Progress and Challenge in Statistically Downscaling Climate Model Outputs

Abstract

Statistical downscaling is a process to build up statistical relationships between large-scale (usually 1˚-3˚ on latitude and longitude) climate model outputs and point/watershed-scale meteorological variables. It is an important technique to conduct climate change impact assessment for a specific site or a watershed. This paper systematically reviewed the recent advances in three fields related to statistical downscaling methods: perfect prognosis, model output statistics, and stochastic weather generator. Merits and drawbacks associated with each downscaling method were summarized. In addition, the challenges in progressing statistical downscaling methods were stated, as well as the potential solutions. The contribution of this review is aimed at pointing out the direction of developing statistical downscaling methods and providing clues for climate change impact studies.

Projections of future precipitation and air temperature over the Tibetan Plateau based on Coupled Model Intercomparison Project Phase 6 multimodel ensembles

“…The TP is the highest plateau in the world, with an average altitude of more than 4,000 m. It has a complex terrain and unique plateau climate characteristics, which have a significant impact on global climate systems (Han et al, 2017;Li et al, 2018;Xu et al, 2019). The annual average air temperature on the plateau is approximately 0 C, and the highest average monthly temperature in most areas is <10 C (Duan et al, 2016;Li et al, 2016;Chen et al, 2016a;2016b). Most of the central and southeastern parts of TP are primarily controlled by the Indian Ocean Monsoon and have relatively abundant precipitation, whereas most of the western and northern parts controlled by the westerlies have less precipitation (Xie et al, 2018;Zhang et al, 2018).…”

“…For example, Su et al (2013) found that most of the 24 evaluated CMIP5 models could capture the spatiotemporal patterns of precipitation and temperature for the TP, while introducing a cold bias of approximately 1.1-2.5 C and an overestimation of 62-183% in the projected precipitation. Jia et al (2019) also demonstrated that the 33 CMIP5 models evaluated in their study could capture the seasonal temperature patterns; however, most of them tended to underestimate temperatures by approximately 2.0 C. Therefore, downscaling methods and multimodel ensembles (MMEs) are of great significance for correcting biases and reducing uncertainty, as well as for retaining the valuable climate change signals of GCMs (Wang et al, 2016;Ngai et al, 2020;Zhang et al, 2019;Wang et al, 2019a;Chen et al, 2016a;2016b;Liu et al, 2019). For example, projected the climate changes over China under different emission scenarios based on seven GCMs downscaled by four statistical downscaling methods.…”

“…(2019) also demonstrated that the 33 CMIP5 models evaluated in their study could capture the seasonal temperature patterns; however, most of them tended to underestimate temperatures by approximately 2.0°C. Therefore, downscaling methods and multimodel ensembles (MMEs) are of great significance for correcting biases and reducing uncertainty, as well as for retaining the valuable climate change signals of GCMs (Wang et al ., 2016; Ngai et al ., 2020; Zhang et al ., 2019; Wang et al ., 2019a; Chen et al ., 2016a; 2016b; Liu et al ., 2019). For example, Yang et al .…”