living with uncertainty: climate change, river flows and water resource management in scotland

The Science of the Total Environment 294(2002) 29–40

0048-9697/02/$ - see front matter� 2002 Elsevier Science B.V. All rights reserved.PII: S0048-9697Ž02.00050-5

Living with uncertainty: climate change, river flows and waterresource management in Scotland

Alan Werritty*

Department of Geography, University of Dundee, Dundee DD1 4HN, UK

Accepted 15 October 2001

Abstract

The recent increased variability of Scotland’s hydroclimate poses major problems for water resource managerscharged with making informed investment decisions given the likely impact of future climate change. Two strategiesare developed in this paper to assist managers faced with this environmental uncertainty. The first involves trendanalysis of precipitation and runoff since the 1960s and 1970s viewed against longer-term variability reported frominstrumental records. The second strategy is based upon current climate change scenarios coupled with GCMs, anddownscaling of precipitation and temperature to provide inputs to rainfall-runoff models. The long-term records ofprecipitation(back to the 1860s) and runoff(back to the 1930s) reveal the late 1980s and early 1990s as the wettestperiod on record for the west but not for the east. Over the period 1961–1996 the precipitation gradient has intensifiedacross Scotland: wetter west; relatively dry east. Changes in runoff over the period 1970–1996 are also reported withincreases in annual flows at 33 out of 38 stations(significantly at 12 stations) and decreases in low flows at 21 outof 38 stations(significantly at one station). The bulk of these flow increases occurred in the south and west especiallyin the autumn and spring. In terms of high flows over the period 1970–1996, four out of 44 stations reported achange in magnitude and 15 reported an increase in the frequency of POT events. In terms of future climate change,Hulme and Jenkins(1998) predict annual precipitation increases of 6–16%(Scotland) and 6–14%(Scottish Borders)from the 2020s to the 2080s based on the Hadley Centre model(HadCM2) medium–high scenario. Seasonal changesare concentrated in the autumn(SON) and winter(DJF) with increases as high as 24 and 29% for the autumn bythe 2080s.(Arnell NW, et al. Institute of Hydrology Report No. 107, Wallingford, 1996), using an earlier transient,

Hadley experiment(IS92a), predict a 5–15% increase in annual runoff across Scotland by the 2050s, locally risingto 25%. Simulation flow duration curves for the 2050s generateQ values up by 5% or less(Rivers Don, Almond95

and Nith) andQ up by 10–24%(Rivers Don, Almond, Nith and Lyne Water). In terms of water resource planning,5

these predicted changes should be regarded as first order approximations, as they take no account of natural climaticvariability, and could generate different absolute values if other scenarios were used. The predictions are, however,broadly consistent with trends in precipitation and runoff for Scotland since the 1970s. Major issues of concern towater resource managers are identified and commented upon in the light of these predictions.� 2002 ElsevierScience B.V. All rights reserved.

Keywords: Climate change; River flows; Water resources; Scotland

*Corresponding author. Tel.:q44-1382-345084; fax:q44-1382-344434.E-mail address: [email protected](A. Werritty).

30 A. Werritty / The Science of the Total Environment 294 (2002) 29–40

1. Introduction

There is mounting evidence that the globalclimate is changing in response to human activityand that the increase in temperature recordedduring the 20th century cannot be attributed entire-ly to natural causes(Intergovernmental Panel onClimate Change, 1996a). Although global warm-ing is the most obvious manifestation, regionalincreasesydecreases in precipitation and stormactivity also provide evidence of climate change.If sustained over the next few decades, it is clearthat global changes in temperature and precipita-tion will have a significant impact on the medium-to long-term management of water resources(Intergovernmental Panel on Climate Change,1996b). But water resource planners, who arerequired to include the likely impacts of climatechange on current operating procedures futurepolicy and capital investment plans, do so againsta background of considerable uncertainty(Arnellet al., 1994; Jowitt and Hay-Smith, in press).Much of this uncertainty arises from the inherentnature of the global climate as a physical system.How then can water resource planners live withsuch uncertainty and make informed decisionsconcerning the impacts of future climate change?Over the last decade two main strategies have

evolved in assessing the hydrological effects ofclimate change. The first is to identify past hydro-climatic variability over appropriate timescales byexamining changes in precipitation and runoff.This strategy is crucially dependent on the availa-bility and quality of appropriate long-term records.Careful analysis of such long-term records, coupledwith trend analysis of the recent past, can providea basis for future water resource planning over theshort-term(e.g. Hisdal et al., 1995). The successof this strategy depends on the identification ofappropriate historical analogues to match recentshort-term trends. The second strategy, which hasdeveloped rapidly over the last decade, involvesthe development of climatic scenarios(based onestimates of future trends in global population,economic and technological developments and theresulting behaviour of the climatic system) coupledwith Global Circulation Models(GCMs) andappropriate downscaling of outputs of temperature

