Maize yield suitability mapping in two major Asian Mega-Deltas using AgERA and CMIP6 climate projections in crop modeling

Asian Mega-Deltas (AMDs) are important food baskets and contribute significantly to global food security. However, these areas are extremely susceptible to the consequences of climate change, such as rising temperatures, sea-level rise, water deficits/surpluses and saltwater intrusion. This study focused on maize crop suitability mapping and yield assessment in two major AMDs: the Ganges Delta, spanning parts of northeast India and Bangladesh, and the Mekong Delta across Vietnam and Cambodia. We investigated the historical climate reanalysis AgERA datasets and climate projections from the Coupled Model Intercomparison Phase 6 (CMIP6) for the periods 2040–2070 and 2070–2100 using PyAEZ-based modeling to estimate maize yields for periods in the near (2050s) and far future (2100s). Province-level yield estimates were validated against statistics reported by the governments of the respective countries. Model performance varied across regions, with R2 values ranging from 0.07 to 0.94, MAE from 0.67 t·ha−1 (14.2%) to 1.56 t·ha−1 (20.7%) and RMSE from 0.62 t·ha−1 (14.6%) to 1.74 t·ha−1 (23.1%) in the Ganges Delta, and R2 values from 0.23 to 0.85, MAE from 0.37 t·ha−1 (12.8%) to 2.7 t·ha−1 (27.2%) and RMSE from 0.45 t·ha−1 (15.9%) to 1.76 t·ha−1 (30.9%) in the Mekong Delta. The model performed comparatively better in the Indian region of the Ganges Delta than in the Bangladeshi region, where some yield underestimation was observed not accurately capturing the increasing upward trend in reported yields over time. Similarly, yields were underestimated in some provinces of the Mekong Delta since 2008. This may be attributed to improved management practices and the model’s inability to fully capture high-input management systems. There are also limitations related to the downscaling of CMIP6 data; the yield estimated using the downscaled CMIP6 data has small variability under rainfed and irrigated conditions. Despite these limitations, the modeling approach effectively identified vulnerable regions for maize production under future climate scenarios. Additionally, maize crop suitability zones were delineated, providing critical insights for planning and policy design to support climate adaptation in these vulnerable regions.