Growing Threats From Swings Between Hot and Wet Extremes in a Warmer World

The abrupt alternation between hot and wet extremes can lead to more severe societal impacts than isolated extremes. However, despite an understanding of hot and wet extremes separately, their temporally compounding characteristics are not well examined yet. Our study presents a comprehensive assessment of successive heat‐pluvial and pluvial‐heat events globally. We find that these successive extremes within a week occur every 6–7 years on average within warm seasons during 1956–2015, about 15% more often than would be expected by chance, and that they have a significant increase in frequency of about 22% per decade due to warming. We further investigate the role of vapor pressure deficit (VPD) and find that heat‐pluvial (pluvial‐heat) events are linked to negative (positive) VPD anomalies. Our results are statistically significant based on moving‐blocks bootstrap resampling and field significance tests, highlighting these methods' importance in robustly identifying compound events under autocorrelation and multiple‐testing conditions.

include the movement of features such as areas of enhanced monsoon convection or the jet stream meandering (Shang et al., 2020;Shimpo et al., 2019;Swain et al., 2016;Z. Wang et al., 2019). There may also be direct linkages: high temperatures are a key factor contributing to atmospheric instability, potentially leading to or enhancing localized precipitation events that terminate previous heat events through strong evaporational cooling (Berg et al., 2013;Fowler et al., 2021;Wang et al., 2017). The other side of the coin is the occurrence of pluvials followed by heat events, which may be associated with tropical cyclone-released diabatic heating effects or region-scale thermal advection Emanuel, 2003;Hart et al., 2007;Parker et al., 2013;Sukhovey & Camara, 1995). Another important physical mechanism is that the elevated moisture fluxing into the atmosphere during a pluvial event can increase atmospheric latent heat content, which may result in higher near-surface wet-bulb temperatures favorable for the occurrence of humid heat (Liu et al., 2017Matthews et al., 2022;Speizer et al., 2022).
When extreme heat is combined with pluvial flooding or vice versa in close succession, adverse impacts may be exacerbated due to the short recovery time. For successive heat-pluvial events, a sequence of heat extremes followed by pluvial flooding occurred in the United States in September 2017 (Cappucci, 2019), in the United Kingdom in August 2020 (ITV Weather, 2020) and in South Korea in July 2020 (Min et al., 2022), leading in each case to severe infrastructure damages, livestock deaths, and flood-related morbidity/mortality. The considerable source of concern is that people are generally not well prepared for high-intensity rainfall during prolonged hot weather; when it does occur, it can be so rapid that people have little time to adjust and safely evacuate (De Ruiter et al., 2020;Matthews et al., 2019;. As an example of a successive pluvial-heat event, Japan experienced heavy rainfall and subsequent extreme heat in July 2018, causing more than 300 fatalities and large economic losses (Kawase et al., 2020;. The landfalls of tropical cyclones Irma and Ida in Florida and Louisiana, respectively, led to notable health impacts on residents who in the storms' aftermath were without air conditioning to combat the typical high heat stress values of late summer (Chatlani & Madden, 2021;Skarha et al., 2021). In such cases, the subsequent extreme heat adds to impacts in affected areas because the damage to infrastructure such as roads and power grids makes it more difficult to avoid heat exposure, and to obtain treatment in the case of heat illness (Issa et al., 2018). These recent examples highlight the importance of investigating the temporally compounding characteristics of heat-pluvial and pluvial-heat extremes more broadly.
Compared to well-understood underlying dependent drivers (e.g., concurrent drought and heatwave), quantifying the relationship between temporally compounding hot and wet extremes remains challenging. Recent work has, however, made progress. For example, analysis in the central United States found that a high percentage of floods are preceded by a heat stress event (Zhang & Villarini, 2020). Elsewhere, assessments of consecutive heat wave and heavy rainfall events in China found an increased probability of hotter and shorter heat waves followed by heavy rainfall compared to heat waves not followed by heavy rainfall (You & Wang, 2021). Consecutive heat and pluvial events have also been investigated in previous studies, again focused on China Liao et al., 2021). Globally, an increasing percentage of floods are likely to be accompanied by hot extremes, using observed dry bulb temperature for the identification of heat extremes and hydrological models for the simulation of flood hazards . Compared to dry heat, however, humid heat measures (that include the effects of both temperature and humidity) better capture the physiological drivers of heat stress and therefore are more likely to reflect dangerous conditions (Mora et al., 2017;Raymond et al., 2021). More importantly, the use of coarse-resolution general circulation models (∼100-300 km) and conceptual hydrological models may lead to considerable uncertainty in projecting spatially resolved flood risks caused by heavy precipitation (Duethmann et al., 2020;Grimaldi et al., 2019;Zhang et al., 2021). Consequently, there is a need for thorough assessments of the spatiotemporal detections of temporally compounding heat and pluvial events, as well as descriptions of the underlying factors.
Here, we present a comprehensive global analysis of changes in the frequency of temporally compounding heat and pluvial events and underlying factors that may affect these events. Temporally compounding humid heat extremes followed by pluvial extremes are referred to as heat-pluvial events; extreme pluvials followed by humid heat events are termed pluvial-heat events. We aim to gain a comprehensive understanding of temporally compounding heat and pluvial events over the historical period by detecting compound events, conducting decomposition analysis of warming or moistening effects, and identifying influential factors.

