Soil water balance models optimized for daily and decadal evapotranspiration and soil moisture estimation in rainfed sugarcane cultivation in Rio Largo, AL.
evapotranspiration, inverse modeling, water balance
Know the dynamics of water in the soil and the amount required by the crop is fundamental for agricultural planning and management. This study aimed to evaluate the performance of the Thornthwaite and Mather and Kc dual (FAO-56) models for estimating soil moisture and evapotranspiration of sugarcane in Rio Largo, AL. Meteorological, crop and soil measurements were obtained in the experimental area of the Campus of Engineering and Agrarian Sciences of the Federal University of Alagoas (UFAL). The adjustments of crop and soil parameters (Kc, Kcb and f) were performed by inverse modeling techniques, in which volumetric moisture data measured by TDR probes and actual evapotranspiration obtained by the Bowen energy balance method were used. The Metropolis-Hastings method was used to adjust the model coefficients. The models were evaluated by the performance indices of modified Willmott's concordance index (dw) (1985), coefficient of determination (r2), root mean square error (RMSE), systematic and non-systematic errors. The total actual evapotranspiration during the cycle estimated by the Thornthwaite and Mather model on a daily scale ranged from 553.6 to 827,6 mm. The actual evapotranspiration for the dual Kc model ranged between 804,7 and 841,6 mm. The adjusted Kc values optimized with actual evapotranspiration and volumetric moisture were lower than FAO-56 values by 5.6, 28 and 69.6%, respectively. The low Kc value obtained with volumetric moisture in the ThM model resulted in lower actual evapotranspiration. The optimized Kcb values with actual evapotranspiration and volumetric moisture were lower than the FAO-56 values by 25.8% and 13.3%, respectively. Both models underestimated volumetric moisture and overestimated actual evapotranspiration under P-ETc ≥ 0 condition. However, better precision and accuracy of the models was observed for ETR estimation in the P-ETc ≥ 0 condition. The best performance of the two models for volumetric moisture estimation was obtained in the P-ETc < 0 condition when optimized with volumetric moisture.