EDITORIAL FUTURE OF DISTRIBUTED MODELLING
SPECIAL ISSUE OF HYDROLOGICAL PROCESSES
This special issue of hydrological processes is the result of a meeting on the future of distributed models held at Lancaster University under the auspices of the British Hydrological Society. I was asked to organize this symposium as a result of the paper published in 1989 in the Journal of Hydrology 'On changing ideas in hydrology: the case of physically-based models'. That paper was a critical review of the process of physically based modelling in which I argued that, although the application of physically-based models might appear to be rigorous in principle, in practice the process was flawed due to the unknowns in the system, overparameterization of the models and the implicit lumping of subgrid processes inherent in the numerical approximations used. It was shown that there was a certain equivalence between distributed physically-based models and lumped conceptual models, and it was suggested that all hydrological models are in general invalidated by detailed comparisons with field observations. However, there are problems that require 'physically-based' predictions, in particular the problems of predicting the future hydrological effects of land use and climatic change when no data is available for model calibration or recalibration. Thus it is necessary to continue to improve the hydrologist's available modelling tools in this area.
In deciding on a format for the symposium I chose to try and avoid any concentration on the contentious issues of the past but rather to look to the future. Thus three papers were invited to address the subject of 'the future of distributed models' (the first three papers published in this special issue), while other contributions of the current state of the art were presented in a poster session to stimulate discussion. The other papers in this issue derive mostly from these poster papers. The symposium finished with a round table discussion of some of the issues that arise in the development and application of physically-based models.
It was clear from the discussion that, even within the small part of the hydrological research community represented at the symposium, there are healthy differences of opinion about the way to proceed in the future. Those differences are reflected in the papers of this issue. The paper by Hegh-Jensen and Mantoglou reviews the problems of distributed models and, in particular, that of defining effective parameter values at the grid scale of distributed models. They suggest that future developments of distributed models must start to take account of the heterogeneous nature of the hydrological system within a stochastic formulation. The paper by Bathurst and O'Connell stresses the importance of gaining experience in the application of distributed models, concentrating on the Systkme Hydrologique Europeen (SHE) model. Beven and Binley tackle the problem of parameter calibration from a different viewpoint, attempting to estimate the predictive uncertainty of distributed models as conditioned on observed data through a computationally intensive Monte Carlo calibration procedure.
Koide and Wheater, in an application of a finite element model of hillslope hydrology to a small plot experiment, have looked at the problems of parameter calibration within a more traditional framework. Their results suggest that the Darcian flow assumptions on which their model is based may not be appropriate to their field plot measurements where some kind of preferential flow may be taking place under some flow conditions. Chappell and Ternan also address the problem of parameter calibration within a Darcian framework, emphasizing the need to take the bulk heterogeneity of the soil in terms of layering and hillslope sequences into account in distributed models. The availability of new techniques in the development of distributed models is represented by the papers by Binley and Beven who discuss some strategies for