Modelling long-term stream acidification in the chemically heterogeneous Upper Severn catchment,...

Ž .The Science of the Total Environment 286 2002 215�232

Modelling long-term stream acidification in thechemically heterogeneous Upper Severn catchment,

Mid-Wales

Timothy Hilla,�, Paul Whiteheada, Colin Nealb

aAquatic En�ironments Research Centre, Department of Geography, Uni�ersity of Reading, Whiteknights, P.O. Box 277,Reading, Berkshire, RG6 6AB, UK

bCentre for Ecology and Hydrology, Maclean Building, Crowmarsh Gifford, Wallingford, Oxfordshire, OX10 8BB, UK

Received 20 August 1998; accepted 25 July 2001

Abstract

A two-box version of the long-term acidification model MAGIC is applied to the Upper Severn catchment,Mid-Wales. Comparison between modelled output and the observed stream- and groundwater chemistry points tothe limitations of modelling, due to the inherent complexity and variability in the catchment hydrology, soils, geologyand chemistry. The MAGIC model is used to produce long-term hind- and forecast predictions of average stream-,soil- and groundwater chemistry, and to simulate long-term changes in sulfate and nitrate deposition, in line withcurrent proposals of reduction. Changes in flow-routing pathways between soil- and groundwater are simulated andthe long-term effects on streamwater quality noted. The use of a long-term acidification model enabled thesimulation of streamwater quality under these different case scenarios. However, the modelled output is insensitiveto depositional and flow routing changes, indicating that catchment processes are not being represented to asufficient degree. Changes in simulated output as a result of increased acidic deposition are not statisticallysignificant, lying within the variance of long-term observed data. Simulated changes in flow routing suggest a lack ofmodel sensitivity, in terms of the effect on stream chemistry. The need for large amounts of measured data to ensurecorrect model representation of the hydrology, chemistry and the heterogeneous�variable nature of uplandcatchments is outlined. It is vital that these long-term data are available to ensure that problems do not arise due toover-reliance by catchment managers on potentially unreliable modelled output. � 2002 Elsevier Science B.V. Allrights reserved.

Keywords: Hydrology; Acidification; Modeling; MAGIC; Variability; Heterogeneity; Severn; Afon Hafren; Hafren

� Corresponding author. Tel.: �44-118-9318733; fax: �44-118-9755865.Ž .E-mail address: [email protected] T. Hill .

0048-9697�02�$ - see front matter � 2002 Elsevier Science B.V. All rights reserved.Ž .PII: S 0 0 4 8 - 9 6 9 7 0 1 0 0 9 7 7 - 9

( )T. Hill et al. � The Science of the Total En�ironment 286 2002 215�232216

1. Introduction

A considerable volume of research has beenamassed in the area of acidification of uplandareas, where surface waters are often both acidic

Žand acid-sensitive Stoner et al., 1984; Neal et al.,.1992, 1997a; Harriman et al., 1994 . The acidifi-

cation of such catchments may be considered onŽtwo time-scales: the short term of the order of

.hours to days , where periodic pulses of acidityresult from changing hydrological conditions; and

Žthe long term of the order of years to decades.and longer , reflecting changes in the buffering

Žcapacity and depositional loading Whitehead et.al., 1988, 1990 . The major causes of acidification

in upland Britain are considered to be as a resultof increased acidic deposition and land use changeŽ .Cosby et al., 1990; Neal et al., 1986, 1992, 1997b .Stream acidification has a detrimental effect onin-stream ecology as a result of decreased waterquality. In particular, increases in H� and Al3�

concentrations and the depletion of base cationsŽhave adverse effects on fish populations Harri-

man et al., 1994; Crisp and Beaumont, 1997; Neal.et al., 1997a .

The problems associated with acidification cre-ate a need to predict the susceptibility of a catch-ment to acidification with respect to changes inhydrology, deposition and land use. Models areused to simulate environmental change scenarios,resulting in modelled output affecting manage-ment decisions, particularly in terms of the effectsof emission reductions. Numerous types of math-ematical models have been produced to describethe processes associated with acidification occur-ring on these different time-scales. Short-term

Žacidification models include: Birkenes Chris-. Žtophersen et al., 1982 ; PULSE Bergstrom et al.,¨

. Ž1985 ; end-member mixing analysis EMMA;Christophersen and Neal, 1990; Christophersen et

.al., 1990; Hooper et al., 1990 ; TOPMODELŽ . Žmodified for stream chemistry Beven and

.Kirkby, 1979; Robson, 1993 ; and for the longterm, model of acidification of groundwater in

Ž .catchments MAGIC; Cosby et al., 1985a,b andŽPROFILE�SAFE Warfvinge and Sverdrup,

.1992 . In this paper, long-term models are con-sidered, with the MAGIC model applied to anupland, forested catchment. Long-term acidifica-tion models are used in the field of acidificationand catchment management for a variety of pur-poses:

1. Prediction of future acidification;2. Simulation of acidification under different de-

positional and land use scenarios;3. Hindcast reconstruction of past catchment

chemistry;4. Simulation of changes in hydrological

processes to determine relative importance;5. Placement of short-term acidification models

and their results in a broader, long-term con-text; and

6. Calculation of critical loads

Long-term acidification models can then beused to link soil and hydrological processes ex-amining the impact of acid deposition on surfacewater. Problems arise with any mathematicalmodel when trying to simulate complex heteroge-neous upland environments, due to model simpli-

Žfication Hill and Neal, 1997; Neal et al.,.1997a,b,c . Many long-term acidification models

have been utilised to determine critical load val-ues in order that thresholds can be determinedfor deposition levels. However, few studies haveestablished the representativeness of such modelsbecause of a lack of long-term field data. Therepresentation of catchment processes by thesemodels is of great importance, as they are used inthe determination of international and nationalemission legislation. In particular, these modelsare used in conjunction with the critical loadsconcept in determining thresholds of acidic depo-sition for surface waters. Therefore, it is neces-sary to evaluate a model in terms of the validity,

Žreliability and accuracy of the output Kirchner,.1990 .

