Accuracy assessment of temperature trends from ERA5 and ERA5-Land

Abstract

Many environmental processes and ecological systems are being affected by the warming temperatures as a result of climate change. To correctly identify and attribute the uncertainty estimates in these systems, an investigation of tem-perature trend signal existing in the datasets that are used to study such systems is necessary. In this study, the trend of widely used ERA5 and ERA5-Land temperature estimates between 1951 and 2020 were validated using temperature trends from ground station-based observations in Turkey. The investigation included datasets obtained over 540 stations, and the seasonality and spatio-temporal variability of the trend accuracy was also investigated. On average, the trends of observations over all stations were 0.82 degrees C/decade and 0.30 degrees C/decade for the periods 2001-2020 and 1951-2020, respectively, indicating strong evidence of climate change. When the model datasets used spatially and temporally continuous datasets, the trends identified were 0.91 degrees C/decade and 0.21 degrees C/decade over the entire Turkey for the years 2001-2020 and 1951-2020, respectively. The difference in the 70-year trends of the two esti-mates was attributed to the missing datasets in observations. The differences between the trends of model estimates and observations were higher for the first decade than for the last two decades, stressing the impact of improved model estimates over time. All products showed heavy seasonality, suggesting that winter trends (1.3 degrees C/decade on average) are much higher than the summer (0.3 degrees C/decade) between 2001 and 2020. The results indicated a high degree of consistency between the trends of ERA5/ERA5-Land and observations, implying they may be used as a replacement to observations.