Liu et al. (2026) High-Resolution Daily Evapotranspiration Estimation in Arid Agricultural Regions Based on Remote Sensing via an Improved PT-JPL and CUWFM Fusion Framework

Identification

• Journal: Remote Sensing
• Year: 2026
• Date: 2026-01-15
• Authors: Hongwei Liu, Xiaoqin Wang, Hongyu Zhang, Mengmeng Li, Qunyong Wu
• DOI: 10.3390/rs18020291

Research Groups

Key Lab of Spatial Data Mining & Information Sharing of Ministry of Education, Academy of Digital China (Fujian), Fuzhou University, Fuzhou 350108, China

Short Summary

This study developed an improved Priestley–Taylor Jet Propulsion Laboratory (PT-JPL) model using the Land Surface Water Index (LSWI) and a novel Crop Unmixing and Weight Fusion Model (CUWFM) to generate daily 30 m resolution evapotranspiration (ET) data in arid agricultural regions, demonstrating superior accuracy and spatial detail compared to existing methods.

Objective

• To develop and validate a high spatiotemporal resolution evapotranspiration (ET) estimation method suitable for arid and semi-arid agricultural regions.
• To improve the Priestley–Taylor Jet Propulsion Laboratory (PT-JPL) model by replacing the relative humidity-based soil moisture constraint with the Land Surface Water Index (LSWI) to enhance performance in water-limited environments.
• To construct a Crop Unmixing and Weight Fusion Model for ET (CUWFM) that integrates crop fraction decomposition, phenology-aware base date selection, hierarchical similar-pixel screening, and similarity-weighted temporal prediction to generate daily 30 m ET products.
• To systematically validate the proposed methods against ground observation data and compare them with existing mainstream ET products and spatiotemporal fusion algorithms.

Study Configuration

• Spatial Scale: Changji Hui Autonomous Prefecture, Xinjiang, China (43.33°N to 45.00°N, 85.28°E to 91.53°E). Output ET data at 30 m spatial resolution. Validation at Daman station (3 × 3 pixel average for 30 m ET) and Changwu station (single 30 m pixel).
• Temporal Scale: Daily ET estimation. Study period for Changji Prefecture: crop growing season from 30 March to 16 October 2020. Validation periods: 2020–2023 for Daman station, and 2021–2023 for Changwu station.

Methodology and Data

• Models used:
  • Improved Priestley–Taylor Jet Propulsion Laboratory (PT-JPL) model.
  • Crop Unmixing and Weight Fusion Model for ET (CUWFM).
  • Benchmark spatiotemporal fusion methods for comparison: Spatial and Temporal Adaptive Reflectance Fusion Model (STARFM), Enhanced Spatial and Temporal Adaptive Reflectance Fusion Model (ESTARFM), and function fitting-based spatiotemporal fusion method (Fit-Fc).
  • Other ET products for comparison: PML-V2 and MOD16.
• Data sources:
  • Satellite Data: Landsat 8 and Sentinel-2 (for Normalized Difference Vegetation Index (NDVI), Land Surface Water Index (LSWI), surface albedo, Fractional Vegetation Cover (FVC), Leaf Area Index (LAI)).
  • Reanalysis Data: Fifth-generation European Centre for Medium-Range Weather Forecasts (ECMWF) atmospheric reanalysis (ERA5) (0.25° × 0.25° daily resolution) for relative humidity (RH), mean temperature (Ta), atmospheric pressure (Pa), solar shortwave radiation (Rs), downward longwave radiation (RL↓), and upward longwave radiation (RL↑).
  • ET Products: PML-V2 ET dataset (500 m spatial resolution, daily) from the National Tibetan Plateau Data Center.
  • Auxiliary Data: 2020 crop distribution map of Changji Prefecture (from "Plot-level Cropping Structure Dataset Based on Sentinel Images and Phenology Information"), crop phenological dates (field surveys, local agricultural archives).
  • Ground Observations:
    • Daman station (HeiHe Watershed Allied Telemetry Experimental Research (HiWATER) network): Eddy covariance observations (daily 30 min latent surface heat flux, converted to daily ET).
    • Changwu station (Changwu National Field Scientific Observation and Research Station for Agricultural Ecosystem in Shaanxi Province): Large weighing lysimeter observations (daily ET).

Main Results

• Improved PT-JPL Model Performance: The improved PT-JPL model, using LSWI instead of relative humidity for soil moisture constraint, significantly enhanced ET estimation accuracy. At Daman station (maize), R² increased to 0.78 (from 0.74), NSE to 0.77 (from 0.71), and RMSE decreased to 1.06 mm/day (from 1.20 mm/day), with PBIAS improving from -7.99% to 0.07%. At Changwu station (winter wheat), R² increased to 0.80 (from 0.70), NSE to 0.69 (from 0.62), and RMSE decreased to 0.74 mm/day (from 0.82 mm/day). The improved PT-JPL-ET also outperformed PML-V2 (correlation coefficient 0.83, RMSE ≈ 1.49 mm/day) and MOD16 (correlation coefficient 0.78, RMSE ≈ 1.73 mm/day).
• CUWFM Fusion Performance: The CUWFM successfully generated daily 30 m ET data with high pixel-level accuracy and superior spatial detail. Validation against ground observations at Daman station for July 2020 yielded an R² of 0.770 and an RMSE of 0.716 mm/day. In comparative spatial pattern recognition against STARFM, ESTARFM, and Fit-Fc, CUWFM-ET achieved the best performance across all All-round Performance Assessment (APA) metrics (RMSE = 0.174 mm/day, AD = -0.051 mm/day, Edge = -0.072, LBP = -0.006), demonstrating an accuracy improvement of 15–65%.
• Crop-Specific ET Dynamics: The fused CUWFM-ET data accurately captured distinct ET dynamics for major crops in Changji Prefecture. Winter wheat ET peaked at approximately 5 mm/day during heading–flowering (late May to early June). Maize ET exhibited a single broad summer peak of approximately 5–6 mm/day during the tasseling–grain filling stage (July–August). Cotton ET showed two pronounced high-value periods, reaching approximately 5–6 mm/day, corresponding to flowering and boll development stages.

Contributions

• Developed an improved PT-JPL model by redefining the soil evaporation constraint using the Land Surface Water Index (LSWI), significantly enhancing its performance and physical consistency in arid and semi-arid regions.
• Introduced a novel Crop Unmixing and Weight Fusion Model (CUWFM) that integrates sub-pixel crop fraction decomposition, phenology-aware base date selection, hierarchical similar-pixel screening, and similarity-weighted regression.
• Generated daily 30 m spatial resolution ET data that exhibits superior accuracy, spatial detail, and texture preservation compared to mainstream spatiotemporal fusion methods (STARFM, ESTARFM, Fit-Fc).
• Provided a robust framework for producing high spatiotemporal resolution ET datasets crucial for field-scale irrigation scheduling and precision water resource management in arid and semi-arid agricultural regions.

Funding

Science and Technology Program of Fujian Province of China, grant number 2023I0007.

Citation

@article{Liu2026HighResolution,
  author = {Liu, Hongwei and Wang, Xiaoqin and Zhang, Hongyu and Li, Mengmeng and Wu, Qunyong},
  title = {High-Resolution Daily Evapotranspiration Estimation in Arid Agricultural Regions Based on Remote Sensing via an Improved PT-JPL and CUWFM Fusion Framework},
  journal = {Remote Sensing},
  year = {2026},
  doi = {10.3390/rs18020291},
  url = {https://doi.org/10.3390/rs18020291}
}