In this paper, the long-term acidification modelMAGIC is applied to the Hafren�Upper Severncatchment and an assessment of model sensitivityrelating to observations of variability in ground-and surfacewater chemistry is made. The impor-

( )T. Hill et al. � The Science of the Total En�ironment 286 2002 215�232 217

tance of groundwater in the Hafren catchmenthas been noted, but until recently, groundwater

Žchemistry has not been directly measured Neal.et al., 1997a . Previous studies using MAGIC on

the Hafren examined long-term acidification withreference to depositional loading and conifer af-forestation, with no direct measurements of

Ž .groundwater chemistry Whitehead et al., 1988 .Other investigations in the area were undertakenon the adjacent Wye catchment, and again didnot directly use measurements from groundwaterŽ .Robson et al., 1991; Forti et al., 1996 . For thisinvestigation, direct measurements of groundwa-ter chemistry are used with MAGIC to calculatelong-term acidification trends. Furthermore, sen-sitivity analysis is performed using changes inSO2� and NO� deposition and hydrological flow4 3routing. These change analyses have implicationsfor both catchment modelling and the operationof hydrological processes within catchments. Forthis application of MAGIC, a semi-distributedchemical data set is used to compare modelledoutput with observed data.

2. Study area

This application of the MAGIC model is basedon the partially forested Afon Hafren�Upper

Ž .River Severn catchment, Mid-Wales UK , beingstudied from the source to the Hafren gaugingstructure. The area has been the subject of inten-sive monitoring by the Institute of HydrologyŽ .Plynlimon since the late 1960s, with an intensivechemical programme initiated in the early 1980sŽ .McCulloch, 1997 . The Severn catchment hasthree major tributaries, the Afon Hafren, AfonHore and the Afon Tanllwyth, with catchmentareas of 3.67, 3.08 and 0.89 km2, respectively, and

2 Ž .a total area of 8.75 km Fig. 1 . The upperreaches originate on an extensive plateau, with atotal altitudinal range for the catchment of320�740 m.

2.1. Geology

The geology of the Plynlimon area is comprised

Fig. 1. Location map of the Hafren�Upper Severn catchment.

of Ordovician and Lower Silurian rocks. The FanŽ .grits Ordovician are tough greywacke grits of

high quartzose content, with exposures on theŽ .upper plateaux area Fig. 2 . Overlaying the gritsŽare the Fan shales blue�grey mudstone and

. Ž .shales , which tend to be cleaved Breward, 1990 .At their junction with the Gwestyn shalesŽ .Silurian , a resistant band, the Rhaeadr mud-stone, is found. Gwestyn shales are grey�blackshales and mudstones containing iron pyrite,weathering to a yellow�brown colour. The Fron-goch mudstones contain little pyrite, and conse-quently are grey in colour. The catchment con-tains a fault band distinguishable by an outcrop of

Ž .Fan grits Breward, 1990 .

Fig. 2. Sketch map of the geology.


Ž .Fig. 3. Sketch map of the soil types adapted from the Soil Survey of England and Wales, 1988 .

2.2. Soils

The area is composed of a complex mosaic ofstagno-podsol, peat, brown earth and stagno-gley

Ž .units Neal et al., 1997b , with the major differ-ences between soil types the result of drainageŽ .Fig. 3 . Impeded drainage on the plateaux andwider interfluves has led to the accumulation of

Ž .peat 1�2m deep . On the more freely drainingslopes, podsols develop. A complex of peat andgley mineral soils develops on the valley bottomsdue to varying water levels and impeded drainageŽ .Kirby et al., 1991 . On the steeper slopes, thereare shallow soils, mainly podsol, with some rockexposures.

2.3. Vegetation

Approximately 68% of the catchment iscommercially managed coniferous forest. Tree va-

Ž .rieties include: Norway spruce Picea abies ; SitkaŽ . Žspruce Picea sitchensis ; lodgepole pine Pinus. Ž .contorta ; Scots pine Pinus syl�estris ; and

Ž .Japanese larch Larix kaempferi , with the ma-Ž .jority being spruce varieties Kirby et al., 1991 .

The main forested area was planted between thelate 1930s and 1960s, parts of which have beenfelled and are at various stages of replanting. Theupper area of the catchment is moorland, vege-

Ž .tated with acidic grassland Nardus and FestucaŽ .and peat mires Kirby et al., 1991 .

2.4. Climate

Average rainfall is approximately 2518 mmyear�1 , with evaporation and transpiration losses500�700 mm year�1 ; snow forms 5% of annualprecipitation. However, these values are highlyvariable. The average annual temperature is 7.3�C,with an average annual minimum and maximumof 3.7 and 11.0�C, respectively.

3. Conceptual basis and application of MAGIC

For reference, a synopsis of the conceptualbasis of MAGIC is included. MAGIC is a lumped,physically based, process-orientated model capa-ble of simulating long-term soil- and surfacewater

Ž .quality changes Cosby et al., 1985a,b, 1986 . Theconceptual basis for MAGIC can be found in theequilibria equations for a soil�soilwater system,

Ž .as proposed by Reuss and Johnson 1985 , whichhave been expanded to include other important

Ž .ions Whitehead et al., 1988 . Full details of themodel structure and equations used in MAGIC

Ž .can be found in Cosby et al. 1985a,b . In brief,model structure can be grouped into three cate-gories:

1. A suite of 24 equilibrium equations, wherethe concentrations of major ions are gov-erned by reactions involving adsorption,cation exchange, dissolution and precipitation


� Ž . �of aluminium as Al OH , speciation of3aluminium and organic acids and dissolutionand speciation of carbonic acid.

