HTML

The datasets used include the following: (1) daily gridded precipitation on 0.5° × 0.5° horizontal resolution from 1 January 1979 to 31 December 2019, obtained from the U.S. Climate Prediction Center (CPC) unified precipitation project (Chen et al., 2008); (2) daily specific humidity, zonal and meridional winds on 0.75° × 0.75° horizontal resolution from 1 January 1979 to 31 August 2019, from the ECMWF interim reanalysis (ERAInterim; Dee et al., 2011); and (3) monthly Niño3.4 index from CPC. El Niño events are identified when the Niño3.4 index is no less than 1°C and persists for three months. Based on the above criteria, we selected nine El Niño events (1982/1983, 1986/1987, 1987/1988, 1991/1992, 1994/1995, 1997/1998, 2002/2003, 2009/2010, and 2015/2016) from 1979 to 2019. In this study, the spring (March–May, MAM), summer (June–August, JJA), autumn (September–November, SON), and winter (December–February, DJF) seasons are considered separately.

The integrated water vapor transport (
$ \mathrm{I}\mathrm{V}\mathrm{T} $ ) is defined and calculated as below:$$ {\boldsymbol{F}}_{\rm{IVT}} = \frac{1}{g}\int _{{p_{\rm{t}}}}^{{p_{\rm{s}}}}q{\boldsymbol{V}}{\rm{d}}p, $$ (1) where
$ g $ is the acceleration due to gravity,$ q $ is the specific humidity,$ {p}_{\rm{s}} $ is the surface pressure,$ {p}_{\rm{t}} $ is set to be 300 hPa, and$ \boldsymbol{V} $ is the horizontal wind vector$ (u,v) $ .We perform AR detections based on Guan and Waliser (2015). The first step is to compute the threshold of IVT at each grid. For each month, the 85th percentile threshold of IVT is calculated within fivecontinuous months centered on that month. The greater value of the 85th threshold and 100 kg m^{−1} s^{−1} are used to obtain contiguous regions. The second step is to consider the requirements on IVT direction. The purpose is to make sure that most of the AR grids’ IVT has similar direction. The third step is to ascertain the geometry of the regions. Those regions with a length > 2000 km or a large ratio of length to width > 2 are retained as AR regions. More details of the detection method can be found in Guan and Waliser (2015). If a grid point is in the detected AR region, this grid is defined as an AR grid, and the corresponding day is recorded as an AR day for this grid. The frequency of AR occurrence for a grid in each season is calculated as the ratio of the number of AR days to the number of total days in each season. The variables [precipitation, IVT, and divergence of IVT (i.e.,
$ \nabla \cdot {\boldsymbol{F}}_{\rm{IVT}}) $ ] at the AR grids in the AR days belong to an ARgroup. In each season, all variables (precipitation, IVT, and$\nabla \cdot {\boldsymbol{F}}_{\rm{IVT}} $ ) for the ARgroup are individually summed up, in order to obtain the corresponding seasonal cumulation. 
The moisture flux divergence
$ \nabla \cdot {\boldsymbol{F}}_{\rm{IVT}} $ can be decomposed as follows:$$\begin{aligned} \underbrace {\nabla \cdot \frac{1}{g}\int_{{p_{\rm{t}}}}^{{p_{\rm{s}}}} q{\boldsymbol{V}}{\rm{d}}p{\rm{}}}_{\rm{T}} = & \underbrace {\frac{1}{g}\int_{{p_{\rm{t}}}}^{{p_{\rm{s}}}} q ({\frac{{\partial u}}{{\partial x}} + \frac{{\partial v}}{{\partial y}}} ){\rm{d}}p}_{\rm{A}} + \underbrace {\frac{1}{g}\int_{{p_{\rm{t}}}}^{{p_{\rm{s}}}} u\frac{{\partial q}}{{\partial x}}{\rm{d}}p}_{\rm{B}} \\ & + \underbrace {\frac{1}{g}\int_{{p_{\rm{t}}}}^{{p_{\rm{s}}}} v\frac{{\partial q}}{{\partial y}}{\rm{d}}p}_{\rm{C}},\\[15pt] \end{aligned}$$ (2) where the term A (horizontal wind divergence), term B (zonal moisture advection), and term C (meridional moisture advection) contribute to the term T (
$ \nabla \cdot {\boldsymbol{F}}_{\rm{IVT}} $ ). 
At each grid, we rank the daily precipitation of wet day (daily precipitation > 1 mm) from small to large for each season during 1979–2019. The statistical 95th percentile is used as the threshold for extreme precipitation (Hagos et al., 2016; Dong et al., 2018). The definition of a 2day extreme precipitation event (2day EPE) is as follows. First, the days with daily precipitation amount exceeding the 95th threshold are marked as extreme days. Then, we compare precipitation amounts on the day before and after the marked day, and the day with larger precipitation amount is selected. A 2day EPE covers the marked day and the day with a larger precipitation amount. The days of all 2day EPE are identified as 2day EPE days (Shang et al., 2020; Xiong and Ren, 2021). If a 2day EPE is accompanied with an AR (at least one day is an AR day), it is defined as an AREPE and the days for this EPE are AREPE days. We also sum up all variables (precipitation, IVT, and
$ \nabla \cdot {\boldsymbol{F}}_{\rm{IVT}} $ ) on 2day EPE days and AREPE days for each season, respectively.