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

Abstract

Mean annual air temperature (MAAT) values for the historic (1961 – 1990) and modern (1991 –2020) reference climatic periods, collected from 91 meteorological stations across the Sakha (Yakutia) Republic were used in an evaluation of the performance of several modern reanalysis products. The model error analysis shows GHCN-CAMS to be the best-performing MAAT reanalysis product, with the lowest absolute error, below 0.15°С, and best concordance with observation data. This product can be used as a ‘baseline climate’ in geospatial modeling and future climate forecasts. Its main limitation is monthly temporal resolution. The ERA5-Land product, having higher spatial and temporal resolution, can be used to track changes in climate-related variables between periods, but its absolute MAAT values deviate systematically and non-concordantly from the observed data. Parameters of the linear regression relating ERA5-Land data to observations are shown to be time-dependent, hence such evaluation yields increased uncertainty limiting the applicability of this approach. All reanalysis products underestimate the observed MAAT increase at stations across central Yakutia, which suggests the appearance of stable ‘urban heat islands’ in the recent decades.