RRRD030: Integrating measurements with models for improved confidence in pollutant loads estimation

Petra M. Kuhnert1, Dan Pagendam1, Rebecca Bartley2, Ross Searle2, Erin Peterson1, William B. Leeds3, Chris K. Wikle4,Rob Ellis5, Ryan Turner5, David Waters6, Cameron Dougall6, Stephen Lewis7, Zoe Bainbridge7 and Jon Brodie7

1 CSIRO Mathematics, Informatics and Statistics

2 CSIRO Land and Water

3 Department of Statistics, University of Chicago, USA

4 Department of Statistics, University of Missouri, USA

5 Department of Science, Information Technology, Innovation and the Arts, QLD

6 Department of Natural Resources and Mines, QLD

7 Centre for Tropical Water and Aquatic Ecosystem Research (TropWATER), James Cook University, QLD

pdfDownload the RRRD030 Research Outcomes Report5.67 MB

Executive Summary

As part of the Reef Water Quality Protection Plan (Reef Plan) there is a requirement to reduce the export of nutrients, sediments and pesticides to the Great Barrier Reef (GBR) lagoon. Underpinning strategies, such as the Reef Rescue Water Quality Grants, the Paddock to Reef Integrated Monitoring, Modelling and Reporting (P2R) Program and the Reef Rescue Research and Development Program are aimed at measuring contaminant export from catchments, and then reducing those exports through the introduction of improved land management practices.

The current approach for loads estimation for the Reef Plan uses Source to model and calculate loads from the GBR, which is calibrated with on ground measurements. There is a direct need to "blend" the model output from Source with measurements obtained through monitoring to provide a level of confidence in the loads that can assist managers in determining where management is having an effect.

We focus on the Upper Burdekin catchment in the GBR catchment area and develop a "blending" methodology that adopts a Bayesian Hierarchical Modelling (BHM) framework. This framework takes the daily measurements of flow and concentration (e.g. total suspended sediment or TSS) at 5 sites monitored in the catchment, and "blends" them with Source output from 411 spatial locations. This "blending" provides estimates (and levels of confidence) of TSS, flow and the load at each daily time step, for each spatial location.

In the context of sediment loads in the GBR catchments, we developed this "blending" methodology to provide:

  • Confidence in the sediment load arising from the Source model;
  • Estimates of the load provided at the daily, monthly, seasonal (wet versus dry) and annual scales with uncertainties;
  • The exceedance of the load as a probability map or curve to highlight spatial and/or temporal regions of exceedances;
  • Forecasts of concentration, flows and loads in the Upper Burdekin catchment one year out with a level of confidence around the forecast; and
  • Sources of erosion that are contributing to the load.

We outline two cases studies to illustrate the power of the BHM: (1) Weany Creek (14 square kilometer basin sampled daily over an 11 year period); and (2) the Upper Burdekin (40,000 square kilometres) consisting of 411 sites along a stream network.

This work is currently being finalised with additional support and funding provided by CSIRO to provide a methodology that can be used to:

  • prioritise areas within a catchment to obtain improvements in the loads
  • identify the sources of loads; and
  • invest more effort in monitoring at a site to have higher confidence in the loads.