Impacts of Future Changes in Heavy Precipitation and Extreme Drought on the Economy over South China and Indochina

As two of the most damaging natural disaster types in the world, heavy precipitation and extreme drought cause serious economic losses and casualties almost every year (Philip et al., 2018; Wang et al., 2019a; Nanding et al., 2020; Nangombe et al., 2020; Zhang et al., 2020; Lu et al., 2021; Min et al., 2022; Qian et al., 2022; Tang et al., 2022b). According to the statistics of the Emergency Events Database, the total economic damage caused by extreme flooding and extreme drought reached approximately 651 and 128 billion during 2000 to 2019, respectively (http://www.emdat.be). For example, a persistent extreme drought event occurred in Yunnan Province, China, from spring to summer 2019, for which the direct economic losses totaled approximately 6.56 billion RMB (Wang et al., 2021). More specifically, around 7 million residents were affected during this event, and crop failure was caused over at least 1.35 × 10⁴ km² of cropland. In the past few decades, severe heavy precipitation and extreme drought events have occurred more and more frequently across the world (Zhang and Zhou, 2019; Chang et al., 2020; Yao et al., 2021; Yu and Zhong, 2021). More importantly, significant changes in heavy precipitation and extreme drought can be expected in the future throughout the world, such as in Europe (Hawcroft et al., 2018; Ruosteenoja et al., 2018), North America (Akinsanola et al., 2020; Chen and Ford 2021), central Asia (Fan et al., 2021; Yao et al., 2021), Southeast Asia (Zhu et al., 2020; Tang et al., 2021b), Australia (Alexander and Arblaster, 2017), South America (Li et al., 2020), and Africa (Ayugi et al., 2021; Nooni et al., 2022). Given that changes in heavy precipitation and extreme drought may pose serious economic threats, anticipating such changes and their impacts on economies in the future is therefore critical for regional sustainable development.

Influenced by its complex terrain and strong air–sea interaction, the historical changes that have taken place in heavy precipitation over the South China and Indochina (INCSC) region are more intricate compared to those in other regions (Li et al., 2019; Tang et al., 2021a). Moreover, the expected future changes in total amount and frequency of extreme precipitation within the INCSC region are quite uneven, manifesting as a larger increase over the Malay Peninsula, Cambodia, southern Myanmar, and southern regions of South China, and a smaller increase over Thailand and Vietnam under three of the Shared Socioeconomic Pathways (SSPs) (Tang et al., 2021b). Dominated by developing countries, the economic growth, industrial structure, and resource allocation of the INCSC region are also highly unbalanced. For example, affected by multiple factors (e.g., geographical conditions), the industrial structure in Vietnam and the Malay Peninsula shows significant differences. It is undeniable that the economic development of the INCSC region is sensitive to changes in heavy precipitation and extreme drought, because a considerable number of people in this region live within 100 km of the coastline, and the region’s economic development is skewed towards coastal areas (Ngai et al., 2020; 2020; Ge et al., 2021; Overland et al., 2021; Qin and Dai, 2022). Against this background, there is an urgent need to clarify the impact of future changes in heavy precipitation and extreme drought on the gross domestic product (GDP) over the INCSC region.

Suppressing global warming to 1.5°C relative to preindustrial levels is the target of the Paris Agreement (IPCC 2018). Besides, informing mitigation and adaptation policy considerations is one of the primary objectives of the Scenario Model Intercomparison Project for phase 6 of the Coupled Model Intercomparison Project (CMIP6) (O’Neill et al., 2016). Assessing the avoided impact by limiting global warming to 1.5°C instead of other specific global warming levels (e.g., 2°C) is also crucial for designing adaptation and mitigation strategies, and therefore has been documented in several existing studies based on CMIP6 outputs (e.g., Zhu et al., 2020; Tang et al., 2022a). Besides, although the possible future changes in heavy precipitation over the INCSC region have been explored in some previous studies (e.g., Ge et al., 2021; Tang et al., 2021b), projecting the future changes in heavy precipitation and extreme drought at the specific levels of global warming using CMIP6 outputs with bias correction remains an as yet unstudied issue.

In addition, changes in extreme events have substantial impacts on demography (Chen and Sun, 2021; Tuholske et al., 2021; Park and Jeong, 2022; Ullah et al., 2022), economics (Gu et al., 2020; Kotz et al., 2022), and ecology (Elad and Pertot, 2014; Bao et al., 2022), and yet the potential impact of heavy precipitation and extreme drought on GDP over the INCSC region at specific global warming levels is still poorly studied. By limiting global warming to 1.5°C instead of other higher global warming levels of 2°C, 3°C and 4°C, the changes in GDP impacts induced by the associated changes in heavy precipitation and extreme drought are deserving of further investigation. This issue is particularly important for risk aversion, especially in the INCSC region, which is densely populated and has a developing economy. Therefore, this study was designed to answer three questions: (1) How will heavy precipitation and extreme drought change with global warming in the INCSC region and what will the economic impacts be? (2) To what extent are the changes in heavy precipitation and extreme drought and their impacts on GDP alleviated by limiting global warming to 1.5°C instead of 2°C, 3°C and 4°C? (3) What are the relative contributions of climate change and GDP change to the effects on GDP?

