Estimation of renewable water resources in the European Union

Size: px
Start display at page:

Download "Estimation of renewable water resources in the European Union"

Transcription

1 FRIEND '97 Regional Hydrology: Concepts and Models for Sustainable Water Resource Management (Proceedings of the Postojna, Slovenia, Conference, September-October 1997). IAHS Publ. no. 246, Estimation of renewable water resources in the European Union H. G. REES, K. M. CROKER, N. S. REYNARD & A. GUSTARD Institute of Hydrology, Wallingford, Oxfordshire OX10 8BB, UK Abstract The over-exploitation of freshwater resources and the associated impact on commerce and the environment is a major pan-european problem. Freshwater is a finite and fragile resource, yet increases in demand by industrial, agricultural, domestic and municipal consumers have induced great stresses in the hydrological system. To prevent the degradation of freshwater resources, it is essential that techniques are developed to accurately estimate their availability across Europe. No consistent method is used to estimate the availability of water or to compare resources between different countries and regions. This paper presents a standard method to estimate the availability of water resources in Europe. The method is applied in a grid based model where the availability of water within each cell is derived from combining observed river flow data with methods for estimating flows at ungauged sites. The benefits of a grid approach is illustrated by comparing water availability with water demand to derive indicators of water stress. INTRODUCTION An accurate assessment of the renewable freshwater resource is essential for effective and sustainable management. Estimating the renewable freshwater resource available is, however, difficult due to the temporal and spatial variability of the hydrological regime and the complex effect of human interventions. This paper presents results of a study, commissioned by the Statistical Office of the European Communities (Eurostat), which investigated appropriate methods for estimating the renewable water resource across Europe. The study was also required to illustrate the regional variability in the renewable water resource over the European Union. To meet these objectives, gridded maps of average annual runoff were developed at a 10 km by 10 km resolution for the whole territory. For many of the grid cells, it was possible to derive estimates of the renewable water resource by using the observed gauged daily flow data and digitized catchment boundaries held on the FRIEND European Water Archive. To estimate the average annual runoff for the ungauged portion of the grid (i.e. those areas not covered by the FRIEND dataset) a critical assessment of four different estimation methods was undertaken. Each method used the same baseline climatological dataset provided by the Climate Research Unit (CRU) of the University of East Anglia. The methods considered were: regional characterization; the Budyko freshwater balance method; the Turc-Pike freshwater balance method; and the Probability Distributed Model, a conceptual model developed at the Institute of Hydrology. Grids of average annual runoff were derived from each method. By

2 32 H. G. Rees et al. comparing the modelled runoff with observed data, it was possible to assess the performance of the methods and ascertain which was most appropriate to apply at the European scale. A composite grid was then developed combining the observed runoff grid with the best of the modelled grids. As well as illustrating the spatial variability of the renewable water resource, the paper will then show how grids can be used to derive indicators of water stress with respect to both agricultural and urban demand. INPUT DATA The pan-european datasets used in the derivation of the grids of renewable water resource included: (a) FRIEND European Water Archive, containing river flow data for over 3500 gauging stations and 2500 digital catchment boundaries; (b) the CEC Soils Map (CEC, 1985), which provides a consistent classification of the soils in 12 countries of the European Union; (c) baseline climatological data, developed by the Climate Research Unit (Hulme et al, 1995), which provides a mean monthly climatology for the period, covering the European Union and beyond, at a resolution of 0.5 latitude by 0.5 longitude. Nine variables are available: minimum, maximum and mean temperature, precipitation, sunshine hours, vapour pressure, wind speed, frost days and rain days. For this study, only the mean temperature and precipitation were used directly, but the sunshine hours, vapour pressure and wind speed were used to calculate the potential evaporation (PE) according to the Penman equation (Penman, 1948). To represent these variables on the 10 km grid, a simple linear interpolation function was used. GRID DEVELOPMENT FOR GAUGED CATCHMENTS The grid development described in this paper was based on data from the FRIEND European Water Archive. All catchments on the archive with both a gauged river flow record and a digitized catchment boundary were included in the analysis. The renewable water resource of each gauged catchment is calculated using all available flow data. Once the renewable resource of each gauged catchment has been determined, the 10 km grid is overlain on to a map of digitized catchment boundaries enabling those cells for which gauged data is available to be identified. The renewable resource for each gauged grid cell is then estimated using the same weighted area technique described by Arnell in the 1993 FRIEND report (Gustard, 1993). GRID DEVELOPMENT FOR UNGAUGED CATCHMENTS The hydrological regime across Europe shows a high degree of spatial and temporal variability. This makes it difficult to ascribe a single method for estimating runoff at an ungauged site. A key requirement of the project was, however, to develop one

