Assessing crop model improvements through comparison of sorghum (sorghum bicolor L. moench) simulation models: A case study of West African varieties

Better defining niches for the photoperiod sensitive sorghum (Sorghum bicolor L. Moench) varieties of West Africa into the local cropping system might help to improve the resilience of food production in the region. In particular, crop models are key tools to assess the growth and development of such varieties against climate and soil variability. In this study, we compared the performance of three process-based crop models (APSIM, DSSAT and Samara) for prediction of diverse sorghum germplasm having widely varying photoperiod sensitivity (PPS) using detailed growth and development observations from field trials conducted in West Africa semi-arid region. Our results confirmed the capability of each selected model to reproduce growth and development for varieties of diverse sensitivities to photoperiod. Simulated phenology and morphology organs during calibration and validation were within the closet range of measured values with the evaluation of model error statistics (RMSE and R-2). With the exception of highly sensitive variety (IS15401), APSIM and Samara estimates indicate the lowest value of RMSE (<7days) against the observed values for phenology events (flowering and maturity) compared to DSSAT model. Across the varieties, there was over-estimation for simulated leaf area index (LAI) while total leaf number (TLN) fitted well with the observed values. Samara estimates were found to be the closet with the lowest RMSE values (<3 leaves for TLN and <1.0 m(2)/m(2) for LAI) followed by DSSAT and APSIM respectively. Prediction of grain yield and biomass was less accurate for both calibration and validation. The predictions using APSIM were found to be closest to the observed followed by DSSAT and Samara models respectively. Based on detailed field observations, this study showed that crop models captured well the phenology and leaf development of the photoperiod sensitive (PPS) varieties of West Africa, but failed to estimate accurately partitioning of assimilates during grain filling. APSIM and SAMARA as more mechanistic crop models, have a higher sensitivity of the adjustment of key parameters, notably the specific leaf area for APSIM in low PPS varieties, while SAMARA shows a higher response to parameters changes for high PPS varieties. (C) 2016 Elsevier B.V. All rights reserved.