This dataset was generated using automatic machine learning algorithms, integrating multiple data sources including the International Soil Moisture Network and soil moisture observation data from the China Meteorological Administration. It provides daily soil moisture data at a resolution of 0.1 degrees over five layers on the Qinghai-Tibet Plateau from 2000 to 2021. Three versions of the data were produced using different algorithms: AMSMQTP_base, AMSMQTP_best, and AMSMQTP_ensemble, stored in three separate folders. Each folder contains data for five layers, representing depths from layer 1 to layer 5, corresponding to 0-10cm, 10-30cm, 30-50cm, 50-80cm, and 80-110cm, respectively. The files are named following the format "%Y_%m.nc", where "%Y" represents the year (2000-2021), and "%m" represents the month (01-12).
The soil moisture observation data is sourced from the International Soil Moisture Network and the China Meteorological Administration. Meteorological data used for model inputs are obtained from ERA5-Land, surface data from ERA5-Land, GLDAS-2.1, and reprocessed MODIS V6.1, soil texture data from GSDE, and terrain data from GTOPO30.
This data is based on the FLAML automatic machine learning framework, utilizing multiple data sources as model inputs. It is trained based on soil moisture observations from monitoring stations, and soil moisture data is obtained through model inference.
AMSMQTP_ensemble performs the best among the three versions, with the following evaluation metrics for soil moisture across five layers on the Qinghai-Tibet Plateau: (R=0.73, RMSE=0.074, ubRMSE=0.047); (R=0.665, RMSE=0.069, ubRMSE=0.042); (R=0.433, RMSE=0.08, ubRMSE=0.045); (R=0.503, RMSE=0.077, ubRMSE=0.035); (R=0.351, RMSE=0.084, ubRMSE=0.038). AMSMQTP_base, AMSMQTP_best, and AMSMQTP_ensemble overall exhibit lower soil moisture errors compared to GLDAS-2.1, ERA5-Land, SMCI1.0_9km, and GSSM1km.
This work is licensed under
CC BY 4.0 (Creative Commons Attribution 4.0 International License).
CSTR:11738.11.NCDC.NIEER.DB6463.2024
10.12072/ncdc.nieer.db6463.2024