The remainder of this paper is arranged as follows. Section 2 describes the data and methods employed in this study. Section 3 presents results for model performance and diversity, future projections of heavy precipitation and extreme drought with bias correction and their impacts on GDP, as well as the differences between the 1.5°C and 2°C, 3°C and 4°C global warming levels, and the relative contributions of climate change and GDP change for the impacts on GDP. Section 4 summarizes the main results and provides some discussion.

Within the range of 0°–30°N and 90°–120°E, the INCSC region (Fig. 1) mainly includes eight provinces in South China and five countries as well as the Malay Peninsula in Indochina. Bordering the Bay of Bengal and the Andaman Sea in the west and the South China Sea in the east, the INCSC region is a bridge between East Asia and the Maritime Continent. Furthermore, climate change in the future may pose great challenges to the economic development of the INCSC region, which is recognized as one of the hotspots of climate change and where economic development is highly sensitive to climate change (Seneviratne et al., 2021).

Figure 1.  Orography (units: m) and geographical location of the INCSC region, including South China (Yunnan, Guizhou, Hunan, Jiangxi, Fujian, Guangdong, Guangxi, and Hainan provinces) and Indochina (Myanmar, Thailand, Laos, Vietnam, Cambodia, and the Malay Peninsula).

The daily gridded precipitation dataset provided by the APHRODITE (Asian Precipitation–Highly-Resolved Observational Data Integration Towards the Evaluation of Water Resources) project (Yatagai et al., 2012) is used as the observational data. The time span and spatial resolution of this dataset are 1979–2014 and 0.25° × 0.25°, respectively. The daily precipitation and monthly surface temperature are derived from the historical (1850–2014) and future SSP5-8.5 (2015–2100) experiments (O’Neill et al., 2016) of 25 CMIP6 climate models (Table 1). To ensure that as many models as possible are covered, only one realization of each model was used in this study. The multimodel ensemble median (MMM) rather than the multimodel ensemble mean is used to guard against CMIP6 models with exceptionally large errors (outliers) unduly affecting our results (Gleckler et al., 2008). It is worth noting that the results under SSP1-2.6 and SSP2-4.5 are similar to that under SSP5-8.5 (figures not shown).

Under the SSPs global framework, Jiang et al. (2017) updated the GDP projection by using the Cobb–Douglas model with localized economic parameters. A dataset of GDP projections with a higher spatiotemporal resolution (0.5° × 0.5° and yearly) for 177 countries around the world from 2010 to 2100 under different SSPs was obtained. This latest dataset has already been used in many studies to conduct risk assessments and future projections (e.g., Huang et al., 2019; Peng and Li, 2021; Zhou et al., 2022) and can be obtained at https://cstr.cn/31253.11.sciencedb.01683.

Two extreme precipitation indices (Table 2) recommended by the Expert Team on Climate Change Detection and Indices are adopted in this study (Zhang et al., 2011): the maximum consecutive 5-day precipitation (RX5day) and the maximum consecutive dry days (CDD). The former (also referred to simply as heavy precipitation) is widely used for flood risk assessments, and the latter (also referred to simply as extreme drought) is usually regarded as a dryness indicator (Seneviratne et al., 2012; Wartenburger et al., 2017; Zhang et al., 2020). As different CMIP6 models have varied spatial resolutions, we calculate the two extreme precipitation indices on their native grids and then use the bilinear interpolation technique to interpolate them to a common 0.5° × 0.5° grid to match the GDP data (Wang et al., 2019b; Liu et al., 2020; Zhai et al., 2020).

Due to the widespread bias that exists in model simulations of extreme precipitation indices, the variance scaling method, which is a bias correction method, is utilized in our study to increase the credibility of the projections (Peng et al., 2020). Details of the variance scaling method can be obtained in the supplementary material.

Following the study of Zhang et al. (2018), the response rate of extreme precipitation indices to global warming is calculated by the following steps. First, we calculate the time series of two extreme precipitation indices at each grid point and the time series of the global mean surface temperature by calculating the 5-year overlapping mean for each decadal period, which means 2015–2024, 2020–2029, till 2090–2099. Then, the linear regression coefficients between the time series of the extreme precipitation indices and the global mean surface temperature are referred to as the response rates in this study.

