Creating 1-km long-term (1980–2014) daily average air temperatures over the Tibetan Plateau by integrating eight types of reanalysis and land data assimilation products downscaled with MODIS-estimated temperature lapse rates based on machine learning

Air temperature (Tair) is critical to modeling environmental processes (e.g. snow/glacier melting) in high-elevation areas of the Tibetan Plateau (TP). To resolve issue that Tair observations are scarce TP western part and at high elevation, many studies have estimated daily air temperatures by using MODIS land surface (LST) various reanalysis datasets. These estimates however inadequate for supporting high-resolution long-term hydrological simulations or climate analysis due cloud cover, short time span low spatial resolution. improve estimation, this study develops a novel machine-learning based method uses Gradient Boosting model efficiently integrate from stations with eight widely used assimilation datasets (i.e., NNRP-2, 20CRV2c, JRA-55, ERA-Interim, MERRA-2, CFSR, ERA5 GLDAS2) downscaled remote sensing-based lapse rates (TLR). This generate new dataset 1-km resolution period 1980–2014. overcome problem TLR derived limited may be unreliable, estimation developed first estimate spatially continuous monthly TLRs LST then downscale mean obtain MODIS-estimated TLRs. The (GB) selected integrating five other auxiliary variables. models trained validated 100 common (i.e. China Meteorology Administration stations) 13 independent (4 on glaciers). results show proposed can reduce exceptional meantime keeping acceptable downscaling accuracy. JRA-55 best among followed CFSR others. Finally, GB-integrated further outperforms root-mean-squared-deviation (RMSD) 1.7 °C versus 2.0 °C, especially RMSD 1.9 2.7 °C. Both training demonstrated significantly accuracy GB stations. also provides framework multiple data elevation correction mountainous regions not restricted TP.