
The dataset used in this study includes the CN05.1 grid data provided by NCC, which includes daily maximum temperature (T_{max}), daily average temperature (T_{m}), and daily precipitation with grids of 0.25° × 0.25° (Wu and Gao, 2013). We also used ERA5 reanalysis data provided by ECMWF, with grids of 2.5° × 2.5°, 37 pressure levels in the vertical direction, and 4 levels of soil data (Hersbach et al., 2020).
The drought data were obtained from the daily meteorological drought composite index (MCI) station data provided by NCC, and we chose 523 stations in the MYRB and used their respective data for analysis (Fig. 1). We calculated the MCI with reference to the Grades of Meteorological Drought 2017 edition (General Administration of Quality Supervision, Inspection and Quarantine of the People’s Republic of China and Standardization Administration of the People’s Republic of China, 2017), and the formula is shown as follows:
Figure 1. Topography over the study area (26°−36°N, 104°−116°E, shading; m). The dots denote the stations chosen for MCI data.
$$ \begin{aligned}[b] {\rm{MCI}} = & \;{\rm{Ka}} \times (a\times {\rm{{SPIW}_{60}}}+b\times {\rm{MI}}_{30}+c\times {\rm{SPI}}_{90}\\ + & d \times {\rm{SPI}}_{150}), \end{aligned}$$ (1) where MI_{30} represents the moisture index for the last 30 days, with the intermediate variable PET (potential evapotranspiration), calculated with the Thornthwaite method (Thornthwaite, 1948); SPI_{90} and SPI_{150} represent the standardized precipitation index for the last 90 and 150 days, respectively;
$ {\mathrm{S}\mathrm{P}\mathrm{I}\mathrm{W}}_{60} $ is the standardized weighted precipitation index for the last 60 days; a, b, c, and d are weighting coefficients, and in this study, a = 0.5, b = 0.6, c = 0.2, and d = 0.1 for Southwest China; and Ka is the seasonal adjustment coefficient, and Ka = 1.2 for Southwest China in July. To evaluate the simulation of drought conditions for each experiment, the daily MCI was calculated grid by grid according to the above definition. The drought levels corresponding to different MCIs are shown in Table 1.Level Type MCI 1 No drought $ 0.5 < \mathrm{M}\mathrm{C}\mathrm{I} $ 2 Light drought (I) $1.0 < \mathrm{M}\mathrm{C}\mathrm{I}\leqslant 0.5$ 3 Moderate drought (II) $1.5 < \mathrm{M}\mathrm{C}\mathrm{I}\leqslant 1.0$ 4 Severe drought (Ⅲ) $2.0 < \mathrm{M}\mathrm{C}\mathrm{I}\leqslant 1.5$ 5 Extreme drought (Ⅳ) $\mathrm{M}\mathrm{C}\mathrm{I}\leqslant 2.0$ Table 1. Different MCIs and their corresponding drought level (General Administration of Quality Supervision, Inspection and Quarantine of the People’s Republic of China and Standardization Administration of the People’s Republic of China, 2017)
In this study, we mainly focus on the MYRB in Southwest China (26°–36°N, 104°–116°E). According to previous studies (Qi et al., 2019), a heatwave event over the MYRB can be identified if there are two consecutive days with area average T_{m} surpassing the 95th percentile of the daily climatology (defined over 1981–2010). Moreover, we adopted the following criterion for drought event. If more than 50% of the stations have an MCI less than −0.5 (having a light drought or more severe), a drought event is identified. As a CDHE event has its heatwave event part and its drought event part, in identifying an event, we combined the two criteria together that in any consecutive 7 days, a heatwave event and a drought event happen, a CDHE event is identified, and the start and end dates are determined by the start and end of both events.
Referring to the method by Hong et al. (2021), the ISO component in this study was extracted by first removing the series mean of summer from the ERA5 meteorological reanalysis data and CN05.1 temperature data and then removing synoptic fluctuations by taking a 5day running mean.
