Evaluation of Eight Current Reanalyses in Simulating Land Surface Temperature from 1979 to 2003 in China

Abstract

Land surface temperature Ts provides essential supplementary information to surface air temperature, the most widely used metric in global warming studies. A lack of reliable observational Ts data makes assessing model simulations difficult. Here, the authors first examined the simulated Ts of eight current reanalyses based on homogenized Ts data collected at ~2200 weather stations from 1979 to 2003 in China. The results show that the reanalyses are skillful in simulating the interannual variance of Ts in China (r = 0.95) except over the Tibetan Plateau. ERA-Interim and MERRA land versions perform better in this respect than ERA-Interim and MERRA. Observations show that the interannual variance of Ts over the north China plain and south China is mostly influenced by surface incident solar radiation Rs, followed by precipitation frequency, whereas the opposite is true over the northwest China, northeast China, and the Tibetan Plateau. This variable relationship is well captured by ERA-Interim, ERA-Interim land, MERRA, and JRA-55. The homogenized Ts data show a warming of 0.34°C decade−1 from 1979 to 2003 in China, varying between 0.25° and 0.42°C decade−1 for the eight reanalyses. However, the reanalyses substantially underestimate the warming trend of Ts over northwest China, northeast China, and the Tibetan Plateau and significantly overestimate the warming trend of Ts over the north China plain and south China owing to their biases in simulating the Rs and precipitation frequency trends. This study provides a diagnostic method for examining the capability of current atmospheric/land reanalysis data in regional climate change studies.

On the suitability of current atmospheric reanalyses for regional warming studies over China

“…1 and Table 2). Similar biases in T a within various regions worldwide have been widely reported by previous studies (Mao et al, 2010;Pitman and Perkins, 2009;Reuten et al, 2011;Wang and Zeng, 2012;Zhou et al, 2017;Zhou and Wang, 2016b). However, we found that the spatial patterns in the differences in T a are well correlated with the elevation differences between models and stations, as reflected by correlation coefficients (r) of 0.85 to 0.94 (Figs.…”

“…In addition, we replicate the procedures above for the sparsely distributed stations over western China and the Tibetan Plateau. The PMTred algorithm and the QM adjustment have recently been used successfully to homogenize climatic time series (Aarnes et al, 2015;Tsidu, 2012;Dai et al, 2011;Siswanto et al, 2015;Wang and Wang, 2016;Zhou et al, 2017). As such, the significant breakpoints are detected and adjusted at a confidence level of 95 % at 1092 of the 2193 …”

“…This process is known as the greenhouse effect on T a and is quantified by L d . Existing studies have suggested that precipitation frequency better represents the interannual variability in soil moisture in China than the precipitation amount (Wu et al, 2012;Piao et al, 2009;Zhou et al, 2017;Zhou and Wang, 2017a); in turn, Figure 4 illustrates the partial relationships among the annual anomalies in T a and R s , the precipitation frequency and L d . The results show that T a is consistently positively correlated with R s (except over the Tibetan Plateau) and L d ; however, it is consistently negatively correlated with precipitation frequency in the observations and the 12 reanalysis products (Fig.…”

“…Extensive assessment studies have reported that most reanalyses display a certain level of performance in terms of their absolute values (Betts et al, 1996(Betts et al, , 1998Zhou and Wang, 2016b), interannual variability (Lin et al, 2014;Lindsay et al, 2014;Wang, 2016a, 2017a;Wang and Zeng, 2012), distributions (Gervais et al, 2014;Heng et al, 2014;Mao et al, 2010) and relationships among variables (Niznik and Lintner, 2013;Cash III et al, 2015;Zhou et al, 2017;Zhou and Wang, 2016b;Betts, 2004) over regions worldwide. However, these aspects of reanalyses still contain certain errors that restrict the general use of reanalyses, especially in climate applications.…”

“…The errors displayed by reanalysis products arise from three sources: observation error, model error and assimilation error (Thorne and Vose, 2010;Parker, 2016;Lahoz and Schneider, 2014;Dee et al, 2014;Zhou et al, 2017). Specifically, observation error incorporates systematic and random errors in instruments and their replacements, errors in data reprocessing and representation error, which arises due to the spatiotemporal incompleteness of observations (Dee and Uppala, 2009;Desroziers et al, 2005).…”