WRF gray-zone dynamical downscaling over the Tibetan Plateau during 1999–2019: model performance and added value

Abstract

The Tibetan Plateau (TP) is an important component of the global climate system, while the characteristics of its climate are poorly represented in most regional climate models at coarse resolutions. In this study, a 20-year (2000–2019) dynamical downscaling simulation at the gray-zone resolution (9 km) using the WRF model driven by the ERA5 reanalysis is conducted over the TP. Based on comparisons against in-situ observations and the Integrated Multi-satellite Retrievals for GPM (IMERG) version 6 satellite precipitation product, the assessment of basic climate variables, such as near-surface air temperature (T2m) and precipitation, is performed to evaluate the model’s performance and understand its added value better. Results show that both WRF and ERA5 can successfully reproduce the spatial patterns of annual mean and seasonal mean surface air temperature. However, significant cold and wet biases are found especially over the southeastern TP in ERA5, which are greatly improved in WRF with reduced RMSEs. Not only the climatological characteristics, but also the inter-annual variability and seasonal variation of T2m and precipitation are well captured by WRF which reduces the cold and wet biases especially in winter and summer compared to ERA5, respectively. Besides, at daily scale, the overestimation of precipitation in WRF and ERA5 is mainly caused by the overestimated precipitation frequency when precipitation intensity changed slightly. Furthermore, WRF outperforms ERA5 in capturing the diurnal variation of precipitation with more realistic peak time in all sub-regions over the TP. Further investigation into the mechanism of model bias reveals that less simulated snow cover fraction plays a crucial role in increasing the surface net energy by affecting surface albedo over the southeastern TP in WRF, leading to higher T2m. In addition, less water vapor transport from the southern boundary of TP leads to reduced wet bias in WRF, indicating that the added value in dynamical downscaling at gray-zone resolution is obtained by representing water vapor transport more realistically.