Contribution of anthropogenic influence to the 2022-like Yangtze River valley compound heatwave and drought event

The study addresses how anthropogenic influences, particularly greenhouse gases (GHGs), contribute to the occurrence and severity of compound heatwave and drought events (CHDEs) in the Yangtze River valley (YRV). CHDEs, defined by simultaneous or consecutive heat and drought, cause greater societal and environmental damage than individual extremes due to the negative thermodynamic relationship between high temperature and low rainfall. Under global warming, CHDE frequency has increased across much of the global land area and is projected to rise further, threatening agriculture, water, energy, and health. The unprecedented August 2022 CHDE in the YRV, lasting 24 days and causing widespread impacts, motivates key questions: whether anthropogenic forcing (and its components, aerosols vs. GHGs) has a detectable influence on an August 2022-like CHDE over the YRV, and how the likelihood of such events will change in the future.

Previous work has defined CHDEs using combinations of precipitation and temperature metrics, including composite indices of rainfall and surface air temperature (SAT), anomalies of meteorological drought with SAT, and combined deficits of potential evapotranspiration and rainfall. Bivariate copula and time-of-emergence methods have been widely used to quantify intensity and frequency for detection and attribution. Studies have shown anthropogenic climate change significantly increases the likelihood of summer CHDEs across continents, with global CHDE trends rising (>0.17/decade from 1951–2010) under anthropogenic forcing while trends are near zero or negative under natural forcing. Attribution analyses separating heatwave and drought often find higher anthropogenic influence on heatwaves than on drought, leading CHDE probabilities to lie between the two when considered jointly. Regionally, 2022 saw record CHDEs across the Northern Hemisphere, including the YRV, western Europe, and North America; over the YRV the heatwave component was rarer than drought in 2022 and more strongly linked to anthropogenic forcing, with heatwave probabilities increased by about 11 times under anthropogenic forcing relative to natural forcing.

Data: Daily maximum temperature (Tmax) and rainfall from 751 stations across the YRV (26°–33°N, 102°–123°E) for 1961–2022 were obtained from the National Meteorological Information Center of China. Model simulations for Tmax and rainfall were taken from nine CMIP6 models under all external (ALL), greenhouse gas (GHG), anthropogenic aerosol (AER), and natural (NAT) forcings, each with at least three ensemble members. Historical ALL runs ending in 2014 were extended to 2022 using SSP2-4.5; other forcing runs ending in 2020 were extended to 2022 by repeating 2019–2020. Simulations were regridded to 2°×2° for analysis. Definition of CHDE: Heatwave (drought) events are identified when daily Tmax exceeds the 90th percentile (daily 5-day running-mean rainfall is below the 10th percentile) for at least five consecutive days during JJA. A joint cumulative probability distribution of heatwave and drought was constructed via bivariate copulas to define a probability-based index (PI) of CHDE severity: PI = P(X≥x, Y≥y) = 1 + C[F(X), G(Y)] − F(X) − G(Y), where F and G are marginal CDFs (exponential for heatwave, gamma for drought). Three Archimedean copulas (Clayton, Frank, Gumbel) were evaluated using RMSE, AIC, and Kolmogorov–Smirnov statistic; Gumbel had the smallest errors and was selected. For characterizing the August 2022 event, normalized maximum 24-day Tmax anomaly (Tx24day) and 24-day rainfall deficit anomaly (Dx24day, negative rainfall accumulation multiplied by −1) during summer were used. Detection and attribution: An optimal fingerprinting framework tailored for extremes was applied using generalized extreme value (GEV) distributions to model PI time series (1961–2022). GEV parameters (location, scale, shape) were estimated by probability-weighted moments. Observed PI was regressed onto model-simulated responses to external forcings to estimate scaling factors β, with uncertainty from a two-stage spatiotemporal block bootstrap. For fingerprinting, fields were regridded to 5°×5°. Risk metrics: Event probabilities were derived from fitted GEVs, with return period T = 1/p. Probability ratios (PR = P1/P0) quantified anthropogenic influence comparing forced (ALL, GHG, AER) to NAT for the present climate (2000–2022). For models unable to simulate the observed PI = 0.06 under NAT, the 1st percentile of observed PI (0.10) over 1961–2022 served as the threshold for PR estimation.