2. A suite of mass balance equations, where themajor flux of ions to and from the soil isassumed to be controlled by atmospheric in-puts, chemical weathering inputs, net uptakein biomass and losses in run-off.

3. Additional equations, relating variables in themass balance equations to variables in theequilibrium equations. These equations con-sist of definitions of alkalinity, total ionamounts within the catchment, lumped physi-cal characteristics of the soil and the sulfateadsorption process.

The model requires inputs for atmospheric de-position, net uptake�release fluxes for basecations and strong anions, and average catchmenthydrological conditions, for example annual pre-cipitation and discharge values.

4. Catchment details, water chemistry and modelcalibration

A two-layer version of the MAGIC model isused, with the upper compartment representingsoilwater and the lower, near-surface groundwa-

Ž .ter Fig. 4 . The model is calibrated using stream-and groundwater chemistry data for the referenceyear, 1996. The stream chemistry data relate tothe Hafren stream at the Hafren flume site, andrepresent a large proportion of the upper Severncatchment. Due to the unavailability of soilwater

Fig. 4. Diagrammatic representation of a two-layer version ofthe MAGIC model.

chemistry data, soilwater chemistry is estimatedusing a nearby ephemeral stream, being represen-tative of acidic gley soilwater. The groundwaterchemistry data are a mean of all 19 shallow

Ž .borehole sites �30 m deep in the catchment.Previous applications of MAGIC to the Severncatchment could not utilise measured groundwa-ter chemical data, as none were availableŽWhitehead et al., 1988; Robson et al., 1991; Forti

.et al., 1996 . Therefore, this application ofMAGIC will allow the use of actual rather thanpredicted chemistry data for groundwater sys-tems. Within the conceptualised two-layer model,different chemical reactions lead to distinctchemistries and are reflected in the parameter set

Ž .for soil- and groundwater Table 1 .

Table 1Soil- and groundwater characteristics used in the two-layered MAGIC model for the Hafren

Variable characteristic Unit Upper layer Lower layerŽ . Ž .soilwater groundwater

Soil depth m 0.8 1.0Porosity Fraction 0.45 0.10

�3Bulk density kg m 1500.0 1457.5�1Cation exchange capacity mEq kg 100 10

Pore volume m 0.36 0.10�2Soil mass kg m 1200.0 1457.5

Mean annual temperature �C 6.1 8.0Mean annual pCO atm 0.05 0.102

�1Mean annual organics �mol l 150 30


Table 2Calibrated parameter set used in MAGIC for the Hafren

Parameter Two-layer model

Upper Lowerlayer layer

2� �3Ž .SO Halfsat mEq m 80 10042� �1Ž .SO Maxcap mEq kg 8 104Ž .pk1 organics 5 5

Base cation weathering2 �1Ž .mEq m year

2�Ca 0 942�Mg 33 55�Na 49 48

�K 0 5

Percentage uptakesŽ .of input to each layer

�NH 98 904�NO 53 683

Calibration of the MAGIC model was per-formed sequentially. First, the concentrations ofCl� and SO2� were calibrated by adjusting occult4

and dry deposition rates. Second, base cationswere optimised to reference-year weathering ratesŽ .Cosby et al., 1990 . Optimisation of uptake ratesand initial saturation, again with reference tostream chemistry, enabled a full calibration of the

Ž . � �sub-compartments Table 2 . NO and NH were3 4calibrated by matching the inputs to the outputsfrom the system. The pCO of ground water was2set to fit field measurements, which were de-termined indirectly from alkalinity determina-tions. The pCO of soilwater was set at values2determined for the soils in the adjacent Wye

Ž .catchment cf. Robson et al., 1991 . Flow propor-tions and routing were derived from the end-

Žmember chemical mixing approach Neal et al.,.1992 . Land use was assumed to be steady-state,

with no tree growth or increases�decreases instock density.

5. Results from long-term modelling using MAGIC

The results produced from the calibration of

Table 3Observed and predicted stream-, soil- and groundwater chemistry for 1996

Rainfall Hafren Predicted Ephemeral Predicted Boreholes PredictedŽ Ž . Ž Ž .flume site surface stream soilwater all shallow groundwater

. .water sites

2� Ž . Ž . Ž . Ž .Ca 5�151 25�117 8�25 56�131224 48 48 18 21 252 252

2� Ž . Ž . Ž . Ž .Mg 0�67 51�81 29�80 48�74617 66 66 62 62 181 181

� Ž . Ž . Ž . Ž .Na 4�299 149�213 135�233 175�35675 179 179 191 191 289 289

� Ž . Ž . Ž . Ž .K 0�39 2�12 0�2 0�844 4 4 0.3 0.3 15 15

� Ž . Ž . Ž . Ž .NH 4�227 0�3 0.5�3.5 1�6.33438 0.5 0.5 1.2 1.2 2.3 2.3

2� Ž . Ž . Ž . Ž .SO 12�283 66�135 82�213 54�396458 88 88 138 138 190 190

� Ž . Ž . Ž . Ž .Cl 11�313 147�220 113�279 152�35591 181 181 201 201 279 279

� Ž . Ž . Ž . Ž .NO 0�165 10�56 0�4 2�21330 24 24 0.25 29 13 13

3� Ž . Ž . Ž .Al 0.2�6.9 3�44 46�91 �

1.7 17 24 61 117 � 2Ž . Ž . Ž . Ž .pH 4.0�5.5 4.4�6.7 3.8�4.1 4.4�7.84.3 4.7 4.7 3.9 4.5 5.2 5.1

Ž . Ž . Ž . Ž .Alkalinity �109�0.2 �38�48 �171��119 �37�4318�29.3 �1 2 �148 �109 169 254

Values in parenthesis are observed ranges, average values in bold; all units, excluding pH, in �Eq l�1 .


