Future Projections of Extreme Integrated Water Vapor Transport and Population Exposure Over the Asian Monsoon Region

Global warming leads to an intensification of the water cycle and an increase in extreme climate events. Most previous researchers have used precipitation to project extreme hydrological changes. However, compared with precipitation, global climate models (GCMs) have a better performance in simulating the integrated vapor transport (IVT). IVT is thus a reliable tool for understanding how hydrological extremes may change in the future. In this study, we first discuss the relationship between IVT and precipitation over the Asian monsoon (AM) region during 1979–2014. We then assess the climatology, variability and extremes of IVT over the AM region and subregions using 16 GCMs from CMIP6. We further apply GCMs to project changes in the magnitude and probability of dangerous extreme IVT events (e.g., 20‐, 50‐ and 100‐year events) in the future. Finally, we quantify the population exposure to extreme IVT events, and explain change in exposure and its uncertainty. Results show that CMIP6 GCMs capture the spatial distribution of IVT well; the intensity of simulated IVT is lower than the ERA5 IVT. Under global warming of 1.5, 2, and 3°C, the magnitude of extreme events, which is independent of the return period (RP), increases over the AM region with relative changes of about 7%, 12%, and 21%, respectively. The probability increases consistently with global warming and the RP. The increase in exposure for the South Asian monsoon (East Asian monsoon) region is both contributed by more extreme IVT events and higher (fewer) population counts; the former has a larger proportion.

Hydrology-related extremes are one of the most impact-related consequences under global warming, especially in the AM region. When investigating future changes in such events, three methods have generally been undertaken. First, hydrology-related indices (e.g., 95th percentile of rainday amounts (P95) and the standardized precipitatio n-evapotranspiration index) using model simulations have been used to explore future changes in floods, droughts and extreme precipitation (Guo et al., 2020;Khan et al., 2020;Zhao et al., 2021). Second, the changing nature of the RPs of extreme precipitation in climate projections, using the generalized extreme value (GEV) distribution, has been evaluated (Frei et al., 2006;Zhang et al., 2018). Third, synoptic-scale characteristics, such as the westerly jet (Yu et al., 2021;Zhao et al., 2014), cyclones (Mori & Takemi, 2016;Zappa et al., 2013) and the extreme IVT Reid et al., 2021;Sousa et al., 2020), have been assessed as we know that these phenomena can lead to extreme precipitation and floods. It has been established that IVT is a strong predictor for precipitation (Lavers & Villarini, 2015;Lavers et al., 2014). Furthermore, compared with precipitation (Dong & Dong, 2021), GCMs have a significantly better performance in simulating IVT ( Figure S4 in Supporting Information S1). This also inspires us more confidence for future projection of extreme IVT.
In addition, the risk or impact of extreme precipitation and flood events is an unavoidable topic, when we focus on future changes in such events. The AR6 from IPCC shows that climate change risks arise from the interaction between hazard, exposure and vulnerability (Guo et al., 2022;Ranasinghe et al., 2021;Zscheischler et al., 2018). Vulnerability is a function of exposure, sensitivity and adaptive capacity (Jones et al., 2015). It is thus necessary to quantify the exposure to such extreme events for different warming levels.
The outline of our paper is the following. Section 2 describes the data sets and methods. Section 3 discusses the relationship between IVT and precipitation over the AM region. Section 4 evaluates the climate properties of IVT in ERA5 reanalysis and CMIP6 GCMs. Section 5 presents future changes in the magnitude and occurrence probability of extreme IVT events over the AM region and monsoon subregions under global warming of 1.5, 2, and 3°C. Section 6 quantifies the population exposure to extreme IVT events, and then explains the change in exposure and its uncertainty. Finally, a general summary and discussion are given in Section 7.

