Aerosols and Deep Convection in the Amazon Basin: A Spectrum of Possible Updraft Responses

Abstract Previous studies have found a wide range of estimates for the impact of aerosols on deep convective cloud (DCC) dynamics, leaving the hypothesized aerosol‐induced invigoration of DCCs an ongoing topic of debate due to unknown pathways and varying magnitude. A central challenge in observational studies is separating the effect of aerosols on convective updrafts from other influences on aerosols and updrafts, so‐called confounders. To address this, we use a causal inference algorithm known as G‐computation, which allows us to determine the causal effect of aerosols on updraft speeds while statistically controlling for confounders. The analysis focuses on isolated DCCs over the Amazon basin during the wet season utilizing data from the GoAmazon 2014/5 experiment, which includes unique in‐cloud vertical velocity retrievals. Additionally, we conduct sensitivity tests to comprehensively account for sampling limitations, including those related to aerosol variability, updraft sampling, convective lifecycle stages, and measurement uncertainties. Results show that updraft speeds are on average only marginally stronger under polluted conditions compared to clean conditions (up to 0.6 m s −1 or 8%). However, 10th and 90th percentile values range from −1.5 to +2.1 m s −1 (−18% to +26%), reflecting substantial uncertainty and implying that both invigoration and enervation of DCCs are possible. Furthermore, we find that the aerosol impact on updraft speeds may be overestimated if convective lifecycle stage is not controlled for. These results highlight that the inferred aerosol effects on updraft speed are highly dependent on how the data are processed and stratified.