Wavelet analysis, which is a common tool for diagnosing timefrequency variations of a time series, has the ability to identify the dominant ISO modes (Mao and Wu, 2006). In this study, we used the Morlet as the wavelet basis function and the
$ {\chi }^{2} $ test for statistical significance to obtain the dominant periodicity of the ISO component.To examine the propagation of intraseasonal wave trains as the basis for numerical experiments, the TN wave flux was calculated phase by phase at 200 and 850 hPa, and the formulation is expressed as (Takaya and Nakamura, 2001):
$$ \begin{array}{c} {\boldsymbol{W}}=\dfrac{1}{2\left{\boldsymbol{U}}\right}\left[\begin{array}{c}\bar{u}\left({{{\psi }}_{x}'}^{2}{\psi }'{{\psi }}_{xx}' \right)+\bar{v}\left({{\psi }}_{x}'{{\psi }}_{y}' {\psi }'{{\psi }}_{xy}' \right)\\ \bar{u}\left({{\psi }}_{x}'{{\psi }}_{y}'{\psi }'{{\psi }}_{xy}' \right)+\bar{v}\left({{{\psi }}_{y}'}^{2}{\psi }'{{\psi }}_{yy}' \right)\end{array}\right],\end{array} $$ (2) where
$ {\boldsymbol{W}} $ denotes the TN wave flux,$ {\boldsymbol{U}} $ is the wind velocity, and$ \bar{u} $ and$ \bar{v} $ are the zonal and meridional winds, respectively. The parameter$ \psi $ represents the stream function, and the subscript indicates the partial derivative.To determine the contribution of different physical processes to the development of this event, we calculated the temperature budget at 925 hPa phase by phase. According to a previous study, the temperature tendency is modulated by horizontal temperature advection, adiabatic processes induced by vertical motion, and the atmospheric apparent heat source, and the formulation can be written as (Yanai et al., 1973):
$$ \begin{array}{c} \dfrac{\partial T}{\partial t}=V \cdot \nabla T+\omega \sigma +\dfrac{{Q}_{1}}{{c}_{p}},\end{array} $$ (3) where the four terms represent the temperature tendency, horizontal advection of temperature, vertical transport, and diabatic heating, respectively. The first three terms are calculated based on ERA5 meteorological data, and the fourth term is obtained from the first three terms (Wang et al., 2017). In Eq. (3),
$ \sigma =\dfrac{RT}{{c}_{p}p}\dfrac{\partial T}{\partial p} $ denotes the static stability with$ R $ as the gas constant and$ {c}_{p} $ as the specific heat at constant pressure, and$ {Q}_{1} $ denotes the diabatic heating containing radiative heating, latent heating, surface heat flux, and subgridscale processes.The Weather Research and Forecasting (WRF) model (version 4.0) is used to carry out the simulation and the PLF experiment in this study. WRFAdvanced Research WRF (WRFARW) is developed by NCAR’s Mesoscale and Microscale Meteorology Laboratory and has been widely used in regional climate simulations (Skamarock et al., 2019). All experiments were initiated on 7 July and ended on 30 July. According to previous studies on simulations of drought events in Southwest China (Wang and Liao, 2015), the following physical parameterization packages were used: the WRFSingleMoment 3class (WSM3) scheme for microphysics (Hong et al., 2004), CAM3 (Community Atmosphere Model version 3) longwave and shortwave schemes (Collins et al., 2004), Mellor–Yamada–Janjić scheme for the planetary boundary layer and surface (Janjić, 2002), Noah’s land surface model (Chen and Dudhia, 2001), and Betts–Miller–Janjić cumulus parameterization scheme (Janjić, 1994, 2000). The simulation area approximately covers a region (23°–38°N, 100°–120°E) and with a horizontal resolution of 30 km. The width of the buffer zone was set to five grid points. The model initial and lateral boundary conditions were obtained from the ERA5 dataset with 6h intervals and 37 vertical pressure levels up to 1 hPa, including geopotential height, horizontal wind, vertical velocity, and specific moisture. The soil data consisted of 4 levels down to a depth of 289 cm underground, and included soil moisture and soil temperature.
To identify the specific roles of the different propagating ISOs in the HW development stage, we introduced PLF experiments and carried out control and sensitivity runs. The boundary forcing and initial field of the control run used real data, while the boundary forcing for the sensitivity run was obtained by removing ISOs from some direction(s) of the original boundary forcing. Based on observational results, we performed three experiments: RMVN, RMVSE, and RMVNSE, representing removing ISOs from the northern, eastern and southern, and all three boundaries, respectively. According to the instructions provided by Dr. Qi Xin, the ISO signals are removed by modifying the wrfbdy file generated in the initialization period. In this file, all the boundary conditions of the four directions throughout the period of the experiment are provided, including wind, temperature, water vapor mixing ratio, etc., and when removing ISO signals from one boundary, all the variables on that boundary should be removed out of ISO. Also, in order to rule out the possible influence from the initial field, we performed 10 extra runs for each of the four experiments using initial fields with some perturbations. The average results of these ensemble runs, which are shown in this paper, were compared to the original run and no obvious deviation was found. The difference between the control run results and the sensitive run results was compared by using the Student’s t test. The physical options and model configuration were kept the same for both the control run and PLF experiments.
Level  Type  MCI 
1  No drought  $ 0.5 < \mathrm{M}\mathrm{C}\mathrm{I} $ 
2  Light drought (I)  $1.0 < \mathrm{M}\mathrm{C}\mathrm{I}\leqslant 0.5$ 
3  Moderate drought (II)  $1.5 < \mathrm{M}\mathrm{C}\mathrm{I}\leqslant 1.0$ 
4  Severe drought (Ⅲ)  $2.0 < \mathrm{M}\mathrm{C}\mathrm{I}\leqslant 1.5$ 
5  Extreme drought (Ⅳ)  $\mathrm{M}\mathrm{C}\mathrm{I}\leqslant 2.0$ 