Data Sets
We select daily specific humidity, meridional and zonal winds on the 1,000, 850, 700, and 500 hPa levels from 16 GCMs in CMIP6 archive (Table S1; Eyring et al., 2016). The horizontal resolutions of models range from 0.7° × 0.7° to 2.0° × 2.5°. For each model, historical simulation over 1979-2014 and future projection under the Shared Socioeconomic Pathway 5-8.5 (SSP585) using ensemble member r1i1p1f1 are considered. It should be noted that the models are chosen based on two principles: (a) daily specific humidity, meridional and zonal wind data availability, (b) the 3°C global warming threshold is reached before the year 2100 under the SSP585 scenario. In an Eulerian framework, we calculate vertically integrated horizontal meridional and zonal atmospheric moisture transports, and then combine it into the total water vapor transport. For the details of IVT, please see Methods in Section 2.2.
To assess the ability of GCMs in reproducing IVT, we use daily specific humidity, meridional and zonal winds, as well as daily precipitation, during 1979-2014 from ERA5 reanalysis (Hersbach et al., 2019) in European Centre for Medium-Range Weather Forecasts with 0.25° × 0.25° resolution. To facilitate comparison, we uniformly convert horizontal resolutions of variables from CMIP6 GCMs and ERA5 reanalysis into a grid of 1° × 1° with a bilinear interpolation scheme; the daily IVT is then calculated. Since GCM has only four pressure layers between 1,000 and 500 hPa, we have to calculate the ERA5 IVT and simulated IVT using four pressure levels uniformly to facilitate comparison. Meanwhile, we conduct a test to compare the ERA5 IVT using the aforementioned four pressure levels with the ERA5 IVT using all 16 pressure levels between 1,000 and 500 hPa from 1 June 2005 to 31 August 2005. The results show that there is little difference between the two ( Figure S1 in Supporting Information S1). The relative bias in region average of IVT over the Asia monsoon region between the two is −2.6% (Figure S1c in Supporting Information S1). We thus calculate the GCM IVT and the ERA5 IVT using four pressure levels uniformly in our study.
We select the AM region as the study area (surrounded by green lines in Figure 1a) where the summer precipitation rate is larger than that in winter with the difference of more than 2 mm/day and the summer precipitation accounts for more than 55% of the total precipitation (Wang et al., 2012). The 1979-2014 precipitation climatology from ERA5 is used to define the AM region; we focus on land monsoon regions. The AM region is further divided into two subregions based on the boundary of the longitude 100°E and the latitude 20°N: the East Asian monsoon (EAM) and the South Asian monsoon (SAM) regions . It's noted that due to the poor performance of GCMs over the Tibetan Plateau (TP), we remove TP from our study region.
The gridded population counts in the year 2000 from the Global Population of the World version 4 (GPWv4; Center for International Earth Science Information Network-CIESIN-Columbia University, 2018) are used to match the historical reference period for 1986-2005. We find that almost 47% of the world's population lives in the AM region. In other words, sufficient monsoon rainfall provides a livable environment for them. Future population estimates under SSP585 between 2010 and 2100 are obtained from National Center for Atmospheric Research (Jones & O'Neill, 2016).

Definition of IVT
IVT within the troposphere based on the daily ERA5 and CMIP6 model simulation is calculated as: where is 9.81 m∕s 2 , denotes the specific humidity, represents the wind, and p denotes the pressure level. We calculate the vertically integrated q*v first, and then obtain IVT by the root sum square of x-and y-components. It should be noted that atmospheric water vapor above 500 hPa accounts for less than 5% . We thus take 500 hPa as the highest level. Furthermore, considering the effect of the atmospheric warming-induced expansion on the increase in geopotential height, we provide the trend in proportion of atmospheric water vapor below 500 hPa over the region (40-180°E, 20°S-70°N) during 1979-2014 ( Figure S2 in Supporting Information S1). The result shows a slightly increasing trend; the value is 0.017% per decade. Thus, it almost has no effect on the calculation of IVT, when we take 500 hPa as the highest level in the future.