Data Sets, Regions, and Seasons
In this study, we identify heat-pluvial and pluvial-heat events using two datasets: the fifth generation of ECMWF (The European Centre for Medium Range Weather Forecasts) global reanalysis (ERA5) (Hersbach et al., 2020) and National Centers for Environmental Prediction (NCEP) (Kalnay et al., 1996), respectively. The choice to use these reanalysis datasets is due to their inclusion of four essential variables (daily mean temperature, precipitation, specific humidity, and air pressure at 2 m from the surface) with global coverage at the daily timescale, which allows for a comprehensive and uniform analysis across different regions and time periods. We use the ERA5 data set for our main analysis, and test the sensitivity of our results with the alternative NCEP data set. To avoid physical inconsistency among different data products, we identify events using variables from each data set independently.
Our analysis spans 1956-2015 and all datasets have been regridded to a common 2.5° × 2.5° grid using the bilinear interpolation. We restrict the analysis to global land areas except for Greenland, Antarctica, and desert regions where annual precipitation is less than 100 mm. We only use data during local summer (May−September in the Northern Hemisphere, November−March in the Southern Hemisphere), as heat and pluvial events occur primarily in the warm season. As with previous global studies (Mukherjee & Mishra, 2021;Perkins-Kirkpatrick & Lewis, 2020), our selections of local summer seasons may miss some events in tropical regions where locally extreme heat may occur at almost any time of year.

Identification of Successive Heat and Pluvial Events, Frequencies, and Trends
We consider locally and seasonally varying thresholds when defining heat and pluvial events, with percentile values calculated from the entire 60-year period . Consistent with previous research, we utilize the wet bulb temperature calculation method based on Bolton's formula as proposed by Davies-Jones to identify heat events (Buzan et al., 2015;Davies-Jones, 2008;Raymond, Matthews, & Horton, 2020). A heat event includes the combined effect of high temperature and high humidity as characterized by the wet-bulb temperature and is defined as the whole available period of data when the daily wet-bulb temperature exceeds the 90th percentile for at least three consecutive days. The weighted average of precipitation is adopted to identify pluvial events (Lu, 2009). The index comprises day-of precipitation as well as the gradually diminishing impact of earlier precipitation by using a weighted average, where the weight declines proportionally with each passing day Liao et al., 2021) (see details in Text S1 and Figure S1 in Supporting Information S1). A pluvial event is defined as the whole available period of data when the daily weighted average of precipitation exceeds the 90th percentile for at least three consecutive days.
Successive heat-pluvial and pluvial-heat events are heat events followed by pluvial events within a 7-day interval, and likewise for pluvial-heat events (see a sensitivity analysis of the time interval in Supporting Information S1). The 7-day interval was selected to balance the trade-off between potential impact and adequate sample size. We have also tested alternative time intervals, which are described in the Supporting Information S1. To handle the occurrence of multiple heat events and pluvial events within a week, we cluster two successive heat or pluvial events into a single event if they are separated by 2 days or less.
We also calculate spatiotemporal changes in successive heat-pluvial and pluvial-heat events between the two 30-year periods (1986-2015 minus 1956-1985). To better compare the trends between different datasets, we use a normalized frequency ratio where the raw frequencies are first normalized by calculating annual values as a fraction of the 1956-2015 mean. The linear trend of annual values of successive heat-pluvial and pluvial-heat events is calculated using the Sen-slope method (unit, decade −1 ) and the related significance is calculated using the Mann-Kendall test.

