient agricultural planning. Ms. Muzava’s work demonstrates how the integration of EO data with physically based models and AI can provide practical decision-support tools for strengthening resilience to flood hazards in vulnerable regions of Africa.

Assessing Coastal Processes and Climate Change Impacts Using Earth Observation Satellite Data: A Case Study of Northern Coastal area of EGYPT:

Dr. Mohamed Abdelfattah (Assistant Professor of Hydrogeology at Port Said University, Egypt) has completed a study using EO data to examine four decades (1985–2025) of coastal change along Egypt’s Mediterranean coastline.
By combining Landsat and Sentinel-2 satellite imagery, DSAS shoreline analysis, Kalman filter forecasting, and GRACE satellite gravimetry with field-validated groundwater sampling, the research revealed 364% urban expansion, severe coastal erosion at Damietta and Rosetta, and groundwater salinization at Port Said linked to seawater intrusion. The findings demonstrate how EO data can reveal these pressures as an interconnected system rather than isolated hazards, offering a transferable framework for coastal management across vulnerable deltas in the Mediterranean and Global South. Crucially, this work showcases how the fusion of multi-sensor EO data with time-series forecasting and field validation can reveal interconnected coastal pressures as a unified system, offering a transferable blueprint for climate-adaptive coastal management across vulnerable deltas.

Monitoring the Swollen Shoot Viral Disease in Cocoa and Its Impact on Cocoa Production in Côte d’Ivoire Using Remote Sensing and Artificial Intelligence:

Dr. Amandine Carine Njeugeut Mbiafeu (researcher at the Centre Universitaire de Recherche et d’Applications en Télédétection (CURAT) in Côte d’Ivoire) developed an innovative project that combines AI and EO data to monitor Cocoa Swollen Shoot Viral Disease (CSSVD).

Côte d’Ivoire is the world’s leading cocoa producer, but cocoa production is increasingly threatened by disease-related decline, environmental stress, and land-use change. The project aimed to combine EO, AI, spatial validation, and analytical tools to support more reproducible and scalable cocoa monitoring. The work developed an AI-based workflow using Sentinel-2 imagery (10m), downscaled S2DR4 data (1m), ALOS PALSAR DEM (12.5 resampled to 10m), field reference samples, and outbreak reference polygons. Machine learning and deep learning models were used to produce annual and seasonal land-cover and cocoa maps from 2015 to 2025. The results made it possible to analyse cocoa dynamics, including stable, gained, and lost cocoa areas.
A second part of the work focused on cocoa anomaly monitoring. Spectral and temporal indicators such as NDVI, NDWI, NDRE, ReCI, MSI, NBR, and NBR2 were used to distinguish healthy and abnormal cocoa conditions, assign interpretable phenology and stress states, and compare outbreak and non-outbreak areas through time.
These methods were integrated into AgroEOFlow, a reproducible EO platform supporting data preprocessing, model selection, spatial validation, mapping, time-series analysis, anomaly detection, and decision-support outputs for agricultural monitoring. More broadly, this integrated approach validates that combining EO data, spectral indicators, and AI within a reproducible platform can enable scalable, near-real-time monitoring of crop diseases and land-use dynamics—delivering critical decision-support tools for agricultural resilience and food security in West Africa’s cocoa monitoring belt.

A High-Resolution Assessment of Water Footprints of Cereal Production Using Earth Observation Satellite Data in Tunisia:

During his fellowship at ESA ESRIN, Dr. Mohamed Rajhi, researcher and project manager at the Tunisian Space Association (TUNSA), developed a satellite-based assessment of the water footprint (WF) of cereal production in Tunisia, a country where agriculture competes for an increasingly scarce freshwater resource under arid and semi-arid conditions. His research addresses a persistent evidence gap, since national water accounting for cereals has relied on aggregated statistics. He designed an integrated processing pipeline combining multi-source Earth Observation and climate reanalysis data: ESA WorldCereal, FAO WaPOR, CHIRPS, MODIS and ERA5 Land, with machine-learning and geospatial modelling, deriving crop phenology from NDVI time series and applying a pixel-level soil water balance to partition actual evapotranspiration into its green (rainfall) and blue (irrigation) components, normalised by remotely sensed yield. Applied to the 2020/21 winter-cereal season across 809,000 hectares of cereal land, the analysis produced a national mean water footprint of about 1,286 m³ per tonne, well below the global wheat average of roughly 1,800 m³/t, and showed that Tunisian cereal production is overwhelmingly rain-fed, revealing a clear inverse relationship between yield and water footprint and pinpointing the irrigated schemes that carry the blue-water cost. Building on these layers, he developed a fuzzy multi-criteria suitability model, weighted through an Analytic Hierarchy Process, that classifies cereal land following FAO suitability standards. The outcome is a transparent, reproducible and decision-ready framework translating satellite observations into actionable indicators for irrigation efficiency, cropland allocation, evidence-based water policy, and transferable to other water-scarce African contexts.