GEV Distribution
Dangerous extreme events, in the aspect of statistical theory, refer to the upper tail of the extreme value distributions. It's thus a small probability event from the view of the probability distributions. In order to quantitatively define the threshold of danger, we use the 20-, 50-, and 100-year return values (RVs) to denote different levels of dangerous extreme events. Here, the RV refers to a specific value of IVT. To estimate the RVs for IVT, a GEV distribution (Coles, 2001) is used to fit the daily IVT at each grid point for each model.
where ξ, σ, and μ are the shape, scale and location parameters, respectively. First, we select a sample of extremes, based on the principle of parameter stability, using the block maxima method. Second, after fitting the sample sequence to the GEV distribution, the parameters are estimated using the maximum likelihood estimation method. Finally, RVs 0 are obtained by inverting the fitted GEV distributions.
where 0 represents the exceedance probability. We thus obtain RPs = 1∕ 0 . Here, except for future changes in the magnitude, we also measure the changes in frequency or probability based on the concept of the probability ratio (PR) (Fischer & Knutti, 2015;Stott et al., 2004). This is defined as PR = f ut ∕ his , where his ( fut ) denotes the probability of a his -year ( f ut -year) RV during the reference period (at different warming levels). It's noted that his -year ( fut -year) means the return period (RP) of his years ( f ut years). We thus obtain: PR = f ut ∕ his = his∕ fut . If PR > 1 (PR < 1) , there is a higher (lower) occurrence probability for extreme events in the future.

Time Windows for Global Warming Levels
Global warming of 1.5, 2, and 3°C is relative to the pre-industrial level . The emission scenario is SSP585. The time sequence of the global average temperature anomaly is filtered by a 21-year running average. The year, in which the threshold is first reached, is then selected. Finally, we form a 20-year time window (Table  S2 in Supporting Information S1) for research, with a 10-year period before the year and a 9-year period after the year. It should be noted that projected changes in extreme events in the following are relative to the historical reference period (1986)(1987)(1988)(1989)(1990)(1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005).

Relationship Between IVT and Precipitation
IVT is one of the factors causing extreme precipitation. For regions with flat topography, such as the AM region, heavy rainfalls usually need strong IVTs. Lavers et al. (2014) shows that the IVT is more predictable than precipitation and could be used to extend the forecast lead time of extreme precipitation by up to 3 days in some locations over Europe. Lavers and Villarini (2015) shows that IVT is a strong predictor for precipitation. To further understand the scientific significance of IVT, we provide correlation coefficients between IVT and precipitation, and the contribution of extreme IVT to extreme precipitation during 1979-2014. As shown in Figure 1a, correlation coefficients over major parts of the AM region exceed 0.4 and pass the significance test, especially in the Indian subcontinent and eastern China. The contribution of extreme IVT to extreme precipitation exceeds 60% in most parts of the AM region ( Figure 1b). Overall, they demonstrate the importance of (extreme) IVT for causing (extreme) precipitation over the AM region quantitatively.
To more directly see the relationship between IVT and precipitation, we provide daily time series of IVT and precipitation of two heavy rainstorms in the historical records (Figures 1c and 1d). One occurs over Sichuan in late June of 1981; the other occurs over Beijing in middle July of 2012. Results show that there is a strong correlation between IVT and precipitation with a correlation coefficient of 0.46 and 0.49, respectively. The two also pass the significance test. The peak IVT and the peak precipitation overlap well. Hence, given the strong relationship between IVT and precipitation, understanding changes to IVT in a warmer world is necessary for preparing for future hydrological extremes over the AM region.

