The mystery of multidecadal precipitation trends in southeastern South America

Abstract

<p>Austral summer precipitation increased by ~25% from 1902-2021 in southeastern South America (SESA; southern Brazil, Paraguay, Uruguay, and northern Argentina); one of the largest observed precipitation trends globally. Previous studies have attributed the SESA precipitation increase to the Atlantic multidecadal variability (AMV), stratospheric ozone depletion, or greenhouse gas (GHG) forcing. While models simulate SESA precipitation increases due to anthropogenic forcing over the projection interval, a precipitation trend as large as that observed is almost never captured in model simulations over the historical period. We first assessed this model behavior by characterizing the relationship between the South American low-level jet and SESA precipitation using analyses of low-level moisture fluxes within the jet region. We use these fluxes as a metric for predicting SESA precipitation in observations, reanalysis, and simulations from phase 6 of the Coupled Model Intercomparison Project from 1951-2014 (the period of overlap between reanalysis data and historical simulations). We find that while the simulated <em>relationships</em> between the jet and SESA precipitation are consistent with observations, only ~2% of the historical simulations yield a SESA precipitation <em>trend</em> that falls within the observed uncertainty range. Therefore, the simulated low-level jet dynamics do not appear to resolve the muted CMIP6 estimates of 20<sup>th</sup>- to 21<sup>st</sup>-century SESA precipitation trends. We subsequently characterize the influence of large-scale dynamics on the SESA precipitation trend. We take particular interest in quantifying the roles of the AMV, stratospheric ozone depletion, and GHG forcing on the SESA precipitation trend across seasons and through time in observations compared to CMIP6. These large-scale dynamics are interpreted in terms of the unexplained portions of the historical trend and interpreted in terms of their representation in the CMIP6 models.</p>