A High-resolution Estimation of Terrestrial Evapotranspiration from Landsat Images and its Applications in a Sparse Vegetation Region

    Accurate estimation of high resolutions of evapotranspiration (ET) is essential to study the variation of water resources in highly heterogeneous regions, but there is a severe paucity of ET products with high spatial resolution for long time series. This research improves the PML_V2 model to estimate a 30 m resolution monthly ET, called the PML_30 model. Furthermore, it is applied to estimate the monthly ET from 2000 to 2020 in the Yarkand Oasis. The method uses a linear transformation to harmonize remote sensing data from the Landsat-5 Thematic Mapper (TM), and the Landsat-7 Enhanced Thematic Mapper (ETM+) to the Landsat-8 Operational Land Imager (OLI), resulting in multi-source Landsat data with long time series. High spatial resolution and long-time series of leaf area index, land surface emissivity, and albedo are derived from the multi-source Landsat data to produce 30 m resolution ET products. The PML_30 model and PML_V2 models were compared to the regional water balance’s multi-year average ET of 380mm. The former is estimated at 344 mm with a relative error of 0.09, whereas the latter is at 304 mm with 0.2. At the point scale, the PML_30 model’s ET was compatible with the water consumption pattern of the related plant, and the variation in groundwater. The average annual ET for the Yarkand Oasis and its lower reaches is 343 mm/yr and 168 mm/yr, respectively. Between 2000 and 2015, the ET of the lower reaches increased by 2.86 mm/yr, but between 2016 and 2020, it decreased. The proposed PML_30 model is easily applicable to a larger scale with increased estimation accuracy and is well suited for areas with high heterogeneity such as areas with sparse vegetation cover.