Problems of calibrating environmental models have long been recognised, and there is an increasing appreciation that model predictions should be associated with some estimate of uncertainty. How to estimate that uncertainty remains problematic, however, since there are multiple sources of uncertainty (including errors associated with the model structures used) that are difficult to estimate given the limitations of the data normally available. It can therefore be difficult to apply statistical methods that depend on specific models of the errors. An alternative approach, GLUE, has been developed at Lancaster, based on the acceptance that there may be many models that provide acceptable fits to the available data. This has now been applied to a wide range of problems including catchment hydrology, atmospheric and water quality, ecology, land surface parameterisations, groundwater and flood modelling. Recent work in this area is funded by a NERC Long Term grant ward to Keith Beven and by the EPSRC/Defra/EA Flood Risk Management Research Consortium.
Some recent references
Contact: Keith Beven