High-resolution spatially explicit land surface model calibration using field-scale satellite-based daily evapotranspiration product

Abstract

High-resolution simulation of water budgets across the agricultural landscape is critically important to a variety of applications, such as precision agriculture, water resources management, and environmental quality assessment. Model-data integration has been shown to be an effective approach to reduce model uncertainties and there is a growing opportunity to improve land surface modeling through spatially explicit calibration with satellite data in recent decades. Recently, a satellite-based daily 30-m resolution evapotranspiration (ET) product BESS-STAIR has been developed, achieving a high performance and well capturing the spatial and temporal dynamics of ET across the U.S. Corn Belt. To explore the potential of high-resolution spatially explicit calibration for advancing land surface modeling at fine scales, we carried out calibration experiments for the Noah-MP land surface model (LSM) over cropland using this newly developed BESS-STAIR ET. We first used Sobol sensitivity analysis to identify the most sensitive parameters for the Noah-MP’s ET simulation. The most sensitive vegetation (minimum stomatal resistance) and soil parameters (saturated hydraulic conductivity, saturated matric potential, and a soil pore size distribution parameter) were calibrated using BESS-STAIR ET to improve model simulation of surface water balance. We conducted calibration experiments at 8 eddy covariance flux tower sites that grew maize and soybean across the U.S. Corn Belt, as well as a regional calibration study on the Spoon River watershed in Champaign, Illinois. When benchmarked with flux tower measurements, the BESS-STAIR ET–calibrated model (driven by flux tower forcing) on average reduced the RMSE of hourly ET from 61 W/m2 to 47 W/m2 for maize, and from 66 W/m2 to 53 W/m2 for soybean, and matched the performance of directly calibrating using flux tower measured ET. The regional study found that calibration using BESS-STAIR ET also improved the simulation of long-term regional water budgets and achieved better performance of ET than traditionally lumped calibration using streamflow. Further analysis revealed that the high-resolution calibration can resolve the spatial variations of ET to a certain extent, and the accuracy of the calibration can be largely attributed to the low bias and excellent long-term correlation of the BESS-STAIR ET data itself. Our study thus demonstrates the effectiveness of high-resolution model calibration and provides important implications in field-scale hydrological modeling and precision agricultural applications.