Hydrological responses to climate change: river runoff and groundwater

River runoff and groundwater

6 — Hydrological responses to climate change:river runoff and groundwater

%

__

L)- 4 .

:.d:‘. ‚ ..‚

-.

•___‘j .- s

0. river runoff. groundwater temperature000

— For all greenhouse gas scenarios, a shiftin runoff regime type is projected to occurin most catchments in Switzerland, withlower summer runoff and higher winterrunoff, hut little change in the total annualvolume.

— Limiting emissions to RCP3PD levelswould reduce the impacts on mean winterand summer runoff by approximately half incomparison with scenario A2 for the period2085 — hut even under scenario RCP3PDrunoff seasonality is likely to be altered.

— For groundwater temperature change,the medium estimates for the period 2085under the three greenhouse gas scenariosrange from +1 to +3°C. The warming, whichcould pose problems for groundwater quality and drinking water production, is accentuated in certain aquifers, especially in themid- to long term (periods 2060 and 2085).Situations of no change in groundwatertemperature, however, cannot be excludedgiven the assessed levels of uncertainty inthe projections.

< Seasonal redistribution and increasing variabilityof runoff make water resource management challenging (Aare river after the dry spe1 in spring201 1; photo: Jröme Wider, PLANAT).

6.1. INTRODUCTIONDue to its natural water storage capacity andits role as a source of water to its downstreamneighbors, Switzerland is sometimes calledthe „water tower of Europe“ . However, theeffects of climate change, je. ‚ glacier melting (Chapter 5) and changes in precipitation(Chapter 3), are calling this role into question.Climate change also affects ecosystems andhuman well-being indirectly through impactson water resources and water quality. Riverrunoff and groundwater reservoirs are crucial for understanding and quantifying theseimpacts: runoff is essential for the functioning of ecosystems and for supplying waterneeded for hydropower production, irrigation,cooling, and other uses; groundwater providesSwitzerland with 80% of its drinking water.

The impact of climate change on river runoff has recently been investigated as part ofa comprehensive assessment for the whole ofSwitzerland (FOEN, 201 2b). Here, this study iscomplemented by considering the effects ofdifferent greenhouse gas scenarios, by conducting a detailed and systematic assessmentof the model uncertainties in runoff projections, and by performing simulations for anintermediate mid-century time period.

Studies addressing the impact of climatechange on groundwater and groundwaterquality in Switzerland are rare. Groundwaterquality is crucially dependent on its temperature, and warming may affect groundwaterbiogeochemistry in a way that reduces itsquality and suitability as a source of drinkingwater. Since the 1980s, at least some aquifersin Switzerland have shown a marked increasein groundwater temperature (on the order of1°C) associated with a change in climate forcing (Figura et aL, 20 1 1). Here, groundwater temperature projections based on the CR201 1 scenarios are presented for a limited selection ofaquifers to provide a preliminary assessmentof potential future groundwater warming.

57


6.2. METHODSIn a first step, runoff from 186 mesoscalecatchments from all parts of Switzerland (20— 1760 km2 in size) is simulated under today‘sclimate and under projected future climateconditions to determine overall hydrologicalchanges and to delineate regions of similarhydrological response. Based on the AlB scenario, these Switzerland-wide simulations areperformed for a reference period (1984—2005,slightly shorter than the CR201 1 reference period) and two time periods in the near (2035)and far (2085) future using the downscaledDAILY-LOCAL dataset (Chapter 3). The hydrological model used is PREVAR-GIUB (ViviroUet aL, 2009a; b). Resulting changes in runoff,precipitation, and air temperature are classified by applying a cluster analysis to groupregions of similar runoff response. Sevendifferent types of region with similar runoff responses are identified, here referred toas response types C1—C7 (Figure 6,1; KöpUn etaL, 20 1 2). The discussion focuses on responsetypes C 1—C6, as C7 is strongly influenced byhydropower production, which follows economic rather than climate processes.

