Changes caused by human activities in the high health-risk hot-dry and hot-wet events in China

The study investigates how human activities influence high health-risk compound hot-dry events (CHDEs) and hot-wet events (CHWEs) in China. Compound events involve concurrent extremes of temperature and humidity, which elevate health risks beyond those associated with temperature alone. Prior assessments have largely focused on single temperature indices and have not explicitly integrated humidity or linked thresholds to health outcomes, leaving a gap in understanding health-relevant compound heat exposures. Given increasing frequencies of CHDEs and CHWEs and rapid anthropogenic changes (e.g., greenhouse gas emissions, urbanization, industrialization), the research aims to (1) define health-based thresholds for compound temperature–relative humidity events using ambulance dispatch data, (2) detect and attribute observed changes in high health-risk CHDEs/CHWEs to anthropogenic versus natural forcings, and (3) project future changes and health implications under contrasting emission scenarios including a carbon-neutral pathway.

Multiple studies have documented rising frequencies and intensities of compound heat events globally and in China, with attribution analyses implicating anthropogenic drivers such as greenhouse gases, urbanization, and industrialization. Traditional detection and attribution methods compare observed changes with model simulations under all forcings (ALL) and natural-only (NAT) to quantify probability ratio (PR) and fraction of attributable risk (FAR). While successful for temperature extremes, applications to compound hot events remain limited and rarely integrate health data. Prior works have used relative meteorological thresholds (e.g., 10th/90th percentiles of T and humidity) to define compound heat and humidity events but have not tied them to health outcomes. Evidence also shows humidity modulates heat-related health impacts, yet results have been debated and often not operationalized into health-relevant thresholds. This study addresses these gaps by coupling ambulance dispatch data with T–RH to derive impact-based thresholds and by conducting detection, attribution, and projections specific to health-relevant compound events.

Health data: Daily emergency ambulance dispatch records for 13 Chinese cities (covering ~79.11 million people) for 2013–2019 were obtained from Ambulance Service Centers (total ~2.5 million emergency cases; annual average ~673,000; average incidence 1302 per ten thousand people). After adjusting for long-term trends, seasonality, day-of-week effects, holidays, and joint T–RH effects, regression analyses related daily mean temperature (T) and relative humidity (RH) to dispatch counts. Relative risk (RR) quantified health risks under different T–RH combinations. Observations: ERA5 reanalysis (ECMWF) provided hourly T and RH at 0.25° resolution for 1979–2022, aggregated to daily metrics and interpolated to 0.5° for analysis. RH was chosen for consistency across temperatures and locations and for practicality in large-scale analyses. Model simulations: Ten CMIP6 DAMIP models provided historical ALL (natural + anthropogenic) and NAT (natural-only) simulations for 1979–2014. For future projections (2021–2100), scenarios included SSP1-1.9, SSP2-4.5, SSP5-8.5; SSP2-4.5nat (natural-only) was available for quantifying anthropogenic contributions in the future. All datasets were bilinearly interpolated to 0.5°×0.5°. Health-based thresholds: Using RR surfaces over 200 T–RH combinations (T percentiles 50–100 by 5th; RH percentiles 0–100 by 5th), high health-risk thresholds were defined as (1) CHDEs: T ≥ 80th percentile and RH ≤ 10th percentile; (2) CHWEs: T ≥ 90th percentile and RH ≥ 55th percentile with a minimum RH of 70%. Sensitivity analyses assessed robustness of thresholds. Event characterization: Focused on warm season (May–September). Computed climatology and linear trends (1979–2022) in days per season meeting high health-risk CHDE/CHWE thresholds. Attribution: Compared observations with model ALL vs NAT. Fitted generalized extreme value (GEV) distributions to pooled warm-season annual data to represent distributions of event frequencies. Two-sample KS tests compared observed and simulated PDFs. Anthropogenic influence quantified via probability ratio (PR = P_ALL / P_NAT) and fraction of attributable risk (FAR = 1 − 1/PR). Confidence intervals (95% CI) obtained via 1000 bootstrap resamples. Regional analyses divided China into eight sub-regions: NWC, NC, NEC, SWC1, CC, EC, SWC2, SC. Future projections: Evaluated occurrence frequencies under SSP1-1.9, SSP2-4.5, SSP5-8.5 (2021–2100) using historical health-based thresholds. Anthropogenic effect in the future computed as difference between SSP2-4.5 and SSP2-4.5nat. Regional and monthly variations assessed for near-term (2021–2040), mid-term (2041–2060), and long-term (2081–2100).