- The YRV experienced a long-lasting CHDE from 2–25 August 2022 (24 days). Regional mean drought persisted 17 days (Aug 8–24). During Aug 2–25, Tmax anomalies exceeded 6.0 °C and accumulated rainfall deficits exceeded 100 mm. - The CHDE severity index PI in 2022 was 0.06, the lowest on record since 1961, corresponding to a return period of ~662 years (5–95% CI: 106 to >10,000 years). - PI correlated more strongly with heatwave intensity than drought: correlation with Tx24day = −0.91, with Dx24day = −0.82 (both p < 0.001), indicating heatwaves dominate CHDE severity in the YRV. PI shows a decreasing trend of −0.013/decade since 1961, consistent with increasing Tx24day (~0.2/decade) and near-zero trend in Dx24day. - Detection and attribution: Scaling factors for PI under ALL forcings were significantly >0 across all nine models, indicating detection of anthropogenic influence. GHG-forcing scaling factors were similar to ALL, while NAT and AER scaling factors were generally not significantly different from zero (eight models), implying a dominant GHG contribution and weak NAT/AER influence. - Anthropogenic influence on risk under present climate (2000–2022): Using threshold PI ≤ 0.10 (observed 1st percentile), probabilities were 2.74% (ALL; 5–95% CI: 1.91–3.80%) vs 0.47% (NAT; 0.24–0.58%), giving PRALL ≈ 9.92 (CI: 6.59–17.00). For GHG, PRGHG ≈ 11.66 (CI: 7.50–21.67). AER probabilities (0.73%; CI: 0.56–0.85%) were close to NAT, indicating little aerosol contribution. PRs increased further for more extreme thresholds (5th→1st percentile). - Future projections (SSP2-4.5): Multi-model ensembles project PI to decrease from ~0.30 in the present climate to ~0.14 at 3 °C global warming (relative to CMIP6 climatology), with accompanying rightward shifts of Tx24day distributions and minimal changes in Dx24day distributions. Estimated PR under SSP2-4.5 at 3 °C relative to NAT is ~38.60 (CI: 24.14–68.28), implying much higher likelihood and severity of CHDEs driven mainly by intensifying heatwaves.

The analysis demonstrates that the August 2022-like CHDE in the YRV reflects an extreme joint occurrence of record heat and severe drought, with severity predominantly controlled by the heatwave component. Optimal fingerprinting of the PI time series clearly detects anthropogenic signals, attributable chiefly to greenhouse gas forcing, while natural variability and aerosol forcing do not show significant contributions. Probability ratio estimates indicate that anthropogenic climate change has already increased the likelihood of CHDEs as severe as 2022 by more than an order of magnitude relative to a natural-forcing climate. Looking forward, projections under a medium-emissions pathway (SSP2-4.5) show a continued decline in PI and rising heatwave intensity, indicating that compound events will become more frequent and extreme; drought variability shows larger uncertainty and smaller trends, reinforcing the dominant role of heatwaves in future CHDE risk. These findings directly address the research questions by quantifying detectability and attribution to different forcings and by projecting future risk escalation in the YRV.

The study develops a probability-based index using a copula framework to quantify CHDE severity and applies an extremes-oriented fingerprinting approach to detect and attribute changes in the YRV. It finds that the 2022 CHDE was the strongest since 1961 (PI = 0.06; ~662-year return period) and that anthropogenic, especially GHG, forcing has significantly increased the likelihood of such events (PRALL ~10; PRGHG ~12) under the present climate. Future projections under SSP2-4.5 indicate PI will decline from roughly 0.30 to 0.14 at 3 °C warming, implying more frequent and intense CHDEs dominated by intensifying heatwaves. Future research should leverage larger large-ensemble datasets that capture the observed threshold under NAT conditions, reduce model biases in drought variability, and use observed heatwave–drought relationships to better constrain future drought and CHDE projections in the YRV.

- Model limitations: The nine CMIP6 models could not reproduce the observed 2022 PI = 0.06 under NAT forcings. Consequently, the 1st percentile of observed PI (0.10) over 1961–2022 was used as the threshold for PR estimates, introducing methodological constraints. - Biases in drought variability: Models show larger biases for simulated drought variability than for heatwaves, affecting CHDE representation and projection. - Uncertainty in automated event identification and sampling variability necessitated regridding to 5°×5° and use of spatiotemporal block bootstrap; residual uncertainties remain. - Limited ensemble and model spread: Reliance on nine models and available ensembles may underrepresent internal variability; more large-ensemble simulations are needed to encompass observed extremes and reduce uncertainty.