Relationships between dynamic response characteristics and physical descriptors of catchments in England and Wales

Abstract

A regionalisation methodology has been applied to catchments in England and Wales enabling estimation of daily flows for any catchment in the region for which physical data and records of rainfall and temperature are available. The rainfall-runoff model IHACRES has been calibrated to 60 catchments to obtain a set of dynamic response characteristics (DRCs) describing the hydrological behaviour within the region. Physical catchment descriptors (PCDs) indexing topography, soil type, climate and land cover were collated and linked to the hydrological model by overlaying catchment boundaries with a geographical information system. Relationships were derived to describe the DRCs in terms of the PCDs so that the model may be used to simulate flow for any catchment in the region, given the driving variables, i.e. rainfall and temperature. In the England and Wales region, rainfall loss parameters have been defined in terms of land use, climate and soil type, whilst hydrograph separation parameters were characterised using topographical and soil variables. The set of DRC–PCD relationships, which were obtained by balancing the dual objectives of hydrological integrity and statistical significance, has been satisfactorily validated on two additional catchments within the region. Analysis of calibration errors was aided by sensitivity tests at one of these catchments in which flow response to variations in DRCs was assessed. Finally, a simple land use scenario demonstrates an application of the methodology in which variation in PCDs may be used to assess the impacts of environmental change.

Introduction

River flow measurements are important data for water resource planning, pollution control, conservation and even recreational use. However, although there are some 1500 gauging stations in Great Britain, data are not always available where a need exists. Given that rainfall data are usually available, rainfall-runoff models provide a technique for the simulation of flows given a set of model parameters. A number of types of models have been developed for this purpose and these may be classified as metric, conceptual and physics-based (Beck, 1991). These three generic model types are expounded by Wheater et al. (1993), who describe metric models as based primarily on observations, seeking to characterise system response from those data alone; conceptual models as seeking to represent all of the component hydrological processes perceived to be of importance in catchment scale input–output relationships; and physics-based models as representing these processes in a more classical mathematical-physics form, through numerical solution of the relevant equations of motion. Whilst application of any of these approaches for modelling a gauged catchment is straightforward, simulation of ungauged river flows demands independent measurement or estimation of the model parameters. Physics-based model parameters may be independently measured in the field, but the models are often data intensive and complex. A hybrid metric-conceptual approach, in which observations are used to test hypotheses about the structure of component hydrological stores, and model parameters may be independently estimated for simulation of flows, is illustrated by Jakeman et al., 1990, Jakeman et al., 1992. A rainfall-runoff model, IHACRES, which calculates component unit hydrographs using an estimate of excess rainfall, is employed to characterise the streamflow regime of individual catchments. Regionalisation by describing these hydrological characteristics in terms of physical descriptors then allows estimation of the unit hydrograph for any catchment in the region. Application of this methodology allows flow series to be constructed and the sensitivity of flow to the hydrological characteristics and to physical descriptors to be investigated. Furthermore, given the future climate and a set of physical descriptors, it is possible to examine the impacts of environmental change for any catchment within the region, giving the methodology application in climate change analysis and land use management.

The estimation of hydrological behaviour using physiographic and climatological catchment attributes has been widely addressed in the literature. Early work used multiple regression to link peak flows to rainfall and topographic factors and to estimate unit hydrograph parameters (Nash, 1960). The multiple regression technique has been widely developed and applied by Heerdegen and Reich (1974)in the USA, Waylen and Woo (1984)in Canada, Ando (1990)in Japan, Mimikou and Gordios (1989)in Greece, Reimers (1990)in Brazil, Nathan and McMahon (1992)in Australia and NERC (1975)in a UK context. Many of these studies aim to estimate event characteristics. Of these, most focus on flood indices such as mean annual flood (Acreman, 1985; Mimikou and Gordios, 1989) or on flood frequency (Burn, 1990; Zrinji and Burn, 1994). Other event and index based studies look at low flows (Nathan and McMahon, 1990, Nathan and McMahon, 1992) and mean annual runoff (Reimers, 1990). A second group of studies estimates unit hydrograph characteristics (Burn and Boorman, 1992; Singh, 1990; NERC, 1975; Heerdegen and Reich, 1974) with a view to reconstruction of flow records. The regression technique has also been used by Nathan and McMahon (1992)and Burn and Boorman (1992)to estimate yield, and by Corderey and Pilgrim (1983)who found a lack of dependence of water loss on physiographic characteristics in humid zones.

In many such studies, catchments are grouped to improve the fit of observed data to the regression model. Originally subregions were delineated geographically (NERC, 1975), before cluster analysis was applied (Tasker, 1982; Acreman and Sinclair, 1986). Burn (1990)appraised the `region of influence' approach, used later by Zrinji and Burn (1994), and various methods for defining homogeneous subregions were discussed by Nathan and McMahon (1990). Alternatives to regression equations for transferring parameters between catchments within a region or subregion have been applied by Vandewiele and Elias (1995), who compared kriging and proximity, and by Burn and Boorman (1992), who explore proximity and regression techniques. In an application by Acreman and Sinclair (1986)no transfer was necessary since the object, the flood frequency distribution, was found to be homogeneous in four out of the five regions.

More recently, focus has been on estimation of water balance model parameters aimed at simulation of continuous records. Bergmann et al. (1990)presented a distributed model describing the interaction between flood hydrographs and basin parameters. Combining loss estimates with modelling in a physically-based stochastic monthly water balance model, Vandewiele and Elias (1995)simulated monthly time series. A next step is the estimation of daily hydrological model parameters, a task begun by Sefton et al. (1995)and Post and Jakeman (1996)with the potential for reconstruction of daily flow records. The application presented is therefore a harsh test for the methodology; the estimation of daily flows for ungauged catchments within a large region, England and Wales, without division into subregions.

Discussion of uncertainties in this type of procedure has covered those which arise in the inputs (Eagleson and Qinliang, 1987; Xu and Vandewiele, 1994) and those in the estimation of parameters (Kuczera, 1983); the latter were demonstrated by Gan et al. (1990)in the impact on catchment storage. In this study, the effect of parameter uncertainty on streamflow is investigated and used to evaluate the flow record reconstruction and to interpret a simple land-use scenario.