Improving crop modeling to better simulate maize yield variability under different irrigation managements

Crop models have been used for investigating crop responses to environmental stresses for decades. The study objectives were to (i) calibrate and validate a simple crop model (Hybrid-Maize) using in-situ measured data from sixteen uniquely managed treatments as part of the University of Nebraska-Lincoln Testing Ag Performance Solutions (TAPS) program in North Platte, NE, and (ii) carry out sensitivity analysis and parameter estimation using a multi-parameter optimization approach. Sixteen Arable Mark 1 sensors and thirty two Mark 2 sensors (Arable Labs Inc., San Francisco, CA) collecting hourly and daily weather and crop information in 2019 and 2020 respectively, were installed in the TAPS subsurface drip irrigation experimental plots which were planted with maize and subjected to different irrigation and nitrogen practices. Hybrid-Maize was used for simulating maize yield on the sixteen treatments in 2019 and 2020. Sensitivity analysis showed that initial light use efficiency (LUE), potential kernel filling rate (G5), potential number of kernels per ear (G2), growth respiration coefficient of grain (GRG), empirical parameter determining the relative contribution of a soil layer to water uptake (SLW), and maximum photosynthetic rate (MPR) were the most sensitive parameters to yield. A novel multi-parameter optimization (MPO) approach based on kriging was used for calibrating these six parameters, and the best parameter sets which were later used for model validation. Calibration results showed that there seemed to be strong linear relationships between total water received (WR) and some of the parameters. By using each year's MPO averages of the six parameters instead of default values, ME, MAE, RMSE, uRMSE, and nRMSE were all reduced by about 69%, 66%, 60%, 27%, and 61% respectively for validation treatments. The advantage of using in-situ sensors, coupled with the suitability of the calibrated model for simulating maize yield under different irrigation management, will make the model more useful in future field-scale research with focus on developing decision support tools for in-season crop management and yield forecasts.