Ronald Katende 

Agricultural productivity in low-income countries (LICs) is highly vulnerable to climate change, a challenge further compounded by the lack of reliable agricultural data. Existing models often assume the availability of comprehensive datasets, an assumption that does not hold true for LICs. This paper introduces a set of innovative frameworks designed to overcome data scarcity by integrating sparse agricultural data, high-resolution climate information, and advanced machine learning techniques. At the core of this approach is a Bayesian hierarchical model that combines satellite-derived climate data with incomplete in-situ agricultural data, enabling probabilistic estimates of productivity in data-limited contexts. The paper also presents a dynamic panel data model to explore long-term interactions between climate and agriculture, capturing sectoral dynamics over extended periods. Additionally, a novel framework for real-time assessment of agricultural productivity loss is introduced, leveraging Bayesian inference to estimate losses based on environmental proxies. These models are intended to improve predictive accuracy and provide practical tools for real-time monitoring and long-term analysis in data-constrained settings, contributing to stronger climate resilience and more informed decision-making in LICs.