To discern the time taken to reach specific warming levels (i.e., 1.5°C, 2°C, 3°C, and 4°C) relative to the pre-industrial period (1861‒90), a 21-year time-slice method is used (Peng et al., 2020). For a certain CMIP6 model, the time taken is determined as the first time when the 21-year running average of the global mean surface temperature change achieves 1.5°C, 2°C, 3°C, and 4°C relative to the pre-industrial level under SSP5-8.5.

According to Zhang and Zhou (2020), population-weighted regional-average change represents the impact of extreme precipitation changes on population. Similarly, GDP-weighted regional-average change is used to detect the impact of heavy precipitation and extreme drought on the economy. The specific steps are as follows: Firstly, we calculate the GDP-weighted change $ {\Delta W}_{i,j} $ at each grid point $ (i,j) $ by multiplying the change in extreme precipitation indices $ {\Delta I}_{i,j} $ by the GDP fraction $ {{G}_{i,j}}/{{G}_{\mathrm{T}\mathrm{O}\mathrm{T}}} $. Then, the GDP-weighted regional-average change $ \overline{\Delta W} $ is the sum of $ {\Delta W}_{i,j} $ across a region, which can be expressed as

where $ {\phi }_{1} $ and $ {\phi }_{2} $ represent the latitudes spanning the specific region, and $ {\lambda }_{1} $ and $ {\lambda }_{2} $ the longitudes. The GDP fraction at each grid point $ {{G}_{i,j}}/{{G}_{\mathrm{T}\mathrm{O}\mathrm{T}}} $ within South China is defined as the ratio of GDP at each grid point $ {G}_{i,j} $ to the total GDP of South China $ {G}_{\mathrm{T}\mathrm{O}\mathrm{T}} $, TOT means the total, and the same is true for Indochina. The calculation of the GDP fraction in each province in South China or country in Indochina is based on the total GDP for the corresponding province or country, which is employed only in section 3.4. Note that we use the GDP fraction instead of the absolute value of GDP in a certain region to avoid covering up countries (or provinces) severely affected by heavy precipitation and extreme drought where the national (or provincial) GDP is low in terms of the whole region.

From Eq. (1), the impact of extreme precipitation indices on GDP is jointly determined by changes in both climate and GDP. If using a fixed future GDP in 2100 as the weight, the combined impact on GDP induced by both climate change and GDP change (hereafter termed the COM impact on GDP) across a region $ \overline{{\Delta W}_{\mathrm{C}\mathrm{O}\mathrm{M}}} $ can be expressed as

The climate-change-alone-induced impact on GDP (hereafter termed the CCA impact on GDP) across a region $ \overline{{\Delta W}_{\mathrm{C}\mathrm{C}\mathrm{A}}} $ is clarified by the regional-averaged change based on a fixed GDP of 2010 as the weight, as follows:

The difference between $ \overline{{\Delta W}_{\mathrm{C}\mathrm{O}\mathrm{M}}} $ and $ \overline{{\Delta W}_{\mathrm{C}\mathrm{C}\mathrm{A}}} $ can be considered as the contribution from GDP change (hereafter termed the GDPC contribution) to the COM-impact on GDP across a region ($ \overline{{\Delta W}_{\mathrm{G}\mathrm{D}\mathrm{P}\mathrm{C}}}=\overline{{\Delta W}_{\mathrm{C}\mathrm{O}\mathrm{M}}}-\overline{{\Delta W}_{\mathrm{C}\mathrm{C}\mathrm{A}}} $).

Climatologically, the observed maxima of persistent heavy rainfall (RX5day) are located in southeastern South China and the east and west coastal regions of Indochina (Fig. 2a). The MMM before correction shows a significant overestimation for RX5day (Fig. 2b), which can be identified by the larger MMM values of RX5day than from the observations over most regions of INCSC. In general, EC-Earth3-Veg and EC-Earth3 present more accurate simulations of RX5day (Figs. S1a11 and a12 in the electronic supplementary material, whereas BCC-CSM2-MR and CanESM5 show severe overestimation when compared with other CMIP6 models (Figs. S1a6 and a7 in the ESM). For the longest dry spell length (CDD), the observed climatological CDD generally increases from the northeast towards the southwest of the INCSC region (Fig. 2d). The MMM before correction can simulate the spatial distribution of CDD in general with an overestimation along the western coast (Fig. 2e). Among the CMIP6 models, CNRM-CM6-1 and CNRM-ESM1-5 tend to demonstrate the best spatial simulation skill for CDD (Figs. S2a8 and a9); however, a severe underestimation of CDD can be found in INM-CM-8 and INM-CM5-0 (Figs. S2a17 and a18 in the ESM). In particular, the area around Yangon in Myanmar is threatened by both persistent heavy precipitation and extreme drought, as indicated by the large values of RX5day and CDD both observationally and in the MMM.

Figure 2.  Climatological spatial distribution (calculated for the period 1979‒2014) of (a‒c) RX5day and (d‒f) CDD for the observation and the MMM before and after bias correction.

