New dataset extension: Caravan-AUS-VIC (Maribyrnong River, Victoria, Australia)

Caravan-AUS-VIC: Maribyrnong River gauging network, Victoria, Australia An extension to Caravan (Kratzert et al., 2023) contributing 12 gauging stations from the Maribyrnong River catchment, Victoria, Australia. Gauge IDs follow the Caravan two-part convention: ausvic_XXXXXX. This is version 2, fully revised to address reviewer feedback from February 2026. All issues raised in kratzert/Caravan#51 have been resolved — see the GitHub issue for the full change log. Gauge network 12 stations spanning headwaters to tidal limit: 7 from the Melbourne Water API and 5 from the Victorian Water Monitoring Information System (WMIS). The network covers tributaries (Deep Creek, Bolinda Creek, Konagaderra Creek, Jackson Creek, Turritable Creek, Main Creek) and the Maribyrnong River mainstem. Three gauges already included in Caravan via CAMELS AUS v2 (230205, 230209, 230210) have been excluded. The longest record is Maribyrnong River at Keilor (ausvic_230200), which starts in 1908. Dataset contents Following the standard Caravan directory layout: timeseries/csv/ausvic/ — 12 daily CSV files, 41 columns each: date, streamflow (mm/day), and 39 ERA5-Land columns (1950-01-02 to present). Pre-ERA5 rows (Keilor 1908–1949) have empty forcing columns. timeseries/netcdf/ausvic/ — 12 CF-1.8 compliant netCDF4 files (float32, _FillValue = -9999.0) attributes/ausvic/attributes_other_ausvic.csv — 14 standard Caravan columns (gauge metadata, catchment area, streamflow period) attributes/ausvic/attributes_caravan_ausvic.csv — 14 Caravan climate indices computed over 1981–2020 using the official caravan_utils.py code attributes/ausvic/attributes_hydroatlas_ausvic.csv — 195 BasinATLAS Level-12 attributes derived using the official Caravan Part-1 GEE notebook. Note: the GEE BasinATLAS dataset contains one additional property compared to the original Caravan publication; all filter lists from the official notebook have been applied correctly. shapefiles/ausvic/ausvic_basin_shapes.shp — single combined shapefile (EPSG:4326), gauge_id column only, polygons derived via HydroBASINS Level-12 BFS upstream traversal licenses/ausvic/license_ausvic.md Meteorological forcing ERA5-Land hourly reanalysis (ECMWF/ERA5_LAND/HOURLY) fetched via Google Earth Engine and aggregated to daily using the official Caravan Part-2 notebook methodology (UTC+11 fixed offset for Melbourne latitude). FAO-56 Penman-Monteith PET computed inline from ERA5-Land inputs using the official Caravan pet.py. Notable caveats Chifley Drive (ausvic_230106) is a tidal gauge. Of 8,070 non-missing days, only 263 exceed the tidal threshold and represent physically valid streamflow. Keilor North (ausvic_230237), Keilor (ausvic_230200), and Chifley Drive (ausvic_230106) are co-located on the lower mainstem. HydroBASINS Level-12 cannot resolve them (all snap to the same cell). Catchment areas use MERIT Hydro 90m for Keilor North and Chifley Drive, and the official Victorian Water figure for Keilor. Bolinda Creek at Clarkefield (ausvic_230211): sensor ceiling at ~500 m³/s. Three days in December 2008 masked as missing. Data sources and license Released under CC-BY-4.0. Streamflow: Melbourne Water (CC-BY-4.0), Victorian Water WMIS / DEECA (CC-BY-4.0). Meteorological forcing: ERA5-Land (Copernicus C3S). Catchment attributes: HydroATLAS/BasinATLAS (CC-BY-4.0). Catchment areas (3 gauges): MERIT Hydro (CC-BY-4.0). Please also cite: Kratzert, F., Gauch, M., Nearing, G., and Klotz, D. (2023). Caravan — A global community dataset for large-sample hydrology. Scientific Data, 10, 61. https://doi.org/10.1038/s41597-023-01975-w