Crop model and weather data generation evaluation for conservation agriculture in Ethiopia - Amsterdam, Netherlands : Elsevier, 2018.
Crop simulation models offer possibilities to evaluate and target agricultural information for sustainable intensification in countries like Ethiopia with inadequate resources for field research. The objectives of this research were to calibrate and evaluate the CERES-Maize, CROPGRO-Dry bean and CROPGRO-Soybean models for practices associated with conservation agriculture and fertilizer N, and to evaluate five generated weather datasets for Ethiopia. Data from multi-year field experiments and additional data obtained from previously conducted national variety trials were used for model evaluation. Generated weather datasets for six agroecologies were evaluated by comparison with observed data and by use of data in the models. Genetic coefficients used in the models for maize, dry bean and soybean were determined by model parametrization and calibration of phenology and yield. The models acceptably simulated the effects of N rate, maize-legume rotation, and crop residue retention plus tillage with average normalized deviation closer to zero, RMSE less or similar to standard deviation of observed data, and with normalized RMSE (nRMSE) < 15%. Both NASA and Global Yield Gap Atlas (GYGA) daily rainfall showed good agreement with observed weather data (RMSE < 9 mm). Daily maximum and minimum temperature of GYGA and WeatherMan datasets had the lowest RMSE of 1.99 and 3.06, and 2.5 and 3.1 °C, respectively. Between 85?100% of simulated grain yields of maize, dry bean and soybean with GYGA and WeatherMan datasets fell within ±10% deviation of mean simulated grain yields with observed weather data, and with the lowest inter-annual variability. It was concluded that model calibrations were satisfactory, and either GYGA or WeatherMan datasets alone or combined could be used to run the crop models in sites which lack observed daily weather data in Ethiopia.
Text in English