Assessing the effects of management and hydro-edaphic conditions on rice in contrasting East African wetlands using experimental and modelling approaches

Lowland rice yields in East Africa remain low despite favourable hydro-edaphic conditions as benefits from improved cultural management vary between and within wetland types and interactions are poorly understood. Hence, multi-year agronomic field experiments were established to assess the differential responses of lowland rice to management (rainfed 0 and 60 kg N ha−1, and irrigated 120 kg N ha−1 + 60 kg PK ha−1) and field position within a floodplain in Tanzania (fringe and middle positions) and an inland valley in Uganda (valley-fringe, mid-valley and valley-bottom positions). We then calibrated and validated the Agricultural Production System Simulator (APSIM), evaluated the importance of external water table data as model input and assessed the relative effects of water and N stress on yield as affected by wetland type and field position. Yields of 3.2–9.2 Mg ha−1 were attained in the floodplain and of 1.9–6.3 Mg ha−1 in the inland valley, highlighting the substantial scope to boost yields beyond current regional means of around 2 Mg ha−1. The model estimated grain yields in both wetlands well within the experimental uncertainty during model validation (n = 12, r2 = 0.76, RMSEa= 0.92 Mg ha−1 in the floodplain; n = 18, r2 = 0.71, RMSEa= 0.72 Mg ha−1 in the inland valley). Results further emphasised the importance of external water table data for sound model performance as they evidently alleviated seasonal droughts. Simulated abiotic stress patterns additionally highlighted hydro-edaphic differences from field positioning within and between both wetlands. While low soil N was generally the main yield constraint, water stress was comparably more pronounced in the inland valley and supplemental irrigation thus more beneficial on yield. Hydro-edaphic field conditions favoured rice production in the floodplain’s fringe with comparably lower N stress, while large spatial-temporal variabilities prevented a distinct delineation based on toposequential field positions in the inland valley.