RESILISENSE : IDENTIFICATEURS SATELLITAIRES OBSERVABLES DE LA RÉSILIENCE DES PETITS EXPLOITANTS AGRICOLES
Resilience has risen to the forefront of IITA’s work and across the CGIAR Centres, yet it remains one of the hardest outcomes to measure. Monitoring, evaluation, learning and impact assessment (MELIA) systems rarely collect resilience data, and where they do, it is fragmented and gathered at project-specific levels, which limits cross-project learning and weakens both upward and downward accountability. ResiliSense, developed and hosted within IITA’s Resilience and Climate Adaptation (RCA) Programme in Nairobi, addresses this gap: it integrates the monitoring data that programmes already collect with advances in remote sensing and machine learning to generate resilience insights at the project, programme, and country levels. In doing so, it provides a bridge from targeted household interventions to landscape-level insight.
THE MEASUREMENT GAP
A SURVEY-ANCHORED, SATELLITE-ASSISTED APPROACH
HOW RESILISENSE WORKS
ResiliSense follows five steps. It begins by importing and parsing existing MELIA survey data into a structured question database. A crosswalk engine then maps each question to the SHARP+ indicator library across its economic, social, environmental, governance, and spatial domains, assigning every match a confidence score and routing uncertain matches to expert review. A gap analysis follows, identifying which resilience dimensions the survey leaves uncovered and, where validated proxies exist, fusing in multi-source earth observation: Sentinel-1 radar and Sentinel-2 optical imagery, including the NDVI, NDMI and NDRE vegetation indices, alongside CHIRPS rainfall, ERA5-Land climate reanalysis, and SRTM terrain data, each calibrated against national reference datasets. The tool then aggregates the available indicators into a household-level resilience score, with per-indicator confidence, data sources, and flagged gaps all made transparent. Finally, ResiliSense supports learning and scaling: it advises how future surveys can be designed to close gaps, and it trains machine-learning models to extend resilience estimates across unsurveyed areas, producing maps that can track resilience over time. Extraction, ingestion, inference and visualisation are automated through the programme’s Farm Intelligence Platform, so that a resilience-agnostic dataset can be turned into a resilience-score report with coverage and confidence transparency.