and precipitation at the regional or catchment scale.Appropriate rainfall-runoff models then provideestimates of changes in runoff under the specifiedclimatic scenarios(Arnell, 1996; Reynard et al.,1998; Kilsby et al., 1998). This provides the basisfor medium- to long-term water resource planning,but requires constant updating as scenarios arerefined, GCMs improved and downscaling proce-dures revised. The successful implementation ofthis second approach is crucially dependent onbetter specified scenarios; increasingly accurateand reliable GCM outputs(especially for precipi-tation); improved procedures for downscaling tomedium-sized catchments; and the calibration ofappropriate rainfall-runoff models. Each of thesesteps involves considerable uncertainty making thefinal estimates of future runoff and evapotranspir-ation liable to substantial errors.Scotland’s hydroclimates have become increas-

ingly divergent from the rest of the UK over thelast 2 decades. In particular, the precipitationgradient across the UK has been accentuated withthe north-west becoming wetter(notably in thewinter) and the south-east drier, especially in thesummer half of the year(Hulme and Jenkins,1998). Similarly, the results from downscalingprecipitation and runoff from GCMs for the 2050sconfirm a strengthened hydroclimatic divide acrossthe UK with Scotland and Southern Englandreporting increasingly divergent trends(Arnell,1996). Indeed, there is strong evidence that thechanging hydroclimate of Scotland is now moreakin to that of southern Norway than that of SEEngland(Roald, 1998).There have been many reports on the changing

hydroclimate of Scotland over the last few years(see Werritty and Foster, 1998 for a summary) butmost have been either narrowly targeted in termsof the time period involved(Mayes, 1995); highlyaggregated in the spatial scale addressed(Smith,1995); very localised in coverage(Curran andRobertson, 1991); or only concerned with changesup to the late 1980s(Arnell et al., 1990). Hitherto,the long-term and short-term trends in precipitationand runoff across the whole of Scotland have notbeen systematically explored.The major aim of this paper is to examine recent

31A. Werritty / The Science of the Total Environment 294 (2002) 29–40

Fig. 1. Location of rain gauges(1892–1996 and 1961–1996).

trends in precipitation and runoff across Scotlandand, by placing them within a longer historicalperspective and comparing them with future pre-dictions (based on climate scenarios and GCMs),assist water resource managers faced with futureclimatic uncertainty over the short- and medium-term. Having explained the databases assembledfor this paper and having outlined the methods ofanalysis, the major findings in terms of long- andshort-term precipitation and runoff trends are pre-sented. Predictions of future precipitation and run-off trends based on current climatic scenarios andoutputs from GCMs up to the 2080s are thensummarised. The paper concludes by assessing thelikely impacts of recent and predicted changes inprecipitation and runoff on Scotland’s waterresources.

2. Data sources and methods of analysis

In order to identify recent trends in precipitationand runoff and place them within a longer histor-ical perspective, a number of databases have beenassembled from existing records held by the UKMeteorological Office and the Scottish Environ-ment Protection Agency. Care was taken in assem-bling these databases to ensure maximum spatialcoverage across Scotland whilst recognising thatcompilation of complete records of daily precipi-tation back into the mid-19th century would provecostly and time consuming. The locations of the13 raingauges selected(including one in north-east England) are given in Fig. 1 and the start andend dates of the four daily and 13 monthly recordsare itemised in Tables 1–3.In terms of long-term runoff records, Scotland

is less well served because river gauging com-menced much later than in other parts of the UK(Werritty, 1987; Marsh and Anderson, in press).Three rivers have flow data which extend back to1930, but the majority of records only date backto the 1960s and 1970s. Two main databases forthe period 1970–1996 have been assembled:

2.1. Daily mean flows

Flow records from 38 river gauging stationswere selected which reflected, as far as possible,

the natural climatic signal and for which the impactof artificial abstractions(surface or groundwater),afforestation or hydro-electric development wasminimal. An attempt was made to choose sites aswidely distributed as possible across Scotland, butthis was constrained by the need to have records,which extended back to 1970. These data weremade available from the Surface Water Archive atthe Centre for Ecology and Hydrology fromrecords supplied by the Scottish Environment Pro-tection Agency and its predecessor bodies.

