Atmospheric rivers (ARs) are filamentary streams of enhanced vapor transport which drive a substantial portion of the global water cycle and impact regional water resources. Climate change studies performed in fully coupled climate models provide some degree of predictability for how these essential moisture conduits will respond to global warming. However, the inherent complexity of these models can make it difficult to isolate the physical drivers behind climate change responses. Given the slow but steady response of sea-surface temperatures (SSTs) to climate change, we wish to investigate the process chain that connects SSTs to AR characteristics (e.g., occurrence, moisture fluxes, etc) and understand if this connection accounts for the majority of the AR response.

This study uses the Community Atmosphere Model version 5 (CAM5) in an aquaplanet configuration with a zonally uniform and equatorially symmetric prescribed SST distribution. We prescribe two unique SST scenarios derived from output produced by fully coupled model runs under present-day and Representative Concentration Pathway 8.5 (RCP8.5) conditions, respectively. We detect ARs with an original, objective algorithm which is conditioned on instances of enhanced vertically integrated vapor transport (IVT) and is available as part of the TempestExtremes software suite.

Aquaplanet ARs are compared to those simulated by fully coupled climate models produced as part of the Coupled Model Intercomparison Project version 6 (CMIP6). We include statistics on AR morphology, occurrence, vapor fluxes, and precipitation, both in terms of mean and extreme (90th+ percentile) events. We contextualize these results by investigating changes in the larger-scale circulation across models. In all, this study will isolate the role of SST as a physical driver of AR characteristics, while simultaneously evaluating the applicability of AR results obtained in an idealized framework to those produced by much more computationally expensive — yet more realistic — fully coupled models.