One-step approach for estimating maize actual water use: part II. Lysimeter evaluation of variable surface resistance models

Abstract

This study evaluated maize variable bulk surface resistance (rs, s m−1) models developed using field/environmental data from non-irrigated fields in a humid climate. The different rs models were reported in the companion paper. Surface resistance values derived from the application of the new models were inserted in the 1965 Penman–Monteith ET equation to calculate actual maize evapotranspiration (ETa). This is the so-called one-step approach. The evaluation was performed with maize water use data measured with a large weighing lysimeter located in the middle of an irrigated maize field (semi-arid climate) near Bushland, Texas, USA. The evaluation was performed for different time scales (semi-hourly to daily) and the rs models bias (MBE) and root mean square errors (RMSE) were determined. Using daytime 30-min rs models in maize ETa (mm 30-min−1) estimation resulted with the lowest relative error of 15.4%. While daytime average rs models developed from daytime average explanatory variables resulted with the lowest relative error of 20.9% in 30-min maize ETa estimation. When rs models from 30-min and daylight average explanatory variables were used to obtain cumulative daytime maize ETa estimations, resulting errors were lower than for semi-hourly time step. In addition, when daytime 30-min and average rs models were applied in conjunction with a fixed rs nighttime value to obtain daily maize ETa, results were similar than those for the daytime time step. Furthermore, another evaluation incorporated rs values obtained adopted ranges of crop evaporative fraction (EF). When this EF-based rs values were used to estimate maize (ETa, mm 30-min−1) the error found was 5.9 ± 21.7%. This result seems high; however, when the EF range-based rs values were applied in the estimation of maize cumulative daytime ETa, results indicated an error of 5.9 ± 11.9%. In this last case, the relative error was much lower than for 30-min ETa estimations. Therefore, it is concluded that some maize rs models, reported in the companion paper, performed well when evaluated in a different climate and agronomic practices and are suitable for the estimation of ETa using the 1965 Penman–Monteith ET equation. In particular daytime 30-min rs models applied to 30-min intervals of ETa estimation and then accumulated over the day performed better than the others rs models and time steps studied.