Moving-Blocks Bootstrap-Resampling-Based Significance Test
As the sequential occurrence of heat and pluvial events at a given location can be relatively rare and largely a matter of chance, traditional methods that estimate compound-event frequency based on event coincidence may struggle to identify causal relationships leading to successive extremes . To address this issue, we use a bootstrap resampling-based significance test to investigate the dependence of two time series, which can test whether the observations are significantly different from what would be expected due to chance alone. In practice, to consider autocorrelation when randomly sampling time series, the moving-blocks bootstrapping is utilized to perform the significance test (Vogel & Shallcross, 1996;Wilks, 1997) using a block size of 3 days. A sensitivity test using alternative block sizes (such as 5 or 10) did not change the significance of our findings. We implement the moving-blocks bootstrap-resampling-based significance test for each grid cell in the following steps: (a) Identify the heat and pluvial event series from 1956 to 2015. The event series are constructed by assigning a label of '1' (indicating occurrence) to the end day of each event in one seriea and to the start day of each event in another series ( Figure S1 in Supporting Information S1); (b) Generate 1,000 resampled event series using the moving-blocks bootstrap, where each resampled series has the same length as the original series. By randomly permuting the event series, rather than the original daily time series, all relevant statistical attributes can be preserved; (c) Compute the occurrence frequencies of heat-pluvial and pluvial-heat events for each pair of resampled series using a pre-determined method (Section 2.2); (d) Compute the empirical distribution of consecutive occurrence frequencies using the 1,000 resampled series; (e) Compute the 95% confidence intervals of the empirical distribution of consecutive occurrence frequencies. (f) Compute the occurrence frequency of heat-pluvial and pluvial-heat events for the original series based on ERA5 and NCEP datasets. (g) Determine whether the occurrence frequency of heat-pluvial and pluvial-heat events for the original series falls within the 95% confidence interval of the empirical distribution of consecutive occurrence frequencies; (h) If the occurrence frequency of heat-pluvial and pluvial-heat events for the original series is outside the 95% confidence interval, then it is statistically significant at the 0.05 level.

Decomposition of Warming/Moistening Effects
To investigate the specific impacts of warming and moistening on the changes observed in temporally compounding heat and pluvial events between the recent  and past  30-year periods, we conducted a decomposition analysis. The goal is to assess the relative significance of warming and moistening effects in shaping the trends of consecutive heat and pluvial events. The methodology involves constructing four distinct time series realizations: (a) with warming and moistening (based on original observational data); (b) without warming and moistening (by removing trends of wet-bulb temperature and weighted average of precipitation); (c) warming alone (by removing the trend of weighted average of precipitation) and (d) moistening alone (by removing the trend of wet-bulb temperature).

Investigation of VPD Anomalies and Field Significance Test
We investigate the potential impact of atmospheric humidity, measured by vapor pressure deficit (VPD) (Massmann et al., 2019;Yuan et al., 2019), on the abrupt alternation between heat and pluvial events. High VPD values are indicative of dry conditions and can exacerbate heat events, while low VPD values are associated with high humidity and may signal a transition from a heat event to a pluvial event. Therefore, VPD anomalies may serve as a key link between heat and pluvial events, making it a relevant diagnostic for our study. For successive pluvial-heat events, we analyze the differences in VPD anomalies between heat events followed by pluvial events (i.e., heat-pluvial) and those not followed by pluvial events (i.e., heat-without-pluvial), specifically focusing on VPD conditions 1 day after the end of heat events. Similarly, for pluvial-heat events, we examine the differences in 1-day VPD anomalies between pluvial events followed by heat events (i.e., pluvial-heat) and those not followed by heat events (i.e., pluvial-without-heat).
To determine whether the observed differences in VPD are statistically significant and not solely due to chance, we conduct field significance tests to address the issue of multiple hypotheses (Wilks, 2006(Wilks, , 2016. Specifically, we control the false discovery rate (FDR) during these tests to minimize the likelihood of identifying false positive results (type I errors) when multiple tests are performed simultaneously. The FDR represents the proportion of rejected null hypotheses that are true. By controlling the FDR, we can increase confidence in the significance of our findings (Benjamini & Hochberg, 1995;Ventura et al., 2004;Wilks, 2006).
Practically, our analysis involves testing the global null hypothesis ( 0 ), which assumes no statistically significant differences in VPD between heat-pluvial and heat-without-pluvial events. To address the multi-hypothesis issue, we perform field significance tests by controlling FDR rates at a certain level . This involves rejecting local null hypotheses whose p-values are no greater than a threshold FDR .
where is the total number of local tests (i.e., grid points) and 0 is the desired level of significance (0.05). To determine the largest K satisfying the equation, we need to order the p-values; any local tests with p-values smaller than or equal to the largest p-value are deemed to be field-significant (Wilks, 2006).

