Assessment of ERA5 Near-Surface Air Temperatures Over Global Oceans by Combining MODIS Sea Surface Temperature Products and In-Situ Observations
Min He, Jun Qin, Ning Lu, Ling Yao
Abstract
ERA5 is a state-of-art reanalysis dataset and has been widely used in climate change analysis and land surface process simulations. However, few assessments have been conducted across the entire ocean for this dataset till now. The motivation of this study is to utilize the in-situ observations and remotely-sensed sea surface temperature (SST) products to propose a new approach to evaluate the precision of ERA5 air temperature over the whole ocean. Firstly, a stacked Multi-layer Perceptron regressor (MLP) model was employed to map global marine air temperature (MAT) by incorporating in-situ records into MODIS SST products. Secondly, a monthly MAT dataset with a spatial resolution of 0.05°×0.05°, was developed and validated for the assessment of ERA5 air temperature particularly in the regions where only satellite observations are available. It demonstrates that the quality of estimated MAT (RMSE = 1.62℃, Bias = 0.03℃) is evidently high in 80.3% of the observation stations, which is much better than that of averaged SST (RMSE = 2.29℃, Bias = 1.14℃). Thus, the estimated MAT is credible enough to be employed to assessing ERA5 air temperature both at the grid and regional scale. Finally, the ERA5 air temperature was compared and validated with the estimated MAT temporally and spatially. The further study suggested that estimated MAT and ERA5 air temperature basically maintain consistency in expressing the persistent warming particularly in the western Pacific, western Atlantic, and northern Indian Ocean. Additionally, the trend of ERA5 air temperature during 50°S∼50°N reaches to 0.016±0.003°C/yr, which is similar to that of estimated MAT (0.017±0.002℃/yr) during 2003−2021.