In a second step, this spatially comprehensiveanalysis is complemented by performing amore detailed analysis of six selected catchments (Rhone at Brig, Vorderrhein at Ilanz,Emme at Wiler, Thur at Andelfingen, Venogeat Ecublens, and Verzasca at Lavertezzo) representing four major response types withincreasing degrees of glaciation (C2, C4, C5,C6, Figure 6. 1 ). Three catchments (Emme, Thur,Venoge) classified mostly as belonging to thesame response type (C2) are selected to account for within-type variability and to testthe validity of the ciassification under different greenhouse gas scenarios. These six catchments are analyzed taking into account allgreenhouse gas scenarios and time periods,the climate modeling uncertainties (Chapters 2and 3), and partly the uncertainty in hydrological impact modeling. The analysis does notcapture the uncertainties inherent to the im-pact model parameters, and only part of thedownscaling uncertainty. For each catchment,the ensemble of 10 DAILY-LOCAL scenarios,as well as the lower, medium, and upper es-timates of the DAILY-REGIONAL dataset(Chapter 3), are used to force four hydrological models of different complexity, structure,and parameterization. These hydrologicalmodels are, in order of increasing complexity:

HBV (Seibert and Vis, 201 2), two versions of themodel PREVAH (PREVAR-GIUB, Vivi roH et al.,2009a; b; PREVAH-WSL, Kobierska et al., 2013),andWaSiM-ETH8.0.1 (Schutta and Jasper, 2007).Data on glacier extent is provided by Linsbaueret at. (2013). The scaling methodology appliedfor transient glacier simulation approximatesthe scenarios RCP3PD and A2 in a way thatmay bias the runoff projections for the highlyglaciated Rhone catchment. Consequentlythese greenhouse gas scenarios are not usedfor the Rhone catchment. Analysis of varianceis used to distinguish between climate uncertainty (including natural variability) and im-pact uncertainty associated with the hydrological models (Chapter 2; climate uncertaintyis not estimated for the Rhone catchment dueto the incomplete coverage of the scenariorange). The assessment focuses on changes inthe annual cycle, as the available climate scenarios do not adequately capture changes inextremes.

Groundwater temperature projections arecalculated for seven aquifers on the SwissPlateau based on the empirical relationshipbetween groundwater temperature and airtemperature (Figure 6,2). Four of the aquifersstudied are recharged mainly by riverbankinfiltration (henceforth river-fed aquifers),and three by precipitation only (precipitation-fed aquifers). Two linear regression mod-eis are employed. One of these focuses onyear-to-year correlations and yields separateestimates of annual mean groundwater temperature and monthly anomalies, whereas theother focuses on seasonality and yields directestimates of monthly mean temperature. Bothregression models are calibrated on historical measurement data that were only recentlyobtained (Schürch, 201 1). The end of the calibration period is set at 2007 for all data sets,while the beginning of the calibration periodvaries depending on the length of the dataavailable: for the three precipitation-fed aquifers (Kaeferberg, Laeufe, andVorem Haag), thecalibration periods begin in 1989, whereasfor three of the the four river-fed aquifersthey begin earlier (1971 for Seewerben, 1978for Signau, 1972 for Weieracker, and 1989 forDistelmatten).

Other factors such as the influence of riverdischarge on the temperature of infiltratingwater are not accounted for and are treatedas statistical uncertainty. Model performance

58 59

Cl L f1 C2 =J C3 C4 C5 C6 2 C7

HH = LCH2014 study catchments

Relative changes in clusters

— very negative 2035

weakly negative2085

± unchanged- Summer

+ weakly positveWinter

.1. very positve

Figure 6.1: The seven runoff response types identified under the Al B scenario using the hydrologicat modetPREVAH-OJUB (KöpUn et al., 2012). Cluster Cl (tight purp{e) corresponds to the Jura Mountains, cLuster C2to the catchments of the Swiss Ptateau (yetlow) and Ticino, and cluster C3—C7 (red, green, blue, and darkpurple) to the alpine region with different degrees of glaciation. The boxes indicate the relative change inrunoff in summer and winter in the scenario periods 2035 and 2085. The black contours indicate the sixcatchments selected tor the in-depth study.


is evaluated by the goodness of fit to subsamples of the measurement dataset (validationdata), after calibration based on the rest ofthe data (training data). The models performbetter for the river-fed aquifers than for precipitation-fed aquifers. This is either becausethe relationship between groundwater temperature and air temperature is tighter inriver-fed aquifers, or because the relevantmeasurement data series are longer. Box andtransfer function models were also tested, butperformed poorly in comparison to the linearregression models . Projections are calculatedbased on monthly mean air temperaturescomputed from the CR201 1 DAILY-REGIONALscenarios (including all three greenhouse gasscenarios and time periods).