2.2. Peaks Over Threshold (POT) series

The POT flood series for Scotland was initiallydeveloped for the Flood Studies Report(NaturalEnvironment Research Council, 1975). In thisinvestigation, Black’s threshold of 45 events dur-ing the 10-year period 1979–1988(Black, 1992)was used to extend each POT series up to the endof 1996 using data supplied by the Scottish Envi-ronment Protection Agency. Only stations withcontinuous records from 1970 which also displayed


Table 1Spearman’s rank correlation coefficients between annual precipitation and years for raingauges across Scotland(1861–1996 and1961–1996) and the whole of Scotland(1861–1992)

Raingauge 1861–1996 Annual Spring Summer Summer Winter

Braemar Criticalr (Ps0.05) y0.0034 0.0679 y0.2766a y0.0775 0.16500.1700

Edinburgh 0.0706 0.0746 y0.1246 0.0768 0.0968Fort William 0.2217a 0.1593 y0.0340 0.3132a 0.0127Inverness 0.0868 0.0838 y0.0506 0.0329 0.1120Islay 0.2136a 0.1713a 0.0178 0.1897a 0.1179Lilburn y0.1605 y0.1153 y0.1231 y0.1986a 0.1038Montrose y0.1656 0.0021 y0.2323a y0.1325 0.0594Portree 0.1779a 0.1586 y0.0390 0.2308a 0.0446Stranraer y0.1083 y0.0439 y0.1609 0.0632 y0.0052Waulk 0.1747a 0.1999a y0.0526 0.2354a y0.0249Wick 0.1984a 0.1157 0.0309 0.1151 0.1751a

Scotland 1861–1992 0.1731a 0.1234 y0.0468 0.1630 0.0692Critical r (Ps0.05)0.1726

Raingauge 1961–1996 Annual Spring Summer Autumn Winter

Braemar Criticalr (Ps0.05) 0.2677 0.1559 y0.1913 0.0710 0.30370.3412

Edinburgh 0.2565 0.1356 y0.1498 0.1336 0.4855a

Fort William 0.4399a 0.1685 y0.0501 0.2446 0.4082a

Inverness 0.2342 0.2873 y0.2736 y0.0286 0.3853a

Islay 0.4129a 0.2692 0.1488 0.0955 0.3511a

Lilburn y0.3089 y0.1115 y0.4438a y0.0889 y0.0183Montrose 0.1040 0.0913 y0.2730 0.1377 0.1719Portree 0.2484 0.1290 y0.1413 0.1346 0.2579Stranraer 0.0573 0.1830 y0.1725 y0.1336 0.1737Waulk 0.3524a 0.1761 y0.2317 0.0470 0.3544a

Wick 0.1490 0.3308 y0.2036 0.1810 y0.0700Dumfries y0.0296 0.0956 y0.2178 y0.0499 0.3725a

Stownoway 0.3918a 0.2497 0.0909 0.2033 0.3305

Significant at the 5% level.a

flow regimes not significantly affected by hydro-electric schemes were eligible for inclusion. Only44 stations met these criteria for the period 1970–1996(Fig. 7).A number of complementary methods of anal-

ysis have been used to generate the results reportedbelow. Trend analysis was undertaken using bothSpearman’s rank correlation coefficient and theIRW smooth algorithm developed by Young(1993). This algorithm has been widely adoptedin analysing econometric and environmental serieson account of its robustness and versatility indepicting temporal trends in very noisy data. Inundertaking correlation analysis, a significance

level of Ps0.05 (or 5%) is used to identifystatistically significant results.

3. Changes in long-term and short-termprecipitation

Correlation coefficients have been calculated forboth long-term (1861–1996) and short-term(1961–1996) annual precipitation series. Raingauges in the north and west(Fort William, Islay,Portree, Waulk and Wick) register significantincreases in annual precipitation for the long-termperiod, whereas rain gauges in the east and south(Montrose and Lilburn) register strong but non-


Table 2Spearman’s rank correlation coefficients between annual maxima for 1-, 2-, 7- and 30-day precipitation and years for Braemar,Edinburgh, Stornoway and Dumfries

Braemar Edinburgh Stornoway Dumfries

Period 1872–1996 1896–1996 1874–1996 1871–1996Critical r (Ps0.05) 0.1775 0.1980 0.1789 0.1767

1-Day maximum y0.2586a 0.1289 y0.0559 0.05652-Day maximum y0.1816a 0.0785 y0.1384 0.02837-Day maximum y0.1387 0.1114 y0.0497 0.003130-Day maximum y0.1530 0.1008 y0.0378 0.1562

1961–1996 Criticalrs0.3412(Ps0.05)1-Day maximum 0.0644 0.2418 0.4367a y0.06372-Day maximum 0.1272 0.2098 0.4425a 0.03807-Day maximum 0.2059 0.0810 0.2409 y0.045630-Day maximum 0.2281 0.0842 0.4157a 0.2329


Table 3Rank correlation coefficients between years and frequency of periods with no precipitation

Braemar Edinburgh Stornoway Dumfries

Start 1872 1896 1874 1871Finish 1996 1996 1996 1996

Critical r (Ps0.05) 0.1775 0.1980 0.1789 0.17671-Day periods y0.5242a y0.0187 y0.1363 y0.04282-Day periods y0.4304a y0.1367 y0.1355 y0.05337-Day periods y0.3503a 0.2088a y0.0241 0.0272

1961–1996 Critical rs0.3412(Ps0.05)1-Day periods y0.7055a y0.0983 y0.4306a y0.21722-Day periods y0.5934a y0.2582 y0.4495a y0.26637-Day periods y0.2961 0.0804 y0.2201 y0.0606