Climate Properties of IVT in ERA5 Reanalysis and CMIP6 GCMs
We evaluate the performance of 16 CMIP6 GCMs to reproduce IVT during 1979-2014 over the AM region and monsoon subregions. We pay more attention to climate properties, including the climatology of IVT, standard deviation of annual variations, and 20-, 50-, and 100-year RVs. It should be noted that we perform the GEV distribution for the individual GCM, but show the result in the form of the multi-model ensemble mean (MME). We find that high IVT occurs along the eastern coast of the AM region; the Indian subcontinent is also significant in the climatology (Figure 2a). These regions are also where the variability is greatest (Figure 2d), a feature relating to EI Nino-Southern Oscillation and monsoons (Kripalani & Kulkarni, 1998;Mariotti, 2007). The spatial patterns of 20-, 50-, and 100-year RVs are all broadly similar to those found in the climatology (Figure 2g and Figure S3 in Supporting Information S1). Furthermore, compared with the ERA5, GCMs well capture the spatial distribution of IVT (middle panels in Figure 2). The largest difference is found in the southern TP (the Himalayas), with a relative bias more than 80% for the mean field and more than 100% for the 20-year RV field, respectively (right panels in Figure 2). In terms of the performance of the individual models, there is a great ability in all GCMs, in which CMCC-ESM2 performs best and INM-CM5-0 performs worst ( Figure S4 in Supporting Information S1).
To compare the performance of GCMs in reproducing the IVT over the monsoon subregions, we provide quantitative statistics (Figure 3). The results show that the intensity and variability of IVT over the SAM region are both greater than those over the EAM region; the intensity (variability) of simulated IVT over the monsoon subregions is slightly lower (larger) than the ERA5 IVT. Specifically, the mean IVT over the SAM region is larger than that over the EAM region; the simulated mean IVT over the AM region and the SAM region is lower than the ERA5 IVT; the simulated mean IVT over the EAM region is rather close to the ERA5 IVT (Figure 3a). There is greater variability over the SAM region than the EAM region; the simulated IVT variations over the monsoon subregions are both larger than the ERA5 IVT ( Figure 3b). The statistical characteristics of the 20-year RV of IVT are similar to those of the average (Figure 3c).

Future Projections of Dangerous Extreme IVT Events
In view of the satisfactory performance of GCMs in simulating the 20-, 50-, and 100-year RVs of IVT over the AM region and monsoon subregions, we apply GCMs to further project future changes in the magnitude and occurrence probability of such dangerous extreme IVT events under global warming. We perform future climate analyses under the SSP245 and SSP585 scenarios. Similar conclusions are obtained under the two scenarios. This is basically consistent with previous works reported in the literature. They have shown that regional-scale responses are almost independent of the precise emission scenarios, but are more closely related to the global warming level (Hu et al., 2017;Zhai et al., 2017). Therefore, we show future projections under SSP585 in the main text, and retain results under SSP245 in Supporting Information S1 (Figures S5-S8).