Global Climatology of Heat-Pluvial and Pluvial-Heat Events
Using two independent reanalysis datasets (ERA5 and NCEP), we quantify the global frequency of successive heat-pluvial and pluvial-heat events during 1956-2015. As shown in Figure 1, successive heat-pluvial events have occurred for almost all global land (with desert and polar regions excluded as noted in Section 2.1). The total number of successive heat-pluvial events observed in the two datasets is about 11 on average during 1956-2015 for each grid cell. Successive pluvial-heat events occur slightly less frequently, with an average of approximately 10 over the 60-year study period. Furthermore, we confirm the significance of the probability of successive events being greater than that expected by chance through the utilization of a moving-blocks bootstrap resampling-based significance test (Figures 1c and 1d). The test shows that the number of detected events based on both ERA5 and NCEP datasets exceed the 95% confidence interval estimates from moving-blocks bootstrap resampling that occur by chance, corresponding to the frequency of heat-pluvial and pluvial-heat of 9.74 and 8.48, respectively. This indicates that, on average globally, successive heat-pluvial events occur about 13%-18% more often than would be expected by chance, likely a signature of correlated heat and precipitation via local thermodynamics (i.e., convection) or colliding contrasting air masses (i.e., weather fronts) (  Spatial maps showing the total number of successive heat-pluvial and pluvial-heat events, respectively. Grid cells that are statistically significant at the 0.05 level according to the moving-blocks bootstrap-resampling-based test are depicted as gray circles. The data set used here is ERA5. Panels (c, d) Significance test of the global-mean consecutive occurrence frequencies using moving-blocks bootstrap resampling based on ERA5 data set for the heat-pluvial events (c) and pluvial heat events (d), respectively. The histogram represents the empirical distribution of global-mean successive events frequency using the 1,000 resampled series based on the moving-blocks bootstrap resampling. The 95% confidence interval is indicated by a vertical dashed line. The red dot and blue dot represent the number of successive events detected based on ERA5 and NCEP datasets, respectively.
Looking at the spatial distribution of the number of events, both successive heat-pluvial and pluvial-heat exhibit clear regional differences globally, as illustrated in Figure 1 and Figure S2 in Supporting Information S1. Such temporally compounding extremes occur most often in West Australia, East North America, Sub-Saharan Africa, and North Asia ( Figure S3 in Supporting Information S1), where they are statistically significant at the 0.05 level according to the moving-blocks bootstrap-resampling-based test (Figure 1a). Compared to the successive heat-pluvial events, the pluvial-heat events occur less frequently, but some hotspots maintain consistency-such as West Australia (Figure 1b). The spatial patterns remain similar even when events are defined using a more extreme percentile, despite having reduced peaks ( Figure S4 in Supporting Information S1).

