dissolved beryllium in rainfall, stream and shallow groundwaters in the upper river severn...

The Science of the Total Environment 314–316(2003) 171–184

0048-9697/03/$ - see front matter� 2003 Elsevier Science B.V. All rights reserved.doi:10.1016/S0048-9697(03)00102-5

Dissolved beryllium in rainfall, stream and shallow groundwatersin the Upper River Severn catchments, Plynlimon, mid Wales

Colin Neal*

Centre for Ecology and Hydrology Wallingford, Maclean Building, Crowmarsh Gifford, Wallingford, OXON, OX10 8BB, UK

Accepted 2 January 2003

Abstract

This paper examines the temporal changes in dissolved beryllium in deposition(rainfall and cloud water), streamwater and groundwater for the upper River Severn catchments at Plynlimon in mid-Wales. There are two main themesto the study. Firstly, time series records are examined to see if anomalous behaviour occurred during 1996, whenremarkably high concentrations were unexpectedly observed in the UK lowland rivers(Neal, Sci Total Environ,2003). The results show(a) Beryllium concentrations in rainfall and stream water remained low throughout theperiod (mean 0.02 and 0.07mg l in rainfall and stream water, respectively) and were often less than the lowesty1

quotable value for a single determination(0.05mg l ). (b) Beryllium concentrations in the streams declined betweeny1

1983 and 1996 from a mean of approximately 0.07 to a mean of 0.04mg l . This was followed by a brief increasey1

in the autumn of 1995(up to values of approx. 0.2mg l and a more sustained increase to approximately 0.12mgy1

l from 1997 to the end of monitoring late in 1998.(c) Rainfall concentrations of Beryllium were indistinguishabley1

from zero throughout most of the monitoring period although concentrations increased late in the study in line withpatterns observed in the stream when concentrations averaged approximately 0.08mg l . (d) Beryllium concentrationsy1

are much lower than observed in the UK lowlands where concentrations as high as 29mg l were recorded. For they1

exceptionally high values occurring in the lowlands, there was the potential for environmental damage to aquaticorganisms such as fish at levels greater than approximately 1mg l . There are no potential problems for the uppery1

River Severn. Secondly, while Neal et al.,wJ Hydrol, 136(1992) 33–49x provided information on the hydrogeochem-istry of beryllium in the upper River Severn using available information at that time(rainfall, cloud water, stemflow,throughfall and stream water), there was no information available on groundwater chemistry. Since the publication ofNeal et al.wJ Hydrol, 136(1992) 33–49x, such information is now available and this paper makes up this shortfall.The results show that beryllium concentrations in groundwater are typically approximately 2–3 times higher thanthose found within the streams(mean 0.14mg l , range 0.06–1.56mg l ). This feature probably reflects they1 y1

increased leaching of beryllium from the bedrock. The findings presented in this study combined with the earlierinformation of Neal et al.wJ Hydrol, 136(1992) 33–49x are used to provide an overview on dissolved beryllium forthe upper River Severn, the most complete and extensive record for the UK.� 2003 Elsevier Science B.V. All rights reserved.

Keywords: Beryllium; Rainfall; Cloud water; Stemflow throughfall; Streamwater; Groundwater; Hafren; Hore; River Severn

*Corresponding author. Tel.:q44-1491-838800; fax:q44-1491-692424.E-mail address: [email protected](C. Neal).