Changes in the Magnitude
The spatial pattern of relative changes in the 20-, 50-, and 100-year RVs of IVT for different global warming levels, relative to 1986-2005, is shown in Figure 4. Changes in 20-, 50-, and 100-year RVs exhibit a similar spatial pattern. The magnitude increases over the AM region at the 1.5°C warming target. The increase is between 4% and 12% (left panels in Figure 4). There is an increase under global warming of 2°C with a value between 8% and 20% (middle panels in Figure 4). The spatial distributions exhibit similar characteristics between 1.5 and 2°C, with a slight increase over the western Indian subcontinent at the 2°C target (middle panels in Figure S9 of the Supporting Information S1). For the higher warming level at 3°C, the increase is more than 20% (right panels in Figure 4). The additional warming from 1.5 to 3°C leads to a significant increase in the magnitude of RVs over the western Indian subcontinent (right panels in Figure S9 of the Supporting Information S1). It should be noted that all 16 GCMs agree with the sign of the changes.
To sum up, the magnitude of the 20-, 50-, and 100-year RVs of IVT all show a gradually increasing trend under global warming of 1.5, 2, and 3°C, especially in the western Indian subcontinent. According to the Clausius-Clapeyron equation, the change in atmospheric moisture content , increases nearly exponentially with temperature; the rate is approximately 7%/°C (Payne et al., 2020).  also show that changes in IVT are mainly due to higher atmospheric water vapor content. Furthermore, this increasing water vapor transport component can, to a certain extent, also explain why future extreme precipitation increases in the literature (Chen et al., 2014;Kharin et al., 2013;Polade et al., 2014;Sillmann et al., 2013;Toreti et al., 2013). It's noted that, although global precipitable water vapor content (PWV) and global IVT generally show increasing trends in the past decades ( Figure S10 in Supporting Information S1; Ren et al., 2023), the effect of global warming on PWV and IVT has significant regional characteristics in the past decades due to large variability on a regional scale. In the future, we thus need to do more work to explore the specific reasons for changes in IVT over the AM region.
To quantitatively measure changes in the magnitude, we provide statistics in the form of box-whisker plots ( Figure 5). The results show that there is little difference among the magnitude of the 20-, 50-, and 100-year RVs for the same warming level for any monsoon region. The amplitude increases by about 7%, 12%, and 21% for the global warming levels at 1.5, 2, and 3°C, respectively. The additional warming from 1.5 to 2°C (3°C) results in an increase of nearly 5% (14%). Our result is nevertheless in accord with Li et al. (2018), who found that regional average changes in extreme precipitation events are independent of the RPs with an increase of about 7% and 11% at the 1.5 and 2°C warming targets. It should be noted that there is little difference among the monsoon  regions. In addition, the spread among models, which refers to uncertainty in future projection, is slightly positively correlated with global warming levels and the length of RPs.

Changes in the Probability
The spatial distribution of probability changes in dangerous extreme IVT events is shown in Figure 6. PR increases consistently with the global warming and RP. This means that 100-year RP events show the largest PR (Figures 6g-6i), or the warming level at 3°C results in the largest PR (right panels in Figure 6; Figure S11 in Supporting Information S1). In spatial terms, the largest increase is located in the tropical region, especially in the southern Indian subcontinent and southern Indochina.
Figure 7 provides the regional mean PR of dangerous extreme IVT events over the AM region and monsoon subregions at the three warming levels. We find that the occurrence probability of extreme IVT events increases with higher warming levels and longer length of RPs. Specifically, the probability of 20-, 50-, and 100-year RP events over the AM region increases by a factor of 1.9, 2.2, and 2.6 under global warming of 1.5°C, respectively. This means that 20-, 50-, and 100-year RP events in 1986-2005 become 11-, 23-, and 38-year RP events under the warming of 1.5°C. For the higher warming level of 3°C, the probability of the three extreme events over the AM region increases by a factor of 4.6, 6.0, and 7.5, respectively. It should be noted that 100-year RP events in 1986-2005 become 13-year events. In terms of uncertainty, there is a significantly positive correlation with the global warming level and the length of RP. Specifically, PR of the 20-year RP events ranges between 1.9 and 3.9 under global warming of 2°C, but for the 100-year RP events, PR ranges between 2.4 and 6.6. PR of the 100-year RP events under global warming of 1.5 and 2°C ranges between 1.6 and 4.1, and between 2.4 and 6.6, respectively. Furthermore, we find that the probability of the three thresholds of extreme IVT events over the SAM region is significantly higher than those over the EAM region in the future (Figures 7b and 7c). For example, the probability of 20-, 50-, and 100-year RP events over the SAM region increases by a factor of 5.4, 7.5, and 9.6 under global warming of 3°C, but over the EAM region, the probability increases by a factor of 4.2, 5.3, and 6.3. Huang et al. (2021) also showed that there is the higher magnitude and frequency risk levels of extreme precipitation over the SAM region under the RCP85 scenario than those over the EAM region.