Spatiotemporal Changes in Successive Events and Warming Effect
We investigate the spatiotemporal changes in the frequency of temporally compounding heat and pluvial events from 1956 to 2015, by comparing the first  and second  30-year periods, as shown in Figure 2a and Figure S5 in Supporting Information S1. In most regions, a higher frequency of successive heat-pluvial events is observed in the later  period compared to the earlier period . A rising trend in event frequency is identified in most parts of South America, Sub-Saharan Africa, South Asia, and North Australia, with 15 or more events during the latest 30-year period. In general, the spatial patterns and temporal trends of ERA5 and NCEP are in good agreement (Figures S5 and S6 in Supporting Information S1). However, there are some noticeable discrepancies over the eastern United States and sub-Saharan Africa, which (a) Spatial change in successive heat-pluvial events between the two 30-year periods (recent, 1986-2015 minus past, 1956-1985). (b) Annual time series of the normalized frequency ratio of successive heat-pluvial events. Black line is the annual normalized frequency ratio based on ERA5; blue dashed line is the 30-year average. dRatio is the difference between the averages during 1956-1985 and 1986-2015; Slope is the linear trend using the Sen-slope method (unit, decade-1). (c) The relationship between the changes in successive events and changes in individual extremes. Color circles show bin-averaged ratios of heat-pluvial events corresponding to ratios of individual extremes. (d) Decomposition of the frequency of heat-pluvial events due to warming/moistening effects. It shows the probability density function of the global mean changes in the frequency of heat-pluvial events between the two 30-year periods, based on raw observational data (black), data with moistening signal removed (red), data with warming signal removed (blue) and data with both warming and moistening signals removed (green). may be attributed to data uncertainty and complex relationship between heat and convection, or potentially related to statistical effects from sequential events. The overall frequency of temporally compounding heat and pluvial events has seen a statistically significant (Mann-Kendall test, p < 0.05) increase for both event subcategories, with an increase of about 20%-25% per decade (Figure 2b and Figure S6 in Supporting Information S1).
To explore further how the increases in individual extremes contribute to successive heat and pluvial events, we analyze the relationship between the changes in compound events and changes in individual extremes at the grid level between the two 30-year periods (Figure 2c, Figures S7 and S8 in Supporting Information S1). In general, the upward trend in heat-pluvial event frequency is an expected consequence of these upward trends in univariate hazard frequencies. Specifically, as the frequency of individual heat events increases, there is a simultaneous rise in the occurrence of compound heat-pluvial events across more areas. However, the frequency of individual pluvial events does not exhibit a corresponding increase in these areas ( Figure S9 in Supporting Information S1). This suggests that the increase in successive heat-pluvial events is primarily affected by changes in heat events, and is consistent with our findings for successive pluvial-heat events ( Figures S5-S9 in Supporting Information S1), as well as with other work suggesting that increases in heat dominate the trends in many compound events involving temperature and another variables Liu et al., 2022;Yin et al., 2022).
To explain the increased trend of successive heat and pluvial events, we conduct a decomposition analysis to disentangle the relative importance of the effects of warming and moistening on the trends of successive heat and pluvial events (Figure 2d and Figure S10 in Supporting Information S1). We find that the observed trends in heat-pluvial or pluvial-heat events can be reproduced by considering the influence of warming, specifically the change of wet-bulb temperature. In other words, the changes in these events can primarily be explained by the effect of warming alone (Figure 2d). The effect of warming is especially prominent in South America, South Asia, and North Australia, which are co-located with hotspots in Figure 2a, while moistening without warming has little effect (Figures S10 and S11 in Supporting Information S1). In other words, once the warming effect has been removed there is no "residual" increase in successive events. Therefore, it is the increased heat extremes under a warming climate that have made successive heat-pluvial and pluvial-heat events occur more frequently in recent decades.

Possible Factor Affecting the Transitions Between Heat and Pluvial Events
In this section, we further examine the influence of VPD on transitions between heat and pluvial events. To address the issue of multiple hypothesis testing resulting from spatial dependence, we have conducted a field significance test to determine whether the differences in VPD between heat-pluvial and heat-without-pluvial events are statistically significant or not (Section 2.5). We find significant differences in VPD exist between heat-pluvial and heat-without-pluvial events for about 85% of grid cells at the p = 0.05 level (Figure 3), as well as a similar number for the comparison between pluvial-heat and heat-without-pluvial events using FDR test ( Figure S12 in Supporting Information S1). Importantly, we reveal that small and negative VPD anomalies are linked to the transition from heat to pluvial (Figures 3a and 3c), while high positive VPD anomalies accompany the transition from pluvial to heat (Figures 3b and 3c). We found reduced VPD anomalies are associated with an increased probability of subsequent pluvial events. This is because lower VPD values, indicating higher humidity, create conditions that are more conducive to precipitation. Low VPD can alleviate some of the effects of extreme temperatures on plant health (Grossiord et al., 2020;Novick et al., 2016), and the high moisture content can supply the fuel for pluvial events, especially of a convective nature. On the contrary, increased VPD anomalies imply high atmospheric aridity, related to the termination of pluvial events and the abrupt onset of heat events. This finding echoes the observational evidence that high VPD enhances atmospheric demand for water, depleting soil moisture and simultaneously heating the atmospheric boundary layer (Teuling et al., 2013;Zhou et al., 2019).

