Reconstruction of a daily gridded snow water equivalent product for the land region above 45° N based on a ridge regression machine learning approach

Abstract. The snow water equivalent (SWE) is an important parameter of surface hydrological and climate systems, it has a profound impact on Arctic amplification change. However, there are great differences among existing SWE products. In the land region above 45∘ N, products associated with limited time span spatial coverage, resolution coarse, which greatly limits application data in cryosphere change studies. this study, utilizing ridge regression model (RRM) machine learning algorithm, we integrated various to generate spatiotemporally seamless high-precision RRM product. results show that feasible utilize based algorithm prepare global scale. We evaluated accuracy product using hemispheric-scale course (HSSC) observational Russian survey data. mean absolute error (MAE), RMSE, R, R2 between observed SWEs 0.21, 25.37 mm, 0.89, 0.79, respectively. dataset improved by 28 %, 22 37 11 % compared original AMSR-E/AMSR2 (SWE), ERA-Interim SWE, Global Land Data Assimilation System (GLDAS) GlobSnow ERA5-Land datasets, respectively, higher resolution. production method does not rely heavily independent product; takes full advantage each dataset, into consideration altitude factor. MAE ranges from 0.16 for areas within <100 m elevation 0.29 800–900 range. best worst Canadian region. RMSE 4.71 mm 31.14 >1000 Finland This good stability, extremely suitable datasets large scales, can be easily extended preparation other datasets. expected provide more accurate support available “A Big Earth Platform Three Poles” (https://doi.org/10.11888/Snow.tpdc.271556) (Li et al., 2021).