Two improved shuttleworth-wallace models for estimating consecutive daily evapotranspiration

Abstract

Remote sensing-based evapotranspiration (ET) models have been widely employed to quantify regional ET. However, mapping consecutive daily ET with high accuracy and fine spatiotemporal resolution remains challenging. This study developed a pre-processing model (SWH-mTSF) and a post-processing model (SWH-ETrF) to enhance the temporal continuity of estimates from an improved Shuttleworth-Wallace (SWH) model. The models were applied to estimate ET in the Loess Plateau of northern Shaanxi (LPNS) from 2002 to 2009 at 1-day temporal and 1-km spatial resolution. The SWH-mTSF model enhanced the temporal resolution and quality of remote sensing-based LAI/FPAR using data assimilation, producing continuous daily ET estimates through physical mechanisms. Conversely, the SWH-ETrF model interpolates the SWH-simulated ET time series by using the alfalfa reference ET (ET_r) as a benchmark. Both models were comprehensively evaluated using flux measurements at the Changwu station and state-of-the-art products (MOD16A2 and PML-V2). Evaluation at Changwu station showed SWH-ETrF (r=0.83, RMSE=0.68 mm/day) outperformed SWH-mTSF (r=0.66, RMSE=0.86 mm/day) in simulating ET at daily and seasonal scales, effectively capturing annual ET fluctuations. Compared to PML-V2 (which overestimated ET) and MOD16A2 (which underestimated ET), the SWH-ETrF model provided more reliable ET estimates for cropland and grassland in the LPNS. Variance-based global sensitivity analysis revealed that meteorological variables had a stronger influence on ET estimates than physiological vegetation parameters (LAI/FPAR) in arid and semi-arid cropland regions. The SWH-ETrF model remained robust even with fewer ET measurements. These advantages make SWH-ETrF well-suited for regional and global applications due to its simplicity and efficiency.