Population Exposure to Extreme IVT Events
Dangerous extreme events, which deviate substantially from the climatology, cause serious losses, because they strongly challenge the carrying capacities of ecosystem and human system. The impact of extreme events on human is thus an essential topic. Since the IVT and the population count over the AM region both have significant spatial characteristic, we have to separately calculate the population exposure to extreme IVT events for each grid point. We first define extreme IVT events as those exceeding the 20-, 50-, and 100-year RVs from the 1979-2014 baseline. The thresholds denoting different levels of danger are used to judge whether extreme IVT events occurred during the present-day (1986)(1987)(1988)(1989)(1990)(1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005) and future period. The total population exposure is the sum of the population that experiences extreme events exceeding the thresholds over the study area. Fraction of exposure is then the ratio of the total exposure to the total population. We present the population exposures to these extreme IVT events at the present-day level, 1.5, 2, and 3°C warming levels over the AM region and monsoon subregions, respectively. It should be noted that the population counts in 2000 and those projected under the SSP585 emission scenario are used at the present-day level and future warming levels, respectively.
Exposure to these extreme IVT events increases consistently with global warming (Figure 8). Specifically, over the AM region, the population exposure to extreme events that exceed the baseline 20-year RVs (EEEB20) increases from 5.0% at the present-day level to 9.0%, 12.6%, and 21.3% at the 1.5, 2, and 3°C global warming levels, respectively (solid line in Figure 8a). Similarly, the exposure to EEEB50 increases from 1.9% at the present-day level to 4.0%, 6.0%, and 11.0% at the three global warming levels. The exposure to EEEB100 increases from 0.9% at the present-day level to 2.2%, 3.4%, and 6.7% at the three global warming levels.
We also compare population exposure between the EAM region and the SAM region. The results show that the exposure to extreme events over the SAM region is significantly larger than that over the EAM region under global warming. For example, the exposure to EEEB100 over the EAM region increases from 0.8% at the present-day level to 2.0%, 3.0%, and 5.6% at the 1.5, 2, and 3°C global warming levels, respectively (solid line in Figure 8b). However, for the SAM region, the exposure to EEEB100 increases from 0.9% at the present-day level to 2.5%, 4.2%, and 8.9% at the three global warming levels (solid line in Figure 8c).
To investigate the reasons for the increase in exposure for the EAM or SAM region, we compare fraction of population exposure obtained based on projected population counts (FPP; solid line in Figures 8a-8c) with that obtained based on population counts in 2000 (FP2000; dashed line in Figures 8a-8c). The latter assumes that population counts remain unchanged in the future. Results show that for the EAM region, FPP is slightly lower than FP2000 (Figure 8b). It means that the increase in exposure mainly results from more extreme events. We further examine the projected population counts. As shown in Figure 8d, the population counts over the EAM region have a generally decreasing trend in the future. The increase in exposure for the EAM region is thus contributed by more extreme IVT events and fewer population counts. However, for the SAM region, FPP is larger than FP2000 (Figure 8c). The difference between the two is less than 5%. The future population counts always exceed the population counts in 2000 ( Figure 8d). Therefore, the increase in exposure for the SAM region is both contributed by more extreme IVT events and higher population counts; the former has a larger proportion.
Furthermore, we find that the spread among models, which refers to the uncertainty, increases with continuous warming, but decreases with the RP (Figures 8a-8c). To explain why the uncertainty decreases with the RP, we review the concept of fraction of population exposure first. It is the ratio of the total exposure to the total population. The difference in fraction of population exposure among three extreme events thus results from the total exposure. This part is the accumulation of population experiencing dangerous extreme IVT events that exceed baseline 20-, 50-, 100-year RVs. We thus provide CDFs of extreme IVT values at any four grid points in the study region at the 3°C global warming level under the SSP585 scenario (Figure 9). Results all show that the difference among GCMs in simulating more extreme values is smaller. It means that occurrence frequencies of extreme IVT events that exceed baseline 100-year RVs among GCMs are closer than those of extreme IVT events that exceed baseline 20-year RVs. Therefore, the uncertainty in fraction of population exposure decreases with the RP. It's noted that we randomly provide four grid points in the study region to show the uncertainty, since the conclusion drawn is robust and consistent.

