Thermodynamically inconsistent extreme precipitation sensitivities across continents driven by cloud-radiative effects

Abstract

Extreme precipitation events are projected to intensify with global warming, threatening ecosystems and amplifying flood risks. However, observation-based estimates of extreme precipitation-temperature (EP-T) sensitivities show systematic spatio-temporal variability, with predominantly negative sensitivities across warmer regions. Here, we attribute this variability to confounding cloud radiative effects, which cool surfaces during rainfall, introducing covariation between rainfall and temperature beyond temperature's effect on atmospheric moisture-holding capacity. We remove this effect using a thermodynamically constrained surface-energy balance, and find positive EP-T sensitivities across continents, consistent with theoretical arguments. Median EP-T sensitivities across observations shift from -4.9%/°C to 6.1%/°C in the tropics and -0.5%/°C to 2.8%/°C in mid-latitudes. Regional variability in estimated sensitivities is reduced by more than 40% in tropics and about 30% in mid and high latitudes. Our findings imply that projected intensification of extreme rainfall with temperature is consistent with observations across continents, after confounding radiative effect of clouds is accounted for.

Fig. 1. Scaling of extreme precipitation with temperature in observation.

A Global map of extreme precipitation (P95) – observed temperature (EP-T) sensitivities estimated from quantile regression method, (B) zonal variation of estimated EP-T sensitivities. Scaling curves between daily extreme rainfall intensity and temperature for (C) tropics, (D) mid-latitudes, and (E) high-latitudes. The light orange lines show the scaling curves for each grid while the dark orange line indicates the mean response of all the grid cells. The black dotted line indicates the CC rate (7%/°C). Note the logarithmic y-axes in (C–E).

Fig. 2. Scaling of cloud-radiative effects with temperatures.

A Global map of net cloud radiative effect (CRE) defined as the difference between “clear-sky” and “all-sky” radiative fluxes including both shortwave and longwave radiation, isolated on the days when rainfall is greater than P95 (95^th percentile). B zonal variation of estimated CRE. Scaling curves between CRE and observed temperatures for (C) tropics, (D) Mid-latitudes, and (E) High-latitudes. The gray lines show the scaling curves for each grid cell while the black line indicates the mean response of all grid cells.

Fig. 3. Scaling of extreme precipitation with temperature after correcting for cloud radiative effects.

A Global map of extreme precipitation (P95) – temperature sensitivities (EP-T) without the effect of cloud cooling, (B) zonal variation of estimated sensitivities. Scaling curves between daily extreme rainfall intensity and temperature (cloud-adjusted) for (C) tropics, (D) Mid-latitudes, and (E) High-latitudes. The light blue lines show the scaling curves for each grid cell, while the solid dark blue line indicates the mean response of all the grid cells. The black dotted line indicates the CC rate (7%/°C). Note the logarithmic y-axes in (C–E).

Fig. 4. Comparison of extreme precipitation-temperature scaling rates with and without cloud-cooling correction across regions and datasets.

Extreme precipitation-temperature (EP-T) scaling rates were estimated using observed temperatures (red) and with temperatures corrected for the cloud-cooling effect (blue) for tropics, mid-latitudes, and high-latitudes for (A) CPC data and (C) GPCP data. B and D same as (A and C) but for the residuals between observations and fitted quantile regression.