3 Estimation of renewable water resources in the European Union 33 consistent method applicable to the whole of Europe. Four different approaches to estimating runoff for the ungauged grid cells were tested: (a) The Probability Distributed Model (PDM), a conceptual water balance approach to rainfall-runoff modelling based on a soil moisture accounting procedure (Moore, 1985). The model provides estimates of the monthly runoff (in mm) generated within each 10 km grid cell independently. There is no routing of runoff through the river network, or from one cell to another. (b) A freshwater balance approach using an empirical formula proposed by Budyko (1961): Runoff = AAR.exp (-PE/AAR) (1) where AAR is the average annual rainfall and PE is the potential evaporation. (c) A second freshwater balance approach first developed by Turc (1954) and modified by Pike (1964) where the actual evaporation (AE) is written as: AE = AAR/(1 + (AAR/PE) 2 )' 72 (2) and Runoff = AAR - AE (3) (d) A regional characterization approach based on multivariate regression techniques (Gustard et al., 1992) in which the actual evaporation is given by the equation: AE = r-pe (4) The value of the conversion factor r increases with rainfall to reflect the proportion of water available as any soil moisture deficit is replenished. When the total annual rainfall reaches a certain threshold, the soil is assumed to be saturated and that the value of r is 1. For the UK, r can be expressed as (Gustard etal, 1992): r = SAAR for AAR < 850 mm r = 1.0 for AAR > 850 mm (5b) By comparison, the equation for the whole of Europe was calculated as: r = SAAR for AAR < 3900 mm r = 1.0 for AAR > 3900 mm (6b) (5a) (6a) VALIDATION OF RESULTS FOR UNGAUGED CATCHMENTS In order to compare the performance of the four models a statistical comparison of observed and modelled runoff was undertaken for each gauged grid cell. The relationship between the modelled and observed data was expressed in terms of a bias, given by: bias = (modelled runoff / gauged runoff)-100% (7) The observed runoff and modelled runoff were calculated for 1257 gauging stations across Europe and the bias calculated. Overall, the mean bias of the PDM model was

4 34 H. G. Rees et al. Table 1 Comparison of bias for countries in Europe. Country All Belgium Switzerland Germany Denmark Spain France Italy Ireland Norway Netherlands Sweden Finland UK No stations Mean percentage PDM bias: Budyko Turc-Pike Regression approximately 78%, which suggested that the model was consistently underestimating the runoff. This compared with a mean bias of 101% using the Budyko equation, 86% using the Turc-Pike equation and 119% using the regression based approach. The mean bias for each country is given in Table 1. The performance of each model was also considered using regression analysis to obtain statistics for the factorial standard error of each model. The relationship between the modelled runoff (RO m ) and the observed runoff (R0 0 ) is considered to be in the form: RO m = a-ro 0 " (8) Therefore, linear regression analysis is undertaken on the equation in the form: log RO m = log a + b. log R0 0 (9) where log a is the intercept and b is the parameter estimate. If the model (PDM, Budyko, etc.) had predicted the flows exactly, the parameters a and b, from the above regression equation, would have equalled 1. In the absence of ideal (predicted) data, these parameter estimates should be as close to one as possible, while maximizing the fit of the model, represented by the R 2 (percentage of variance explained) value, and reducing the factorial standard error. From Table 2, it can be seen that, for all models, the variance and factorial standard errors are very similar. Using this information with that for the mean bias, the Budyko method would seem the most effective method for estimating the runoff in ungauged catchments. As Table 1 shows, the four models considered all demonstrate variations in the accuracy with which the annual runoff can be predicted across Europe and each model performs well in some regions, but not so well in others. Numerous other studies have been undertaken to identify the most appropriate method for deriving runoff from rainfall and evaporation across Europe. Arnell et al. (1990), compared model sensitivity to changes in rainfall and evaporation and concluded that the Turc- Pike model approximates runoff reasonably well (Budyko was not considered in Arnell's study). In Spain, Estrela et al. (1995), looked at the relative performance of