Table 4Observed and simulated streamwater chemistry

Observed Simulated1996 1860 1996 2136

Streamwater chemistry2�Ca 48 37 48 522�Mg 66 62 66 64�Na 179 179 179 179

�K 4 3 4 4�NH 0.5 0 0.5 0.54

2�SO 88 33 88 914�Cl 181 174 181 181

�NO 24 1 24 2433�Al 17 5 24 25

pH 4.7 5.0 4.7 4.7Alkalinity �1 73 2 0

Soilwater chemistry2�Ca 18 7 21 312�Mg 62 57 62 59�Na 191 191 191 191

�K 0.3 0.2 0.3 0.3�NH 1.2 0 1.2 1.24

2�SO 138 37 138 1364�Cl 201 193 201 201

�NO 0.3 0.8 29 2933�Al 61 42 117 109

pH 3.9 4.7 4.5 4.5Alkalinity �148 17 �109 �98

Groundwater chemistry2�Ca 252 218 252 2502�Mg 181 173 181 176�Na 289 289 289 289

�K 15 12 15 14NH � 2.3 0.1 2.3 2.34

2�SO 190 113 190 2144�Cl 279 271 279 279

�NO 13 0.4 13 1333�Al � 1 2 3

pH 5.2 5.2 5.1 5Alkalinity 169 308 254 223

All units, excluding pH, in �Eq l�1 .

MAGIC compare well with both stream-, soil-Ž .and groundwater chemical data Table 3 . For the

streamwater chemistry, a good fit is found betweencalibrated and observed data, indicating that rep-resentation of within-catchment processes arereasonable, or that inadequacies with the processrepresentation are masked in parameterisation.Within the calibration of soil- and groundwater,

there are a few minor exceptions. The most domi-nant feature is the large difference betweenobserved and modelled soilwater nitrate concen-trations. The soilwater data are compared to anephemeral stream, which may not be wholly rep-resentative of soilwater composition. Nitrate up-take from soilwater before entering the stream orthe gley soil-type behaving differently to the soilmass of the catchment could account for thisanomalous feature. There are also slight discrep-ancies between the predicted and modelled alka-linity in the soil mass, again related to soilwatercomposition differing from the composition of theephemeral stream. Despite these differences, thechemistry data for the ephemeral stream are thebest representation of soilwater chemistry dataavailable. For the groundwater, a good fit isobserved, with the exception of the alkalinity,which has a notable difference, despite majorcation and anion concentrations agreeing wellwith observed data. Likely reasons for the differ-ences include the possibilities that the variabilityin observed data is not represented fully by themodel, and that the mean values taken fromselected observation sites are not totally repre-sentative of the groundwater chemistry. MAGICis capable of simulating mean annual values forsurface runoff, soil- and shallow groundwaterchemistry to a reasonable degree for the refer-ence year. As the calibration of the modelled datais reasonable when compared with the observedchemistry, the model can be used to examine thelong-term effects of acidification on catchmentchemistry, with particular emphasis on surfacewater quality.

6. Long-term acidification of the Upper Severncatchment

The long-term hindcast simulation of stream-,soil- and groundwater chemistry was performedusing a reconstructed sulfate deposition sequencebased on data from the Department of the Envi-

Žronment Warren Spring Laboratory, 1983; Vin-.cent et al., 1996; Vincent et al., 1998 . Forecasting

was undertaken assuming maintenance of sulfurdeposition at 1996 levels. Long-term simulated


Fig. 5. Long-term simulated chemistry in the Hafren catchment.

chemistry suggests the Upper Severn catchmenthas acidified from pre-industrial times to the pre-sent, with future acidification showing slight re-

Ž .covery Table 4 and Fig. 5 . The effects of acidi-fication are reflected in the pH, alkalinity andaluminium concentrations, although they have notchanged substantially. Modelled base cation andstrong anion concentrations have not altered sig-nificantly, despite the changes in depositionchemistry. Large increases in sulfate and nitrateconcentrations are predicted in streamwaters fromthe pre-industrial periods. The simulated changesin streamwater chemistry reflect depositionalmodifications. Historical changes prior to 1996exhibit the greatest changes in constituents.

With respect to forecasts for 1996�2136, main-tenance of deposition at 1996 levels results inminimal effects on concentrations, although aslight acidifying effect is simulated. The patternsin modelled streamwater chemistry are mirrored

Žin the soil- and groundwater chemistries Table 4.and Fig. 5 .

Table 5The effects of decreased sulfate deposition scenarios onstreamwater chemistry for the year 2136

2�SO deposition4

Original Reductionscenario 30% 60% 80% 100%

2�Ca 48 50 49 47 462�Mg 66 64 64 63 63�Na 179 179 179 179 179

�K 4 4 4 4 4�NH 0.5 0.5 0.5 0.5 0.54

2�SO 88 74 57 46 364�Cl 181 181 181 181 181

�NO 24 24 24 24 2433�Al 24 18 12 10 8

pH 4.7 4.8 4.8 4.9 4.9Alkalinity 2 16 30 40 49



Fig. 6. Scenario analysis for decreased sulfate deposition in runoff. The percentages expressed on each chart represent reductionsin the total SO2� deposition to the catchment.4

6.1. Simulated forecasts of the effects of reduced�increased sulfate deposition

A method capable of remediation of catchmentacidification is to reduce sulfate emissions fromanthropogenic sources, such as power stations,thereby decreasing acidic depositions on acid-sen-

sitive areas. The use of modelling techniques en-ables simulation of the effects of reduced deposi-