6.3. RESULTSThe changes in runoff in Switzerland inducedby climate change can be grouped into sevendifferent response types (C1—C7, Figure 6.1) thatmainly refiect the major geographical regionsof Switzerland. Cluster Cl corresponds to theJura Mountains, cluster C2 to the catchmentsof the Swiss Plateau and Ticino, and clusterC3—C7 to the alpine region, with different degrees of glaciation. Changes in runoff can besummarized as an increase in winter and a decrease in summer, with the total annual volurne of runoff remaining approximately thesame. Changes in runoff in autumn and springare far less pronounced. These hydrologicalresponses are summarized for each cluster inFigure 6. 1 . For catchments in clusters C3—C7, thethermally controlled melting of glaciers andsnow governs the seasonal runoff. Here, the

Figure 6.2: Map showing the ocation of the aquifersstudied. Blue: river—fed aquifers. Oreen: precipita—tion—fed aquifers.

projected warming affects seasonality by in-creasing the proportion of ram in winter precipitation, by resulting in earlier snow melt,and by decreasing the amount of snow and icemelted during the summer. Accordingly, thesecatchments shift from a snow-controlled (nival) regime to a more rain-controlled (pluvial)regime. This shift is less pronounced at higherelevations with a higher degree of glaciation(regions C4 and C7). The rainfall-controlledcatchments in the Jura Mountains (Cl) and onthe Swiss Plateau (C2) show similar changesin seasonality, with the projected reductionof precipitation in summer and an increase ofliquid precipitation in winter directly alteringthe runoff. In general, projected changes inSwitzerland are more pronounced in alpine areas and in the distant future.

The robustness of the Switzerland-wide response signal is evaluated based on an uncertainty analysis for the six catchmentsselected for an in-depth study. Figure 6.3 presents the projected annual cycle and relateduncertainty bands for these six catchmentsin comparison to the reference period for theAlB scenario in the distant future (2085). Forall catchments, the uncertainty is large, butthe salient hydrological responses describedabove remain valid: (i) earlier melting of snowand ice in both alpine catchments (Rhone andVorderhein); (ii) less summer runoff; and (iii)greater winter runoff in most catchments.Although the projected changes in the perialpine catchments (Emme, Thur) and in thecatchment on the Swiss Plateau (Venoge) arecomparatively small in absolute terms, therelative changes are considerable (Figure 6.3,bottom row). In the alpine Rhone catchment, themelting of residual glaciers prevents the occurrence of a decrease in summer runoff untilthe end of the century under the non-intervention scenario AlB.

The effect of climate change mitigation on theprojected runoff is seen by comparing the different greenhouse gas scenarios (excluding thestrongly glaciated Rhone catchment, for whichonly the AlB projections are available; Figure6.4). Changes in runoff are projected to appearby the first scenario period (2035) irrespectiveof the greenhouse gas scenario. Differencesbetween the mitigation (RCP3PD) and thenon-intervention (AlB and A2) greenhouse gasscenarios become clear in the mid-century period (2060), when the effects of climate changeare also greater. For the end of the century

(2085), a further increase in the impacts of cli-mate change is projected, and differences appear between AlB and A2. Limiting emissionsto RCP3PD levels would reduce the impacts onrunoff in summer and winter, when the changeis strongest, by approximately a factor of twoin comparison to the impacts projected forscenario A2 for 2085. This potential reductionof the impact is independent of the responsetype (Figure 6.4, bottom row).

The uncertainty in the runoff projections forthe non-glaciated catchments is dominatedby the uncertainties in the climate modelsand by the natural variability of the climate(Figure 6.5). Within high alpine catchments, differences in the complexity of the glacier andsnow melt routines, as well as differences inthe representation of reservoirs in the hydrological models, are crucial and associated withhigh uncertainties, but the sign of the change

remains consistent across all model chainsof the DAILY-LOCAL data set (Chapter 3). Thisfinding is relevant for future studies and forthe interpretation of past studies that rely on asingle hydrological model. In the lowlands, theensemble mean indicates an increase of winterrunoff but high uncertainties make the signof the change inconclusive. In the mountaincatchments, bw winter runoff implies highuncertainties in the relative runoff change. Theuncertainties in the summer runoff projectionsremain relatively stable whereas uncertaintiesin the winter runoff tend to increase with time.