5 Estimation of renewable water resources in the European Union 35 Table 2 Regression analysis. Regression statistics Parameter estimate (b) Intercept (log a) a R 2 Factorial standard error PDM Budyko Turc-Pike Regression the Budyko and Turc-Pike models and found Budyko to be more consistent with observed data, although Turc-Pike was better for low evaporation conditions. These findings would suggest that the use of a simple empirical model over Europe can indeed be justified. COMPOSITE MAP OF RENEWABLE WATER RESOURCE Figure 1 shows a composite map of renewable water resource, expressed in terms of the average annual runoff. The gridded map was generated from flow records, in gauged areas, and by applying the Budyko equation in the ungauged areas and includes estimates of runoff for all major drainage basins affecting the European Union. WATER STRESS INDICATORS In developing a grid based model of the renewable water resource, water demand information can be readily superimposed to derive grids (or maps) of water stress. Such grids were developed for the two most significant types of water demand: agriculture and urban. KEY : Runoff 0-100mm mm mm mm M mm mm > 1000mm Fig. 1 Composite map of average annual runoff.

6 36 H. G. Rees et al. KEY 0-50% 51-80% 81-90% M % >100% Agricultural water demand Fig. 2 Irrigation demand as a proportion of average annual runoff. According to the World Resources Institute (WRI, 1990), agricultural (irrigation) demand across Europe amounts to approximately 110 km 3 year" 1. In several areas, particularly in southern Europe, the demand for water is fast approaching the limits of the resource. Grids of agricultural demand were developed using the soil use information available within the CEC Soils Map. For each relevant soil use type, typical figures for the crop water requirement were assumed according to FAO guidelines (FAO, 1977). With irrigation assumed to be confined to the summer months from April to September, the net irrigation requirement was calculated by subtracting crop water requirement, from estimates of summer runoff derived from the PDM. It was further assumed that only 20% of each irrigable grid cell is irrigated using the sprinkler irrigation method at 75% efficiency. The resulting grid, representing irrigation demand as a proportion of the average annual runoff, is shown in Fig. 2. Despite the assumptions made, the grid reaffirms the existence of problems of water stress brought about by agriculture in certain areas of southern Europe. Urban water demand For Europe as a whole, urban water demand, which includes industrial and domestic use, accounts for over two thirds (67%) of freshwater abstractions (WRI, 1990). To derive grids of urban water stress, the Eurostat Degree of Urbanization data coverage was used. This presents three classes of urban density: thinly (< 100 heads of population per km 2 ); intermediate ( heads of population per km 2 ); and (> 500 heads of population per km 2 ). A daily per capita consumption rate was assumed for each urban class on the basis of data presented in the Dobris Assessment (EEA, 1995). The resulting urban water demand grid can be directly compared with

7 Estimation of renewable water resources in the European Union 37 KEY 0-50% 51-80% % H % M >ioo% Fig. 3 Summer urban water demand as a proportion of summer runoff. the grids of renewable water resource. Figure 3 compares an increased summer demand (120% of average demand) with summer runoff. It clearly shows there is an important need for the effective management of freshwater resources and for the provision of artificial storage or transfer facilities, especially in southern Europe and the densely populated regions of the north. CONCLUSIONS With the ever growing demands for freshwater, it is vitally important that reliable methods are used to assess the availability of the resource. However, few improvements can be made without better and more reliable data. The four methods considered all used the baseline climatology for Europe supplied by the Climate Research Unit (CRU). There is a recognized tendency for the CRU data to underestimate precipitation, especially in regions affected by snowmelt, while the potential evaporation, calculated according to the Penman equation, is generally over-estimated. Despite offering the best data currently available at a pan-european scale, there is general agreement that, with better station coverage and harmonized definitions of the variables, significant improvements could be made to the CRU dataset and hence estimation of the renewable water resource. All methods demonstrated variations in the accuracy of prediction, with each performing well in some countries and not so well in others. This problem occurs as a result of using a single method for the whole of Europe, ignoring any regional discrepancies in climate and flow regime. It should be noted that the Budyko method, which was developed to represent just the annual freshwater balance only, is considered unsuitable for estimating the renewable water resource at any finer temporal resolution. With its ability to run at a daily or monthly time step, the PDM model demonstrates a versatility which is ideal for deriving estimates of the resource

