H. S. Wheater, V. S. Isham, C. Onof, R. E. Chandler, P. J. Northrop, P. Guiblin, S. M. Bate, D. R. Cox, and D. Koutsoyiannis, Generation of spatially-consistent rainfall fields for rainfall-runoff modelling, 7th National Hydrology Symposium of the British Hydrological Society, Newcastle, doi:10.13140/RG.2.1.4315.4163, British Hydrological Society, University of Newcastle, 2000.
In the last decade, radar data have become routinely available for the UK, providing a means of observing spatial rainfall. Although radar has limitations with respect to performance and long records of continuous data are not yet available, it represents an important source of information which allows, for the first time, the continuous spatial distribution of rainfall to be studied. In parallel, new research into spatial rainfall modelling has produced a range of tools with potential for hydrological application. However, most methods of representing rainfall for hydrological design and simulation are relatively primitive. There is a need, therefore, to combine the strengths of new data sources and new modelling methods to produce a new generation of rainfall modelling tools to support hydrological practice. In a major study for the Ministry of Agriculture, Fisheries and Food, a comprehensive suite of rainfall modelling tools has been developed, with wide applicability to provide inputs to distributed or lumped rainfall-runoff models based on radar or raingauge data. These tools include spatial-temporal models, generalised linear models and hybrid modelling approaches. In spatial-temporal models, rainfall is modelled in continuous space and time and hence can be aggregated to any required spatial or temporal scale. The generalized linear model approach represents point rainfall at a number of locations by what is essentially an extension of a multiple regression approach. In this way, any important explanatory variables can be included (for example elevation, rainshadow effects, distance from the sea) as well as temporal dependence (e.g. previous rainfall). The model is thus extremely flexible, and can incorporate spatial non-stationarity as well as long-term climate effects. The hybrid approach, which has been developed for situations of limited spatial data, uses the concept of spatial-temporal disaggregation.
Full text (67 KB)
See also: http://dx.doi.org/10.13140/RG.2.1.4315.4163