172 C. Neal / The Science of the Total Environment 314 –316 (2003) 171–184

1. Introduction

Beryllium (Be) is a divalent metal of hightoxicity in aquatic environments at concentrationsapproximately 1mg l (Reeves, 1979; Heath ety1

al., 1988; Haines et al., 1991; Edmunds andTrafford, 1993; Edmunds and Smedley, 1996).Beryllium hydrolyses easily due to its small sizeand high surface charge density. Under circumneu-tral and alkaline conditions Be precipitatesyco-precipitates as oxideyhydroxide phases(especiallywith aluminium) resulting in dissolved Be concen-trations in the range of a few ng l to a fewmgy1

l (Edmunds and Trafford, 1993; Vesely et al.,y1

2002). Beryllium is mobilised under acidic con-ditions (Edmunds and Trafford, 1993; Edmundsand Smedley, 1996; Vesely et al., 2002) andconsequently it is under such conditions that envi-ronmental concerns are normally the greatest(Nealet al., 1992; Vesely et al., 2002). For example, aregional freshwater survey in the Czech Republicshowed that Be concentrations increased a hundredfold between pH 7 and pH 4: the median concen-trations are approximately 0.03mg l and 1mgy1

l , respectively, the corresponding ranges beingy1

0–0.1 mg l and 0.3–3.0mg l (Vesely andy1 y1

Majer, 1996; Vesely et al., 2002). Furthermore,increased Be mobility under acidic conditions isof concern in relation to atmospheric deposition ofacidic oxides from industry. Indeed, it has beenshown that Be concentrations increased 100% instream water at the artificially acidified watershedat Bear Brook, Maine, USA(Kaste, 1999; Nortonet al., 1999; Vesely et al., 2002). However, therelationship between Be and acidity is not straight-forward and factors such as soil acidification, soilstores and bedrock sources come into play.

Within the UK, considerable information hasbeen accumulating over the past decade on region-al levels of Be in ground and surface waters aspart of regional surveys undertaken by the BritishGeological Survey(Edmunds and Trafford, 1993).However, long term data for Be levels are sparsefor the UK with most of the information comingfrom two sources: the upper River Severn catch-ments in mid Wales(Neal et al., 1992) and thelowland eastern UK rivers entering the North Sea(Neal, 2003). These data sources are of national

significance and they provide a major body of datathat, in part, makes up for a shortfall of informationon Be owing to the difficulties of measuring lowlevels on a regular basis(a problem which hasonly gradually been overcome over the past 10–20 years).

With an increasing interest in Be in the environ-ment (Vesely et al., 2002), Neal (2003) investi-gated a major hydrochemical dataset for thelowland UK collected as part of a major commu-nity research programme, the Land Ocean Inter-action Study(LOIS). The study came up with astartling observation that in OctoberyNovember1995, there was an anomalous and short-livedincrease in Be concentration(up to 29 mg l )y1

linked to drought conditions and sewageyindustri-ally related discharges.

For the upper River Severn study of Neal et al.(1992), the monitoring programme has continuedup to the present time(Neal et al., 1997a) andnew information on temporal changes in Be con-centrations in rainfall and stream water have beenaccumulating. Furthermore, with the introductionof boreholes in the area(Neal et al., 1997b,c) newinformation on groundwater has been collectedthat was not available at the time of the Neal etal. (1992) study of Be.

In this paper, the data collected on Be for theupper River Severn is examined in the light of thenew information. This is done to● Examine if there is an increase in Be for streams

in the uplands which mirrors the anomalousperiod observed within the UK lowlands.

● Examine if rainfall Be concentrations have alsochanged during the anomalous period(there isno equivalent data for the UK lowlands).

● Provide new information on Be concentrationsin groundwater for the upper River Severncatchments.The contribution provides the most detailed

study of Be hydrochemistry for a UK upland‘endmember case’ to complement earlier work onBe for the upper River Severn, regional surveysfor the UK and the eastern UK rivers.

2. Study area

The Be concentration data reported in this papercomes from wide ranging studies at the Centre for

173C. Neal / The Science of the Total Environment 314 –316 (2003) 171–184

Ecology and Hydrology catchment research sitesat Plynlimon in mid-Wales(Neal et al., 1997a):the headwater catchments of the River Severn,with three main tributaries, the Afon Hafren, theAfon Hore and the Nant Tanllwyth(8.7 km area2

in total). The upper River Severn drains a hilltopplateau dominated by acid moorland in the upperportion of the catchment(Pumlumon Fawr) andthe Hafren Forest on the plateau edge, intermediateslopes and valley bottom. The moorland and forestcatchments represent a mixture of upland acid soiltypes dominated by peaty podzols with subsidiarypeaty gleys while deep peat deposits are alsoimportant in the moorland plateau area. The bed-rock comprises fractured Lower Palaeozoic mud-stones, shales and grits.