1970–1996 Critical rs0.4000(Ps0.05)1-Day periods y0.5424a 0.1210 y0.3955 0.01882-Day periods y0.5195a y0.0635 y0.2618 y0.24197-Day periods y0.2683 0.1397 y0.2633 0.0229


significant decreases over the same period(Fig. 1,Table 1). A similar west–east contrast is found inthe seasonal trends with the west reporting signif-icant spring and autumn increases(Islay, Waulk),the north significant winter increases(Wick), andthe east significant summer decreases(Braemar,Montrose). For Scotland as a whole, the period1861–1992 just registers a significant increase atthe 5% level. During the short-term period(1961–1996) 10 sites register an increase in annualprecipitation, significantly so for sites in the westand north(Fort William, Islay, Waulk and Stor-

noway: Fig. 1, Table 1). Seasonally significantwinter increases are recorded across much of Scot-land (Edinburgh, Fort William, Inverness, Islay,Waulk and Dumfries). In the summer, most sitesregister a decrease, especially in the east, althoughonly that at Lilburn is significant. Graphical trendsfor Fort William (representing the west) and Brae-mar (representing the east) using the IRW smoothalgorithm (Young, 1993) are shown in Fig. 2 forthe long-term period. Fort William reports anincrease of approximately 600 mm in annual pre-cipitation since the early 1970s, whereas over the


Fig. 2. Comparison between the annual precipitation trends(using IRW smooth algorithm) for Fort William (west coast)and Braemar(eastern Scotland) 1892–1996.

same period, Braemar registers a slight rise fol-lowed by a recent decline. Most of the FortWilliam increase occurred in the winter and spring,whereas Braemar’s recent decline is mainly attrib-utable to a reduction in summer precipitation. Incomparing these results with earlier ones(to 1994)reported in Foster et al.(1997), it is noteworthythat the apparent inexorable rise in precipitation atFort William (to values never experienced duringthe instrumental record) now seems to have endedexcept in the spring. More generally, these resultsconfirm the findings of Mayes(1995) and Smith(1995) based on records to the early 1990s. Asignificant increase in long-term precipitation inthe north and west of Scotland has been coupledwith a significant decrease in the south and east.This pattern has intensified in the recent past(1961–1996) with the northwest becoming verymuch wetter. Over that period, winters havebecome wetter and summers drier, especially inthe east.Correlation analyses have been undertaken on

the four daily precipitation records to detect chang-es in the annual maxima of 1-, 2-, 7- and 30-dayprecipitation(Table 2). Over the long-term period,

Braemar and Stornoway report decreases for allfour precipitation maxima(significantly so for 1-and 2-day maxima in Braemar), whereas Edin-burgh and Dumfries reported increases(none beingsignificant). By contrast over the short-term peri-od, all stations except Dumfries record increasesin precipitation maxima(significantly so for Stor-noway for 1-, 2- and 30-day maxima). Turning tothe opposite precipitation extreme, a similar anal-ysis using the four daily records was undertakenbased on the frequency of 1-, 2- and 7-day periodseach year with no precipitation(Table 3). Duringthe long-term period, there is a general decreasein the frequency of dry spells(significantly so for1-, 2- and 7-day periods in Braemar) althoughEdinburgh reports a significant increase in thefrequency of 7-day dry spells. Over the short-termperiod all stations(except Edinburgh) report strongdecreases in the frequency of dry spells(especiallyBraemar and Stornoway for 1- and 2-day periods).Summarising the changes in extremes over theshort-term period, Stornoway has clearly experi-enced an increase in precipitation extremes(withhigher maxima and fewer dry spells), whilst Brae-mar shows no significant changes in annual pre-cipitation maxima but a strong decrease in shortduration dry spells(i.e. more wet days over theyear). These findings are generally consistent withthe earlier conclusion that the northwest to south-east precipitation gradient across Scotland hasbecome more accentuated since 1961.

4. Annual and seasonal changes in river flows(1970–1996)

Annual and seasonal changes in runoff arenow analysed in terms of low, mean and highflows derived from Daily Mean Flow and Peaksover Threshold(POT) databases for each year1970–1996. Of the 38 river gauging stations forwhich mean and low flows could be derived, allexcept three report a positive correlation(i.e. anincrease in mean flows) and in 12 cases thecorrelation is significant at the 5% level. Themajority of stations exhibiting significant increasesin mean flows are located in the south with nonelocated in the north-east(Fig. 3). Each recordwhich registered a significant increase in mean


Fig. 4. Trends in mean annual flows for all river flow stationswhich registered a significant increase 1970–1996. Each recordhas been scaled by its 1970–1996 average and filtered usingthe IRW smooth algorithm. Occasional breaks within individ-ual records occur where the smoothing algorithm failed toachieve a continuous plot.