Žtion on streamwater chemistry Whitehead et al.,.1988 . Several different sulfate deposition scenar-

ios are considered, decreasing linearly for a pe-riod of 20 years from 1996, until a predetermined

Žreduced level of deposition is reached 30, 60, 80


.and 100% reduction . Thereafter, deposition re-mains constant at this reduced value until the end

Ž .of the modelling period 2136 . All reductions inSO2� deposition have an effect on stream-, soil-4

Žand groundwater chemistry Table 5, Figs. 5 and.6 . Initially, reduced sulfate deposition improves

streamwater quality, but the rate of improvementdeclines towards the end of the modelling period.A significant feature is produced by a 30% reduc-

tion in SO2� deposition, resulting in an ecologi-4cally significant change in alkalinity, which in-creases it to a positive value. However, other thanCa2�, SO2� and Al3�, there are no significant4changes in streamwater cation and anion concen-trations in response to the reduced levels of sul-fate deposition. Streamwater acidity shows slightrecovery since the reduction emission protocols ofthe 1980s, and maintenance of SO2� deposition4

Fig. 7. Scenario analysis for decreased sulfate deposition in soilwater. The percentages expressed on each chart represent simulatedreductions in the total SO2� deposition to the catchment.4


Fig. 8. Scenario analysis for decreased sulfate deposition in groundwater. The percentages expressed on each chart representsimulated reductions in the total SO2� deposition to the catchment.4

at 1996 levels results in slight water quality im-provements post-1996. Different levels of sulfatereduction produce similar patterns of water qual-ity improvement, although the concentrations ofconstituents are altered. Decreasing SO2� depo-4sition by 100% enables streamwater chemistry torecover, but not to pre-industrial levels, by theend of the modelled period. For the soil- and

groundwater, the patterns of recovery match thatŽ .of the surface run-off Figs. 7 and 8 . As with the

streamwater, 100% reduction in the levels of sul-fate deposition does not cause a return to thepre-industrial levels by the end of the modelledperiod. Also, with the exception of Ca2�, SO2�

4and Al3�, there are no other significant changes

Ž .in the cation and anion concentrations Table 5 .


Table 6The effects of increased and decreased NO deposition3scenarios on streamwater chemistry for 2136

�NO deposition3

15% Original Reductionincrease scenario 15% 30% 60%

2�Ca 52 48 52 51 512�Mg 64 66 64 64 64�Na 179 179 179 179 179

�K 4 4 4 4 4�NH 0.5 0.5 0.5 0.5 0.54

2�SO 91 88 91 91 914�Cl 181 181 181 181 181

�NO 28 24 21 17 1033�Al 27 24 23 22 19

pH 4.7 4.7 4.7 4.8 4.8Alkalinity �3 2 4 7 14


The greatest recovery is recorded in the soilwater,indicating the importance of the soil mass inacting as a store for buffering the acidity.

6.2. Simulated forecasts of the effects of reduced�increased NO depositionx

Current proposals for the reduction of NOxreleased during power generation indicate that apotential 78% could be removed from the processŽ .UNECE, 1994 . NO production levels are to bexreduced within the European Union from 1980levels by 10 and 30% for 1993 and 1998, respec-

Ž .tively Hultberg and Skeffington, 1998 . For thepurpose of the modelling exercise, NO� is used3to represent a significant component of NO .xNO� concentrations are both increased and de-3creased to represent changes in depositional load-ing of NO . Nitrate deposition has been increasedxby 15% and decreased by 15, 30 and 60%, thedecreases being in line with the proposed reduc-tions. The decreases and increases in NO� depo-3sition did not significantly alter the streamwaterchemistry, although slight changes are predictedŽ .Table 6 and Fig. 9 .

6.3. Effects of flow-routing changes

Changes in the contribution of groundwaterhave been highlighted as a possible method for

improving streamwater quality, in terms of acid-Ž .ity, at Plynlimon Neal et al., 1997a . Therefore,

flow-routing changes in groundwater contribu-tion, with respect to soilwater, are modelled toexamine the effects on streamwater chemistry.Groundwater contributions are increased to dou-ble and decreased to half the present contribu-tion with respect to the contribution from soilwa-ter. The aim of this is to examine the long-termeffects of hydrological changes on water qualityŽ .Table 7 and Fig. 10 . Despite the hydrologicallysignificant changes in flow contribution, notableeffects on streamwater chemistry were not simu-lated in modelled output.

7. Discussion

Using the long-term acidification model,MAGIC, stream-, soil- and groundwater chem-istry have been simulated in the past and future

Ž .under current 1996 deposition scenarios. Theuse of MAGIC enables simulation of the mean,annual streamwater chemistry using a simpletwo-component model. Comparative studies in theWye catchment have indicated that a two-layermodel provides a good prediction of long-termwater quality, even when soil- and groundwater

Ž .are heterogeneous Forti et al., 1996 . For thisstudy, a two-component model does predictstream-, soil- and groundwater chemistry. How-ever, from modelling depositional scenariochanges, a lack in model sensitivity is noted, hav-ing major implications for modelled results. Sev-eral features of catchment hydrochemistry canexplain the lack of sensitivity of the model tochanges in inputs. First, the variability and het-erogeneity within upland catchments cause aver-age values to be used in models that may or may

Žnot be representative Neal and Robson, 1994;.Neal, 1996; Neal et al., 1997a . Second, end-mem-

ber identification proves difficult, being related tothe complex nature of the hydrochemical interac-tions. Third, inaccuracies arise because of modelsimplification within chemical reactions, hydrolog-ical routing and ground- and soilwater interac-tions. Unidentifiability within component end-