The Hafren Forest comprises mainly Sitkaspruce (Picea sitchensis) with some Norwayspruce(Picea abies), larch (Larix spp) and lod-gepole pine(Pinus contorta) planted in variousphases from the mid 1940s through to the late1960s onto acid moorland. Harvesting has beenundertaken in some areas and this has mainlyinvolved ‘stem only’ removal from site leaving thefelling debris (stumps, branches and needles)behind.

Mean rainfall is 2518 mmyyear with evapotran-spirational losses of approximately 500–700 mmyyear and it is typical for the upland UK(Hudsonet al., 1997). The rainfall is relatively unpollutedexcept for moderate long distance transport ofacidic oxides and for all the streams. For all thestreams, the flow responses to storm events arevery flashy(Neal et al., 1997a).

Stream chemistry is variable and predominantlylinked to inputs of water from two hydrochemi-cally distinct parts of the catchment(Neal et al.,1997a). Under baseflow conditions, stream watersare approximately pH 7. They are calcium andbicarbonate bearing as they are mainly derivedfrom shallow groundwater areas(weatheringreleases calcium from the soil and bicarbonatecomes from biogenic sources of carbon dioxide asweathering consumes hydrogen ions). For theAfon Hore, calcium and bicarbonate concentra-tions are higher than for the Afon Hafren and NantTanllwyth since the bedrock contains a higherproportion of calcium carbonate, which weathers

relatively quickly. Stormflow waters are mainlyderived from the acidic soils of the area. Thisprovides runoff that has low pH and low calciumconcentrations but it is enriched in aluminium dueto mobilisation under acidic conditions.

Groundwater is dominated by fracture flow andthe chemistry ranges from acidic waters character-istic of the soil zone to moderately alkaline andcalcium bicarbonate bearing waters characteristicof bedrock weathering(Neal et al., 1997b,c). Thegroundwaters show a highly dynamic responsesimilar to the streams(Neal et al., 1997b,c).

3. Monitoring points and sampling programme

There are four types of site monitored forberyllium.

Firstly, there are rainfall, cloud water and themain tributaries of the headwaters of the RiverSevern that have been monitored weekly for up to18 years. Rainfall was collected from open gaugesat the top and bottom of the HafrenyHore catch-ment while cloud water was collected near the topof the catchment, both on a weekly basis. Thethree main tributaries of the upper River Severnwere monitored(two each on the Afon Hafren andAfon Hore and one on the Tanllwyth) on a weeklybasis using grab samples. Only one site has acatchment confined to the moorland area(upperHafren) while the remaining sites have catchmentspartially to completely forested. They are the lowerHafren, which includes the moorland drainagefrom the upper Hafren, the entirely forested Tan-llwyth, the upper Hore, and the lower Hore. Theforest area in the lower Hore catchment wasclearfelled between 1985 and 1989. For the othersites, some localised felling has taken place overmany years to thin and in some cases harvestsmall areas of the crop, and approximately half ofthe Tanllwyth catchment was felled in February1996.

Secondly, seven small catchment areas(-15ha) were monitored to examine the interactionbetween soil type and forest harvesting on waterquality. The monitoring has been weekly or fort-nightly for 4–12 years. A paired catchmentapproach with control and manipulated forestedcatchments has been used at two locations(i.e.


two times two monitoring points). All the sites arewithin the Hafren Forest and represent two of themain soil types of concern, peaty podzols(south2Hore, south east 1, 2 and 3) and peatygleys (Tanllwyth 1 and 2). The manipulated sitesat Plynlimon are south2Hore, south east 1 andTanllwyth 1: they drain into the Afon Hafren withthe exception of south2Hore, which drains into theAfon Hore.

