Abstract
Urbanization, climate change, and agricultural stress are increasing thedemand for precise and timely environmental monitoring. Land SurfaceTemperature (LST) is a key variable in this context and is retrieved fromremote sensing satellites. However, these systems face a trade-off betweenspatial and temporal resolution. While spatio-temporal fusion methods offerpromising solutions, few have addressed the estimation of daily LST at 10 mresolution. In this study, we present WGAST, a Weakly-Supervised GenerativeNetwork for Daily 10 m LST Estimation via Spatio-Temporal Fusion of TerraMODIS, Landsat 8, and Sentinel-2. WGAST is the first end-to-end deep learningframework designed for this task. It adopts a conditional generativeadversarial architecture, with a generator composed of four stages: featureextraction, fusion, LST reconstruction, and noise suppression. The first stageemploys a set of encoders to extract multi-level latent representations fromthe inputs, which are then fused in the second stage using cosine similarity,normalization, and temporal attention mechanisms. The third stage decodes thefused features into high-resolution LST, followed by a Gaussian filter tosuppress high-frequency noise. Training follows a weakly supervised strategybased on physical averaging principles and reinforced by a PatchGANdiscriminator. Experiments demonstrate that WGAST outperforms existing methodsin both quantitative and qualitative evaluations. Compared to thebest-performing baseline, on average, WGAST reduces RMSE by 17.18% and improvesSSIM by 11.00%. Furthermore, WGAST is robust to cloud-induced LST andeffectively captures fine-scale thermal patterns, as validated against 33ground-based sensors. The code is available athttps://github.com/Sofianebouaziz1/WGAST.git.