As can be seen in Fig. 2, the value of absolute error for both RX5day and CDD is significantly reduced after employing bias correction (Figs. 2b, c; 2e, f), with specific values shown in Tables S1‒S4 (in the electronic supplementary material). Meanwhile, the spatial correlation coefficient has also significantly improved (Tables S1‒S4). Therefore, bias correction is used to constrain the future projections of heavy precipitation and extreme drought, and all subsequent results are based on these corrected indices.

The CMIP6 models and their MMM also vary considerably in their timings of the different levels of global warming (i.e., 1.5°C, 2°C, 3°C, and 4°C) relative to the pre-industrial period under SSP5-8.5 (Fig. 3). For the MMM, the years when warming of 1.5°C, 2°C, 3°C and 4°C occurs are roughly 2024, 2034, 2051 and 2066, respectively. Moreover, the timing also varies among individual CMIP6 models; for example, CanESM5 and EC-Earth3-Veg reach the 2°C global warming level before 2020, while NorESM2-LM does not reach it until 2050.

Figure 3.  Year that each CMIP6 model and the MMM reach the different global warming levels of 1.5°C, 2°C, 3°C and 4°C under SSP5-8.5. The years are defined by the mid-year of a 21-year running mean period.

The response rates of heavy precipitation and extreme drought to future global warming are investigated based on future data in the period of 2015‒2100. Under SSP5-8.5, heavy precipitation events (i.e., RX5day) intensify robustly across the whole INCSC region, as indicated by the response rates over the whole region being positive (Fig. 4a). The area-averaged MMM response rate of RX5day to global warming is about 6.85% K⁻¹ (interquartile model range: 3.92%‒9.47% K⁻¹), which approximately fits the Clausius–Clapeyron relation. A larger response rate for RX5day can be expected in western Myanmar and the Malay Peninsula. In other words, compared with other regions, western Myanmar and the Malay Peninsula may experience more intense heavy precipitation in the future. As for CDD, it intensifies over the whole region except in central Myanmar (Fig. 4e). Area-averaged over the INCSC region, the increase in CDD in response to global warming is about 3.41% K⁻¹ (interquartile model range: 1.81%‒7.41% K⁻¹). Larger response rates appear over the Malay Peninsula, Cambodia, and the provinces of Guangdong and Fujian in China, indicating that the duration of consecutive dry days may increase in these regions in the future. Besides, different from RX5day, the response rate of CDD to global warming varies greatly among individual models, with lower model agreement, as indicated by the smaller stippled areas in Fig. 4e.

Figure 4.  MMM responses of (a) RX5day and (e) CDD to global warming over the INCSC region during 2015–2100 and relative changes of (b–d) RX5day and (f–h) CDD from global warming of 1.5°C to 2°C, 3°C and 4°C under SSP5-8.5 relative to 1995–2014. Stippling denotes areas where at least 2/3 of the models agree on the sign of the change. The slants indicate statistical significance at the 90% confidence level by employing the two-tailed Student’s t-test.

Compared to the 1.5°C global warming level, the 3°C and 4°C levels are projected to cause a comprehensive increase in RX5day (Figs. 4c and d). Whereas, the changes in RX5day at the 1.5°C and 2°C warming levels lack spatial consistency, showing lower model agreement (Fig. 4b). This may be caused by model nondeterminacy and internal climate variability (Hawkins and Sutton, 2009, 2011; Lehner et al., 2020). The former is due to structural differences between models, while the latter is due to natural periodic oscillations within the climate system, and both cannot be ignored when considering an additional 0.5°C warming (Zhang and Zhou, 2020). For extreme drought, the changes in CDD between the 1.5°C and other warming levels generally show a higher model agreement in South China but a lower model agreement in Indochina (Figs. 4f‒h) compared with RX5day. Besides, the CDD in central Myanmar will increase for an additional warming of 0.5°C (Fig. 4f), which is different from the result obtained in Ge et al. (2019) and Supari et al. (2020).

To conclude, heavy precipitation tends to intensify across the whole INCSC region, especially in western Myanmar and the Malay Peninsula, while the duration of consecutive dry days will increase over the whole region, except central Myanmar, against the background of global warming. Both heavy precipitation and extreme drought show a significant strengthening over most of the INCSC region at the 4°C global warming level relative to the 1.5°C level.

Figure 5 displays the spatial distribution of GDP and GDP fraction in 2010 and the MMM responses of GDP weighted changes in heavy precipitation and extreme drought to global warming under SSP5-8.5. In 2010, the areas with larger GDP are concentrated in Thailand, the Malay Peninsula and South China (Fig. 5a). The distribution of GDP fraction in 2010 is closely consistent with that of GDP (Fig. 5b). When taking the GDP fraction distribution in 2010 into account, the response rates of GDP weighted changes in two extreme precipitation indices to global warming are generally larger in Guangdong Province, Hunan Province, Fujian Province, Jiangxi Province, Hainan Province, the Malay Peninsula, and southern Cambodia (Figs. 5c and d). This implies that the changes in heavy precipitation and extreme drought will exert a greater CCA impact on GDP in these regions.

