On the Suitability of Current Atmospheric Reanalyses for Regional Warming Studies over China

Abstract

Reanalyses have been widely used because they add value to the routine observations by generating physically/dynamically consistent and spatiotemporally complete atmospheric fields. Existing studies have extensively discussed their temporal suitability in global change study. This study moves forward on their suitability for regional climate change study where land-atmosphere interactions play a more important role. Here, surface air temperature (Ta) from 12 current reanalysis products were investigated, focusing on spatial patterns of Ta trends, using homogenized Ta from 1979 to 2010 at ~2200 meteorological stations in China. Results show that ~80% of the Ta mean differences between reanalyses and in-situ observations are attributed to station and model-grid elevation differences, denoting good skill in Ta climatology and rebutting the previously reported Ta biases. However, the Ta trend biases in reanalyses display spatial divergence (standard deviation=0.15-0.30°C/decade at 1°×1°grids). The simulated Ta trend biases correlate well with those of precipitation frequency, surface incident solar radiation (Rs), and atmospheric downward longwave radiation (Ld) among the reanalyses (r=-0.83, 0.80 and 0.77, p<0.1) with their spatial patterns considered. Over southern China, the Ta trend biases (by order of -0.07°C/decade) are caused by the trend biases in Rs (by order of 0.10°C/decade), Ld (by order of -0.08°C/decade) and precipitation frequency (by order of -0.06°C/decade). Over northern China, the Ta trend biases (by order of -0.12°C/decade) jointly result from those in Ld and precipitation frequency. Therefore, improving simulation of precipitation frequency and Rs helps to maximize regional climate signal component. Besides, the Ta trend biases show negative spatial correlations (approximately r=-0.26, p=0.00) with inverted trend in NDVI (Normalized Difference Vegetation Index) implying that incorporating vegetation dynamics can advance regional warming modeling. Inclusion of accurate aerosol information in MERRA-2 (Modern-Era Retrospective Analysis for Research and Applications, version 2) helps improve regional climate simulation. ERA-20CM (a twentieth century atmospheric model ensemble without assimilating observations) presents a comparable pattern of the Ta trend biases (standard deviation=0.15°C/decade) to ERA-Interim and JRA-55 (the Japanese 55-year Reanalysis) that assimilating some Ta observations, which indicates perturbed physical ensemble technique significantly narrows regional warming uncertainties in reanalyses.