Daily 1 km Surface Flow for Catchments Draining to the Great Barrier Reef - Grid to Grid model (eReefs, BOM, UKCEH)

Full description

This dataset provides modelled hourly, 0.01° (~1 km) resolution NetCDF grids of surface flow rates (water flowing over land and in rivers) from 1 Jan 2007 through to 31 Dec 2023 in the catchments of the Great Barrier Reef. This dataset was generated with the Grid‑to‑Grid (G2G) distributed hydrological model configured jointly by the Bureau of Meteorology (BOM) and the UK Centre for Ecology & Hydrology (UKCEH). Flows were produced based on blended gridded rainfall and potential evapotranspiration with mapped landscape data (terrain, soils, land cover, derived river channel dimensions). The G2G model was applied across the whole GBR catchment area to supply consistent, hourly freshwater discharge inputs (river plus ungauged overland runoff) at 69 river mouths for driving the eReefs hydrodynamic and biogeochemical (v4.0) models. The surface flows from these models (this dataset) were also made available so users can trace and visualise regional areas that contributed to resulting flood plumes. It should be noted that the G2G model was tuned for generating correct river mouth flow rates, not flow rates far from the coast. This dataset was produced using the Grid‑to‑Grid (G2G) hydrological model. G2G works on a regular 0.01° (~1 km) grid so it can estimate freshwater flows not only where river gauges exist but also across ungauged land down to the coastline. About 22% (~76,000 km²) of the land area draining to the Great Barrier Reef (GBR) has no river gauges; without this approach the run off from these ungauged land areas would be poorly characterised. Various types of spatial data were prepared to set up the G2G model for the GBR region. These include the flow direction, flow accumulation, mean slope of grid cells, real area of each grid cell, main river length and slope, land/river designation, land cover (urban and suburban percentage), the distance between the centroid of grid cells along the river network, dominant soil class and soil hydraulic properties. These data have been prepared with the guidance and help from the UKCEH team. G2G takes hourly rainfall and potential evaporation grids and combines them with mapped landscape information (terrain slopes and flow paths, soils and their water holding capacity, land cover including urban areas, and derived river channel dimensions) to estimate how much rain becomes quick surface runoff, slower subsurface flow, and finally river discharge moving cell‑to‑cell to the coast. Water is routed from grid-cell to grid-cell, with flow paths configured using a digital terrain model. The G2G model integrates both observed stream flows with estimated stream flows predicted from rain run off. Hourly flow data from 317 streamflow gauges (Bureau's internal database, also available online from https://www.bom.gov.au/waterdata/) were injected into the model grid before routing downstream so measured peaks and recessions were preserved; the model then added simulated contributions from upstream ungauged areas. A gentle adjustment maintained slower groundwater/baseflow states in line with observations without abrupt shifts. Major storages (Fairbairn, Burdekin Falls and others) were not explicitly represented by calibrated reservoir operation parameters in this configuration due to limited consistent operational data. Their regulating influence enters the modelling indirectly via downstream gauge flow insertion rather than through a detailed damping scheme. Rainfall and PET inputs: rainfall surfaces were generated by blending hourly gauge records (2209 stations; gauge density decreases inland) with daily Australian Gridded Climate Data (AGCD) gridded rainfall using inverse‑distance weighting within 50 km of each grid cell; in data‑sparse western areas daily totals were distributed evenly across hours. Potential evapotranspiration (PET) was sourced from AWRA-L (Australian Water Resources Assessment – Landscape, available from https://awo.bom.gov.au/) daily products (0.05° grid). This was mapped to the 0.01° G2G grid at an hourly time step by the model using linear interpolation. Spatial variation in input data density, especially sparse hourly rainfall coverage inland, is the main source of variation in reliability. Calibration and ungauged areas: model calibration was done in two stages so the behaviour is consistent where measurements exist and still reasonable where none are available. First, a small set of 'global' parameters that influence how rain becomes runoff everywhere (linked to soil, slope and land cover) were adjusted using one representative year within 2007–2014 while inspecting how well the model reproduced observed flow patterns at many gauges. Second, simple 'local' tweaks were applied at individual gauged reaches to better match the timing and gradual fall of peaks (by adjusting how fast water moves along channels and how much slow groundwater returns to the stream). This keeps changes modest and avoids over‑fitting. After these steps the same parameter set was left unchanged and checked ('verified') against later years (2015–2023) to confirm stability. Areas without gauges, including coastal strips and inland headwaters, use only the mapped landscape inputs with no custom tuning; performance there depends mainly on rainfall data quality and how representative the soil and slope maps are. Overall skill, expressed as a median Nash–Sutcliffe Efficiency of about 0.7 for hourly flows across 317 gauges over 2007–2023, shows the calibrated model reproduces most flow variability while acknowledging that accuracy is lower in inland regions with sparse hourly rainfall and in very small flashy catchments. Format: NetCDF files - Daily files each containing a single 24 hour time period, averaged from the source hourly data. Variable: g2gflow - units: cumecs (m^3/s), type: Float32, NoData value of -999 Timestamp convention: the flows labelled at e.g. 01:00 am 01-Jan-2007 are the average flows in the internal 01:00 am to 02:00 am, with the timestamp marking the start of the time interval. Grid size is 1146, 1866 CRS: EPSG:4326 Origin = (142.269993896596134,-10.640000493265688) Pixel Size = (0.009999999776483,-0.009999999776483) Upper Left ( 142.267, -10.64) Lower Left ( 142.267, -29.30) Upper Right ( 153.73, -10.64) Lower Right ( 153.73, -29.30) An example on how to to programmatically download the data files can be found in the Python script "03-download-daily-g2g-data.py" in the code repository https://github.com/open-AIMS/ereefs-g2g-prototype-visualisation . Visualisations: MP4 visualisations for Queensland (full extent), North, Central, and South regions with synchronized surface flow and salinity plots. These visualisations highlight the influence of high surface flows on the GBR marine eco system by showcasing the variation in salinity. During intense rain events the increase in surface flow leads to a clearly visible decrease in salinity in coastal areas. Limitations: The model was tuned to get total hourly discharge right at downstream river mouths, not to perfectly map every upstream cell. Surface flow patterns away from gauges, and in small fast‑responding coastal creeks or headwaters, have high uncertainty, especially where only daily rain was spread evenly across hours. Flows upstream of gauges where observed data are inserted are estimates, not measurements. Large dams are only indirectly represented. Use the grids to view broad patterns and relative source areas, not to derive precise small‑scale water balances. eAtlas Processing: The original hourly data converted to daily data by averaging the source hourly data. This was done to minimise the size of the model dataset, and to align better with the intended visualisation products. The full hourly data is stored in the eAtlas data repository. It is not available online, but is available on request. Location of the data: This dataset is filed in the eAtlas enduring data repository at: data\custodian\2025-2029-eReefs\QLD_BOM_eReefs-g2gflow_2011-2023 Change log: As updates to this dataset are published, the changes will be recorded here. 2026-03-20: Initial publication 2026-03-25: Add links to data and code

Lineage

Maintenance and Update Frequency: notPlanned

Notes

Credit
The work described in this record was developed as part of eReefs Phase 5, funded by the Australian Government's Reef Trust, managed by the Great Barrier Reef Foundation and co-funded through in-kind support and co-investment from project partners and collaborators.