- Health-relevant thresholds: High health-risk CHDEs occur at T ≥ 80th percentile with RH ≤ 10th percentile; high health-risk CHWEs at T ≥ 90th percentile with RH ≥ 55th percentile and RH ≥ 70% absolute. Relying on temperature alone underestimates health risks of compound events. - Observed spatial-temporal patterns (1979–2022, ERA5, May–September): High health-risk CHDEs most frequent in Northwestern China and middle–lower Yangtze (≈10–20 days/year); high health-risk CHWEs most frequent on the Qinghai–Tibet Plateau and Northeastern China (≈20–30 days/year). CHDEs peak in May–June; CHWEs peak in July–August. High-incidence areas also show significant positive trends. - Model performance: Multi-model ensemble ALL and NAT spatial correlations >0.80. Time series correlation (1979–2014) between observations and ALL: r = 0.65 for CHDEs and r = 0.72 for CHWEs (P < 0.01). Two-sample KS tests show observed vs simulated PDFs are consistent (p = 0.46 for CHDEs; p = 0.41 for CHWEs). - Nationwide attribution (1979–2014): For high health-risk CHDEs, P_NAT ≈ 9.2% (95% CI: 9.2–9.4%), P_ALL ≈ 21.5% (21.4–21.7%), PR = 2.34 (2.28–2.36), FAR = 0.43 (0.42–0.44) → anthropogenic activities increased occurrence by ~2.34×. For high health-risk CHWEs, P_NAT ≈ 12.8% (12.7–12.8%), P_ALL ≈ 8.0% (7.9–8.2%), PR = 0.63 (0.62–0.65), FAR = −0.60 (−0.55 to −0.62) → anthropogenic activities decreased occurrence to ~0.63×. - Regional/monthly attribution: Human activities significantly increase high health-risk CHDEs in NWC (notably May–August, peaking in August), and in CC/EC in July–August, while decreasing high health-risk CHWEs in SWC1 (July–August) and broadly across CC/EC (July–August). Contrasting impacts are especially evident in the middle and lower Yangtze. - Future projections (using historical thresholds): Relative to past climate, national average frequencies up to 2060: SSP1-1.9: CHDEs 1.41×, CHWEs 1.71×; SSP2-4.5: CHDEs 1.92×, CHWEs 2.03×; SSP5-8.5: CHDEs 3.05×, CHWEs 2.13×. Carbon-neutral pathway (SSP1-1.9) reduces anthropogenically driven frequencies by about one-half (CHDEs) and over one-fifth (CHWEs) compared to SSP5-8.5. - Future anthropogenic impact (SSP2-4.5 minus SSP2-4.5nat): Increased CHDEs particularly in NC (June) and NWC (July; shifting to August in long term). Decreased CHWEs mainly in SC (June), extending to EC and NEC (July–August); timing window narrows in SC from June–August to July–August in the long term. - Health impacts: Emergency cases attributable to high health-risk CHDEs/CHWEs are projected to increase substantially, expanding from southeastern concentrations to include northwestern regions under higher emissions.

By deriving health-based thresholds from ambulance dispatch data, the study directly links compound temperature–humidity exposures to acute health outcomes. Attribution results show anthropogenic activities have significantly increased high health-risk CHDEs while decreasing high health-risk CHWEs at the national scale, with particularly strong and opposing regional impacts in the Yangtze River basin. The findings imply that assessments based solely on temperature underestimate population health risks during compound events, underscoring the need to include humidity. Projections indicate that ambitious mitigation (carbon neutrality-like SSP1-1.9) would substantially curb the anthropogenic contribution to high health-risk compound events compared to high-emissions futures, especially reducing CHDE increases in northern and northwestern China and CHWE occurrence in eastern China. These insights are relevant for public health planning, heat-health warning systems, water and energy management, and urban design tailored to regional compound hazards.

The study introduces impact-based, health-relevant thresholds for compound hot-dry and hot-wet events using ambulance dispatch data and demonstrates that anthropogenic activities have increased high health-risk CHDEs (~2.34×) and decreased high health-risk CHWEs (~0.63×) in China since 1979. Clear regional and seasonal contrasts emerge, particularly in the Yangtze River region. Future projections show strong scenario dependence: carbon-neutral pathways substantially reduce anthropogenic contributions to health-risk compound events relative to high-emission scenarios. The work advances detection and attribution for compound hazards by integrating health data and highlights the importance of considering humidity in heat-health risk assessments. Future research should expand multi-center studies, refine health-relevant thresholds for diverse outcomes (mortality, hospitalization), investigate modifying roles of environmental and socioeconomic factors, and elucidate mechanistic pathways to inform targeted regional prevention and control strategies.

- Health data were limited to ambulance dispatches from 13 cities (2013–2019), which may constrain generalizability across China and to other countries with different climates, social contexts, and behaviors. - Thresholds derived for ambulance-dispatch outcomes may not transfer directly to other health endpoints (e.g., mortality, hospitalizations) or populations without careful validation. - RH-based thresholds, while practical, abstract from absolute humidity and other moisture metrics that may influence physiological responses; data availability constrained broader humidity metrics. - CMIP6 model and scenario availability limited ALL vs NAT future comparisons to SSP2-4.5 vs SSP2-4.5nat, potentially underrepresenting uncertainty across other pathways. - Potential inconsistencies or reporting constraints in ambulance data and lack of public availability may limit reproducibility; code is available upon request.