Observational datasets used in this study include: SST from the NOAA 1/4° monthly Optimum Interpolation Sea Surface Temperature (OISST) version 2.1 (Reynolds et al., 2002), precipitation from both the gridded monthly station precipitation in China (Version 2.0, Zhao et al., 2014) and the CPC Merged Analysis of Precipitation (CMAP; Xie and Arkin, 1997), and pressure level wind, geopotential height and relative humidity data from the NCEP-NCAR Reanalysis-1 (Kalnay et al., 1996). Monthly anomalies were obtained by subtracting the monthly mean climatology for the period 1983−2012. We adopted the East Asian summer monsoon index (EASMI, Wang et al., 2008) to represent the large-scale summer monsoonal circulation, which is defined by the difference of 850 hPa zonal wind (U850) between (5°−15°N, 90°−130°E) and (22.5°−32.5°N, 110°−140°E).
The fully coupled ocean-atmosphere model used in the study is the Climate Forecast System version 1.0 of Nanjing University of Information Science and Technology (NUIST CFS1.0, He et al., 2020), which is developed from the coupled model SINTEX-F (Luo et al., 2003, 2005a, b; Masson et al., 2005). The atmospheric component is the latest version of ECHAM4 with a high horizontal resolution (T106) of about 1.1°×1.1°. A hybrid sigma-pressure vertical coordinate (19 levels in total) is used with the highest vertical resolution near the earth’s surface. The ocean component is the reference version 8.2 of OPA with the ORCA2 configuration, and the horizontal resolution is 2° (longitude) × 2° cosine (latitude) with meridional resolution increased to 0.5° near the equator. The ocean model has 31 vertical levels, 19 of which are in the upper 400 meters. The coupling variables (i.e., water, heat, and momentum flux, and SST) without flux correction are exchanged every two hours between the OPA and ECHAM4 models by means of the OASIS 2.4 coupler (Valcke et al., 2000).
To evaluate the model’s performance in predicting the East Asian summer monsoon and precipitation over China, ensemble hindcasts initiated from 1 March of each year during 1983−2017 were performed. The hindcasts consist of 9-member ensemble predictions generated by three different initial conditions for each of three model versions with modified coupling physics (Luo et al., 2008). Only the observed weekly NOAA OISST values are assimilated into the coupled model to generate realistic and well-balanced initial conditions required for the hindcasts.
To explore the possible impacts of SST anomalies in the tropical oceans on the MLYRV precipitation, a set of ensemble sensitivity experiments were conducted. To obtain robust results, we added another 18 ensemble members in addition to the 9 members used in the hindcast experiments. As for the additional 18 members, the initial conditions for the NUIST CFS1.0 were obtained by assimilating the 6-hourly zonal and meridional wind, air temperature and surface air pressure of the JRA-55 reanalysis data (Kobayashi et al., 2015) into the atmosphere model and NOAA OISST data into the ocean model in a coupled manner. Similarly, the additional 18 members were generated by various combinations of different initial conditions and modified coupling physics.
To evaluate the relative contributions of SST anomalies in different tropical oceans to the MLYRV precipitation, we conducted five groups of ensemble experiments from 1 May to 31 July. Each group consists of two sensitivity experiments. In the first group of the experiments, we specified the observed monthly climatological SST of 1983−2012 in one sensitivity experiment (EXP_IOCLM) in the tropical Indian Ocean (IO, 20°S−20°N, 50°−100°E), and keep elsewhere free ocean-atmosphere coupling. In the other sensitivity experiment (EXP_IOOBS), we specified the observed monthly SST values during April-August 2020 in the IO region, and keep elsewhere free ocean-atmosphere coupling. Thus, the differences between the two sensitivity experiments (EXP_IOOBS-minus-EXP_IOCLM) represent the impacts of the Indian Ocean SST anomalies on the MLYRV precipitation.
Similarly, in the second, third, and fourth group of the sensitivity experiments, we specified the observed monthly climatological SST in the Maritime Continent (MC, 10°S−10°N, 100°−130°E, EXP_MCCLM), central and eastern equatorial Pacific (CEP, 5°S−5°N, 180°−80°W, EXP_CEPCLM), and North Atlantic Ocean (NAT, 0°−25°N, 60°−15°W, EXP_NATCLM), respectively, and keep elsewhere free ocean-atmosphere coupling. And in the counterpart experiments, we specify the observed monthly SST values during April-August 2020 in the MC (EXP_MCOBS), CEP (EXP_CEPOBS), and NAT (EXP_NATOBS), respectively, and keep elsewhere free ocean-atmosphere coupling. Thus, the differences between EXP_MCOBS and EXP_MCCLM (EXP_MCOBS-minus-EXP_MCCLM), between EXP_CEPOBS and EXP_CEPCLM (EXP_CEPOBS-minus-EXP_CEPCLM), and between EXP_NATOBS and EXP_NATCLM (EXP_NATOBS-minus-EXP_NATCLM) represent the impacts of the SST anomalies in the MC, CEP, and NAT on the MLYRV precipitation, respectively. In addition, we conducted a fifth group of the sensitivity experiments, in which we specified observed monthly climatological SST in one experiment (EXP_ALLCLM) and observed monthly SST values in the other experiment (EXP_ALLOBS) in the above four regions. And the differences between the two experiments (EXP_ALLOBS-minus-EXP_ALLCLM) represent the combined impacts of the SST anomalies in the above four regions on the MLYRV precipitation. The groups of the sensitivity experiments are summarized in Table 1.
Indian Ocean (20°N−20°S, 50°−100°E) Maritime Continent
Central and Eastern
North Atlantic (0°−25°N,
Sum of these four regions CTL climatological SST (1983−2012) SEN observational SST
Table 1. Experiments conducted to investigate the relative contributions of SST anomalies in different tropical oceans to the precipitation anomaly in June−July 2020 over the MLYRV Ensemble simulations (27 members) are conducted starting from 1 May 2020
High-quality observations have shown that the tropical IO has experienced rapid surface warming over the past few decades (Luo et al., 2012; Roxy et al., 2014). At the same time, the June−July averaged IO SST anomaly in 2020 reached its highest value in the last 40 years (Fig. S2a in the ESM). In order to examine the respective role of the multi-decadal warming trend and interannual variations of the IO SST in the 2020 summer MLYRV extreme precipitation, two additional ensemble prediction experiments are conducted. In the first model simulations from 1 May to 31 July, we specified observed monthly climatological SST plus detrended SST anomalies in the tropical IO, and kept other regions fully coupled (EXP_IODET). Thus, the differences between EXP_IODET and EXP_IOCLM (EXP_IODET-minus-EXP_IOCLM) represent the impacts of interannual anomalies of the tropical IO SST anomalies on the MLYRV precipitation in June−July 2020. In the second experiment, we specified observed monthly climatological SST plus multi-decadal warming trend component in the tropical IO (EXP_IOTRE). The differences between EXP_IOTRE and EXP_IOCLM (EXP_IOTRE-minus-EXP_IOCLM) therefore represent the impacts of the IO multi-decadal warming trend on the MLYRV precipitation. The additional two experiments are summarized in Table 2.
Indian Ocean (20°N−20°S, 50°−100°E) CTL climatological SST (1983−2012) SEN1 climatological SST plus the multi-decadal warming trend induced SST anomalies SEN2 climatological SST plus detrended SST anomalies
Table 2. Experiments conducted to investigate the relative contributions of the Indian Ocean SST multi-decadal warming trend and interannual variations to the precipitation anomaly in June−July 2020 over the MLYRV. Ensemble experiments (27 members) are conducted starting from 1 May 2020