The ARTMIP Tier 2 Polar Synthesis: A comparative analysis of polar atmospheric river detection tools in historical and future climate states

Abstract. Atmospheric rivers (ARs) are the conduit for the majority of atmospheric moisture transport to the polar regions and influence the evolution of polar ice sheets and sea ice. Their impacts on the polar cryosphere are expected to intensify as atmospheric moisture content and temperatures increase in a warming climate. Previous Atmospheric River Tracking Method Intercomparison Project (ARTMIP) studies show that quantitative and qualitative conclusions about the global climatology, impacts, and future changes of ARs depend on the choice of AR detection tool (ARDT) used in the analysis. There is a community need for a similar comparison of ARDTs in the polar regions (both the Arctic and Antarctica), where the unique atmospheric conditions require different parameters for AR detection relative to global ARDTs. Here, we present the ARTMIP polar synthesis experiment, a collaborative effort to provide community guidance on best practices for ARDT selection in polar studies. We build upon the existing ARTMIP framework to curate and compare AR catalogues from eleven ARDTs that have been developed for polar AR identification. We first analyze Arctic and Antarctic AR climatology across ARDTs in the historical record using MERRA-2 atmospheric reanalysis, focusing on similarities and differences among ARDTs linked to their algorithm design, and the effect of inter-ARDT differences on AR-attributed ice sheet and sea ice impacts. We then extend our analysis to future AR projections by applying a subset of ARDTs to three representative members of the CESM2 large ensemble under the SSP3-7.0 emissions scenario up to the year 2099. Finally, we use a subset of five ARDTs in a case study to assess the sensitivity of existing ARDTs to changes in detection thresholds. Our results show that polar ARDTs can generally be grouped into two categories based on their detection of AR frequency and impacts: (1) "global heritage" ARDTs that were originally developed for global AR studies and subsequently tuned for polar AR detection; and (2) "Antarctic heritage" ARDTs that were developed from inception for AR detection in the interior of Antarctica. The global heritage ARDTs generally show polar AR frequency and impacts that peak in the oceanic storm tracks of both the Northern and Southern hemispheres and decrease toward the poles, with little to no ARs identified in the interior of the Antarctic Ice Sheet. The Antarctic heritage ARDTs, conversely, show a less pronounced meridional gradient in AR detection in both hemispheres, and attribute a substantial proportion of precipitation in the Antarctic interior to ARs. Global heritage ARDTs show a large spread in future AR projections due to their varied algorithm design parameters, while Antarctic heritage ARDTs show greater cross-algorithm consistency in projecting small increases in AR frequency. Our polar synthesis leads us to recommend that researchers carefully consider their research questions and study region when choosing which ARDT(s) to use, and following the guidance from previous ARTMIP experiments, we recommend using multiple ARDTs to constrain uncertainty related to AR detection where possible.