Thirdly, there are a series of six boreholes drilledto monitor groundwater chemistry and ground-water levels within the Hafren catchment. Thesesites were monitored weekly to fortnightly for upto six years. The exploratory boreholes were estab-lished in 1984 throughout the catchment. Theycover a range of geomorphologic environments.Four monitoring boreholes were installed near thepaired catchments to provide data on groundwaterquality changes at felling and control sites(southeast 1 and 3 and Tanllwyth 1 and 2). Two addi-tional paired borehole sites were also monitored(Hafren 4 examined the influence of felling, whilethe quarry borehole provided the control).

Fourthly, stemflow and throughfall were sam-pled between 1984 and 1991 on a weekly basiswithin a mature Sitka spruce stand near the nowclosed CEH Dolydd office at Staylittle.

Details of the catchment characteristics andsampling periods are provided in a companionpaper(Neal et al., 2003).

4. Chemical analysis

For the chemical analysis, the samples were firstfiltered using 0.45mm cellulose nitrate mem-branes. In the case of the stream waters, thesamples were filtered in the field while all othersamples were filtered soon after return to thelaboratories and stored in polyethylene bottles. Thebottles were cleaned before use using 10% vyvconcentrated hydrochloric acid followed by dis-tilled water and the bottles were thoroughly rinsedin several aliquots of filtered sample prior tostorage. On return to the Plynlimon laboratory, thefiltered samples were acidified with aristar gradeconcentrated nitric acid(to 1% vv acid) to elimi-nate the potential for Be precipitation onto thesample containers. The acidified samples were then

kept in the dark at 48C (both at the Plynlimonoffice and on return to CEH Wallingford) the toavoid potential biological deterioration prior to Beanalysis.

Samples were analysed for Be using inductivelycoupled plasma mass spectrometry. For the deter-mination, mixed quality control standards wereincluded in the routine analyses. These standardswere checked weekly against international qualitycontrol references prepared by the United StatesGeological Survey National Water Quality Labor-atory(SRM 1643c). Each new batch of calibrationstandards was always crosschecked with the pre-vious batch and the international control refer-ences. With each batch of samples, CEHWallingford’s own quality control standards werealso included. If there were significant discrepan-cies between the old and new calibration standards,a complete remake of new standards was carriedout. The CEH Wallingford laboratory takes part inUK and international inter-laboratory comparisonexercises, a rigorous quality assurance system ismaintained.

The lowest quotable values for single determi-nations were taken as six times the standarddeviation obtained for a blank solution(0.05 mgl ). Within this paper, Be concentrations are ofteny1

close to or below the lowest quotable value forindividual determinations. However, in order toplot the time series data and to calculate mean andmedian Be concentrations for the various sites,values above and below the lowest quotable valueneed to be used. Furthermore, the lowest quotablevalue for an individual determination is higherthan that for multiple determinations that havethen been averaged. For the present purpose, themeans and medians were calculated using the ‘rawdata’ that included the less than values. The rawdata were also used for the plotted data, and hencethere appear both positive and negative concentra-tions below the lowest quotable value. For all theplots, the concentration scale includes the negativevalues. The lowest concentration set was minusthe lowest quotable value(i.e. y0.05 mg l ) toy1

show the range of scatter near and below thelowest quotable value and provide a visual indi-cation of the number of points that lie within thelowest quotable value for individual data points


(i.e. "0.05 mg l ). For the remainder of they1

paper, where information is presented in the textfor numbers less than the lowest quotable value,the numbers are italicised.

