Landscape feedbacks, Climate Change, and Food Production Risk in East Africa

Land Cover/ Land Use Change (LCLUC) and climate change due to elevated levels of greenhouse gases (GHG) both exert strong influence over agricultural production in developing countries. In addition, LCLUC is a first-order climate forcing capable of mitigating or exacerbating GHG effects on climate. LCLUC and GHG forcings that alter climate may then in turn influence agricultural systems, thereby creating a feedback to LCLUC. These human-land-climate interactions are complex- particularly in heterogeneous landscapes like East Africa- where subsistence farming and highly variable growing season conditions are excellent for investigating the nature and magnitude of these interactions. Here we present results of a sensitivity test exploring these feedbacks in Kenya, Tanzania, Uganda, Rwanda, and Burundi. First, we assess the impacts of "aggressive" agricultural expansion and generally warmer, wetter boundary conditions due to large increases in GHG for 2050-2059 using the Regional Atmospheric Modeling System (RAMS). These results are contrasted to simulations conducted for 2000- 2009 conditions. Second, we examine simulated changes in agricultural yield using the process-based crop model CERES-Maize; we use maize as a proxy for all East African agricultural productivity. These simulated yields are imported into the Land Transformation Model (LTM), which, informed by socioeconomic constraints, develops a new projection of land cover under the new yield conditions. Third, by comparing the original "aggressive" landscape with the new yield-feedback landscape, we show areas exhibiting sensitivity to LCLUC under elevated GHG conditions. Although this is a mere sensitivity test, we can identify areas prone to potential food production disturbances. These sensitive areas, while unlikely to see climate and landscape changes this dramatic, can serve as "canaries in the coal mine" for predicting larger changes in East African food production. Additionally, we discuss key challenges of propagating error in coupling these systems.