Figure 5.  (a) Spatial distribution of observed GDP (units: billion USD). (b) Spatial distribution of observed GDP fraction in 2010 (units: 10⁻⁴). (c) MMM responses of GDP weighted changes in RX5day to global warming (units: %² K⁻¹). (d) MMM responses of GDP weighted changes in CDD to global warming (units: %² K⁻¹). Stippling denotes where at least 2/3 of the models agree on the sign of the change.

As mentioned above, changes in precipitation extremes might be occluded by the noise at regional scales. The probability distributions of the GDP fraction with specific changes in RX5day and CDD are employed here, as this shows strong signals of change in precipitation extremes in the whole region, which can provide more information than the simple regional average. In this part, we investigate the probability distributions of GDP fraction affected by the certain changes in heavy precipitation and extreme drought at specific global warming levels over South China and Indochina under SSP5-8.5 (Fig. 6). The probability distributions were calculated as follows. Firstly, we calculated the change in heavy precipitation and extreme drought at each grid point. Secondly, we obtained the GDP fraction of each grid point. Thirdly, for a specific change (x-axis), we could calculate the sum of the GDP fractions at grid points where this change occurred (y-axis), and then form the distribution.

Figure 6.  Probability distributions of GDP fraction when experiencing certain changes in (a, b) RX5day and (c, d) CDD at global warming levels of 1.5°C, 2°C, 3°C and 4°C under SSP5-8.5 over (a, c) South China and (b, d) Indochina. The changes in RX5day and CDD are given as percentages relative to 1995–2014 and the observed GDP fraction in 2010 is used to calculate the probability distributions. Solid lines represent the MMM and shading represents the interquartile model range. The box-and-whisker plots denote the MMM changes to the 10th, 25th, 50th, 75th, and 90th percentiles of GDP fraction, and the dots denote the GDP weighted average changes.

For heavy precipitation (RX5day), the distribution of GDP fraction at the 3°C and 4°C warming levels stretches more in the upper tail over both South China and Indochina compared to other warming levels (Figs. 6a and b). Therefore, more intense heavy precipitation events are expected to influence a larger proportion of GDP over South China and Indochina with global warming. According to the MMM estimates, one quarter of the GDP in South China (Indochina) will suffer from an intensification of RX5day, with more than 4.58% (2.38%), 7.99% (3.34%), 10.61% (6.33%), and 16.44% (9.49%) at 1.5°C, 2°C, 3°C and 4°C warming levels, respectively (box-and-whisker plots in Figs. 6a and b). For extreme drought (CDD), there is no consistent shift in the GDP fraction distribution with global warming over both South China and Indochina. If the GDP distribution remains as that in 2010, 10% of the GDP in South China suffers from an increase in extreme drought by more than 2.86%, 9.66%, and 4.80% at 2°C, 3°C and 4°C warming levels, respectively (box-and-whisker plot in Fig. 6c). In Indochina, 10% of the GDP suffers from an increase in extreme drought by over 2.39%, 0.40% and 2.61% at 1.5°C, 2°C and 4°C warming levels, respectively (box-and-whisker plot in Fig. 6d). As the CDD in nearly the whole of South China at the global warming level of 1.5°C is projected to be shorter (Fig. S5e), a 90% share of the GDP in South China experiences a decrease in dry spells at that level of warming (Fig. 6c). Similarly, a 90% share of the GDP in Indochina experiences a decrease in dry spells at the 3°C warming level (Fig. 6d).

