Annual 30 m leaf-on daytime land surface temperature composites across China from 2000 to 2024

Erzheng Liang , Luoma Wan, Yaotong Cai, Feiyi Gao, Ziheng Zhu

Long-term, spatially detailed land surface temperature (LST) records are needed to compare thermal environments across heterogeneous landscapes, but combining fine spatial detail with nationwide coverage over decades remains challenging. We present Chinatherm30, annual 30 m leaf-on daytime LST composites across China for 2000--2024. Quality-screened Terra MODIS thermal targets were combined with annual Landsat optical indices, topography and geographic coordinates in a cross-year random forest, followed by low-frequency residual adjustment. Spatial-block validation of the random forest yielded a correlation of 0.904 and an RMSE of 3.967°C. The median annual nationwide correlation with Landsat 8 LST during 2013--2024 was 0.864. Ground evaluation across 23 sites and 131 site-years, predominantly in the Heihe basin, yielded a correlation of 0.840 and an RMSE of 5.397°C. Independent unmanned aerial vehicle thermal imagery further evaluated local spatial patterns. The record supports long-term comparisons of leaf-on daytime thermal environments in urban, agricultural and vegetated landscapes.