DARK (formerly Buffet v1.0) AR detection catalog (CESM2-LE, 1990-2009 and 2080-2099)

Description This dataset provides yearly binary atmospheric-river (AR) masks derived from the CESM2 Large Ensemble (CESM2-LE), produced with the DARK (formerly called Buffet v1.0) detection algorithm.Although originally developed with a polar (Antarctic) emphasis, the implementation operates globally on the full CESM2 grid.Output files follow the ARTMIP Tier-1 conventions (variable naming, attributes, dimensions, CF metadata, and filename structure). Method Input field.The algorithm uses the CESM2 total Integrated Vapor Transport (IVT) directly from the CESM2-LE atmospheric output (cam.h2.IVT) for each ensemble member (LE2-1011.001 , LE2-1031.002 , LE2-1051.003) and experiment (BHISTsmbb and BSSP370smbb). Overview of the steps 1. Rolling seasonal threshold (98th percentile baseline).For each leap day-of-year (1–366), we compute a 15-day rolling-window climatology of IVT at each grid cell and take the 98th percentile across all days and years from the historical reference period (1990–2009) for the historical period, and from the future period (2080–2099) for the future period.Non-leap years are mapped to leap-year DOYs by shifting days ≥ 29 Feb (+1). 2. Exceedance mask.A grid cell is assigned 1 if its local IVT exceeds the corresponding climatological threshold for that day-of-year and location; otherwise 0. 3. Area filtering (< 20 000 km²).The algorithm applies 8-connected labeling with periodic longitude boundaries to ensure dateline continuity, removing components smaller than 20 000 km².Grid-cell areas are computed on WGS-84 using geocentric radius and local Δlat/Δlon. 4. Length filtering (< 2 000 km) with seam-free centerlines.For each remaining component (AR candidate) and time step: A skeleton is computed, converted to a pixel-adjacency graph, and all endpoint-to-endpoint paths are enumerated. The path best aligned with the component’s PCA axis is selected, extended to the edges, and smoothed with a B-spline. Sub-grid geodesic segment lengths (WGS-84) are summed to obtain total AR length.Candidates shorter than 2 000 km are discarded. 5. Output assembly.Each yearly NetCDF file contains the binary field ar_binary_tag(time,lat,lon) with 1 = AR, 0 = no AR. Files & Format One NetCDF per year and ensemble member: b.e21.BHISTsmbb.f09_g17.LE2-1051.003.ar_tag.Buffet_v1.<freq>.<YYYYMMDD-YYYYMMDD>.nc b.e21.BSSP370smbb.f09_g17.LE2-1051.003.ar_tag.Buffet_v1.<freq>.<YYYYMMDD-YYYYMMDD>.nc where <freq> is inferred from the input CESM2-LE data (e.g., 6hr). Variables ar_binary_tag(time, lat, lon) — byte; 1 = AR, 0 = no AR. Attributes: description = "binary indicator of atmospheric river" scheme = "Buffet" version = "1.0" credits = "Developed by Victoire Buffet, Vincent Favier, and Benjamin Pohl" time(time) — CF-compliant time coordinate withstandard_name = "time", long_name = "Time",units = "minutes since <year>-01-01 00:00:00",calendar = "standard", axis = "T". lat(lat) — latitude in degrees_north;CF/ARTMIP attributes: standard_name, long_name, axis = "Y". lon(lon) — longitude in degrees_east;CF/ARTMIP attributes: standard_name, long_name, axis = "X". Compression:Data variable uses NetCDF-4 deflate (zlib=True, complevel=1, shuffle=True).The time dimension is unlimited. Spatial / Temporal Coverage Spatial: Global CESM2-LE atmospheric grid (0.94240838° × 1.25°).Although the algorithm was designed with a polar focus (Antarctic and Arctic ARs), it evaluates the entire globe. Temporal: Historical baseline: 1990–2009 (b.e21.BHISTsmbb experiment) Future scenario: 2080–2099 (b.e21.BSSP370smbb experiment) Time resolution: 6-hourly (cam.h2 files) Parameters (defaults) Rolling window: 15 days Percentile: 98th Minimum area: 20 000 km² Minimum length: 2 000 km Reproducibility Inputs CESM2-LE ensemble IVT files with variable IVT: b.e21.BHISTsmbb.f09_g17.LE2-xxxx.xxx.cam.h2.IVT.<YYYYMMDDHH>-<YYYYMMDDHH>.nc b.e21.BSSP370smbb.f09_g17.LE2-xxxx.xxx.cam.h2.IVT.<YYYYMMDDHH>-<YYYYMMDDHH>.nc Two stages Build rolling‐DOY percentiles (once): b.e21.BHISTsmbb.f09_g17.LE2-xxxx.xxx.ivt98pctl.rolling15D.1990-2009.nc Tag each year to produce ar_binary_tag yearly files using the same 98th-percentile baseline. Software Python ≥ 3.11Key libraries: xarray, numpy, scipy, scikit-image, geopy, networkx, matplotlib. Geodesic distances computed with geopy (WGS-84). Files written CF-compliant with netCDF-4. Credits Developed by Victoire Buffet, Vincent Favier, and Benjamin Pohlat the Institut des Géosciences de l’Environnement (Grenoble) and Biogéosciences (Dijon). Keywords Atmospheric Rivers; CESM2; ARTMIP; IVT; Climate Change; Antarctic; Arctic; Polar Meteorology; Large Ensemble Related Material ARTMIP format and Tier-1 conventions CESM2-LE documentation