A 30-m daily soil temperature dataset for alpine grasslands in central Tibetan Plateau during 2001–2024

This dataset provides 30-m daily mean soil temperature (ST) at -5 cm for the Nyainqentanglha Mountains region (central Tibetan Plateau) during 2001–2024. The dataset was generated using a satellite-based fusion and ground-based calibration method implemented in Google Earth Engine and R. Specifically, the fused 30-m daily LST was derived as the product of annual maximum LST from Landsat and the ratio of daily LST to annual maximum LST from MODIS, based on the assumption that the ratio of daily LST to annual maximum LST would not differ between 1-km and 30-m scales at a location. The fused LST was then converted to daily mean soil temperature at -5 cm using month-specific linear calibration models derived from ground observations at 10 weather stations (4300–5500 m) in the region during 2006–2010.The dataset covers the Nyainqentanglha Mountains and surrounding areas (29°30′–31°N, 90°–91°30′E) and is provided in an Albers Equal Area Conic projection (WGS 84). The dataset is distributed as quarterly multi-band GeoTIFFs (96 files, 24 years × 4 quarters, approximately 120 GB). Each file is named ST_<YEAR>_<QUARTER>.tif (e.g., ST_2020_Q1.tif) and contains one band per day in that quarter, with bands named ST_<DATE> (e.g., ST_20200101). Pixel values represent daily mean soil temperature at -5 cm, stored as signed 16-bit integers with a scale factor of 0.01. Units are degrees Celsius (℃), and the NoData value is -32768.Independent validation against ground observations indicates robust performance in both temporal and spatial evaluations (R2, 0.81–0.88; RMSE, 2.70–3.07 ℃; mean biases, 0.17–0.27 ℃), accurately tracking the seasonal soil temperature dynamics. Missing values may occur at very high elevations (typically > 5600 m), where persistent cloud/fog conditions can reduce the availability of valid Landsat observations and hinder the estimation of annual maximum LST. A unified calibration strategy applied across heterogeneous terrain may introduce elevation-dependent biases (slight underestimation at lower elevations and slight overestimation at higher elevations), although biases remain within an acceptable range for regional studies. In addition, transient snow accumulation and melt can alter the surface-subsurface thermal relationship and may contribute to increased errors during freeze-thaw periods.