Forecasting the water flows draining to the Great Barrier Reef using the G2G distributed hydrological model

Wells, Steven C.; Cole, Steven J.; Moore, Robert J.; Black, Kevin B.; Khan, Urooj; Hapuarachchi, Prasantha; Gamage, Nilantha; Hasan, Mohammad; MacDonald, Andrew; Bari, Mohammed; Tuteja, Narendra. 2018 Forecasting the water flows draining to the Great Barrier Reef using the G2G distributed hydrological model. Wallingford, UK, NERC/Centre for Ecology & Hydrology, 72pp. (Technical Report (contract no. 112-2015-16), CEH Project no. C05793) (Unpublished)

Before downloading, please read NORA policies.
[img] Text (Contract Report (submitted and in review))
Forecasting the water flows draining to the Great Barrier Reef using the G2G Distributed Hydrological Model V2 rev4.pdf - Submitted Version
Restricted to NORA staff only

Download (3MB) | Request a copy


This document reports on a CEH (UK) and Bureau of Meteorology (Australia) collaborative project to apply the G2G (Grid-to-Grid) distributed hydrological model for forecasting water flows to the Great Barrier Reef (GBR). It forms a component of the ‘Catchment Monitoring and Modelling’ stream of the eReefs project concerned with protecting and preserving the Reef environment. The primary purpose here is to quantify the water discharges along the coastline of the Reef. These in turn, through modelled water quality constituent concentrations, can be used to forecast sediment and nutrient loads for input to Coastal Ocean Models and other Marine models of the Reef marine environment. The G2G model of the land area draining to the Reef also has utility for forecasting water availability and flood hazard in Queensland, including assessing impacts on transport. G2G (Grid-to-Grid) is a physical-conceptual grid-based distributed hydrological model used by environment agencies in Britain to support 24/7 flood guidance at countrywide scale out to 5 days. It is also used to create Hydrological Outlooks months ahead using predictions of seasonal precipitation, to produce spatial flood risk assessments under climate change, and to assess water discharges into the sea through combining gauged and modelled river flows. Its area-wide formulation allows G2G to provide real-time flow forecasts everywhere on a 1km model grid and thus is particularly well suited for application to areas that are ungauged. This is achieved through use of a model formulation that is underpinned by spatial datasets on terrain, land-cover and soil/geology properties. These landscape properties act to shape a storm pattern into a flow response providing a coherent space-time map of river flows from catchment to coast and from flood to drought. Depending on application, G2G can be configured for a range of model domains (catchment, river basin, regional, national, global) and grid-resolutions (from 50 m through 1 km and 25 km have been used). G2G has previously been evaluated in Australia for water flow forecasting at a range of time-scales, for flood warning and water availability applications up to seven days ahead, through a pilot case study on the Upper Murray. The present study builds on this experience. This report provides background to the G2G model and its use in the UK. Challenges in setting up G2G using the spatial datasets available for Australia are discussed in some detail as part of configuring the G2G to the land area draining to the Reef. An approach to model setup, calibration and assessment is developed that is tailored to the hydrometric network of the Reef land area and the impact of reservoir controls judged to be significant. About 22% (76,000 km2) of the Reef land area is ungauged, making the G2G ungauged modelling capability particularly important. This document concerns the G2G model configured to the Northern Domain, as work led by CEH, with the Bureau reporting separately on the Southern Domain. The domain separation provides an opportunity to assess the ungauged capability of G2G through use of the Northern Domain G2G model parameters as a starting point for the Southern Domain G2G model. Over a seven year period 2007 to 2016, hourly flows for a total of 130 gauged catchments supports assessment of the Northern Domain model, compared to 162 (with some overlap of gauges) for the Southern Domain. Only records for 2010 are used for calibration of G2G model parameters: the global model parameters are few in number with the main model configuration support coming from spatial datasets on terrain, soil and land-cover. Local model parameters, controlling channel conveyance and return flows (from subsurface to surface), are obtained through a “model reach calibration” where gauged inflows to each gauged catchment are treated as known. Reservoirs at Fairbairn and Burdekin Falls dams are treated as damped outflows within the G2G Norther Domain configuration to obtain improved performance without recourse to modelling the detail of reservoir operation. The performance of the G2G model is first assessed in simulation-mode, with a