8 38 H. G. Rees et al. at varying time intervals. Previous studies (Arnell & Reynard, 1996) have shown the PDM to be conceptually sound and, therefore, further work in developing the model on a pan-european basis should be considered. A major advantage of a grid based approach is the ability to combine and compare, on a cell by cell basis, the renewable water resource with other spatially referenced data. The lack of detailed data on the extent of human activity posed considerable difficulties to this particular exercise with the demand estimates derived on the basis of some very broad assumptions. Despite these problems, the project succeeded to illustrate those area prone to water stress. While good progress has been made in the course of the study, there still remains considerable scope for further work. In considering the long-term renewable water resource, no attempt has been made within the project to account for the inter-annual availability of the resource or the complex, and ever changing, influences of man. To protect freshwater resources from unsustainable exploitation, it is important that such issues are considered. With freshwater becoming an increasingly precious resource, methods of estimating the renewable water resource must continue to develop and improve. Acknowledgements The research presented in this paper was undertaken as part of the Eurostat SUP.COM 95 project, "Estimation of Renewable Water Resources in the European Union", funded by the European Commission through the Fourth Framework Programme. The authors would like to thank both Eurostat, for allowing the results of the project to be published in this paper, and also the Climate Research Unit of the University of East Anglia, who supplied the baseline climatological data. Various other spatial data sets were kindly provided by the Eurostat GISCO service. REFERENCES Arnell, N. W, Brown, R. P. C. & Reynard, N. S. (1990) Impact of climatic variability and change on river flow regimes in the UK. Inst. Hydro!., Wallingford, Report no Arnell, N. W. & Reynard, N. S. (1996) The effects of climate change due to global warming on river flows in Great Britain. /. Hydrol. 183, Budyko, M. I. & Zubenok, L. I. (1961) The determination of evaporation from the land surface. Izv. Akad, Nauk SSSR, Ser. Geogr. no. 6, CEC (1985) Soil Map of the European Communities. 1: CEC-DGVI., Luxembourg. Estrela, T., Ferrer, M. & Ardiles, L. (1995) Estimation of precipitation-runoff regional laws and runoff maps in Spain using a geographical information system. Proc. FRIEND-AMHY Conference (Thessalonika). European Environment Agency (1995) Europe's Environment The Dobris Assessment (ed. by D. Stanners & P. Bourdeau) FAO (1977) Guide for Predicting Crop Water Requirements. FAO Irrigation and Drainage Pap. no. 24C, Revised Gustard, A., Bullock. A. & Dixon. J. M. (1992) Low flow estimation in the United Kingdom. Inst. Hydrol., Wallingford, Report no Gustard, A. (ed.) (1993) Flow Regimes from International Experimental and Network Data (FRIEND). Vol. I: Hydrological Studies. Institute of Hydrology, Wallingford, UK. Hulme, M., Conway, D., Jones, P. D., Jiang, T., Barrow, E. & Turney, C. (1995) Construction of a European climatology for climate change modelling and impact implications. Int. J. Climatol. 15, Moore, R. J. (1985) The probability-distributed principle and runoff production at point and basin scales. Hydrol. Sci. J. 30(2), Penman, H. L. (1948) Evaporation in nature. Rep. Progr. Phys. XI, Pike, J. G. (1964) Estimation of annual run-off from meteorological data in tropical climate. J. Hydrol. 2(2), Turc, L. (1954) Le bilan d'eau des sols, relation entre les précipitations, l'évaporation, et l'écoulement. Ann. Agron. 5, World Resources Institute (WRI) (1990) World Resources