Reliable yield of reservoirs and possible effects of climatic change

Statistics of seasonal and daily rainfall in nine regions covering the United Kingdom are presented, together with the best currently available predictions of changes in rainfall and evaporation, based on various general circulation model (GCM) assessments of the climate 40 years ahead. On that basis, runoff sequences are generated for the region of northwest England with north Wales (NWE & NW, humid temperate) and for southeast England (SEE, temperate but less humid), both for the current climate and for the year 2030 estimate, the latter incorporating higher evaporation losses. The future annual runoff is reckoned as 8% less than the current average in the SEE region and 4% less than that in the NWE & NW region. Direct supply reservoir yield versus storage graphs are then derived in each case for a range of duration-based probability of emptiness, their yields (i.e. sustained draw-off rates) and storage capacities being expressed as proportions of current annual runoff. The year 2030 graphs fall systematically below those for the current climate, representing 8% to 15% loss of yield from existing storage in the SEE region and 4% to 25% loss of yield in the NWE & NW region. Débit garanti d'un réservoir et effets éventuels d'un changement climatique Résumé On donne des statistiques sur les précipitations saisonnières et journalières dans neuf régions couvrant le Royaume-Uni ainsi que les meilleures prévisions disponibles sur les changements des précipitations et evaporations estimées à partir de divers modèles GCM pour les 40 années à venir. Sur cette base on produit des séries de débits journaliers pour le nord-ouest de l'Angleterre et le nord du pays de Galles (région humide et tempérée), ainsi que pour le sud-est de l'Angleterre (région tempérée mais moins humide), soit avec le climat Open for discussion until 1 June 1992 579 /. A. Cole et al. 580 actuel, soit avec un climat estimé pour 2030, ce dernier comportant des pertes par evaporation plus élevées. Par conséquent, les débits annuels dans l'avenir sont calculés comme étant de 8% inférieurs aux débits actuels dans le région sud-est de l'Angleterre et de 4% inférieurs dans la région nord-ouest de l'Angleterre et le nord du pays de Galles. Pour les réservoirs d'alimentation directe on présente des graphiques des prélèvements en fonction de la capacité des réservoirs pour une série de durées de vidange, les prélèvements constants dans les réservoirs et leurs capacités étant exprimés en pourcentages du ruissellement annuel actuel. Les graphiques pour l'an 2030 se couchent systématiquement en-dessous de ceux qui caractérisent le climat actuel; en effet ils représentent une perte de prélèvement de 8% à 15% (cas du sud-est de l'Angleterre) ou de 4% à 25% (cas du nord-ouest de l'Angleterre et du nord du pays de Galles) par rapport aux capacités actuelles. AIMS OF THE INVESTIGATION The yield reliability of reservoirs is a much-studied topic. The prime object of a water reservoir is to sustain a highly reliable supply, so that a given draw-off rate may continue despite the natural variability of inflow to the system. All such systems are liable to some risk of shortfall or deficit in supply as the reservoir approaches emptiness or actually empties. It is the task of the hydrologist to estimate such risks and to evaluate what yield can be supplied at very low risk, Le. high reliability. What then might be the effect of climatic change on the yields of reservoirs? To attempt to answer this question in regional terms, one requires data on runoff variability for the area of interest for both the current climate and for plausible changes in climate. Since runoff can be regarded as the residual or effective precipitation, after deduction of evaporative and other losses, work proceeds as follows: (a) investigate the statistics of seasonal and daily rainfall for regions that are spatially homogeneous; (b) quantify the changes in rainfall amounts for each season according to best available forecasts of GCMs; (c) quantify the increased evaporative losses linked to the forecast increase in air temperature; (d) generate runoff sequences for both the current and forecast climates; (e) simulate the operation of direct supply reservoirs of various capacities and for a range of draw-off rates for current and forecast inflow regimes; and (f) assess the effect of climate change upon yield for specified reliability and storage volume and upon required storage volume for specified reliability and yield. The reservoir examples given here relate to southeast England, which is where climatic change is likely to have the most impact upon water resources 581 Reliable yield of reservoirs and effects of climatic change within the United Kingdom, and to a region covering north Wales and northwest England, which will probably suffer a lesser impact. VARIABILITY OF PRECIPITATION IN NINE UK REGIONS In a series of reports commissioned by the Water Research Centre (Wigley & Jones, 1987; Gregory et al, 1991) the Climatic Research Unit (CRU) of the University of East Anglia has studied the spatial and temporal variability of precipitation in nine regions covering the whole United Kingdom for the period 1931 to 1989. Defining homogeneous regions In order to define regions that have been homogeneous in terms of their precipitation, Wigley et al. (1984) examined the principal components (PCs) of annual and seasonal rainfall across the British Isles. Figure 1 shows an example of how the first three components of winter rainfall are distributed. PC 1 largely reflects the north-south gradation. PC 2 identifies well-defined zones, separating north Wales from south Wales; also the grouping of western Scotland and the northern half of Ireland. PCs 2 and 3 separate the East Anglian and northeastern regions from the rest of England. Similar PC analyses for the other seasons enable coherent regions to be formed. The time series of data from a central site in each region is then correlated with all other sites over the British Isles. Maps of the correlation decay with each central site allow the definition of the coherent regions to be refined. Further details of the method are given in Wigley et al. (1984). The nine regions adopted for separate study are shown in the fourth quadrant of Fig. 1. The mean daily rainfall observed at seven dispersed locations in each region served to represent the corresponding regional average rainfall. It has been shown that this seven-gauge mean accounted for some 90% of the daily variability of areal rainfall and 95% of the 30-day rainfall variability in each region (Wigley & Jones, 1987). Statistics for monthly and longer time scales The reports have established for each of the nine regions shown in Fig. 1, the means, standard deviations and skewness of monthly and annual totals, listed in Table 1. Some of the most informative results coming from those studies are in the form of time-plots of a given season's variability over the 59-year span of record. For England-and-Wales, Scotland and Northern Ireland the "national averages" in Figs 2 and 3 present examples for spring (March to May) and summer (June to August): both raw and smoothed data are shown. Spring rainfall in England and Wales in the 1980s J. A. Cole et al. 582 Fig. 1 Winter principal component patterns (PCI, PC2 and PC3) and the nine regions chosen for rainfall statistics. showed an enhanced variability with biennial oscillation. No trend away from the mean was seen, however, and that is also true when looking at annual rainfalls for all nine regions, as shown in Fig. 4. Structure of daily time series The daily rainfall series have a value going beyond that of a mere record of past events. By dissecting their structure, one may proceed to generate longer sequences of statistically similar data for the purpose of evaluating water resource systems. Furthermore, the sensitivity of such systems to possible climatic change can then be tested by suitable changes in the model parameter structure. The most widely used approach to interpreting such series is to couple a renewal process for deciding whether each day is wet or dry with an amount model to choose the amount of rainfall on 583 Reliable yield of reservoirs and effects of climatic change Table 1 Statistical summary of monthly and annual precipitation totals for nine regions of the United Kingdom