Summer climate in Asia is complex with a great variety of thermodynamic and dynamic processes occurring simultaneously and often interacting with each other. Examining its spatial variability including, for example, how 2020 compares to other years, is thus surprisingly difficult, as evidenced by the great variety of methods used in the literature. Examples include indices, such as the Wang-Fan index (Wang and Fan, 1999) or the East Asian Summer Monsoon Index (Wang et al., 2008) which are often used to quantify monsoon strength. These are usually single numbers, rather than spatial fields and are consequentially often too blunt. Clark et al. (2021) showed, for example, that specific values of the Wang-Fan index can occur during days with a wide variety of circulation patterns. Alternative spatial fields, such as those of the variability of surface pressure, geopotential height, and wind velocity can be difficult to interpret in terms of impacts and are rarely spatially coherent.
To handle these types of issues, we use the approach of clustering as described in Clark et al. (2021) in which daily June, July, and August fields of sea level pressure from the ECMWF ERA5 (Hersbach et al., 2020) reanalysis are combined into eight collections (hereafter “clusters”) using the k-means clustering technique. With new data now available from 2019 and 2020, the clusters of Clark et al. (2021) have been updated, using ERA5 output from 1979 onwards. Throughout this article, fields from ERA5 will be used as a representation of the observations of 2020 due to their spatial coverage and corresponding associated fields, for which gridded observations (of moisture fluxes, for example) are not yet readily available.
To examine how summers like 2020 might play out in a future warming world, we use model simulations from two ensembles of fully coupled ocean-atmosphere models. The first (denoted PPE), consists of a collection of 15 perturbed physics variants of the Met Office HadGEM3-GC3.05 model (Yamazaki et al., 2021) driven by observed greenhouse gas concentrations up to 2005 and following the (Relative Concentration Pathway) RCP8.5 (Moss et al., 2008) scenario to 2100. The second ensemble consists of simulations contributed to the CMIP6 (Eyring et al., 2016) archive, following a (Shared Socioeconomic Pathway) SSP5-8.5 (O'Neill et al., 2016) pathway, very similar to that of RCP8.5. Both are representative of a world of future rapid economic development and convergence. CMIP6 models, whose simulations were used for the analysis (based on data available at the time of doing the analysis) are shown in Table 1. Models were divided according to their resolution to gain an insight into whether their resolution had any effect on results. A single realization from each model was used.
Horizontal resolution Model names Total number of models Sea level pressure (SLP) Greater than approximately 250 km CNRM-CM6-1-HR, NorESM2-MM, EC-Earth3, GFDL-CM4, CMCC-CM2-SR5, TaiESM1, MRI-ESM2-0, BCC-CSM2-MR, CESM2-WACCM, CNRM-CM6-1, CNRM-ESM2-1, GFDL-ESM4, CESM2, HadGEM3-GC31-MM, EC-Earth3-Veg 15 Approximately 250 km NorESM2-LM, INM-CM5-0, MPI-ESM1-2-LR, IPSL-CM6A-LR, FGOALS-g3, NESM3, MIROC6, ACCESS-CM2,ACCESS-ESM1-5,UKESM1-0-LL 10 Precipitation Greater than approximately 250 km As for SLP, but without CMCC-CM2-SR5, TaiESM1, CESM2-WACCM, CESM2, HadGEM3-GC31-MM, EC-Earth3-Veg 9 Approximately 250 km As for SLP, but without INM-CM5-0 9
Table 1. CMIP6 models contributing to Fig. 6 and supplementary Figs. S6 and S7.