Does MODIS Water Vapor Matter? An Evaluation of Its Influence on Landsat 9 LST and Urban Thermal Patterns

Abstract

Land Surface Temperature (LST) is a crucial biophysical parameter in climate change analysis, urban heat island studies, and ecosystem dynamics. LST estimation from Landsat thermal sensors is highly sensitive to atmospheric water vapor characterization, where erroneous representation of this parameter can induce temperature biases of several Kelvin, particularly in tropical regions with high humidity. Daily MODIS water vapor products from Terra (MOD05_L2) and Aqua (MYD05_L2) are widely used as atmospheric correction inputs for Landsat LST retrieval; however, differences in overpass timing between the two sensors may produce distinct water vapor characteristics owing to diurnal atmospheric dynamics. This study aims to analyze the influence of water vapor derived from MODIS Terra and MODIS Aqua on Landsat 9-based LST estimation and its implications for urban thermal characterization in Sukasada, Buleleng, Bali. Methods employed include water vapor extraction using the MODIS near-infrared radiance ratio approach, and LST retrieval using the Split-Window Algorithm adapted from Skokovic et al., with NDVI Threshold-based emissivity estimation. Results revealed contrasting water vapor characteristics between MODIS Terra (mean: 0.612 g/cm², SD: 0.006) and MODIS Aqua (mean: 0.387 g/cm², SD: 0.054), reflecting distinct diurnal atmospheric dynamics. Nevertheless, both water vapor sources yielded statistically identical LST outputs, with a mean of 298.11 K and standard deviation of 3.98 K. Cross-validation against Landsat Level 2 LST products produced an RMSE of 1.46 K and R² of 0.96 for both scenarios. Future research is recommended to integrate vertical atmospheric transmissivity profiles derived from ERA5 reanalysis data and to develop locally calibrated Skokovic algorithm coefficients tailored to Bali’s tropical atmospheric characteristics.