Summary and Discussion
We first discuss the relationship between IVT and precipitation over the AM region during 1979-2014. We then assess the performance of 16 CMIP6 GCMs to reproduce the climatology, variability and extremes of IVT during 1979-2014 over the AM region and monsoon subregions compared with ERA5 reanalysis. The 20-, 50-and 100-year extreme IVT events are defined using the GEV distribution. We further apply GCMs to project changes in the magnitude and occurrence probability of those extreme events under global warming of 1.5, 2, and 3°C. Finally, we quantify the population exposure to extreme IVT events, and explain the change in exposure and its uncertainty. The conclusions are summarized as follows: 1. Over the AM region, there is a strong correlation between IVT and precipitation, especially in the Indian subcontinent and eastern China. The contribution of extreme IVT to extreme precipitation exceeds 60% in most parts of the AM region. In terms of two heavy rainstorms in historical records, the peak IVT and the peak precipitation overlap well. Overall, extreme IVT is a strong predictor for extreme precipitation. 2. GCMs capture the spatial distribution of climate characteristics of IVT during 1979-2014 over the AM region well. The intensity (variability) of simulated mean IVT is slightly lower (larger) than the ERA5 mean IVT. The largest difference is located in the Himalayas, with a relative bias more than 80% for climatology and more than 100% for the 20-year RV, respectively. 3. There is little difference among changes in the magnitude of the 20-, 50-and 100-year RVs. Under global warming of 1.5, 2, and 3°C, the magnitude of these extreme IVT events shows a gradually increasing trend over the AM region, especially in the western Indian subcontinent. The increase is thus about 7%, 12%, and 21% over the AM region for the different warming levels, respectively. In terms of probability changes, PR of extreme IVT events increases consistently with the RP and global warming. The probability of 20-year (100-year) RP events over the AM region increases by a factor of 1.9 (2.6), 2.7 (3.9), and 4.6 (7.5) at the three global warming levels, respectively. In spatial terms, the largest increase in PR is located in the tropical region, especially in the southern Indian subcontinent and southern Indochina. The probability of extreme IVT events over the SAM region is greater than those over the EAM region. Furthermore, the uncertainty of PR is significantly positively correlated with the global warming level and RP. 4. Population exposure to these extreme IVT events increases consistently with global warming. The population exposure to EEEB20 (EEEB100) over the AM region increases from 5% (0.9%) at the present-day level to 9.0% (2.2%), 12.6% (3.4%), and 21.3% (6.7%) under global warming at 1.5, 2, and 3°C, respectively. Furthermore, the population exposure to extreme events over the SAM region is significantly greater than that over the EAM region. This is because that the increase in exposure for the SAM (EAM) region is both contributed by more extreme IVT events and higher (fewer) population counts; the former has a larger proportion. In terms of the uncertainty in exposure, the spread among models increases with continuous warming, but decreases with the RP. We find that occurrence frequencies of extreme IVT events that exceed baseline 100-year RVs among GCMs are closer than those of extreme IVT events that exceed baseline 20-year RVs. Therefore, the uncertainty in fraction of population exposure decreases with the RP.
In fact, larger IVT may be likely to result in more extreme precipitation, as well as larger and more frequent floods. However, it's noted that an increase in IVT could not alone result in extreme precipitation. Similarly, the intensification of the westerly jet or the generation of tropical cyclones may not alone cause strong winds. Another necessary condition is that the water vapor is transferred to the troposphere by a lifting motion, and then condenses into rain. In a word, this work presents a simple and useful diagnostics tool based on IVT to investigate precipitation-related extremes and their impacts.  Jones and O'Neill (2016). It should be noted that all data sets above are freely available for the public. After registering and logging in an account, everyone can download the data set directly.