The process-based crop model selected is the Decision Support System for Agrotechnology Transfer

(DSSAT) [115, 116]. DSSAT integrates crop physiology and phenotype, weather and soil data, and crop

management strategies. It includes different crop growth modules, each dealing with a specific set of crops.

For example, CROPGRO simulates the growth and yield for grain legumes (e.g., Soybean, cowpea), CERES for

grain cereals (e.g., rice, maize, and millet), and SUBSTOR for potatoes and cassavas etc. New crops may also be

added to the model. DSSAT has been applied to several regions of Africa in the past with success [117, 118], including

a recent study conducted by proposal member Wang for 13 African countries [119]. The response of crop yield to anomalies

in seasonal climate varies depending on crop type, cultivar, fertilizer application, and other crop

management. The cultivars and varieties most widely used in the BNB will be chosen for the simulations,

with tailored fertilizer application, manure application, irrigation prevalence, and planting time, from the

grid-level reanalyzed ISRIC-WISE 1.1 soil database [120] and the geographically explicit fertilizer

applications dataset of [121]. Model calibration will be based on approximately two thirds of the years for

which data is available (provided by IFPRI), and validated on the remaining one third, based on historical

remote sensing observations and SURFEX land surface model simulations. The calibrated model will

subsequently be used to simulate crop yield responses to forecast information (from the disaggregated

seasonal forecasts and from WaterGAP3/groundwater model predictions) through a hindcast

assessment. Potential adaptations will also be explored, including changes in agricultural input, cultivars,

management strategies, etc., to help guide farmers’ decision-making strategies.