Selection of the best-performing climate reanalysis model for the Republic of Sakha (Yakutia) based on mean annual precipitation

Regional scenario-based projections require a comprehensive understanding of baseline climatic conditions, particularly the spatial distribution of total mean annual precipitation within the region. Precipitation data from 40 meteorological stations across the Republic of Sakha (Yakutia) for the period from 1961 to 2020 were used to evaluate the performance of modern reanalyses: CRU TS, ERA5-Land, GPCC, NCEP-NCAR, PREC/L, and JRA55. The performance of each model was assessed by comparing the observed mean annual precipitation to the reanalysis field values in pixels corresponding to the locations of the observation points. The statistical assessment employed Lin’s coefficient of concordance, Wilmott’s index of agreement, Kendall’s tau, and the root mean square error. Interpolation-based models (CRU TS, GPCC, PREC/L) demonstrated a superior ability to reproduce observed total precipitation, whereas modeling-based reanalyses tended to overestimate it by more than 100 mm/year, or by 30% to 50%. The GPCC reanalysis exhibited the best performance when compared to observations; however, it appeared to be significantly overfitted, as evidenced by a substantial negative spatial correlation between total precipitation coverages for the periods from 1961 to 1990 and 1991 to 2020. Consequently, the interpolation uncertainty associated with overfitting precludes the use of GPCC data as a reliable benchmark. Ultimately, the CRU TS 4 reanalysis was determined to be optimal as a baseline for total precipitation coverage. According to CRU TS 4 data, the mean annual precipitation across the Republic of Sakha (Yakutia) was 285 ± 81 mm for the period from 1961 to 1990 and 293 ± 92 mm for 1991 to 2020, indicating an insignificant change of 8 ± 18 mm. Thus, between the two climatic periods, the annual precipitation in the Republic of Sakha (Yakutia) increased by 8 ± 18 mm, a change that is not statistically significant.