Given the low concentrations being measuredwithin this study, an analysis of intermediate stan-dards and blanks was undertaken using the rawanalytical records available. For the intermediatestandards, only the data for the lowest one(1 mgl ) was looked at in detail as this was close toy1

the maximum concentration measured in the sam-ples analysed. In total, there were 2300 blanksanalysed as well as 850 1mg l intermediatey1

check-standards. The results showed that for theblanks, the mean concentration(with an error barof twice the standard deviation) is 0.01"0.03 mgl with a range of y0.01–0.06mg l . They1 y1

corresponding values for the 1mg l intermediatey1

check-standard is mean 1.01"0.11 mg l with ay1

range of 0.86–1.11mg l . With regards to they1

blanks, the values were slightly higher during theearly phase of the work, but the values are none-theless not significant in relation to the measure-ments presented in this paper. Correspondingly, theestimated concentration for the 1mg l interme-y1

diate check-standard is close to the actual value.Thus, the measured values are reliable within theremit of the study remembering that in many casesthe values measured are close to or below thelowest quotable value.

5. Hydrological data

Hydrological information on rainfall, cloudwater, stream-flow and groundwater-level was col-lected to assess the influence of hydrology on Beconcentrations.

The rainfall and streamflow data comes fromthe Plynlimon hydrological network, maintainedby the catchment section of the Centre for Ecologyand Hydrology, Wallingford(Hudson et al., 1997).Rainfall measurement was based on an integratednetwork of gauges located throughout the catch-ment while flow was measured at four flumeswithin the catchment: the upper and lower Hore,the lower Hafren and the Tanllwyth. There are nogauging structures for the other streamwater qual-ity monitoring sites and so flow data had to be

extrapolated from the nearest, most suitable site.In all cases, the flow data used was expressed interms of area-weighted values(mmy15 min) toallow for the different size of the catchments andthe absolute differences in flow. For cloud water,stemflow and throughfall, volumes of catch wereused to provide an indication of the relative levelsof deposition although, of course, in the case ofcloud water, there was no direct measure of theactual transfer of water flux. Correspondingly, forgroundwater, the depth to the water table wasdetermined to provide an indication of changes inlevel: the levels quoted have not been normalizedto a standard datum level such as m.a.s.l.

6. Results

The results of this study are presented as asummary in Table 1 and as three sets of time-series graphs: Fig. 1, rainfall, cloud water and themain upper River Severn tributaries; Fig. 2, theintermediate to small streams; Fig. 3, the ground-waters. For Table 1, information on throughfalland stemflow was included for completeness(thedata were described by Neal et al., in 1992).Preliminary analysis showed that there was nodiscernable felling effect and hence the plots,tables and forthcoming interpretation of the resultsdo not deal with this issue to simplify matters.The results for Be are presented in a sequenceatmospheric inputsycanopy, stream water andgroundwater to provide an overview of the resultsbefore the findings are brought together within adiscussions section.

6.1. Beryllium in rainfall, cloud water, throughfalland stemflow

The levels of Be in rainfall, cloud water,throughfall and stemflow are low with mean con-centrations of less than 0.1mg l and a maximumy1

concentration of 1.47mg l (in cloud water).y1

The salient features are:

● Rainfall: In general, rainfall Be concentrationswere less than the lowest quotable value andfor the full period of record, the mean concen-


Table 1Summary statistics of beryllium concentrations in rainfall, cloud water and groundwater within the upper River Severn

Mean Median Maximum Flowweightedmean

AtmosphereycanopyRain 0.02 0.01 0.24 0.02Cloud 0.07 0.02 1.47 0.04Throughflow 0.01 0.01 0.13 0.01Stemflow 0.02 0.02 0.17 0.02

Large streamsUpper Hafren 0.04 0.03 0.27 0.05Lower Hafren 0.06 0.06 0.30 0.08Upper Hore 0.06 0.05 0.27 0.09Lower Hore 0.07 0.06 0.33 0.11Mean main streams 0.06 0.05 0.29 0.08Minimum main streams 0.04 0.03 0.27 0.05Maximum main streams 0.07 0.06 0.33 0.11