Fig. 9. Scenario analysis for changes in nitrate deposition in the Hafren catchment. The percentages expressed on each chartrepresent simulated reductions�increases in the total NO� deposition to the catchment.3

members has also been found when using acid-Ž .neutralising capacity ANC for hydrograph split-

Ž .ting Neal et al., 1997b . Despite these problems,MAGIC was used to attempt to model averagestream-, soil- and groundwater chemistry for thecatchment, in both hind- and forecasting situa-tions, although as with other long-term acidifica-tion models, it is difficult to confirm the accuracy

Ž .of model predictions Reynolds, 1997 . Modelling

of streamwater chemistry into the future indicatesa slight reversal of acidification within the catch-ment, based on 1996 levels of deposition. Rever-sal of the acidification process is also predicted ifSO2� deposition is decreased; a 30% decrease4will increase water quality above important eco-logical thresholds. Streamwater quality improvedto near pre-industrial levels with a total reductionin anthropogenic SO2�, indicating the potential4


Table 7The effects of simulated changes in flow routing on streamwater chemistry

Groundwater contribution

Original Halved Doubled

1860 1996 2136 1860 1996 2136 1860 1996 2136

2�Ca 37 48 52 37 47 52 37 50 522�Mg 62 66 64 62 66 64 62 66 64�Na 179 179 179 179 179 179 179 179 179

�K 3 4 4 3 4 4 4 5 5�NH 0 0.5 0.5 0 0.5 0.5 0 0.7 0.74

2�SO 33 88 91 33 89 91 33 88 914�Cl 174 181 181 174 181 181 174 181 181

�NO 1 24 24 1 26 26 Q1 22 2233�Al 5 24 25 5 27 26 5 22 23

pH 5.0 4.7 4.7 5.0 4.7 4.7 5.0 4.8 4.7Alkalinity 73 2 0 73 �4 �2 73 7 4


for reversing the anthropogenic acidification ef-fect over long time-scales. However, these resultsremain highly speculative given the problems out-lined. As for the simulation of changes in NO�

3deposition, no significant differences wereobserved in modelled output. There are severalpossibilities as to why only small changes wereobserved in the stream, despite the reduction andincreases in NO�:3

1. Nitrate deposition may only account for asmall proportion of the total NO� flux within3the catchment.

2. Nitrate deposition is not a significant processdriving run-off chemistry.

3. Soil processes dominate run-off chemistry, in-dicating they are critical in determining theNO� concentrations in the stream.3

4. Model sub-compartments are not wholly rep-resentative of complex interactions betweensoil- and groundwater, producing under-sensi-tised results.

If it is the case that NO� deposition does not3significantly alter stream water quality, currentproposals for reductions in output from powergeneration may not have as great an impact as ishoped. Representation of the nitrogen dynamicswithin the model may be improved if the model

Ž .MAGIC-WAND see Jenkins et al., 1997 or

MAGIC version 7 is used. Future investigationsneed to assess the effects of land use change inresponse to forest growth, changes in stockingdensity, clear-felling and other management prac-tices. These changes may significantly alter bothNO� uptake and SO2� depositional loads. Simu-3 4lated changes in flow routing also produced unex-pected results. Large changes in streamwaterchemistry were not simulated in response to sig-nificant flow routing alterations, which does notconcur with field observations. In the adjacentTanllwyth sub-catchment, the introduction of aborehole near to the stream opened fractureroutes into the streambed, thus increasinggroundwater contributions to the stream. Subse-quently during stormflow, Ca2� concentrationswere elevated by 30% of original levels and pH

Ž .was increased by 0.4 pH units, Neal et al., 1997a ,despite the increase in hydrological contributionof groundwater being small. Again, this modelledresult implies a problem with the model in repre-senting highly heterogeneous and variable soilwa-ter�groundwater interactions.

The results from this modelling exercise indi-cate that for these inherently complex environ-ments, the MAGIC model is capable of modellinggross catchment chemistry based on averagedmeasured values. The extent to which the resultscan be relied on in these heterogeneous catch-ments is under scrutiny. The lack of model sensi-


Fig. 10. Scenario analysis for changes in flow routing in the Hafren catchment.

tivity in determining catchment chemistry, as aresult of depositional and hydrological changes,suggest that there is a problem with the represen-tation of this type of model in complex heteroge-neous systems. The long-term acidification modelMAGIC has simulated streamwater chemistry toa reasonable degree based on average depositio-

nal, flow and calibration values. However, there isan underlying problem with the use of averagevalues as a representation of the catchment hy-drochemistry. Furthermore, there is anotherproblem identified due to parameter uncertaintyand over-parameterisation, which may producesimilar unrepresentative output. In this paper and


Table 8Variation in streamwater chemistry compared with simulatedacidification

Streamwater chemistry

Ž .Simulated Observed 1983�1998

1860 1996 2136 Min. Max. Mean S.D.

2�Ca 37 48 52 16 119 44 10.52�Mg 62 66 64 14 121 64 9.9�Na 179 179 179 99 292 177 19

�K 3 4 4 2 25 4 1.7�NH 0 0.5 0.5 0 52 1 2.54

2�SO 33 87 91 45 255 87 194�Cl 174 181 181 110 344 201 33.5

�NO 1 24 24 0 97 20 10.533�Al 5 24 25 2 120 21 15.4

pH 5.0 4.7 4.7 4.1 6.8 4.4 �Alkalinity 73 2 0 �63 69 �2 19.2


Žother publications Hill and Neal, 1997; Hill, 2000;.Neal et al., 1997d , both spatial and temporal

variations in rainfall, soilwater, groundwater andstreamwater quality have been shown to be sig-nificant features of catchment hydrochemistry.These features, despite being of large signifi-cance, are not represented within the modelstructure or processes in the MAGIC model.Variation occurs on a variety of levels within the

