Global projections of short-duration rainfall extremes using temperature-covariate models

Short-duration, high-intensity rainfall extremes associated with convective storms pose a growing risk to urban areas under a warming climate, yet their future evolution remains difficult to quantify at the global scale using existing modelling approaches. Local projections are often constrained by the lack of long high-resolution observations and by the limited ability of climate models to accurately simulate sub-daily precipitation processes at the global scale. Here, we present a globally applicable framework for projecting changes in rare, short-duration rainfall extremes using temperature as a covariate in a non-stationary extreme value framework building on the TENAX model, driven entirely by global climate model output and without reliance on local observational data. The focus on rare, short-duration extremes directly targets the class of events responsible for a disproportionate share of climate-related impacts.

The approach links changes in rainfall intensity distributions to projected shifts in wet-day temperature distributions from CMIP6 models, integrating over the full temperature distribution rather than relying on uniform scaling or mean-shift assumptions. Dew-point temperature is employed as a proxy for atmospheric moisture availability, allowing thermodynamically constrained intensification of convective rainfall extremes to be represented consistently across climates. In an initial multi-regional application, the framework projects robust intensification of hourly-scale rare rainfall events, with increases of order 10–20% by late century under intermediate emissions scenarios and substantially larger changes under high-emissions pathways. Accounting for changes in the full temperature distribution shows that the strongest intensification occurs for the rarest events, which is underestimated when intensities are scaled only by mean temperature changes.

We further extend the framework to a global scale to assess spatial patterns and key structural uncertainties in projected short-duration rainfall intensification. Results highlight that methodological choices, including the selection of temperature covariate (dew-point versus surface air temperature), can introduce differences comparable to inter-model climate uncertainty in some regions, particularly in moisture-limited and continental climates. Treating these choices explicitly as structural uncertainties provides a clearer interpretation of projection robustness across diverse hydroclimatic regimes and highlights uncertainties beyond inter-model spread alone.

Overall, this work demonstrates that temperature-covariate approaches, when carefully formulated and driven by global climate models, offer a transferable and physically grounded pathway for projecting rare, short-duration rainfall extremes worldwide. The framework enables consistent global assessments in data-scarce regions and supports climate-change impact studies and urban adaptation planning by explicitly quantifying the uncertainties that matter most for short-duration rainfall risk.