Intermediate and small streamsTanllwyth 0.11 0.11 0.47 0.14South2Hore 0.06 0.05 1.14 0.05South east 1 0.08 0.07 0.33 0.11South east 1 0.11 0.11 0.39 0.17Tanllwyth 1 0.09 0.08 0.36 0.09Tanllwyth 2 0.07 0.07 0.40 0.08Mean intermediateysmall streams 0.09 0.08 0.52 0.11Minimum intermediateysmall streams 0.06 0.05 0.33 0.05Maximum intermediateysmall streams 0.11 0.11 1.14 0.17

GroundwatersSouth east 1 0.13 0.12 0.50 –South east 3 0.28 0.32 0.82 –Hafren 4 0.13 0.12 1.56 –Quarry 0.08 0.07 0.56 –Tanllwyth 1 0.06 0.03 0.66 –Tanllwyth 2 0.15 0.13 0.72 –Mean groundwaters 0.14 0.13 0.80 –Minimum groundwaters 0.06 0.03 0.50 –Maximum groundwaters 0.28 0.32 1.56 –

tration was only 0.02mg l (0.02 mg l ony1 y1

a flow weighted basis, with a median of 0.01mg l and a maximum of 0.24mg l ). Aparty1 y1

from the occasional outlier point, the highestconcentrations occur towards the end of therecord (autumn 1995). This increase startedwith a short-lived peak of approximately 0.16mg l at the end of November 1995(therey1

was a gradual increase from near zero concen-trations in mid September 1995 and a subse-quent decay to near zero concentrations at thebeginning of January 1996). After this time,

there was a more sustained increase from lessthan the lowest quotable value in May 1997 toapproximately 0.10mg l mid-August 1997y1

and the mean concentration from the initialincrease to the end of record, late 1998, was0.07mg l .y1

● Cloud water: Cloud water contained more Bethan rainfall, with a mean of 0.07mg l (flowy1

weighted mean 0.02mg l , median 0.02mgy1

l and maximum 1.47mg l ) but many ofy1 y1

the values are less than the lowest quotablevalue. As with rainfall, there was an increase in


Fig. 1. Temporal variations in beryllium concentration for rainfall, cloud water and the major tributaries(upper and lower AfonHafren and upper and lower Afon Hore) of the upper River Severn catchment. N.b. the Be concentration range plotted was keptthe same for all the sites represented for comparative purposes. In the case of cloud water, one anomalously high point(1.47 mgl ) has been excluded. Note also that the lowest quotable value for individual data points is 0.05mg l . This corresponds to any1 y1

area"0.05mg l about thex-axis.y1

concentration in 1997, but the initial peakobserved in rainfall late in 1995 was not clearlyrepresented. However, there was generally ahigher scatter in data for cloud water comparedto rainfall and the pattern of increase with timewas less well marked for cloud water. As themonitoring period for cloud water was from1992 onwards whereas the rainfall data coversan extended early period of record where con-centrations are low(1983 to 1992), the differ-ence in concentrations and ranges may in partreflect the timing differences.

● Throughfall and stemflow: Throughfall andstemflow have concentrations similar to thosein rainfall. The mean, median, maximum andflow weighted mean for throughfall was 0.01,0.01, 0.31 and 0.01mg l , respectively. They1

corresponding values in stemflow are moderate-ly higher at 0.02, 0.02, 0.17 and 0.02mg l .y1

As with cloud water, there was an issue ofsampling timing as the stemflow and throughfalldata was collected during the early part of therecord when Be concentrations were particularlylow in rainfall. Thus, the stemflow and through-fall was enriched relative to rainfall during theearly period of record, but it is unclear if thereis enrichment with respect to cloud water aswell since data on cloud water were missingfrom the early part of the record.

