Uncertainty and Trade-Off Analysis

Through the projects we have worked on and our published research, we have considerable experience in the analysis of the uncertainties and sensitivities in models of complex systems. To help explore uncertainty and trade-offs within ensemble model scenarios, taking account of costs of monitoring and mitigations, we have developed UNCOVER.

What is UNCOVER?

We can help you harness your existing model or build a new one to generate the required scenarios in UNCOVER.  You can then use the interactive map and model parameter space to explore uncertainties and trade-offs between near-optimal solutions based on a range of measures and techniques.  It includes:

  • Interactive mapping to alter your monitoring regime (removing sites / altering sampling frequencies)
  • Intuitive mitigation scenarios such as ‘tree planting’ (based on sub-setting or ‘brushing’ of model parameter space)
  • User-defined costs for both monitoring and mitigation scenarios
  • Pareto set trade-off analysis using multiple performance measures to find near-optimal solutions
  • Uncertainty in model predictions using the GLUE (Generalized Likelihood Uncertainty Estimation) methodology first developed by Beven and Binley in 1992.
  • Quantile-quantile plots to assess significance of mitigations

It allows you to make a more informed decisions based on uncertainties, whole catchment performance, business information and trade off analysis, through the exploration of different scenarios for different monitoring strategies or mitigations.

Our Scientific Approach

We have always appreciated that the underlying uncertainty in the model parameters can influence and propagate into model predictions and (should) impact decision making.  We have applied a range of approaches from GLUE to Pareto front analyses to a wide range of environmental and flood risk studies, and recently had the opportunity to develop these techniques into a single framework through the EU SWITCH-ON project. Whilst familiar with a range of sensitivity, scenario and uncertainty approaches, it is also important to retain a real-world focus on how decisions are made.  Typically this involves quantifying costs and benefits (or dis-benefits) of different scenarios both in terms of the monitoring effort required to calibrate or validate models, and the cost of introducing mitigations.

Our Project Experience

Modelling for Monitoring, Environment Agency, 2014/15

Investigating the impact of a changing monitoring regime on different integrated catchment models built to understand diffuse pollution in the Weaver-Dane catchments.  A detailed SIMCAT model was constructed and particular reaches were calibrated depending upon an assumed monitoring regime, and the catchment wide performance was assessed on the basis of all available monitoring data.  A HYPE model was also set up to assess the performance of this coarse-scale, open-source European model, alongside two detailed INCA-P and INCA-N models. SIMCAT and INCA were run in a Monte Carlo framework and a GLUE uncertainty analysis was undertaken for SIMCAT.  Key conclusions were that the head water sampling regime is critical to modelling diffuse pollution.

Fieldmouse: Modelling impacts of Catchment Sensitive Farming, EA, 2014/15

Working with EA Evidence and CEH to improve an S-P-R framework GIS based model to assess the effectiveness of agri-environment measures.  This is a farm-scale model whereby emissions to land are transported across the landscape taking into account topography to the watercourse network.  The loads are combined with point sources and routed along the river network.  We enhanced the model capability and we will be working with CEH and Lancaster University to improve the model functionality based on the application of the SWAT model to a test catchment, and to calibrate the model in an uncertainty (GLUE) framework.

Leave a Reply