Discussions and Conclusions
Humid heat and pluvial flooding are serious weather extremes by themselves, but when occurring sequentially at the same location, they can cause more severe consequences than an isolated extreme event. While extreme heat or pluvial flooding alone has attracted considerable attention over the past decades Martin, 2018;Sun et al., 2021;Wang et al., 2021), the global climatology of successive heat and pluvial events remains unclear. In this study, we perform a comprehensive global assessment of heat-pluvial and pluvial-heat events. The order of these events matters: when heat precedes, the soil becomes parched, hindering water absorption and resulting in flash floods during subsequent rainfall. Conversely, when heavy rainfall occurs first, it damages infrastructure, making it challenging to cope with subsequent heatwaves. Through analyzing both event sequences, our study helps gain a deeper understanding of the risks associated with each sequence. Based on two datasets, we reveal the baseline frequencies and spatiotemporal changes of successive extremes. Our findings demonstrate the increased risk of rapid transition between heat and pluvial events in a warmer climate in recent decades. Hotspots are centered in West Australia, East North America, Sub-Saharan Africa, and North Asia. We find that more frequent heat extremes due to a warming climate have resulted in a higher incidence of heat-pluvial and pluvial-heat events. Furthermore, our findings demonstrate that notable VPD anomalies are typically observed in the transitions between heat and pluvial events.
To ensure the consistency and robustness of the analysis, we conduct multiple sensitivity analyses related to data sources, time intervals and thresholds, and alternative event definition (see details in Text S2 in Supporting Information S1). Although the frequency of successive heat-pluvial and pluvial-heat events is very sensitive to the choice of extreme thresholds (Figures S13 and S14 in Supporting Information S1) and time lags (Figures S15-S18 in Supporting Information S1), our main results of increased trends in successive heat-pluvial and pluvial-heat events are robust (Figures S13 and S14 in Supporting Information S1). Disagreement between datasets has the largest effect on our results for eastern United States and sub-Saharan Africa regions.
Extreme temporally compound events are often rare, and their occurrence can be coincidental due to chance. Our study highlights the value of using bootstrap resampling-based significance tests, a method that has been overlooked in previous studies Zhang & Villarini, 2020). Specifically, it is crucial to consider autocorrelation when randomly sampling time series of precipitation and temperature. This consideration is particularly significant when defining events based on a sequence of consecutive hot or wet days. The moving-blocks bootstrapping method used in this study accounts for the uncertainty in temporally correlated Represents VPD anomalies between heat events followed by pluvial events (heat-pluvial) and heat events not followed by pluvial events (heat-without-pluvial). (b) Represents VPD anomalies between pluvial events followed by heat events (pluvial-heat) and pluvial events not followed by heat events (pluvial-without-heat). Grid points that meet local statistical significance at the 0.05 level are shown as gray circles, while the ones that meet the false discovery rate (FDR) criterion by having a sufficiently small p-value are marked by gray points. (c) Is the probability density function of the map a (green line) and b (red line) for the VPD anomalies causing the transition between heat and pluvial events. (d) Is the FDR for testing field significance of VPD anomalies (Pa) between heat-pluvial and heat-without-pluvial events. event coincidence by generating a set of surrogate time series that have the same statistical properties, including temporal covariance structure or temporal correlation, as the original time series (Vogel & Shallcross, 1996).
We also show VPD has crucial but different impacts on successive heat-pluvial and pluvial-heat events, respectively. Small and negative VPD anomalies are linked to successive heat-pluvial events, while pluvial-heat events are more likely to exhibit a positive VPD anomaly. We confirm that observed differences in VPD are not due to random chance or measurement error, but rather reflect real differences between the two types of events based on field significance test. It is noteworthy that our findings reveal significant difference in VPD between heat-pluvial and heat-without-pluvial events for approximately 85% of grid cells. This reinforces the importance of VPD as a meaningful metric for understanding and distinguishing between these event types. To ensure the robustness of our analysis, we also use two alternative methods: Walker's test and moving block bootstrapping-based multivariate test, both of which provide consistent results (see details in Text S3 and Figures S19-S22 in Supporting Information S1). In our study, we find that while 1293 grid elements achieved local statistical significance at the 0.05 level, 1282 grid elements have p-values that meet the FRD criterion, indicating statistically field significance. Although the difference is small, the latter approach using field significance tests provides a more accurate and reliable statistical significance assessment and helps prevent overestimating results.
Our study indicates that VPD plays a vital role in temporally compounding heat and pluvial events, which are often ignored in previous compound events. Importantly, we highlight the asymmetric impacts of VPD on the rapid transitions between heat and pluvial extremes, which could provide a reference and insight into early warning and anticipation of emerging temporally compounding hydrological hazards. The physical mechanisms underlying compound heat and pluvial events are complex. While detecting and presenting a global assessment of two emerging compound extremes is our priority and focus in this study, identifying the process-based evolution and underlying mechanisms of the rapid transition from extreme heat to pluvial or vice versa from a physical standpoint is an important and challenging task. Future studies should be undertaken to further investigate these mechanisms and help advance our understanding of compound hazards.