Lars Radtke

Atmospheric rivers [PDF]
finished 2026-03
supervised by Michael Schmiedeberg
with the Potsdam Institute for Climate Impact Research

Abstract

Atmospheric rivers (ARs) play a crucial role in the global distribution and transportation of moisture and are often linked to extreme precipitation events. While extreme precipitation intensifies with global warming, the relative contribution of atmospheric rivers remains uncertain. This study applies the PIKART algorithm to catalog projected atmospheric rivers from simulated integrated water-vapor transport (IVT) in a multi-model ensemble. A spatial and temporal resolution test showed a strong sensitivity of results, with substantial reductions in AR metrics for coarser grids. Atmospheric river characteristics are then analyzed with 1◦ spatial and 6 h temporal resolution to determine differences in the future scenarios SSP1-2.6 and SSP5-8.5 to the historical simulation. Results indicate a strong intensity increase, a poleward shift of mid-latitude atmospheric river tracks and a slight decrease in the number of detected annual tracks. Size, shape and lifetime showed no significant differences. Impact-relevant changes of atmospheric rivers are evaluated by their link to extreme precipitation intensity and frequency. Precipitation events are classified as AR-associated when detected footprints spatially overlap on the same day. A low-frequency component analysis on global near-surface air temperature and the extreme precipitation metrics is applied to separate forced radiative trends from internal variability. Land-averaged changes of extreme events, relative to a 1950-1970 historical baseline, are evaluated against global-mean temperature. Exponential scaling rates are estimated and compared to theoretical expectations, resulting in ensemble mean intensity rates of (5.8 ± 1.3) %/K for all events, (7.5 ± 2.7) %/K for AR-associated events and (5.2 ± 1.4) %/K for non-AR events. All rates align with a strong thermodynamic contribution. The frequency of extremes scales with (12.9 ± 1.9) %/K for all events, (18 ± 7) %/K for AR-associated events and (12.8 ± 1.2) %/K for non-AR events. Faster scaling of AR-associated compared to non-AR events is projected in all models for both extreme intensity and frequency, but large uncertainties remain from model spread and low AR event statistics.