Coupling of VIIRS and Sentinel-2 observations to generate Landsat-scale TIR imagery for estimating landscape evapotranspiration using the Simplified Surface Energy Balance (SSEB) model

: Satellite thermal infrared (TIR) data are essential for remote sensing of energy exchange processes at the Earth-atmosphere interface. Current satellite sensors are subject to a trade-off between spatial resolution and temporal coverage. Daily TIR observations from MODIS and VIIRS are limited to coarse spatial resolutions of ∼400–1000 m, whereas higher-resolution Landsat TIRS observations (∼100 m) provide infrequent temporal coverage (∼8 days). Downscaling methods that correlate TIR imagery with higher-resolution multispectral data can improve the availability of fine-scale TIR observations. Combining VIIRS I5 TIR (∼400 m) and Sentinel-2 multispectral (∼10–60 m) imagery is promising for enhancing the coverage of Landsat-level TIR observations, as Sentinel-2 provides complementary multispectral data without TIR sensors, but has largely been overlooked to date. We therefore evaluated a Random Forest-based downscaling framework to generate 100 m TIR observations from VIIRS/Sentinel-2 data and tested the implementation of downscaled LST to estimate ET rates using the widely applied Simplified Surface Energy Balance (SSEB) model. We found that downscaling could produce TIR imagery with spatial detail comparable to Landsat reference data (R 2 = 0.80–0.91, RMSE = 0.77–1.28 K). VIIRS/Sentinel-2 ET estimates were in excellent agreement with Landsat reference ET, as implementing downscaled LST in the SSEB model buffered uncertainty in the TIR data (RMSE = 0.11–0.29 mm/day, R 2 = 0.90–0.95). Our findings highlight the potential of VIIRS/Sentinel-2 downscaling to bridge the spatial–temporal resolution trade-off in thermal remote sensing and demonstrate the effective integration of downscaled VIIRS LST data within the SSEB ET model.