2026-10-13T21:41:26Z https://www.scidb.cn/oai

10.57760/sciencedb.00zzg 2026-09-03T09:26:54Z

2026-09-03 ChinaTherm30: a 25-year record of annual 30 m daytime land surface temperature composites during the leaf-on season across China doi:10.57760/sciencedb.00zzg en ChinaTherm30 is a 25-year record of annual 30 m daytime land surface temperature (LST) composites across China during the leaf-on season from 2000 to 2024. The annual thermal target was derived from the Terra MODIS MOD11A2 Version 6.1 daytime LST product. Eight-day composites with nominal start dates between day of year (DOY) 150 and 300 were retained, pixels with Mandatory QA bits of 00 or 01 were accepted, and the available eight-day LST values were combined using a pixel-wise annual median.A single cross-year random forest model related the annual MODIS thermal targets to 30 m leaf-on NDVI, WET and NDBSI derived from the China Earth Observation Data Cube, elevation and slope from Copernicus DEM GLO-30, and geographic coordinates. The model first estimated the annual 30 m temperature pattern. Differences between this estimate aggregated to 0.01° and the MODIS annual target were spatially smoothed using a normalized Gaussian filter with a standard deviation of 10 grid cells. The smoothed field was then added to the 30 m estimate to form the released ChinaTherm30 product.The dataset contains 2,275 single-band GeoTIFF files, comprising 25 years and 91 regular 4° × 4° tiles per year. The ScienceDB release is organized into 25 annual TAR archives, accompanied by an English README and data dictionary. The GeoTIFFs use EPSG:4326 with a pixel spacing of approximately 0.0002695°. LST is stored as Int16 with a scale factor of 0.01 °C and a NoData value of −32768.ChinaTherm30 represents the daytime thermal background during the defined annual leaf-on window. It should not be interpreted as an instantaneous observation, daily mean, nighttime temperature or full-year mean LST. The scientific domain is the land area of China. The files retain regular rectangular tile geometry for direct mosaicking, so users should apply an appropriate national boundary and land mask when calculating land-area statistics. The dataset supports long-term comparisons and cross-scale analyses of land-surface thermal conditions across China. land surface temperature; remote sensing; statistical downscaling; random forest Erzheng Liang PUBLIC https://creativecommons.org/licenses/by/4.0/ dataset Science Data Bank