Metastatistical Extreme Value Framework Reveals Robust Improvement in Characterizing Humid Heatwave across Indian Urban Agglomerates

Intensifying humid heatwaves (HHWs) across the Indian subcontinent highlight the growing risk of elevated temperatures and increased humidity in urban agglomerates, characterized by dense populations and complex sustainability demand. While conventional Extreme Value Theory (EVT) approaches are commonly used to estimate design events of heat-humidity compound extremes, they are constrained by limited sample sizes and unstable upper tail estimates in data-sparse regions. Recent advances in Metastatistical Extreme Value (MEV) theory demonstrate significant skill improvements for hydroclimatic extremes such as rainfall and flash droughts, but its applicability to compound heat–humidity, i.e., HHWs, remains unexplored for Indian sub-continent, with diverse climate types, primarily dictated by monsoonal circulations. Here, we present the first observational assessment showing the skill of MEV probabilistic framework that can capture the year-to-year variability of the HHW distributions. Using in-situ observations from 54 urban and peri-urban sites across India for over four decades (1980–2025), we develop the MEV-based probabilistic estimates of extreme HHW magnitude, which we compare against the conventional EVT-based distribution. HHW events are identified based on daily observed maximum wet-bulb temperature exceeding the 90^th percentile daily variable threshold that persists for two consecutive days or more, while HHW magnitude is estimated as the positive anomaly of daily extreme wet-bulb temperature, exceeding the 90^th percentile variable threshold, normalized by its interquartile range. Our results show that MEV consistently outperforms conventional EVT in estimating design events for approximately, 68% of sites, indicating improved representation of moderate to rare HHW events.  The uncertainty bounds (indicated by the interquartile range) of the MEV versus EVT design events suggest that MEV offers lower uncertainty, represented by narrower interquartile ranges, compared to conventional EVT across approximately 70% of sites. For example, for a representative site across eastern coastal India, the quantification of the record HHW event during July 2020, with a 64-year return period, is illustrated. The MEV estimated quantile provides error estimates of ~1%, whereas the conventional model underestimates the design events by ~4%, suggesting the MEV model offers improved representation of compound heat and humidity design events, which have implication towards public health and ecosystem sustainability. Our study provides the first application of MEV models to understand the heat-humidity nexus across urban agglomerates of India, and demonstrates its potential to define impact-relevant metrics in a warming climate.