DARK AR Detection Catalog (ERA5, 1979–2023)

This dataset provides yearly binary atmospheric-river (AR) masks derived from the ERA5 reanalysis, produced using the DARK (formerly Buffet v1.0) detection algorithm.Although originally developed with a polar (Antarctic) focus, the implementation operates globally on the full ERA5 grid.All outputs follow ARTMIP Tier-1 conventions, including variable naming, attributes, dimensions, CF metadata, and filename structure. Method Input field ERA5 vertically integrated water vapor flux components:viwve (eastward) and viwvn (northward) Data source: ERA5 reanalysis (ECMWF), typically 6-hourly fields. Overview of the steps Rolling seasonal threshold (98th percentile baseline)For each leap day-of-year (1–366), compute a 15-day rolling climatology of Integrated Vapor Transport (IVT) and extract the 98th percentile.Non-leap years are mapped to leap-year DOYs by shifting days ≥ 29 Feb (+1).The 98th-percentile field is precomputed from 1979–2023. Exceedance maskA grid cell is marked 1 if the local IVT exceeds its day-of-year threshold; otherwise 0. Area filtering (< 200 000 km²)Connected regions are identified using 8-connectivity and periodic longitude boundaries (dateline continuity).Components with total area < 200 000 km² are removed.Grid-cell areas are computed on WGS-84 using geocentric radius and local Δlat/Δlon. Length filtering (< 2 000 km)Each connected component is skeletonized and represented as a graph.The endpoint-to-endpoint path best aligned with the principal component axis is selected, extended to the edges, and smoothed using a B-spline.Sub-grid geodesic segment lengths (WGS-84) are summed; components shorter than 2 000 km are discarded. Output assemblyEach yearly NetCDF file contains the binary field ar_binary_tag(time, lat, lon) with values 1 = AR and 0 = no AR. Files & Format One NetCDF file per year: ERA5.ar_tag.DARK.<freq>.<YYYYMMDD-YYYYMMDD>.nc where <freq> is inferred from input ERA5 temporal resolution (e.g. 6hourly). Variables Variable Type Description ar_binary_tag(time, lat, lon) byte 1 = AR, 0 = no AR Attributes: description = "binary indicator of atmospheric river" scheme = "DARK" version = "1.0" credits = "Developed by Victoire Buffet, Vincent Favier, and Benjamin Pohl" Coordinates: time — CF-compliant (minutes since <year>-01-01 00:00:00, calendar = "standard", axis = "T") lat — degrees_north (axis = "Y") lon — degrees_east (axis = "X") Compression: NetCDF-4 deflate (zlib=True, complevel=1, shuffle=True); time unlimited. Spatial / Temporal Coverage Spatial: Global ERA5 native 0.25° × 0.25° grid.(Algorithm tuned for polar ARs but operates globally.) Temporal: 1979 – 2023, using ERA5 6-hourly timesteps. Parameters (defaults) Rolling window: 15 days Percentile: 98th Minimum area: 200 000 km² Minimum length: 2 000 km Reproducibility Inputs ERA5 u/v flux files:vertical_integral_of_eastward_water_vapour_flux_{year}.ncvertical_integral_of_northward_water_vapour_flux_{year}.nc Percentile baseline:percentile_98_IVT_rolling_15D_era5.nc (1979–2023) Outputs Yearly AR masks:ERA5.ar_tag.DARK.6hourly.YYYYMMDD-YYYYMMDD.nc Software Python ≥ 3.11 Key libraries: xarray, numpy, scipy, scikit-image, geopy, networkx, matplotlib Geodesic distances computed with geopy (WGS-84) NetCDF written CF-1.6-compliant using netCDF4 Credits Developed by Victoire Buffet, Vincent Favier, and Benjamin Pohlat Institut des Géosciences de l’Environnement (Grenoble) and Biogéosciences (Dijon). Keywords Atmospheric Rivers; ERA5; ARTMIP; IVT; Antarctic; Polar Meteorology; Reanalysis; Climate Diagnostics Related Material ARTMIP format and Tier-1 conventions ERA5 reanalysis documentation