Notes on Stedinger et al. (2008)

Stedinger, J. R., Vogel, R. M., Lee, S. U., & Batchelder, R. (2008). Appraisal of the generalized likelihood uncertainty estimation (GLUE) method. Water Resources Research, 44(12).

The generalized likelihood uncertainty estimation (GLUE) method has been in use for watershed models since 1992 following the inaugural paper by Beven and Binley (1992). Thus, there was an early realisation of the need to gauge uncertainty in the predictions made my models within this field and also account for the uncertainty in input data.

Unfortunately, this paper (and others before it) relay that the GLUE methodology may be significantly flawed for this purpose when compared to more contemporary uncertainty processes, although careful application could improve the quality of the estimations when using GLUE.


  1. Introduction


There is a reference to the Predictions in Ungauged Basins (PUB) initiative of the International Association of Hydrological Sciences, which might be worth looking at. These guys seems to be doing stuff in almost total isolation to the rest of the V&V community, which is kind of interesting.

There were over 500 citations to the Beven and Binley (1992) GLUE paper by the time of writing of this paper, for the reasons given as:

“GLUE’s popularity can be attributed to its simplicity and its applicability to nonlinear systems, including those for which a unique calibration is not apparent.”

Yet, there appear to be severe problems with the methodology:

“Recent evaluations of GLUE by Christensen [2004], Montanari [2005], Mantovan and Todini [2006] and this study clearly demonstrate that prediction limits derived from GLUE can be significantly different from prediction limits derived from correct classical and widely accepted statistical methods.”

The main problem appears to relate to the choice of likelyhood function (to be defined properly later):

“if one wants to correctly understand the information content of the data, one needs to use a likelihood function that correctly represents the statistical sampling distribution of the data.”