Dynamical downscaling of temperature extremes over China using the WRF model driven by different lateral boundary conditions

Highlights

• The WRF can skillfully reproduce the spatial distributions of extreme temperature indices.

• The accuracy of WRF in simulating interannual variation is consistent with that of the driving data.

• The WRF simulation driven by ERI appeared best in terms of the bias frequency distributions for most indices.

Abstract

Extreme temperature events have considerable impact on society and natural ecosystems, which receive less attention. This study examined the impact of lateral boundary conditions (LBCs) on the downscaling simulations of the Weather Research and Forecasting (WRF) model for temperature extremes over China for the period 1981–2010. The driving data comprised two reanalysis datasets from two different institutes, ERA-interim and NECP-R2. Twelve extreme temperature indices were calculated from the WRF simulations and compared with both reanalysis products and observations. The comparison includes their climate mean and interannual variation. Results indicated that WRF can skillfully reproduce the spatial distributions of extreme temperature indices, despite some biases in certain regions. The large biases in the warmest and coldest days, warm spell duration and cool days over eastern China from reanalysis data were reduced significantly. The simulation driven by ERA-interim outperforms NECP-R2. In terms of interannual variation, the WRF model was able to capture the observational trends except the number of days with warm spell and days with moderate warm. Furthermore, the accuracy of WRF in simulating interannual variation was consistent with that of the driving data. The WRF simulation driven by ERA-interim appeared best in terms of bias frequency distributions for most indices. Overall, the results revealed that WRF dynamical downscaling of temperature extremes is sensitive to selection of LBCs, and that performance could be improved by adopting driving data such as ERA-interim.

Introduction

Globally averaged near surface air temperature has increased by 1.09 °C in 2011–2020 than in 1850–1900. As in other regions of the world, China has recorded a substantial trend of warming with magnitude of 0.9–1.5 °C (IPCC, 2021). Certain climate extremes, including increases in extreme temperature events (such as hot days and heat waves), have been exacerbated and triggered by the increased warming (Chen and Li, 2017). The climate of China is influenced by the East Asian monsoon and complex terrain that make the region especially vulnerable to change in temperature extremes. Extreme temperature events have considerable impact on society and natural ecosystems. Therefore, it is essential to understand the climatic behavior of such events in China, and to improve their simulation by numerical models.

Global climate models (GCMs) are considered important tools for examining the past climate and future changes of climatic extremes. Temperature extremes in China have been investigated in many previous studies using the output of GCMs involved in the Coupled Model Intercomparison Project (CMIP) (Zhou et al., 2014; Chen and Sun, 2015; Dong et al., 2015; Sun et al., 2016; Guo et al., 2017; Luo et al., 2020; Wang et al., 2020). Dong et al. (2015) and Luo et al. (2020) indicated that CMIP5 models showed substantial biases in terms of the duration and frequency of certain extreme temperature indices. Although the spatial pattern of extreme temperature is captured better by CMIP6 models than by CMIP5 models, difficulties remain in capturing the features of cold nights and warm days when large cold biases are evident over the Tibetan Plateau (e.g. Luo et al., 2020). To some extent, these biases might occur because GCM resolution is coarser than the scale of such extreme events.

Regional climate models (RCMs) have been adopted widely for dynamical downscaling of GCM output, in which the resolution and the physical processes related to the impact of regional-scale regimes on climatic characteristics and extreme events can be refined (Gao et al., 2012; Liang et al., 2019). The skills and biases of RCMs in simulating extreme events including temperature extremes have been well demonstrated for China and East Asia (e.g., Ji and Kang, 2013, Ji and Kang, 2015; Park et al., 2016; Li et al., 2018; Wang et al., 2019). For example, Tang et al. (2017) examined the capability of the Weather Research and Forecasting (WRF) model in terms of simulating temperature extremes over the Coordinated Regional Climate Downscaling Experiment-East Asia (CORDEX-EA) domain, and confirmed that the spatial distribution of both the hottest days and the coldest nights can be simulated accurately by the WRF model. Kong et al. (2019) generated temperature extremes over China using the output of two RCMs, and reported on the capability of each in simulating the spatiotemporal variation of extreme indices. Yu et al. (2021) also found that RCMs are able to skillfully reproduce the spatial patterns and probability density functions of maximum and minimum temperatures.

Although RCMs demonstrate greater potential that GCMs in simulating temperature extremes, uncertainties can be introduced depending on the different lateral boundary conditions (LBCs). Systematic biases that are unavoidable in GCM or global reanalysis LBCs can be propagated through RCMs, introducing far reaching implications on the simulation products (Jones et al., 1995; Gong and Wang, 2000; Diaconescu et al., 2007). Biases induced by LBCs are generally larger than those of the initial conditions, and even larger than the deficiencies of the various parameterization schemes used to represent physical processes (Vukicevic and Errico, 1990; Szépszó, 2011; Chikhar and Gauthier, 2017). Yang et al. (2012) found that WRF simulations forced by different reanalysis data produce large differences in the mean circulation and interannual variation of seasonal precipitation in the China region. Oh et al. (2013) indicated that the simulation skills of RegCM4 for the extreme climate events defined by absolute thresholds are significantly different according to LBCs over South Korea. Subsequently, Yang et al. (2016) examined the sensitivity of the Laboratoire de Météorologie Dynamique Model with Zoom Capability (LMDZ4) in the simulation of mean temperature and its climatic trend over East Asia using three GCMs, and found that the simulation using Flexible Global Ocean-Atmosphere-Land System Model: Grid-point Version 2 (FGOALS-g2) for the LBCs produced the best results. Huang and Gao (2018) showed that the WRF simulation of precipitation and extreme precipitation was sensitive to the different reanalysis data boundaries over China. Most previous studies have focused on RCM sensitivities to LBCs on climatological surface air temperature and precipitation. However, the temperature extremes in China, where the climate of most regions is controlled primarily by the East Asia summer monsoon, have received less attention. Furthermore, the effects of LBCs on the interannual variability of extreme temperature events are also unclear.

In parallel with the aims of the CORDEX-EA, this study used the WRF model to conduct dynamical downscaling of different reanalysis data for the period 1981–2010. The two sets of reanalysis data used were the National Centers for Environmental Prediction Department of Energy global reanalysis (NECP-R2) and the European Centre for Medium-Range Weather Forecasts (ECMWF) Interim reanalysis (ERA-interim), both of which are used widely for providing LBCs. The overall aim of this article is to examine the impact of different LBCs on the downscaling simulations of temperature extremes in terms of climatology pattern and interannual variation over China. The remainder of this paper is organized as follows. The data and experimental set up are described in Section 2. Section 3 presents the extreme temperature indices considered in this study. The results of the simulation of the temperature extremes over China are presented in Section 4. Finally, Section 5 provides our main conclusions.