The three lowland catchments of the same response type show similar behavior. This finding supports the clustering derived from thespatial analysis (Köptin et aL, 2012; Figure 6.1)and justifies the extrapolation from the selected set of six catchments to the responsetypes identified in the Switzerland-wide study.

60

Rhone Vorderrhein Thur Emme Venoge Verzasca

Seewerben

Mean runoff (mm/day)

E;:I

:::::::: ::zzz: :::Jan Ju Jan Jan Jul Jan Jan Jul Jan Jan Ju Jan Jan Jul Jan Jan Jul Jan

Absolute change in mean runoff (mm/day)

[::::: ih

61

Jan Jul Jan Jan Jul Jan Jan Jul Jan Jan Jul Jan Jan Jul Jan Jan Jul Jan

Relative change in mean ranaff (%)\::zf t_;r [ZZZZZ ZZ ZZZZZZ [4ZZZZJ,: : w :z : EE ::zE:z :E-100 --100 --100 --100 --100 -100

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1Jan Jul Jan Jan Jul Jan Jan Jul Jan Jan Jul Jan Jan Jul Jan Jan Jul Jan

Figure 6.3: Mean runoff (ftp row) over the reference period (btack) and the scenario period 2085 (co[ored)tor scenario Al B, and the corresponding absolute change (middle row) and relative change (bottom row).For projected runoff, the mean over the uncertainty range is shown (bold colored line) along with the stan—dard deviation (shaded) and the minima/maxima (thin lines). Colors indicate the runoff response types toreach catchment according to Figure 6.2.


Based on the CR2011 climate projections,groundwater warming can be expected inall aquifers studied, hut (with the exceptionof Distelmatten) there are marked differencesbetween river-fed and precipitation-fed aquifers (Figure 66). According to the medium es-timates, by the end of the century the latterare projected to warm by < 1°C on averagewhile the former are projected to warm by1—3.5°C. However, the results do not indicateclearly whether the relatively slight warm-ing projected for precipitation-fed aquifersresults from a weak coupling of groundwater temperature to air temperature or fromthe comparative shortness of the data sets.Taking into account the uncertainty in theprojections, it is possible that groundwatertemperatures will increase by up to 7°C inriver-fed aquifers and by up to 2°C in precipitation-fed aquifers in any given year of theprojection period 2085. However, as a resultof the large uncertainties involved, the pro-jections do not exclude a situation in whichno warming, or even a slight cooling, mighttake place. Comparison of the groundwatertemperature projections under the three different greenhouse gas scenarios shows thatthey are very similar for the 30-year periodcentered on 2035, but diverge later (Figure 6.6).With regard to seasonality, the two regressionmodels are not always consistent; however,there is a clear tendency for warming to bestrongest in summer and autumn. The uncertainty associated with the projections resuitsmainly from the uncertainty inherent in thestatistical predictability of groundwater temperature from air temperature, which is responsible for 70—80% of the total projectionuncertainty. The remaining 20-30% of the totalprojection uncertainty results from the difference between the lower and upper estimatesof the CR2011 projections. The impact uncertainty is larger in most river-fed aquifers thanin precipitation-fed aquifers because of thestronger coupling with the naturally varyingair temperatures.

6.4. IMPLICATIONSThe total annual runoffvolume in Switzerlandis projected to remain approximately the sameas it is now. Rowever, at the regional scale,eg. ‚ in highly glaciated alpine valleys like theRhone catchment, this will be partly at theexpense of retreating glaciers. This general

statement is especially true for the northernpart of the Alps. In the Ticino and the southern Valais however, the annual runoff volumewill decrease (FOEN, 20 1 2b). The relative stabil-ity of the long-term annual runoff over timeis related to the smallness of the projectedchanges in annual precipitation rates, supplemented by the contribution of glacier melt inglaciated catchments.

A seasonal redistribution ofrunoffis projectedunder all greenhouse gas scenarios, and willaffect summer and winter runoff in all catchments. This confirms and corroborates previous findings (IPCC, 2007b; FOEN, 2012b) and hasalready been described qualitatively twentyyears ago (VAW, 1 990; OcCC, 2007). In the presentstudy, the hydrological responses to climatechange have been quantified using the latestand most comprehensive modeling approach.Furthermore, this study demonstrates that theresponse patterns are robust with respect tothe considerable uncertainties that result fromthe choice of climate and hydrological models.