Fig. 3. River flow stations in operation by 1970 with significantincreases in mean annual flows 1970–1996.

Fig. 5. Trends in mean annual flows for all river flow stationswhich failed to register a significant increase in flow 1970–1996. Each record has been scaled by its 1970–1996 averageand filtered using the IRW smooth algorithm. Occasionalbreaks within individual records occur where the smoothingalgorithm failed to achieve a continuous plot.

flows was then scaled(by dividing by the 1970–1996 mean) and a trend generated using the IRWsmooth algorithm. A strikingly consistent temporalsignature emerges, in which the mean flows risefrom 70–80% of the mean in 1970 to a peak of110–115% in the late 1980s, falling back to 100–110% by 1996(Fig. 4). By contrast, the 26 stationswhich did not register a significant increase inmean flows report a different signature with arapid rise from 70–90% of the mean in 1970 to105–120% in the early 1980s falling back to 85–105% by 1996(Fig. 5). In terms of seasonalflows, 17 of the 38 stations show significantincreases in the spring and autumn and 15 showsignificant increases in mean winter flows. It isnoteworthy that of the 12 stations reportingincreases in the autumn and spring and 11 increas-es in the winter, the majority are in the south-west.Twenty stations(more than half) report decreasesin mean summer flows, but none of these issignificant.Changes in flow extremes are now examined

beginning with low flowsQ : the flow exceeded95

95% of the time. Twenty-one of the 38 gauging

stations report a decrease over the period 1970–1996, but only one of these(the highly urbanisedWhite Cart Water) is significant at the 5% level.When the low flows for each station are scaledand a trend derived using IRW smooth, mostrecords show values below the mean until 1980,


Fig. 6. Trends in low flows(Q ) for all river flow stations95

1970–1996. Each record has been scaled by its 1970–1996average and filtered using the IRW smooth algorithm. Occa-sional breaks within individual records occur where thesmoothing algorithm failed to achieve a continuous plot.

Fig. 7. River flow stations in operation by 1970 with significantincreases and decreases in flood magnitude(above POT thresh-old) and significant increases in flood frequency(POT events).Period of record: 1970–1996.

rising above the mean in the years 1985–1990,but thereafter registering a strong decline to 1996when virtually all stations are at 60–100% of themean(Fig. 6). Turning now to high flows, changesin both the magnitude and frequency of POTevents have been analysed over the period 1970–1996, POT magnitude being expressed as the meanvalue of all the flow events that have exceededthe threshold in a given year. Only four of the 44stations register significant changes in magnitude:two increasing and two decreasing(Fig. 7). Bycontrast, 15 stations report significant increases infrequency (typically from approx. two to fourevents per year), the majority being in the centraland southern part of Scotland.Summarising these results, it is clear that mean

flows have increased most consistently in the southand west, these increases being particularly pro-nounced during the 1970s to mid-1980s. Sincethen, mean flows have stabilised in the south andwest, but have declined in the north and east.Summer flows have generally declined throughoutScotland, but not sufficiently at any station togenerate a significant change. These results aregenerally in agreement with the findings of Smithand Bennett(1994), which only covered the period1970–1989. The changing pattern of low flowsregisters a general rise from 1970 to 1985 followed

by a relatively steep decline to 1996 when virtuallyno station exceeded its 1970–1996 average. Interms of high flows, the results confirm those ofGrew and Werritty(1995, 1995), i.e. there hasbeen a generally sustained increase in the frequen-cy of POT events from the mid-1980s onwards.However, given a POT threshold of 4.5 events peryear, the physical significance of this increasedfrequency should not be overstated. Although therehas been a series of major catastrophic floods since1989(Black, 1996), there is no consistent increasein the size of moderately high flows(POT events)across Scotland. Black and Burns(in press) pro-vide a more detailed analysis of the current floodrisk within Scotland.

5. Climate change, river flows and waterresources 2020–2080

The most extensive analysis of the impact offuture climate change on river flows and waterresources in the UK is that undertaken by Arnellet al. (1997) and summarised in Arnell(1996).


Using the output from the Hadley Centre transientclimate change experiment IS92a(adopted by theUK Climatic Change Impacts Review Group,1996), coupled with a lumped conceptual rainfall-runoff model, Arnell et al. generated annual waterbalances and flow duration curves for 21 represen-tative catchments across Britain under the antici-pated climate of the 2050s. Only four of thecatchments(R Don, R Almond, Lyne Water andR Nith) were located in Scotland, and they exhib-ited strikingly different responses to catchmentsfrom southern England. Under this scenario, annu-al runoff for all four Scottish catchments is pre-dicted to increase by 8.9–11.6%, in contrast todecreases(of up to 30%) for most English catch-ments. Applying a water balance model across thewhole of the UK and reporting the results for a0.5=0.58 grid, the predicted increase in annualrunoff across the whole of Scotland for the 2050sis 5–15%, but locally exceeds 25%. Turning tothe simulated flow duration curves for the 2050s,only the R Don, R Almond and R Nith register anincrease inQ of 5% or less. By contrastQ95 5

