An Overview of ARTMIP's Tier 2 Reanalysis Intercomparison: Uncertainty in the Detection of Atmospheric Rivers and Their Associated Precipitation
Atmospheric rivers (ARs), or long but narrow regions of enhanced water vapor transport, are an important component of the hydrologic cycle as they are responsible for much of the poleward transport of water vapor and result in precipitation, sometimes extreme in intensity. Despite their importance, much uncertainty remains in the detection of atmospheric rivers in large datasets such as reanalyses and century scale climate simulations. To understand this uncertainty, the Atmospheric River Tracking Method Intercomparison Project (ARTMIP) developed tiered experiments, including the Tier 2 Reanalysis Intercomparison that is presented here. Eleven AR detection tools (ARDTs) submitted hourly tags--binary fields indicating the presence or absence of atmospheric rivers--of detected atmospheric rivers in the Modern Era Retrospective Analysis for Research and Applications, version 2 (MERRA-2) and European Centre for Medium-Range Weather Forecasts’ Reanalysis Version 5 (ERA5) as well as six-hourly tags in the Japanese 55-year Reanalysis (JRA-55). Metrics such as AR frequency, in the context of ARTMIP mean versus consensus, precipitation attributable to ARs, and seasonal cycle are examined.
The ARDTs participating in this study employed a variety of techniques including absolute and relative thresholding, geometry requirements, and machine learning to detect ARs in MERRA-2, ERA5, and JRA55. More ARs tended to be detected in MERRA-2 due to elevated values of precipitable water (TPW) and integrated vapor transport (IVT) relative to ERA5 and JRA55. An additional difference between the reanalyses is that due to the finer resolution, ERA5 has finer ARs with stronger gradients. ARDTs that used seasonally, and spatially varying thresholds have better agreement between the reanalysis products compared to ARDTs that use a fixed threshold. The variance in the seasonal cycle of global percent area with an AR highlights the importance of geometry thresholds as well. Differences due to ARDTs are larger than differences due to reanalysis.
Due to a higher climatological mean for integrated water vapor transport in MERRA-2, atmospheric rivers were detected more frequently relative to the other two reanalyses, particularly in algorithms that use a fixed threshold for water vapor transport. The finer horizontal resolution of ERA5 resulted in narrower atmospheric rivers and an ability to detect atmospheric rivers along resolved coastlines. The fraction of hemispheric area covered by ARs varies throughout the year in all three reanalyses, with different atmospheric river detection tools having different seasonal cycles.