Ž .model structure, from the inputs rainfall , theŽ .stores soil- and groundwater and the outputs

Ž .streamwater . The variation observed in the fielddata poses fundamental questions as to themethod of determining critical loads and the ef-fects of long-term deposition on catchment acidi-fication. The acidification effect simulated byMAGIC for the Hafren can be placed into per-spective when compared with the total variation

Ž .observed for the period 1983�1998 Table 8 .From Table 8, it is evident that the use of averagevalues to represent a catchment system is farfrom adequate, with the true extent of the varia-tion demonstrated by the large ranges found inobserved data. Furthermore, random noise andcyclical patterns in deposition are also not repre-sented in the model, and again can be highlysignificant in determining the hydrochemistry of a

Ž .catchment Neal et al., 1997d . It is thereforeimperative that within catchment studies, the

range of surface-, soil- and groundwaterchemistries are assessed fully before any mod-elling exercise is undertaken. Using both mod-elled and observed data, it is possible to reachconclusions concerning the effects of acidifica-tion. However, any variability in stream chemistryposes many questions as to the representation ofa model not containing a component of spatialvariation, particularly when the model is used torepresent regional changes in water quality.

From this modelling study, the value of a largelong-term data set is evident, not only to providedata to calibrate and test a model, but also toexamine the conceptualisation of catchmentchemistry within the model structure. Althoughlong-term models are difficult to validate, a long-term chemical data set can provide much detail.In the case of acidification of the Hafren catch-ment, no trends were recorded either in the in-crease of anions or decrease in cations as a result

Žof acidic input over the last 20 years Robson and.Neal, 1996 . Prediction using MAGIC indicates

no further acidification based on current deposi-tion, concurring with observed long-term data.

It is evident that long-term models have theiruses in determining the long-term stream chem-istry from reconstructed depositional scenarios.However, the large data requirements for thistype of model are also noted. Lumped hydro-chemical models are capable of simulating thegross chemical behaviour of highly complex het-erogeneous environments, although no account istaken of the spatial variation. Furthermore, tem-poral variability is not simulated at a high enoughresolution to account for the variation in observedchemistry data. However, due to the complexheterogeneity and variability, both spatial andtemporal, found in the hydrochemistry at Plynli-mon, doubts as to the validity of such lumped

Žmodels remain Neal, 1996; Neal et al., 1997e;.Hill and Neal, 1997 . Future hydrochemical catch-

ment-research requires a combined approach,with a modelling framework integrated with bothlong-term environmental monitoring and datacollection. An assessment of the validity of mod-els can then be made with reference to actualobserved data, ensuring that models remain asuseful tools for catchment managers.


Acknowledgements

The authors would like to thank the Centre forEcology and Hydrology and its staff at Plynlimonfor their support during the fieldwork stages ofthis paper.

References

Bergstrom S, Carlsson B, Sandberg G, Maxe L. Integrated¨modelling of runoff, alkalinity and pH on a daily basis.Nord Hydrol 1985;16:89�104.

Beven K, Kirkby M. A physically based, variable contributingarea model of basin hydrology. Hydrol Sci Bull

Ž .1979;24 1 :43�69.Breward N. Geochemical Cycling Processes Involving Major

and Trace Elements at Plynlimon, Mid-Wales. Leeds, UK:Department of Earth Sciences, University of Leeds, 1990.PhD thesis, 565 pp.

Christophersen N, Neal C. Linking hydrological, geochemicaland soil chemical processes on the catchment scale: aninterplay between modelling and field work. Water Resour

Ž .Res 1990;26 12 :3077�3086.Christophersen N, Seip HM, Wright RF. A model for

streamwater chemistry at Birkenes, Norway. Water Resour1982;18:977�996.

Christophersen N, Neal C, Hooper RP, Vogt RD, Andersen S.Modelling streamwater chemistry as a mixture of soilwaterend-members � a step towards second generation acidifi-cation models. J Hydrol 1990;116:307�320.

Cosby BJ, Hornberger G, Galloway J. Modelling the effects ofacid deposition: assessment of a lumped parameter modelof soil water and streamwater chemistry. Water Resour Res

Ž .1985a;21 1 :51�63.Cosby BJ, Wright R, Hornberger G, Galloway J. Modelling

the effects of acid deposition: estimation of long-termwater quality responses in a small forested catchment.

Ž .Water Resour Res 1985b;21 11 :1591�1601.Cosby BJ, Whitehead P, Neale R. A preliminary model of

long-term changes in stream acidity in South-western Scot-land. J Hydrol 1986;84:381�401.

Cosby BJ, Jenkins A, Ferrier R, Miller J, Walker T. Modellingstream acidification in afforested catchments: long-termreconstructions at two sites in central Scotland. J Hydrol1990;120:143�162.

Crisp DT, Beaumont W. Fish populations in the PlynlimonŽ .streams. Hydrol Earth Syst Sci 1997;1 3 :541�548.

Forti MC, Neal C, Robson A. Modelling the long-term changesin stream, soil and ground water chemistry for an acidmoorland in the Welsh uplands: the influence of variationsin chemical weathering. Sci Total Environ 1996;180:187�200.

Harriman R, Likens GE, Hultberg H, Neal C. Influence ofmanagement practices in catchments on freshwater acidi-fication: afforestation in the United Kingdom and North

America. In: Steinberg C, Wright R, editors. Acidificationof Freshwater Ecosystems: Implications for the Future.Wiley and Sons, 1994:83�101.

Hill TJ. Dynamic Modelling and Analysis of Hydrochemistryin Upland Forested Catchments. The University of Read-ing, 2000. PhD thesis, 200 pp.

Hill TJ, Neal C. Spatial and temporal variation in pH, alkalin-ity and conductivity in surface runoff and groundwater forthe Upper River Severn. Hydrol Earth Syst Sci 1997;Ž .1 3 :697�717.