Although the long-term annual runoff will remain approximately constant, the year-to-yearvariability and the seasonal redistribution ofdischarge suggest some challenges forwater resources management. Year-to-year variability,which may significantly affect water management, has not yet been investigated thoroughlybecause of the limitations of the downscalingmethod (Chapter 3). Concerning seasonal redistribution, the FOEN report (FOEN, 20 1 2b) as wellas Meyer (2012), showed for the Swiss Plateauthat an increase in the duration of dry spellsin summer might lead to water shortages likethose that occurred in the summer of 2003. Asshown in the present study, these water shortage situations are further exacerbated by theprojected groundwater temperature increase,which can affect groundwater quality. In all regions of Switzerland, water scarcity in summerhas additional implications for agriculture andecosystems, drinking water supply and hydropower production (SGHL and CHy, 201 1). For in-stance, Holzkämper et al. (2013a) showedthatthepositive effects of climate warming on agricultural yield are suppressed by water stress inextremely dry years like 2003. Still, in a continental context, the threat of drought stressis less for Switzerland than for southern andeastern Europe (IPCC, 2007b).

Climate change, and especially the higher temperatures associated with climate change, willcause a shift in the runoff regime from nival topluvial in catchments that are not strongly in-fluenced by glacial meltwater. Accordingly, anincrease in the variability of runoff throughout the year and from year to year, resultingin lower stability and lower predictability, isanother challenge that will need to be met inthe future. A more flexible and adaptive water management will be needed to deal with amore irregular and longer flood season (KöpLinet aL., 20 1 4) and an expected increase in thefrequency of occurrence of extreme eventssuch as droughts and floods. Joint regionalgovernance of water resources and watermanagement across political boundaries can

balance water demand and water availabilityat the local scale. Furthermore, multifunctional storage that — apart from hydropowerproduction — can be used for drinking watersupply, irrigation, artificial snow production,and flood retention might be a viable solution.Again, regional and multi-user agreementshave to be established, which is not an easytask also from a political point of view.

The implications for hydropower productionhave been studied by Hänggi et aL. (201 la; b).They considered the direct effect of changesin runoff on hydropower production acrossSwitzerland from the reference period 1980—2009 to the period 2035 (under scenario AlB),assuming current production schemes and

Rhone* Vorderrhein

2035


1I!;N NVenoge VerzascaThur Emme

_

62

Jan Jul Jan Jan Ju Jan Jan Jul Jan Jan Jul Jan Jan Jut Jan Jan Jul Jan

2060


_

Jan Jut Jan Jan Ju Jan Jan Jul Jan Jan Jul Jan

2085

Jan Jul Jan Jan Jul Jan


j[EE iz fJan JuL Jan Jan Jul Jan Jan JuL Jan Jan Jul Jan Jan Ju[ Jan Jan Jut Jan

Figure 6.4: Mean runoff over the reference period (black) and the three CH2O1 1 time periods (coLored)and greenhouse gas scenarios A2 (soUd), Al B (dashed), and RCP3PD (dotted). The Lines represent meansover 24 simuLations; the uncertainty range is not shown. CoLors indicate the runoff response types for eachcatchment according to Figure 6.2.*Scenarjos A2 and RCP3PD are not shown for the Rhone catchment due to potentiaL bias (section 6.2).

63


electricity markets. For winter they find an in-crease of 10% in hydropower production, forsummer a decrease of 4 to 6%, and overall forthe whole year a slight increase of 0.9 to 1.9%.Despite these encouraging resuits, adverseimpacts of climate change, especially in themid- (2060) and long term (2085) for single hydropower stations, cannot be excluded.

Although this study presents detailed andcomprehensive resuits, more research is needed to determine how changes in climatevariability and climate extremes will affectrunoff, as changes in climate variability arenot captured in the climate change scenariosused (Chapter 3). Furthermore, uncertaintiesrelated to the internal model parameters anda broader spectrum of downscaling methodsshould be addressed in future studies.

lt is worth emphasizing that the presentstudy shows that mitigation of greenhousegases can reduce the change in runoff in winter and summer by about a factor of two. Withregard to the implications summarized above,mitigation can heip considerably in attenuating the adverse effects of climate change.Still, some changes will occur and will require adaptation. Thus, future modeling stud-ies should couple hydrological impact modelswith models for water resource management,irrigation, hydropower production, etc. ‚ toquantify the impact of hydrological change onsociety, the economy, and ecosystems.