values(the flow exceeded 5% of the time) reportincreases of 10%(R Don); 16% (R Almond);24% (Lyne Water); and 11% (R Nith), again,typically much higher than for catchments inEngland.Since the study by Arnell(1996), there have

been significant advances in modelling future UKclimates, these being summarised in Conway(1998) and Hulme and Jenkins(1998). A majordevelopment has been the use of four land gridboxes(covering most of the land mass of the UK)for reporting GCM outputs and conveniently gen-erating separate values for Scotland(Wick to justsouth of Montrose, Fig. 1) and the Scottish Borders(north of Edinburgh into northern England, Fig.1). Further refinements have included modellingseasonal as well as annual trends and reportingthese results against the historic background oftemperature and precipitation from 1900 and1961–1990 means. Annual, winter and summerprecipitation for the Scotland and Scottish Bordersgrid boxes are reported for the 2020s, the 2050sand the 2080s using four contrasting climaticscenarios. The absolute range of percent changes(with respect to 1961–1990 means) for the medi-

um–high HadCM2 model outcomes are itemisedin Table 4. The mean increases in annual precipi-tation from the 2020s to the 2080s for the Scotlandand Scottish Borders land grid boxes are 6–16%and 6–14%, respectively. The seasonal distributionof these increases is concentrated in the autumn(SON) and winter (DJF) with increases as highas 24 and 29% in the two grid cells for autumnmonths by the 2080s. By contrast predicted sum-mer (JJA) changes in precipitation vary fromy5to q5%.In terms of water resource planning, these values

can only be regarded as broad-brush trends. AsHulme and Jenkins(1998) are at pains to pointout, they take no account of natural climaticvariability and could generate different absolutevalues if other climatic models were used. Downs-caling these outputs to individual catchments, andcoupling these precipitation estimates to rainfall-runoff models for a representative range of Scottishconditions, has yet to be attempted. Thus the workof Arnell (1996), based on an earlier climaticmodel (see above), provides the best current esti-mates of changes in annual water balances andrunoff regimes necessary for water resourceplanning.

6. Water resource planning and climate change

In summarising the potential impacts of climatechange on water resources, the UK ClimateChange Impacts Review identified a number ofissues(Climatic Change Impacts Review Group,1996). Only those itemised below are of particularconcern to water resource managers in Scotland:

● An increase in river flows in the winter and adecrease during the summer. Predicted increasesin winter precipitation and associated wettercatchment conditions are likely to result in ahigher frequency of riverine floods(Black andBurns, in press).

● In terms of water quality, higher water temper-atures will accelerate self-purification but thiscould also increase the risks of algal bloom andaccelerate eutrophication in rivers and lakes. Inwinter, increased flows will increase the dilutionof treated effluents and pollutants, whilst in


Table 4Changes in mean annual and seasonal precipitation(with respect to 1961–90) for 30-year periods centred on the 2020s, 2050s and2080s(medium–high HadCM2 scenario)

2020s 2050s 2080s

Mean Range Mean Range Mean Range

AnnualScotland q6% q4 toq9% q5% q2 toq12% q16% q10 to 21%Scottish Borders q6% q3 toq8% q5% 0 toq15% q14% q8 toq18%

Winter (DJF)Scotland q11% q8 toq15% q10% q7 toq15% q18% q10 toq22%Scottish Borders q10% q7 toq14% q11% q7 toq19% q19% q11 toq27%

Spring (MAM)Scotland y4% y9 to 0% y2% y9% toq4% q9% q1 toq13%Scottish Borders y2% y9 toq9% 0% y6 toq7% q11% q1 toq18%

Summer (JJA)Scotland q5% q2 toq7% y2% y5 toq2% q5% y2 toq10%Scottish Borders q2% 0 toq4% y5% y10 toq3% q1% y5 toq4%

Autumn (SON)Scotland q10% q6 toq13% q15% q8 toq26% q29% q20 toq45%Scottish Borders q10% q5 toq16% q13% q3 toq30% q24% q14 toq38%

Mean changes from ensemble-mean of HadCM2 experiment. Range from the four ensemble members of HadCM2 experiment(after Hulme and Jenkins, 1998).

summer, lower flows will reduce the dilution(Ferrier and Edwards, in press).

● Across the UK, climate change could increasethe demand for public water supplies by 5%over the period 1990–2021. The demand forspray irrigation across the UK(predicted to riseby 69% without climate change) could rise by115% with predicted increases in temperature.This latter figure is likely to be much lower forScotland, but significant in impact, especially ifsurface and sub-surface abstractions continue tobe largely unregulated(Fox and Walker, inpress).