6.2. Beryllium in stream waters

The levels of Be in the streams were relativelyhigh compared to rainfall(a factor of two to threetimes higher, although in absolute terms the con-centrations are still low). For the larger streams,the mean varies from stream to stream in the range0.04–0.07mg l with an average for all they1


Fig. 2. Temporal variations in beryllium concentration for intermediate(Nant Tanllwyth) and small(south east 1 and 2 and Tanllwyth1 and 2) streams of the upper River Severn catchment. N.b. the Be concentration range plotted was kept the same for all the sitesrepresented for comparative purposes. In the case of South2Hore, one anomalously high point(1.14 mg l ) has been excluded.y1

Note also that the lowest quotable value for individual data points is 0.05mg l . This corresponds to an area"0.05mg l abouty1 y1

the x-axis.

streams of 0.06mg l . Correspondingly, for they1

smaller streams the mean concentrations are slight-ly higher (range 0.06–0.11mg l , with an aver-y1

age across the streams of 0.09mg l ). There isy1

no clear reason why there are differences from siteto site. However, as with the atmospheric inputsand canopy information, there was a difference insampling period and the data for the smallerstreams come from a later period of record whenatmospheric inputs of Be are higher. This differ-ence in timing may well account for their highermean and flow weighted mean values comparedwith the larger streams. The lowest concentrationsoccurring within the streams are for the upperHafren and the difference cannot be accounted forby timing difference.

The longest stream water data records relate tothe larger ones, the upper and lower Hafren andHore. There are two main features.

Firstly, from 1983 up to late 1995, there was aprogressive decline in Be concentration althoughthe data were highly scattered. This decline isillustrated for the two longest datasets available(the lower Hafren and the lower Hore) using linearregression methodologies. As shown from earlierstudies, the beryllium concentrations are related toflow with highest concentrations occurring underhighflow conditions: this factor accounts in largepart for the high degree of scatter to the datawithin each year. Studies by Kirchner et al.(1993)have shown that the flow effect can be removedfrom trend analysis using multiple regressionswhen the flow term was allowed for using thelogarithm of flow as one of the independentvariables. Using this methodology, multiple regres-sion analysis was based on:

w xBe sa*timeqb*log (flow)qc10


Fig. 3. Temporal variations in beryllium concentration for groundwater in the upper River Severn catchment. Note also that thelowest quotable value for individual data points is 0.05mg l . This corresponds to an area"0.05mg l about thex-axis.y1 y1

For this relationship,wBex is beryllium concen-tration in mg l units, time is in calendar yearsy1

and flow is in mm 15 min , while ‘a’, ‘ b’ andy1

‘c’ are regression constants. Application of thisregression to the lower Hafren reveals:

w xBe sy0.0019"0.0005*timeq0.050"0.004*log (flow)q3.90"0.00410

For this regression,r s0.569 andNs628: the2

" sign represents twice the standard error. Corre-spondingly, for the lower Hore,

w xBe sy0.0029"0.0006*timeq0.069"0.004*log (flow)q5.85"0.05510

For this regression,r s0.620 andNs630. For2

both these equations, there is high statistical sig-nificance for the three regression coefficients.Thus, Be concentrations decline in a statisticallysignificant manner over time while concentrationsincrease as flow increases.

Secondly, concentrations in all the streamsincreased around the same time as rainfall concen-trations increased. The increase started with ashort-lived peak(approx. 0.3–0.5mg l ) at they1

end of November 1995. A more sustained increaseoccurred around mid-August 1997(from s0.05mg l to approximately 0.15–0.25mg l ,y1 y1

depending upon catchment). This increase declinedby 20–50% by the end of the monitoring periodlate in 1998. The increase was higher than that inrainfall.

