We used the weekly self-calibrated Palmer Drought Severity Index (scPDSI) and daily maximum temperatures derived from two observation products (the ERA5 reanalysis31 and the MSWEP dataset32) to identify summer CDHWs during the period 1981 to 2020 (see Methods). We find that the CDHWs have been increasing significantly (P < 0.05) over approximately 57% of global land areas, and these regions are home to approximately 3.94 billion people, mainly located in Africa (40%), East Asia (33%), Latin America (8%), and the Middle East (4%) (see Fig. 1a generated from the ERA5 reanalysis and Figure S1 from the MSWEP dataset). Our results also suggest a prominent spatial difference in the changing rate of CDHW occurrence and inequality globally. Low-income regions are experiencing significantly (P < 0.05) more rapid increases in CDHWs than high-income regions (following the WorldBank’s 2020 income classification) (Fig. 1b). CDHWs in low-income regions have increased by 317% [257%~377%] between the earlier and later period (1981‒2000 and 2001‒2020) in the last 40 years, but the increase is only 127% [70%~184%] in high-income regions, especially in the United States (USA) and Europe where the increases in the frequency of CDHWs are only 55% [14%~87%] and 73% [39%~113%], respectively. The difference is statistically significant at the 5% significance level based on 500 bootstrap resampling and a two-tailed t-test (see Figure S2 and Methods). In addition, there is a strong dependence between historical changes in CDHWs and the fraction of population living in extreme poverty (i.e., on less than $1.90 per day), with a Pearson correlation coefficient of 0.653 (P = 0.002; see Fig. 1c). This is estimated through aggregating locations by ventiles (i.e., 20-quantiles) of the poverty rate from the latest edition of the World Bank’s Global Subnational Atlas of Poverty (GSAP) and by graphing the changing rate of CDHW occurrences. CDHWs show an approximately twofold increase (222% on average) in locations where the share of population living in extreme poverty is lower than 10% between the early and later halves of 1981 − 2020. In comparison, locations where the share of population living in extreme poverty is higher than 90% are experiencing a nearly fivefold increase (482%), on average, in the frequency of CDHWs. Such a dependence becomes even stronger, with Pearson correlation coefficients of 0.787 and 0.790 (see Figure S3), if we consider the less stringent poverty definitions of $3.20 and $5.50 per day, respectively. In addition to the frequency of CDHWs, similar inequalities also exist for the historical changes in the severity of CDHWs (see Figure S4).
We also analyzed the impacts of anthropogenic climate change on the frequency of CDHWs. We quantified the impacts by estimating the likelihood of CDHW occurrences that can be attributed to anthropogenic climate change. The probability ratio concept was used to quantify the likelihood of CDHWs occurring in historical conditions relative to natural-only conditions33. Figure 2a displays probability distributions of the probability ratio derived using land pixels for each income group, and shows a general increase in the frequency of CDHWs under historical conditions relative to historical natural-only conditions (i.e., probability ratios generally greater than 1). We find a contrasting inequality in the effects of anthropogenic climate change on the frequency of CDHWs between poorer and wealthier countries. Specifically, 31% [14%~51%] of land pixels in low-income regions show more than double increases in the frequency of CDHWs due to anthropogenic forcing, versus only 4.7% [0.9%~8.3%] of land in high-income regions (Fig. 2a). We find a statistically significant (P < 0.05) association between the magnitude of anthropogenic impacts on CDHWs and the share of population living in extreme poverty (living under $1.90 per day), with a Pearson correlation coefficient of 0.764 (Fig. 2b). Such a dependence remains if one considers a higher poverty threshold of $3.20 (or $5.50) (see Figure S5).
The upward trends in the frequency of CDHWs over low-income and high-income countries are well reproduced by the state-of-the-art CMIP6/ALL multi-model ensemble simulations (Figs. 3a and 3b). The ALL simulations consider both anthropogenic climate forcings, such as the emission of greenhouse gases and aerosols, and natural climate forcings like solar and volcanic activities. The upward trends over high-income countries are also captured by the CMIP6/NAT ensemble simulations, which solely account for natural climate forcings (P = 0.003). However, the trends over the low-income countries are not replicated by the CMIP6/NAT ensemble simulations (P = 0.279). This indicates that the increase in the frequency of CDHWs in low-income countries during the past 40 years is driven predominantly by anthropogenic climate change.
We further used an optimal fingerprinting approach (see Methods) to estimate contributions from anthropogenic (ANT) and natural forcings (NAT) to the observed changes in the frequency of CDHWs. The statistically significant non-zero scaling factors for ANT (ALL-NAT; anthropogenic forcings) indicate that the ANT signal is detectable across all four income groups (Fig. 3c). However, the NAT signal shows different results for poorer (LIC and LMIC) and wealthier (UMIC and HIC) countries. The inclusion of zero within the uncertainty range of scaling factors suggests a failure to detect NAT forcings in poorer countries, whereas the exclusion of zero in wealthier countries implies the detectability of NAT forcings. In other words, although anthropogenic climate change has contributed to the past increases in CDHW occurrences, the magnitude of such contributions is larger in regions with lower income levels. Specifically, the ANT-forced increases in the frequency of CDHWs are 19% [2%~33%], 18% [6%~26%], 17% [9%~23%], and 12% [5%~18%] per decade for low-income, lower-middle-income, upper-middle-income, and high-income countries, respectively, during the period of 1981–2020. We therefore conclude that anthropogenic climate change has made a larger contribution to the increases in the frequency of CDHWs in poorer countries than in wealthier countries.