The research team, led by Ruixin Duan of the National Institute of Natural Hazards under China’s Ministry of Emergency Management, together with colleagues from Beijing Normal University, the University of Regina, Beijing University of Technology, and the University of Waterloo, constructed what they describe as a dual-dimension attribution framework. Rather than relying on a single statistical lens, the framework integrates two complementary approaches: multi-threshold probability ratios, which compare the likelihood of extreme events in a world with human influences against a hypothetical world without them, and nonlinear scaling analysis, which examines how the strength of that human influence changes as events grow progressively rarer and more severe.

To anchor their analysis in reality, the researchers used the CN05.1 observational dataset, a high-resolution gridded product built from weather station records across mainland China. Against this observational baseline, they deployed simulations from the Coupled Model Intercomparison Project Phase 6, known as CMIP6, the international ensemble of state-of-the-art climate models that underpins much of modern climate science, including the assessments of the Intergovernmental Panel on Climate Change. By comparing simulations that include all human and natural forcings with simulations driven only by natural factors such as solar variability and volcanic eruptions, the team could isolate the signature of human influence on rainfall extremes.

The observational record itself tells a compelling story. Between 1961 and 2020, two widely used extreme precipitation indices — Rx1day, which captures the maximum rainfall falling in a single day, and Rx5day, which measures the heaviest five-day accumulation — exhibited widespread upward trends across China. The study found that 61.4 percent of all grid cells showed positive trends in Rx1day, while 56.9 percent showed positive trends in Rx5day. In other words, across most of the country’s territory, the wettest days and wettest weeks of each year have been getting wetter, a pattern consistent with the fundamental physics of a warming atmosphere.

That physics is worth spelling out. For every degree Celsius of warming, the atmosphere can hold roughly seven percent more water vapor, following the Clausius–Clapeyron relationship. A moister atmosphere provides more fuel for storms, and observations and models alike show that extreme precipitation tends to intensify faster than average rainfall, because the extra moisture is disproportionately funneled into the heaviest events. In some circumstances, particularly for short-duration convective storms, intensification can even exceed the seven-percent-per-degree benchmark — a phenomenon known as super-Clausius–Clapeyron scaling that recent research has linked to shifts from stratiform to convective rain types within storm systems.

The heart of the new study, however, lies in its probability ratio calculations — the currency of modern event attribution science. The probability ratio expresses how many times more likely an event of a given severity has become under human influence compared with a counterfactual climate shaped only by natural forces. For 100-year return period events — downpours so severe that, in a stable climate, they would be expected only once per century — the results are sobering. Under the ALL-forcing scenario, which includes both anthropogenic and natural drivers, the probability of 100-year Rx1day events was approximately 1.84 times that under the NAT-forcing scenario, which includes natural drivers alone. For five-day extremes of the same rarity, the probability ratio was approximately 1.38. A once-in-a-century deluge has effectively become a once-in-54-years event in the case of the one-day extreme, according to these multi-model estimates.

But the truly novel finding emerges when the researchers examined how these probability ratios change with return period. Using a log–log scaling analysis that plots probability ratios against return periods on logarithmic axes, the team uncovered a consistent tendency: probability ratios tend to rise as return periods lengthen. In practical terms, human influence is not merely shifting the entire distribution of rainfall upward — it appears to be disproportionately amplifying the extreme tail, where the most catastrophic and least frequent events reside. Higher probability ratio values were generally associated with longer return periods, suggesting that the rarest, most destructive storms are precisely where anthropogenic forcing leaves its deepest statistical mark.

This nonlinear relationship between forcing and event rarity carries profound implications for risk management. Infrastructure in China — and indeed worldwide — is designed around return periods: dams, urban drainage networks, and flood defenses are typically engineered to withstand 50-year, 100-year, or occasionally 1,000-year events. If those events are becoming substantially more probable, the engineering assumptions embedded in decades-old design standards are quietly eroding. The 2021 record-breaking rainfall around Henan Province, which the study’s authors have examined in earlier work, and the 2023 Beijing–Tianjin–Hebei extreme rainfall event both serve as vivid reminders of what happens when precipitation exceeds the thresholds that infrastructure was built to handle.

The authors are careful to emphasize that the magnitude of the anthropogenic response varies across regions and remains subject to uncertainty — a candid acknowledgment that reflects the genuine challenges of regional attribution science. China spans tropical, subtropical, temperate, and alpine climate zones, and the response of precipitation extremes to forcing differs markedly among them. Confounding factors, including anthropogenic aerosols that can locally suppress rainfall even as greenhouse gases enhance it, internal climate variability, and the coarse resolution of global models in resolving complex topography such as the Tibetan Plateau, all contribute to the uncertainty envelope. Previous studies have documented both detectable human influence on precipitation extremes across China and locally divergent responses, underscoring that attribution is as much about quantifying confidence as about delivering headline numbers.

Even so, the study’s central message lands with force: human activities appear to be enhancing the likelihood of extreme precipitation events in China, particularly the rarer ones, and assessments of climate risk must account for changes in this tail risk rather than focusing solely on shifts in average conditions. As global temperatures continue to climb, the framework developed here — combining multi-threshold probability ratios with nonlinear scaling — offers a template that can be applied beyond China’s borders. For adaptation planners, the implications are clear: the storms once dismissed as statistical outliers are becoming statistical neighbors, and planning for the climate of the coming decades means planning for a distribution whose extremes are moving faster than its center.

Subject of Research: Anthropogenic influence on extreme precipitation events and their return periods in China

Article Title: Anthropogenic forcing enhances extreme precipitation with increasing return period in China

Article References: Duan, R., Zhong, L., Huang, G., Wang, F., Zhang, S., & Tian, C. (2026). Anthropogenic forcing enhances extreme precipitation with increasing return period in China. Climate Dynamics, 64(10), Article 422. https://doi.org/10.1007/s00382-026-08334-6

Image Credits: AI Generated

DOI: 10.1007/s00382-026-08334-6

Keywords: extreme precipitation, anthropogenic climate change, CMIP6, event attribution, China, return period, Rx1day, Rx5day, probability ratio, nonlinear scaling, tail risk, climate adaptation