Hooper RP, Peters NE, Christophersen N. Modellingstreamwater chemistry as a mixture of soilwater end-mem-bers � an application to the Panola mountain catchment,Georgia, USA. J Hydrol 1990;116:321�344.

Hultberg H, Skeffington R, editors. Experimental Reversal ofAcid Rain Effects. Chicester: John Wiley and Sons, 1998.466 pp.

Jenkins A, Ferrier RC, Cosby BJ. A dynamic model forassessing the impact of coupled sulfur and nitrogen deposi-tion scenarios on surface water acidification. J Hydrol1997;197:111�127.

Kirby C, Newson M, Gilman K, editors. Plynlimon Research:The First Two Decades. IH Report No 109, 1991. 188 pp.

Kirchner JW. Heterogeneous geochemistry of catchment acid-ification. Geochim Cosmochim Acta 1990;56:2311�2327.

McCulloch JSG. Foreword: an historical overview from theŽ .watershed. Hydrol Earth Syst Sci 1997;1 3 :381�384.

Neal C. Towards lumped integrated models of complex het-erogeneous environmental systems. Sci Total Environ1996;183:115�125.

Neal C, Robson AJ. Integrating soil water chemistry variationsat the catchment level within a cation exchange model. SciTotal Environ 1994;144:93�102.

Neal C, Whitehead P, Neale R, Cosby BJ. Modelling theeffects of acidic deposition and conifer afforestation onstream acidity in the British uplands. J Hydrol1986;86:15�26.

Neal C, Robson A, Reynolds B, Jenkins A. Prediction offuture short-term stream chemistry � a modelling ap-proach. J Hydrol 1992;130:87�103.

Neal C, Hill T, Alexander S, Reynolds B, Hill S, Dixon A,Harrow M, Neal M, Smith C. Stream water quality inacid-sensitive UK upland areas: an example of potentialwater quality remediation based on groundwater manipula-tion. Hydrol Earth Syst Sci 1997a;1:185�196.

Neal C, Hill T, Hill S, Reynolds B. Acid-neutralization capac-ity measurements in surface and ground waters in theUpper River Severn, Plynlimon: from hydrograph splittingto water flow pathways. Hydrol Earth Syst Sci 1997b;Ž .1 3 :687�697.

Neal C, Robson AJ, Christophersen N. Towards couplinghydrological, soil and weathering processes within a mod-elling perspective. In: Saether OM, Caritat P, editors.Geochemical Processes, Weathering and GroundwaterRecharge in Catchments. Rotterdam: A.A. Balkema, 1997c.395 pp.


Neal C, Wilkinson J, Neal M, Harrow M, Wickham H, Hill L,Morfitt C. The hydrochemistry of the headwaters of theRiver Severn, Plynlimon. Hydrol Earth Syst Sci

Ž .1997d;1 3 :583�618.Neal C, Robson AJ, Christophersen N. Towards coupling

hydrological, soil and weathering processes within a mod-elling perspective. In: Saether OM, Caritat P, editors.Geochemical Processes, Weathering and GroundwaterRecharge in Catchments. Rotterdam: A.A. Balkema, 1997e.395 pp.

Reuss J, Johnson D. Effect of soil processes on the acidifica-tion of water by acid deposition. J Environ Qual1985;14:26�31.

Reynolds B. Predicting soil acidification trends at Plynlimonusing the SAFE model. Hydrol Earth Syst Sci

Ž .1997;1 3 :717�728.Robson AJ. The Use of Continuous Measurement in Under-

standing and Modelling the Hydrochemistry of the Up-lands. University of Lancaster, 1993, p. 278. PhD thesis

Robson A, Neal C. Water quality trends at an upland site inWales. Hydrol Proc 1996;10:183�203.

Robson A, Jenkins A, Neal C. Towards predicting futureepisodic changes in stream chemistry. J Hydrol 1991;125:161�174.

Stoner J, Gee A, Wade K. The effects of acidification on theecology of streams in the upper Tywi catchment in WestWales. Environ Pollut 1984;35:125�157.

United Nations Economic Commission for Europe. Effectsand Control of Long-Range Transboundary Air Pollution.Geneva: UNECE, 1994. No 10, ECE�EB.AIR�39.

Vincent K, Campbell G, Downing C, Hasler S, Davies M,Stedman J, Sansom L, Briscombe C, Page H. Acid Deposi-tion Monitoring in the United Kingdom: The First TenYears. Department of the Environment, 1996.

Vincent KJ, Downing CE, Halser SE, Smith M, Sansom LE,Page HM, Campbell GW. Acid deposition monitoring inthe UK: 1986�1996. Department of the Environment, 1998,pp. 27.

Warfvinge P, Sverdrup H. Calculating critical loads of aciddeposition with PROFILE� a steady-state soil chemistrymodel. Water Air Soil Pollut 1992;63:119�143.

Warren Spring Laboratory. Acid Deposition in the UnitedKingdom. Stevenage: Warren Spring Laboratory, 1983.72 pp.

Whitehead PG, Reynolds B, Hornung M, Neal C, Cosby J,Paricos P. Modelling long-term stream acidification trendsin upland Wales at Plynlimon. Hydrol Proc 1988;2:357�368.

Whitehead PG, Musgrove T, Cosby B. Hydrochemical mod-elling of acidification in Wales. In: Edwards RW et al.,editor. Acid Waters in Wales. Dordrecht: Kluwer AcademicPublishers, 1990.

Modelling long-term stream acidification in the chemically heterogeneous Upper Severn catchment,...

Documents

Transcript of Modelling long-term stream acidification in the chemically heterogeneous Upper Severn catchment,...