● Although many fish species will not be affectedby the increases in water temperatures andchanged river flows, cold water lake fish(nota-bly the native brown trout) and migratory fish(especially salmon) could be affected. Butchanging river flows which reduce habitat suit-ability and the availability of breeding sites mayprove to be more significant(Gilvear et al., inpress).

● Very little is known about the sensitivity ofBritish aquatic ecosystems to climate change,

but river corridors and wetlands are generallymaintained by water levels in either an adjacentriver or shallow groundwater. It is likely thatsuch water levels will not be sustained, resultingin damage to more fragile wetlands and rivercorridors(Acreman, 2000; Bragg, in press).

In addressing these issues the CCIRG noteddifferent adaptive strategies for water suppliers,users and regulators. Given that much water man-agement in the past has been based on fixedstandards of service agreed by the water suppliersand regulators, responses to climate change in theshort term are likely to involve maintaining currentstandards at minimum additional cost. However,new financial imperatives may, in the longer-term,lead suppliers to consider new sources, more fullyintegrated use of existing sources and demandmanagement. Water users are likely to be drivento using water more efficiently(especially non-consumptive use) and may be required to upgradetreatment works for licensed discharges to maintaindownstream quality. Regulators will need to revisecurrent discharge consents as river regimes adjust


to climate change and, if introduced, will have theduty of authorising abstraction licences. Changesin the amount and seasonal distribution of inputsto hydro-electric schemes will require the devel-opment of new operating systems, whilst for floodprotection agencies, current models for estimatingflood risk may need to be revised. However, thereare parts of the water sector where adaptation toclimate change is undesirable or too expensive. Inparticular, the management of highly valued aquat-ic ecosystems is likely to prove unsustainable.

7. Conclusions

Claims that Scotland’s climate has become morevariable over recent decades, especially in termsof precipitation, are substantiated. A relatively dryperiod in the 1960s and 1970s has been succeededby the wettest period on record in the late 1980sand early 1990s, although that now seems to beending. This increase in precipitation has beenconcentrated in the winter half of the year and hasbeen especially pronounced in the north and west.By contrast, the drier east has experienced adecrease in summer precipitation over the sameperiod. This clear climatic signal is also registeredin Scotland’s rivers with those in the west reportingincreases in mean flows 1970–1996, this responsebeing concentrated in the earlier part of that periodand in the winter half of the year. Low flows overthe same period have declined, especially sincethe mid-1980s. By contrast high flows(POTevents) have increased in number across most ofScotland, although there is no consistent trend inthe size of these events. Predicted changes inprecipitation and river flows to the 2050s, basedon current climate change scenarios and outputsfrom GCMs, point to a continuation of many ofthese recent trends. Thus, it is likely that Scotlandas a whole will become wetter than at present, andaverage river flows will increase, notably in theautumn and winter months. However, low flowsare unlikely to show the dramatic declines pre-dicted for parts of south and eastern England,indeed, Q levels could even register modest5

increases. High flows could become more frequentincreasing the likelihood of valley floor inundation.If, as seems likely, Scottish catchments generally

become wetter, the frequency and severity ofmoderate flooding is set to increase over the nextfew decades.It is difficult to judge the implications of these

findings in terms of managing Scotland’s waterresources in the medium term as the future trendsare only specified in broad-brush terms. Moreprecise timing and specification of regional pat-terns will require further research designed toproduce more precise climatic scenarios and betterphysically based downscaled outputs from GCMs.In the interim, and in response to such uncertainty,water managers are urged to adopt the precaution-ary principle and when a new water managementscheme is being designed, the sensitivity of thesystem to plausible climate change should beinvestigated. In addition, during its implementa-tion, the system should be robust in accommodat-ing changes to the external environment includingclimate change. Given that the uncertainty andenvironmental risk associated with Scotland’shydroclimates looks set to increase in the mediumterm, improved understanding of the vagaries ofScotland’s past climate constitutes a major chal-lenge to the scientific community.

Acknowledgments

The results reported in the first part of this paperare based on a project funded by the Scottish andNorthern Ireland Forum for EnvironmentalResearchwcontract SR 97(8)x and jointly undertak-en with Keith Smith and Miranda Foster. I amgrateful for comments on an earlier draft byAndrew Black and delegates at the CIWEM Scot-tish Annual Conference in 1998.

References

Acreman M. The Hydrology of the UK—A Study of Change.London: Routledge, 2000.

Arnell NW. Global Warming, River Flows and WaterResources. Chichester: Wiley, 1996.

Arnell NW, Jenkins A, George DG. The Implications ofClimate Change for the National Rivers Authority, R&DReport 12. Bristol: National Rivers Authority, 1994.

Arnell NW, Brown RPC, Reynard NS. Impact of ClimaticVariability and Change on River Flow Regimes in the UK,Institute of Hydrology Report No. 107. Wallingford, 1990.


