Strategies of Parameter Optimization and Soil Moisture Sensor Deployment for Accurate Estimation of Evapotranspiration Through a Data-driven Method
Yuanyuan Chai,Hu Liu,Yang Yu,Qiyue Yang,Xiaoyou Zhang,Wenzhi Zhao,Li Guo,Omer Yetemen
DOI: https://doi.org/10.1016/j.agrformet.2023.109354
IF: 6.2
2023-01-01
Agricultural and Forest Meteorology
Abstract:A soil moisture data-driven method through inversely solving the Richards' equation was tested, and proved to be a reliable alternative estimator of actual evapotranspiration (ETa) for relatively homogeneous soils in an arid region. Few studies have examined the feasibility of the method using soil moisture for soils with more heterogeneities, which may lead to biased parameters and varied root depths, and thus inaccurate estimates of ETa. We used a genetic algorithm to optimize the parameters most sensitive to ETa estimation, then estimated daily and seasonal ETa using hourly soil moisture data, measured from four sites (from surface to 2-m depth at 20-cm intervals) with different soil textures and crop species in desert-oasis farmlands in northwest China. The seasonal ETa values were 487, 574, 436, and 377 mm at the sites cropped with field maize, spring wheat, seed maize, and seed maize, respectively, in 2018. The daily rates from one site were validated against the eddy-covariance measurements, and the daily rates from all the sites were compared against the estimations from a simple water balance method, and both showed good consistency (R2 = 0.72-0.93). The inverse method has great potential for relatively accurate estimates of ETa in the middle and late growing stages but may induce a slight underestimation in the initial growing season. To determine the minimum monitoring depths of sensors for relatively accurate estimates, ETa values were estimated using soil moisture data at various observation depths, and were compared with eddy-covariance measurements and hypothetical ETa-observed depth curves. The results showed that inaccurate stratification of soil types in the profiles may cause great uncertainty in the estimates, and if texture-contrasting interlayers exist in the profiles, the lowest soil moisture sensor should be placed sufficiently deep, with all the interlayers fully included.