Next, we investigate to what extent the change in extreme precipitation indices and CCA impact on GDP can be alleviated if the 1.5°C warming is achieved rather than the other higher warming levels of 2°C, 3°C and 4°C (Fig. 7). As can be seen, when limiting to the 1.5°C warming target, RX5day is weaker and CDD shorter in South China, i.e., the differences in the area-averaged changes of these indices between 1.5°C and the three higher global warming levels are all positive in this region (Figs. 7a and c). As for Indochina, more intense heavy precipitation can be expected at the three higher global warming levels, and longer consecutive dry days can be expected at the 4°C warming level (Figs. 7a and c). When considering the distribution of GDP fraction, the CCA impact on GDP ($ \overline{{\Delta W}_{\mathrm{C}\mathrm{C}\mathrm{A}}} $) related to RX5day over both South China and Indochina is consistently reduced at the 1.5°C warming level compared to the higher warming levels of 2°C, 3°C and 4°C (Fig. 7b). This means that a larger proportion of the GDP in the INCSC region will be affected by the increased heavy precipitation under global warming. For extreme drought, the additional $ \overline{{\Delta W}_{\mathrm{C}\mathrm{C}\mathrm{A}}} $ over South China is positive from global warming of 1.5°C to 2°C, 3°C and 4°C, while the additional $ \overline{{\Delta W}_{\mathrm{C}\mathrm{C}\mathrm{A}}} $ over Indochina is negative at the 2°C and 3°C level but positive at the 4°C level compared with 1.5°C of warming (Fig. 7d). This indicates that more prolonged dry spells over South China can be expected and will influence more of the local GDP share at higher warming levels, while extreme drought in Indochina is more likely to increase at the 4°C level, thus affecting the GDP proportion there. It is worth noting that there is an obvious regional difference between Indochina and South China for the change in CDD when limiting global warming level from 2°C and 3°C to 1.5°C. The area-average changes in CDD from global warming of 2°C and 3°C under SSP5-8.5 show increases compared with that of 1.5°C in South China, while they show decreases in Indochina (Figs. 7c and d). This is closely related to the enhanced water vapor flux divergence in South China and weak divergence in Indochina under 2°C and 3°C global warming levels compared to 1.5°C (Fig. S6), leading to the increase and decrease of CDD. In addition, model consistency for the reduced changes and $ \overline{{\Delta W}_{\mathrm{C}\mathrm{C}\mathrm{A}}} $ related to heavy precipitation is higher than for extreme drought. This can be partly explained by the stronger coupling of heavy precipitation than extreme drought with global mean temperature, as the scatter distribution in Fig. S3 being more concentrated than in Fig. S4 demonstrates.

Figure 7.  (a) Additional area-average changes in RX5day from global warming of 1.5°C to 2°C, 3°C and 4°C under SSP5-8.5 over South China and Indochina based on the MMM (%, relative to 1995–2014). (b) As in (a) but for additional CCA impacts related to RX5day on GDP (%²). (c, d) As in (a, b) but for CDD. The distribution of GDP fraction in 2010 is used as the weight. Dots indicate where at least 2/3 of the models agree on the sign of the change.

Briefly, when considering the GDP fraction in 2010, the changes in both heavy precipitation and extreme drought under global warming bring a larger economic impact to Guangdong Province, Hunan Province, Fujian Province, Jiangxi Province, Hainan Province, the Malay Peninsula, and southern Cambodia, as compared with other regions. Probability distribution analysis indicates that more intense heavy precipitation events will influence a larger proportion of the GDP over both South China and Indochina under global warming, while the GDP fraction affected by extreme drought varies with different levels of global warming and different regions being considered. From the subregional-average perspective, more than two thirds of the CMIP6 models agree that reduced economic impacts induced by heavy precipitation can be expected over South China and Indochina if global warming is limited to 1.5°C instead of 3°C or 4°C, while the economic impacts caused by extreme drought over South China will be lowered if global warming is limited to 1.5°C rather than 2°C or 3°C.

As shown above, large spatial variability of changes in both precipitation extremes and CCA impacts on GDP can be observed. Therefore, a more detailed and in-depth analysis at national and provincial scales is needed. Specifically, we investigate more thoroughly the influence of an increase in heavy precipitation and extreme drought on GDP fractions in each province of South China and each nation of Indochina (Fig. 8). Besides, Tables S5 and S6 present the GDP faction in each province of South China and each country of Indochina affected by specific magnitudes of changes in RX5day and CDD (expressed as a percentage) relative to 1995‒2014.

Figure 8.  The affected GDP faction in each province/country when the change in the MMMs of RX5day and CDD under SSP5-8.5 relative to 1995–2014 exceeds zero.

For RX5day, there is a consistent projection that a greater proportion of GDP in each province of South China or each country of Indochina will be affected by more intense heavy precipitation at the 4°C warming level (Fig. 8d) compared with the 1.5°C warming level (Fig. 8a). The number of provinces in South China (nations in Indochina) where the GDP fraction influenced by stronger RX5day amounts to more than 50% is projected to increase from 4 (3) at the 1.5°C warming level to 6 (5), 7 (4) and 7 (6) at global warming of 2°C, 3°C and 4°C, respectively (Figs. 8a‒d). Furthermore, the GDP fraction affected by stronger RX5day in Guangxi Province of South China and the Malay Peninsula of Indochina keeps expanding. For example, the GDP fractions affected by more intense RX5day in the Malay Peninsula at the 1.5°C, 2°C, 3°C and 4°C warming levels are 42.84%, 62.65%, 66.94% and 77.52%, respectively (Figs. 8a‒d and Table S5). For CDD, the number of provinces in South China (nations in Indochina) where the GDP fraction affected by longer CDD is more than 50% is also projected to increase at the 4°C warming level (Fig. 8d) compared with the 1.5°C warming level (Fig. 8a). However, the GDP fraction affected by the longer consecutive dry days in different provinces of South China (nations of Indochina) seems not to show consistent changes with global warming (Figs. 8e‒h and Table S6). As we can see from Tables S5 and S6, the affected GDP fractions in all provinces of South China and all nations of Indochina are less than 50% as the two extreme precipitation indices increase by more than 20% relative to 1995‒2014.