focus on how well G2G transforms gridded rainfall and potential evaporation (PE) to river flow. Observed rainfall provided by the Bureau is in the form of a hourly 1km grid of values derived from a network of raingauges providing either daily or hourly values: this acts as a governing constraint on G2G performance for areas where rainfall is sourced from daily values only. The performance of G2G in forecast-mode using full data assimilation (a combination of state-updating, direct flow insertion and error-prediction) of river flows is then assessed in terms of the three metrics R2 Efficiency, relative bias and cumulative volume error. The forecast-mode assessment shows how performance varies with lead-time and the dependence on rainfall forecasts is illustrated using thee alternative input scenarios: perfect foreknowledge of raingauge rainfall, ACCESS-R weather model rainfall and zero rainfall. Against this evidence-base of performance, and an understanding of issues needing to be addressed, the overall performance of G2G is seen as encouraging. The G2G model at 1km grid resolution is applied to produce hourly time-series of water discharge volumes along the Reef coastline in a form suitable for use with the eReef marine models. This exploits the ungauged capability of G2G to use direct insertion of observed flows at the lowest gauged location on each river, together with modelling of ungauged area runoff from rainfall and PE for areas downstream, and routing of these gauged flows and runoffs down to the Reef coastline. G2G flows are used to estimate the fraction of ungauged to gauged area flows for the 35 marine cluster points on a year by year basis: this highlights the importance of ungauged area flow modelling using G2G. Finally, an assessment of the performance of G2G for ungauged area modelling is illustrated for three sites in the lower Pioneer catchment treated as ungauged. This emulates the practical situation of forecasting catchment water flows at the coast, downstream of the last river flow gauge(s) in the catchment. The report closes with a summary of its content, a setting down of the project’s main achievements and a list of recommendations - both of a research and operational nature. A key achievement, beyond configuring and assessing G2G over the Reef’s gauged and ungauged land area, is the knowledge exchange realised between CEH and the Bureau. Bureau scientists are now able to configure and run G2G (and associated scripts for data processing and analysis) on the NCI computer in Australia. Recommendations focus on guiding the Bureau of Meteorology in its future planning of G2G activity, especially in relation to Great Barrier Reef and Queensland applications. Some future opportunities for improvement are set down below. (i) Improving the quality of the hourly gridded rainfall estimates used as input to G2G. (ii) The PE dataset used here extended from January 2007 to April 2016. An extended PE dataset to February 2017 is now available and should be used in future work. (iii) Learning from the Bureau-led assessment of G2G over the GBR Southern Domain and the Bureau’s use of SWIFT catchment models over the gauged land area of the Reef. (iv) Considering improvements to the spatial datasets on soil/geology, terrain and land cover supporting G2G configuration. One example is the better treatment of braided rivers (delineation and channel width estimation) to improve flow routing in flat, downstream reaches. (v) G2G applications in more semi-arid areas of Australia would benefit from trialling, and possibly improving upon, the G2G formulation of infiltration capacity excess runoff, complementing the saturation excess formulation employed here. Recommended future operational developments point to considering the following. (i) The steps needed to make G2G 24/7 operational over the Reef land domain, supporting the eReefs marine modelling requirement for coastal discharge forecasts as well as providing flow forecasts for flood warning and water availability guidance over Queensland. (ii) Gaining benefits through using G2G for spatial risk-based probabilistic flood forecasting, exploiting ensemble rainfall forecast products available to the Bureau to map the temporal evolution of spatial flood risk in real-time.

Item Type: Publication - Report
UKCEH and CEH Sections/Science Areas: Hydro-climate Risks (Science Area 2017-)
Funders/Sponsors: Bureau of Meteorology, Australia, Centre for Ecology & Hydrology
Additional Information. Not used in RCUK Gateway to Research.: Contract Report for Bureau of Meteorology, Australia.
Additional Keywords: water forecasting, Great Barrier Reef, hydrological model, distributed, G2G, coastal discharge, flood, rainfall-runoff
NORA Subject Terms: Hydrology
Marine Sciences
Meteorology and Climatology
Atmospheric Sciences
Date made live: 20 Feb 2018 12:57 +0 (UTC)

Actions (login required)

View Item View Item

Document Downloads

Downloads for past 30 days

Downloads per month over past year

More statistics for this item...