| Summary: | Accurate crop yield estimation remains a persistent challenge in smallholder farming systems, particularly in Sub-Saharan Africa, where fragmented fields, diverse cropping practices, and limited ground-truth data hinder the scalability and precision of models. This study proposes an integrated framework combining Earth Observation (EO), Artificial Intelligence (AI), and process-based Crop Modelling (CM) to address these limitations and improve yield prediction in data-scarce environments. The research aims to assess the potential of high-resolution EO data for capturing spatial variability in smallholder landscapes, calibrate crop models using region- specific parameters and agroclimatic data, and develop and benchmark AI models for scalable, cross-regional yield prediction. A multi-source data fusion strategy will integrate satellite imagery, weather records, and field observations to support the development of models. Calibration and validation will be conducted using field-level yield data from diverse agroecological zones in Kenya to ensure robustness and transferability of the results. The study aims to develop a validated, scalable framework for yield estimation that can inform climate- resilient agricultural planning, support early warning systems, and enhance digital advisory services tailored to the needs of smallholders. This research will contribute to the broader agenda of food security by advancing context-sensitive, data-driven tools for monitoring crop productivity in under-resourced regio
|
|---|