Overall, more intense heavy rainfall is projected to affect more GDP fractions in each province of South China or each country of Indochina at the 4°C warming level compared with the 1.5°C warming level. The 4°C warming level increases the number of provinces in South China or nations in Indochina where more than 50% of the GDP fraction suffers from stronger heavy precipitation and longer consecutive dry days compared with the 1.5°C warming level.

Figure 9 shows the change in the absolute GDP and GDP fraction in 2100 relative to 2010 under SSP5 [note that SSP5 describes alternative society evolutions in the future without considering climate change or climate policy, whereas SSP5-8.5 also includes stabilized radiative forcing (8.5 W m⁻²) by the end of the 21st century]. In 2100, the absolute GDP is projected to have increased over the whole INCSC region under SSP5, with the largest increase in Hunan Province, Guangdong Province, Fujian Province, and Hainan Province (Fig. 9a). In South China, the GDP fraction in 2100 increases in Fujian Province and Guangdong Province under SSP5 (Fig. 9b). In Indochina, the GDP fraction in 2100 increases in Myanmar, Laos, and Vietnam, but decreases in Thailand and the Malay Peninsula under SSP5 (Fig. 9b). These results mean that, although the absolute GDP will likely increase over the whole INCSC region in 2100 compared with that in 2010, there is a process of GDP fraction change among the provinces in South China and among the nations in Indochina. The provinces of Guangdong and Fujian will likely increase their share of South China’s GDP by 2100, while the nations of Laos and Vietnam will likely increase their share of Indochina’s.

Figure 9.  Spatial distribution of (a) GDP change (units: billion USD) and (b) GDP fraction change (units: 10⁻⁴) in 2100 relative to 2010 under SSP5.

As we can see from Eq. (2), the COM impact on GDP is determined by both climate change and GDP change. Figure 10 presents the relative contributions of these two factors (climate change $ \overline{{\Delta W}_{\mathrm{C}\mathrm{C}\mathrm{A}}} $ and GDP change $ \overline{{\Delta W}_{\mathrm{G}\mathrm{D}\mathrm{P}\mathrm{C}}} $) to the COM impacts on GDP. Note that the bars denote the MMMs rather than the multimodel means, and thus the sums of the dotted and solid bars do not need to be equal to the values of the starred bars. In South China, the GDPC contribution dominates the COM impacts on GDP for heavy precipitation (RX5day) at the 1.5°C level, but is almost negligible at the 2°C, 3°C and 4°C global warming levels (Fig. 10a). For extreme drought (CDD), the GDPC contribution in South China is insignificant at all global warming levels, as $ \overline{{\Delta W}_{\mathrm{G}\mathrm{D}\mathrm{P}\mathrm{C}}} $ is relatively smaller than $ \overline{{\Delta W}_{\mathrm{C}\mathrm{C}\mathrm{A}}} $ (Fig. 10b). In Indochina, the contributions of climate change and GDP change related to RX5day and CDD are both positive or negative at the 1.5°C and 2°C warming levels, which jointly promote the change in COM impacts on GDP (Figs. 10c and d in the ESM). Whereas, at the 4°C warming level, the sign of $ \overline{{\Delta W}_{\mathrm{G}\mathrm{D}\mathrm{P}\mathrm{C}}} $ related to RX5day and CDD is opposite to that of $ \overline{{\Delta W}_{\mathrm{C}\mathrm{C}\mathrm{A}}} $, which means the GDPC contribution offsets part of the CCA impact on GDP.

Figure 10.  Contributions of climate change and future GDP change to the COM impacts on GDP at global warming levels of 1.5°C, 2°C, 3°C and 4°C under SSP5-8.5 (units: %, relative to 1995–2014): (a, c) for RX5day; (b, d) for CDD. Starred bars represent the COM impacts on GDP, while dotted and solid bars represent the CCA impact on the GDP and GDPC contribution, respectively. The bars denote the MMM, and the vertical lines indicate the interquartile model range. Stars at the bottom indicate where at least 2/3 of the models agree on the sign of the change.

To conclude, a process of GDP fraction change will occur among the provinces in South China and among the nations in Indochina from 2010 to 2100. During that process, Guangdong and Fujian provinces (Myanmar, Laos and Vietnam) are more inclined to increase their share of South China’s (Indochina’s) GDP under SSP5-8.5 in the future. The relative contributions of climate change and GDP change to the COM impact on GDP vary at different global warming levels. For instance, the GDPC contribution dominates the COM impacts on GDP related to heavy precipitation (RX5day) in South China at the 1.5°C level, but the CCA impact dominates at the three higher global warming levels. As for Indochina, the change of GDP enhances the CCA impact at the 1.5°C and 2°C warming levels, or offsets it at the 4°C warming level.

