第七届青年地学论坛

In data-sparse regions as the Arctic, atmospheric reanalysis is one of the key tools to understand the rapid climate change at regional and global scales. The utility of reanalysis datasets is affected by their accuracy and biases. Therefore, it is important to evaluate their performance. Here we conducted an inter-comparisons of two atmospheric variables, namely 2 m air temperature (Ta) and skin temperature (Ts), from the widely used ECMWF reanalysis datasets ERA-I and ERA5 against in situ observations from three international Arctic buoy programs (IABP, MOSAiC and CRREL) during 2010-2020 in the Arctic. Overall, results showed that both ERA-I and ERA5 agreed well with buoy observations, with the highest correlation coefficient reaching 0.98. On one hand, there is a general warm bias of Ta, 2.27±3.33°C from ERA-I and 2.34±3.22°C from ERA5, when compared with more than 3000 match-up pairs of daily buoy observations. The warm biases of Ta from both reanalysis showed seasonal variations, which reached a peak of 3.73±2.84°C in April, and a minimum of 1.36±2.51°C in September. For Ts, both ERA-I and ERA5 exhibited good consistency with buoy observations, but showing higher amplitude of biases compared with that of Ta, indicating general negative biases of -4.79±4.86°C for ERA-I and -4.11±3.92°C for ERA5. For both reanalysis methods, the largest bias of Ts (-11.18±3.08°C) occurred in December. In the warm summer, the bias is rather small (less than -3°C) during April to October. The cold biases of Ts from ERA-I and ERA5 were probably overestimated with values between -3.33°C and -4.51°C. The bias is affected by the vertical location of the surface temperature sensors on buoys. Biases of both Ta and Ts varied from different buoy programs, yet exhibited same trends.

Arctic; Buoy Obesrvations; Reanalysis; Temperature