Although the models do not perform equallywell for all aquifers, the projections clearlyindicate that groundwater temperatures inriver-fed aquifers, which account for approximately 30% of Swiss drinking water production, will increase strongly. The main reasonfor this groundwater warming is the warmingof the rivers that feed the aquifers. Variousstudies show that higher temperatures affectmicrobiological activity during the infiltration of river water (Sprenger et aL., 201 1; Figura

et al., 20 1 3), and in Switzerland during theunusually hot, dry summer of 2003, oxygenconsumption was observed to increase withincreasing groundwater temperature to suchan extent that groundwater anoxia resulted(Hoehn and SchoLtis, 2011). This could poseproblems not only with respect to groundwater quality (Sprenger et aL, 201 1), hut also withrespect to drinking water production as a result of the possible clogging of pumping wellswith manganese and iron precipitates (Huntet aL, 2002). To assess the impact of climatewarming on oxygen concentrations and on theredox state of groundwater, monitoring at riverbank infiltration sites needs to be continuedand intensified in the future. Furthermore,becaus e the groundwater temperature projections rely solely on statistical relationshipsand are subject to relatively large uncertainties, further studies need to focus on constructing adequate models for groundwatertemperature based on field experiments andlong-term monitoring.

According to the present study, Switzerlandwill retain its role as Europe‘s water tower, hutas summer runoff and drinking water becomemore vulnerable in the course of the next century, adaptation measures, such as improvingthe efficiency of water usage, storage, anddistribution, will need to be implemented toprevent water shortages. Immediate reductionof greenhouse gas emissions as expressed inthe RCP3PD scenario can greatly reduce theseimpacts.

Winter

Relative change in mean runaff (‘1,)Rhone* 250 —

200 —

150 —

100 —

:

—1 00 —

1980-2009 2035 2060 2085

Vorderrhein 1 00

:—1 00 —

1980—2009 2035 2060 2085

Thur 1 00

:64

1980—2009 2035 2060 2085

Emma 100

:—1 00 —

1980—2009 2035 2060 2085

Venoge 100:—1 00 —

1980—2009 2035 2060 2085

Verzasca 250

0

—100 —

Summer

100

50

:—100

1980—2009 2035 2060 2085

1 00

:

—1 001980—2009 2035 2060 2085

100

:

—1 001980—2009 2035 2060 2085

100

:

—1 00 —1980—2009 2035 2060 2085

100

:

—100 —1980—2009 2035 2060 2085

1 00 —_:

—100 —

65

< Figure 6.5: Relative change in mean runoff projected tor winter (DJF, [eft) and summer (JJA, right).Oreenhouse gas scenarios are indicated by the cotors yetlow (RCP3PD), grey (Al B), and purpte (A2).The black boxes and the horizontal tine in the middte of each box show the range and the mean of theensemble, respectively. The shaded range indicates the ctimate uncertainty including the effect of naturatvariability. Note the wider scale of the y-axis for the Rhone and Verzasca tor winter.*Scenarios A2 and RCP3PD and the ctimate uncertainty range for Al B are not shown for the Rhone catchment because of potentiaL bias (section 6.2).

1980—2009 2035 2060 2085 1980-2009 2035 2060 2085


2035 2060 2085

A2

Temperature change (CC)

5 —

AlB


66 67

RCP3PD


Figure 6.6: Projected groundwater temperature change tor each of the three CH2O 1 1 c{imate scenarios andtime periods. Shown are the averaged projections of the two linear regression modeLs. Blue: river—fed aqui—fers. Green: precipitation—fed aquifers. The coored region of each box shows the uncertainty associated withthe climate projections, and the black cutUne shows the combined uncertainty of the temperature change inany given year of the corresponding future peFiod, including both c{imate and impact uncertainty.

Hydrological responses to climate change: river runoff and groundwater

Documents

Transcript of Hydrological responses to climate change: river runoff and groundwater