Arnell NW, Reynard N, King R, Proudhomme C, Branson J.Effects of climate change on river flows and groundwaterrecharge: guidelines for resource assessment. EnvironmentAgency Technical Report W82, Bristol. 1997.

Black AR. Modelling the seasonality of flooding in Scotland,Unpublished. Ph.D, University of St Andrews, 1992.

Black AR. Major flooding and increased flood frequency inScotland since 1988. Phys Chem Earth 1996;20:463–468.

Black AR, Burns JC. Re-assessing flood risk in Scotland. Thisvolume, in press.

Bragg, OM. Hydrology and peat-forming wetlands in Scotland.This volume, in press.

Climatic Change Impacts Review Group. Review of thePotential Effects of Climate Change in the United KingdomDepartment of the Environment. London: HMSO, 1996.

Conway D. Recent climate variability and future climatechange scenarios for Great Britain. Prog Phys Geogr1998;22:350–374.

Curran JC, Robertson M. Water quality implications of anobserved trend of rainfall and runoff. J Inst Water Eng Man1991;5:419–424.

Ferrier RC, Edwards AC. Sustainability of Scottish waterquality in the 21st century. This volume, in press.

Foster M, Werritty A, Smith K. The nature, causes and impactsof recent hydroclimatic variability in Scotland and NorthernIreland. Proc Sixth National Hydrol Symposium. Salford:British Hydrological Society, 1997. p. 8.9–8.17.

Fox IA, Walker S. Abstraction and abstraction control inScotland. This volume, in press.

Gilvear DJ, Heal KV, Stephen A. Hydrology and the ecologicalquality of Scottish river ecosystems. This volume, in press.

Grew H, Werritty A. Changes in flood frequency and magni-tude in Scotland. Proc Fifth National Hydrology Symposi-um. Edinburgh: British Hydrological Society, 1995. p. 3.1–3.9.

Hisdal H, Erup J, Gudmundsson K, Hiltunen T, Jutman T,Oversen NB, Roald L. Historical runoff variations in theNordic countries. Nordic Hydrological Programme, NHPReport 37, Nordic Hydrological Council, Oslo, Norway.1995.

Hulme M, Jenkins CJ. Climate change scenarios for the UK:scientific report. UKCIP Technical Report No 1. Norwich:Climatic Research Unit, 1998.

Intergovernmental Panel on Climate Change. In: Houghton JT,Meira Filho LG, Callander BA, Kattenburg A, Maskell K,

editors. Climate Change 1995: the Science of ClimateChange. Cambridge: Cambridge University Press, 1996. .

Intergovernmental Panel on Climate Change. In: Watson R,Zinqowera M, Moss R, editors. Climate Change 1995:Impacts, Adaptations and Mitigations: Scientific-TechnicalAnalysis. Cambridge: Cambridge University Press, 1996. .

Jowitt PW, Hay-Smith D. Reservoir yield assessment in achanging Scottish environment. This volume, in press.

Kilsby CG, Ewen J, Sloan WT, O’Connell PE. Modelling thehydrological impacts of climate change at a range of scales.Proceedings of the Second International Conference onClimate and Water, 3, Espoo, Finland 1998. p. 1402–1411.

Marsh T J, Anderson JL. Assessing the water resources ofScotland—perspectives, progress and problems. This vol-ume, in press.

Mayes J. Spatial and temporal fluctuations of monthly rainfallin the British Isles and variations in the mid-latitude westerlycirculation. Int J Climatol 1995;16:585–596.

Natural Environment Research Council. Flood Studies Report.London: HMSO, 1975.

Reynard NS, Proudhomme C, Crooks SM. The potentialimpacts of climate change on the flood characteristics of alarge catchment in the UK. Proc of the Second InternationalConference on Climate and Water, 1, Espoo, Finland 1998.p. 320–332.

Roald LA. Changes in runoff in the Nordic countries: a resultof a changing circulation pattern? Proc of the SecondInternational Conference on Climate and Water, 3, Espoo,Finland 1998. p. 1099–1109.

Smith K. Precipitation over Scotland, 1757–1992: someaspects of temporal variability. Int J Climatol 1995;15:543–556.

Smith K, Bennett AM. Recently increased river discharge inScotland: effects on flow hydrology and some implicationsfor water management. Appl Geogr 1994;14:123–133.

Werritty A. WN McClean: pioneer of river gauging in Scot-land. Hydrological Data UK 1985 Yearbook. Wallingford:Institute of Hydrology, 1985. p. 49–54.

Werritty A, Foster M. Climatic variability and recent changesin rainfall and river flows in Scotland. Proc of the SecondInternational Conference on Climate and Water, 3, Espoo,Finland 1998. p. 1110–1119.

Young P. Time variable and state dependent modelling of non-stationary and nonlinear time series. In: Subba Rao T, editor.Developments in Time Series AnalysisChapman and Hall,1993. .

living with uncertainty: climate change, river flows and water resource management in scotland

Documents