Empirical and Mechanistic Prediction of Plant Available Water at Sowing for Wheat in the Southern Great Plains.

Plant available water at sowing (PAW_s) can impact wheat stand establishment, early crop development, and yield. Consequently, PAW_s is an essential input in crop simulation models and its estimation can improve agronomic decisions. Our objective was to predict PAW_s in continuous winter wheat (Triticum aestivum L.) by modeling the soil moisture dynamics of the preceding 4-mo summer fallow. The mechanistic soil water balance models dual crop coefficient (dual K_c) and Simple Simulation Modeling (SSM), were calibrated, validated, and tested using soil moisture datasets collected from 2009 to 2013 in Oklahoma totaling 29 site-years. Additionally, PAW_s was predicted using empirical non-linear models based on cumulative fallow precipitation and the soil’s plant available water capacity (PAWC). Both the dual K_c and SSM models resulted in normalized root mean squared error (RMSE_n) below 12% (20 mm) for the calibration and validation datasets. Modeled PAW_s for the prediction dataset was within ±30% of field observations in 67% of the site-years for both dual K_c and SSM models, with RMSE_n of 27 and 32%, respectively. An exponential and a logarithmic model of PAW_s using cumulative fallow precipitation and PAWC both resulted in RMSE_n = 23 and 29% in the calibration and validationdatasets, respectively. The dual K_c model was slightly superior to empirical models based on non-linear regression analysis, and was superior to the SSM model. Initializing the dual K_c at the start of the preceding fallow or using empirical relationships allow for acceptable predictions of PAW_s, eliminating the need for subjective PAW_s values.