Urban extreme heat risk in China, 2000–2024: identification with historical city-specific thresholds and trend analysis

Technical Summary: Urban Extreme Heat Risk in China, 2000–2024

Problem Statement
Under global warming, extreme heat events have intensified, posing significant risks to urban systems, public health, and social resilience. While previous studies have extensively documented changes in heatwave frequency and intensity, methodological limitations persist in long-term, city-scale risk assessments. Existing approaches often rely on relative thresholds (e.g., ETCCDI indices), which define extremes based on local seasonal distributions. While effective for detecting anomalies, these methods may identify anomalously warm days outside the warm season, weakening the practical interpretation of "extreme heat" in a long-term urban context. Conversely, absolute thresholds (e.g., fixed 35°C) lack sensitivity to local climate backgrounds. Furthermore, many studies rely on raw event counts, which may fail to capture the true evolution of risk by ignoring observation exposure and the probability structure of event occurrence. There is a need for a framework that is both temporally consistent for long-term comparison and sensitive to city-specific historical conditions.

Methodology
This study analyzes daily temperature data from 370 Chinese cities spanning 2000–2024, utilizing a baseline period of 1973–1999 for threshold construction. The research employs a novel city-specific fixed-threshold approach:

1. Threshold Definition: For each city, the 95th percentile of the full-year daily maximum temperature distribution during the 1973–1999 baseline is calculated. This single value ($TX95_i$) serves as the fixed threshold for the entire study period. A day is classified as an extreme heat day if the daily maximum temperature exceeds this city-specific threshold.
2. Risk Metric Construction: To address the limitations of simple event counts, the study constructs an annual occurrence risk rate ($R_{i,t}$), defined as the ratio of extreme heat days ($N95_{i,t}$) to the number of observed days ($ObsDays_{i,t}$) in a given year.
3. Trend Estimation: A log-transformed risk intensity measure ($Z_{i,t}$) is derived to facilitate trend analysis. Long-term trends are estimated using the Theil-Sen median slope estimator, which is robust to outliers and non-normal data. Statistical significance is assessed using the Hamed–Rao modified Mann-Kendall test to account for serial autocorrelation.
4. Validation: The study compares these findings against a relative-threshold approach and conducts robustness checks by varying the metric specification (using raw counts), the smoothing parameter in log transformation, and sample completeness criteria.

Key Contributions

• Methodological Innovation: The study develops a city-specific fixed-threshold framework that ensures temporal consistency in long-term urban comparisons. It moves beyond raw event counts by introducing an occurrence risk rate that accounts for observation exposure.
• Empirical Evidence: It provides a systematic assessment of the spatiotemporal evolution of urban extreme heat risk across 370 Chinese cities over the 25-year period from 2000 to 2024.
• Policy Relevance: The framework links long-term risk trend identification with urban climate governance needs, offering a quantitative basis for differentiated adaptation planning.

Results

• Temporal Evolution: The national annual mean risk rate exhibited a fluctuating upward trend from 2000 to 2024, with a marked acceleration after 2020. The risk rate in 2024 was approximately 2.4 times that of 2000. While 2000–2019 showed moderate fluctuations, the post-2020 period indicates a sustained intensification of risk.
• City-Level Trends: 94.05% of the 370 cities showed positive decadal risk growth. Among cities with statistically significant trends (63.78% of the total), 99.6% experienced increasing risk. The decadal risk growth rates ranged from -39.33% to 175.42%, highlighting substantial heterogeneity.
• Spatial Heterogeneity: Risk growth was most pronounced and spatially continuous in central-eastern and southwestern China, particularly in the Yangtze River Basin, the eastern coast, and the Sichuan Basin. Northeastern and northwestern China showed lower but still predominantly positive growth rates.
• Method Comparison: The fixed-threshold approach identified events highly concentrated in the warm season (June–August accounting for ~85% of events), whereas the relative-threshold approach produced a more even seasonal distribution of identified events and generally higher estimates of risk growth, while both approaches showed highly consistent trend directions.
• Robustness: The main conclusions remained stable across alternative metric specifications (raw counts vs. log risk intensity), different smoothing parameters, and stricter sample completeness criteria (retaining only years with >95% valid days).

Significance
The study establishes a temporally consistent and practically interpretable framework for monitoring urban extreme heat risk. By demonstrating that risk is increasing across the vast majority of Chinese cities but with significant spatial variation, the findings suggest that adaptation strategies cannot be uniform. The research supports the shift from retrospective event counting to the monitoring of long-term risk accumulation. It provides quantitative indicators that can help identify emerging risk areas—particularly those not traditionally considered heat hotspots—and inform forward-looking urban planning, infrastructure development, and public service systems. The authors note that while the study focuses on meteorological heat frequency, future research should integrate humidity, event duration, and population exposure to fully characterize human heat stress.