Heavy precipitation and extreme drought have caused severe economic losses during recent decades. Therefore, it is indispensable to investigate the future changes in precipitation extremes and associated impacts on GDP, particularly in sensitive locations with developing and unbalanced economies, like the INCSC region studied here. In this study, we first analyze the responses of the bias-corrected extreme precipitation indices to global warming and the corresponding climate-change-alone-induced impact (CCA impact) on GDP under SSP5-8.5 by employing the outputs of state-of-the-art CMIP6 models. Then, the reduced economic impacts by limiting global warming to 1.5°C instead of higher global warming levels (2°C, 3°C and 4°C) are examined. Finally, the relative roles of climate change and GDP change in the COM impact on GDP are further investigated. The findings are as follows:

(1) The CMIP6 models vary in their simulation performances for persistent heavy rainfall (RX5day) and longest dry spell length (CDD). Thus, a bias-corrected method is applied to constrain the future projections of RX5day and CDD. The projections under SSP5-8.5 show that heavy precipitation will intensify throughout the INCSC region under global warming, especially in western Myanmar and the Malay Peninsula, while the dry spell duration will lengthen over the whole INCSC region except in central Myanmar. Compared to the 1.5°C global warming level, both heavy precipitation and extreme drought are more enhanced at the 4°C global warming level.

(2) If only the contribution of climate change to the COM impact on GDP under SSP5-8.5 is considered, the changes in both heavy precipitation and extreme drought bring a larger economic impact to Guangdong Province, Hunan Province, Fujian Province, Jiangxi Province, Hainan Province, the Malay Peninsula, and southern Cambodia in the future. Probability distribution analysis suggests that more intensified heavy precipitation events will affect a larger proportion of the GDP over the whole of the INCSC region under global warming, while the GDP fraction affected by extreme drought will vary with different levels of global warming and different regions being considered. In terms of the subregional average, more than two thirds of the CMIP6 models indicate that maintaining a lower global warming level will reduce the economic impact induced by heavy precipitation over South China and Indochina.

(3) To be more specific, the national/provincial economic impacts induced by climate change alone under SSP5-8.5 are also examined. Results demonstrate that more intense heavy rainfall is projected to affect more GDP fractions in each province of South China or each country of Indochina at the 4°C warming level compared with the 1.5°C warming level. Furthermore, the number of provinces in South China or nations in Indochina where more than 50% of the GDP fraction suffers from stronger heavy precipitation and longer consecutive dry days is projected to increase under 4°C of warming relative to 1.5°C.

(4) The COM impact on GDP is contributed by both climate change and GDP change. From 2010 to 2100, a process of GDP fraction change will occur among the provinces of South China and among the countries of Indochina under SSP5-8.5, during which time the provinces of Guangdong and Fujian (the countries of Myanmar, Laos, and Vietnam) are more likely to increase their share of South China’s (Indochina’s) GDP. The relative roles of climate change and GDP change in the COM impact on GDP are found to vary under different global warming levels. For instance, the GDPC contribution dominates the COM impacts on GDP related to heavy precipitation in South China at the 1.5°C level, but the CCA impact dominates at the three higher global warming levels. The GDP change enhances the CCA impact in Indochina at the 1.5°C and 2°C warming levels, or offsets it at the 4°C level.

Our study points out that the GDPs of Guangdong Province, Hunan Province, Fujian Province, Jiangxi Province, Hainan Province, the Malay Peninsula, and southern Cambodia are more likely to be influenced by changes in heavy precipitation and extreme drought under SSP5-8.5. Thus, timely regional adaptation strategies are urgent for these regions. However, it should be mentioned that only one dataset of future projected GDP is used in this study. Therefore, more efforts are needed to reduce the uncertainty of projections. Some recent studies suggest that extreme daily rainfall (Kotz et al., 2022) and heatwaves (Callahan and Mankin, 2022) have a significant impact on global economic growth. Therefore, whether extreme precipitation and drought significantly influence global economic growth deserves further investigation. Besides, our results show some differences for the future change in CDD in South China and Indochina; other possible reasons for this result (e.g., evapotranspiration with temperature and soil moisture), aside from changes in atmospheric circulation, are certainly worthy of investigation in future work.

Acknowledgements. The work was jointly supported by the National Natural Science Foundation of China (42288101), the Strategic Priority Research Program of Chinese Academy of Sciences (XDB40030204), and the National Natural Science Foundation of China (42275032).

Electronic supplementary material: Supplementary material is available in the online version of this article at https://doi.org/ 10.1007/s00376-023-3158-7.