6.3. Beryllium in groundwater

On average, beryllium concentrations in thegroundwaters are approximately twice those in thestreams(the mean concentration across the sitesvaries between 0.06 and 0.28mg l with a nety1

mean of 0.14mg l ). The temporal changes iny1

Be concentration match those observed within therainfall and the streams. Thus, the groundwaters


have a short-lived peak(approx. 0.30–1.60mgl ) at the end of November 1995 and a morey1

sustained increase starting mid-August 1997(fromapprox. the lowest quotable value, 0.05 g l , toy1

within a range of 0.15–0.40mg l , dependingy1

upon the particular borehole). Unlike the streamwaters, there was no evidence of a decline towardsthe end of the monitoring period late in 1998. Theincreases observed within the groundwaters arehigher than those occurring within the streams.

6.4. Beryllium relationships with other elementswithin stream and groundwaters

As shown by Neal et al.(1992), for the firsthalf of the record presented here, Be concentrationswithin the upper River Severn streams are highlycorrelated with aluminium(Al), pH, alkalinity andflow. They explained the relationships in terms ofa two-component mixing model. Under baseflowconditions, the stream waters reflect groundwatersources of relatively high pH and alkalinity butlow Al and Be concentrations while stormflow ismainly derived from the acidic soils where pH andalkalinity is low and Al and Be are mobilised.Thus, Be concentrations are positively correlatedwith Al concentration and flow but negativelycorrelated with pH and alkalinity. Fig. 4 illustratesthe relationship observed for the streams using allthe data now available. For the figure, only thelower Hafren is represented to avoid repetition asall the streams show similar patterns and they arealso similar to those presented by Neal et al.(1992) except for four outlier points at fairly lowflows where Be concentrations are approximatelytwice those normally found. These outlier pointscorrespond to the short-lived peak at the end ofNovember 1995.

For the groundwaters, there were no data avail-able for the Neal et al.(1992) analysis, but itmight be expected that similar patterns should beobserved for the streams allowing for the fact thatthe streamwater chemistry would be expected tobe nearer to that for baseflow chemistry than thatfor stormflow. The patterns observed are illustratedin Fig. 5 where Be is plotted against Al, pH,alkalinity and level. The salient features of thegroundwater chemistry are:

● pH was relatively high in the groundwater witha range of 5.5–6.8 compared with the stream at4.2–6.8. Indeed, the pH of the groundwaterwould be even higher if carbon dioxide wasallowed to degas prior to determination(unlikethe streams which are around atmospheric pres-sure, the groundwaters are oversaturated withrespect to carbon dioxide by up to a 100-fold).

● Alkalinity was relatively high in the ground-water with a range of 50–700mEq l com-y1

pared to a range ofy70–60 mEq l in they1

stream. The groundwater alkalinity was evenhigher than in the lower Hafren during baseflowconditions and this reflects a higher degree ofweathering which can be variable through theupper Severn catchment.

Despite the correlations observed within thestreams, there were no clear Be relationships withAl, pH alkalinity and level comparable to thecorresponding patterns for the streams. However,the two highest Be concentrations correspond tothe highest levels of Al. This feature implies thatacidic Al and Be bearing soil water has enteredthe system and although there has been time forthe alkalinity and pH to increase, Al and Be havenot had time to do so. N.B. there is ample evidencethat ‘slugs’ of soil water can pass through thegroundwater system in the upper Severn as shownby Neal et al.(1997c).

7. Discussion

Beryllium concentrations in rainfall are low inthe upper Severn as becomes an area far fromindustrial sources of contamination. The flowweighted mean concentration of Be in rainfall was0.02mg l and this is less than a tenth the valuey1

found in polluted areas(e.g. 0.22mg l in they1

most polluted area of the Czech Republic for1985–1989; Vesely et al., 2002). Correspondingly,Be concentrations in the upper Severn streams arealso relatively low with a mean of approximately0.07 mg l with maximum concentrations acrossy1

the streams typically approximately 0.3–0.4mgl (1.14 mg l is the highest concentrationy1 y1


Fig. 4. Plots of beryllium against aluminium, pH, alkalinity and flow for the lower Hafren. Note that the lowest quotable value forindividual data points is 0.05mg l . This corresponds to an area"0.05mg l about thex-axis.y1 y1

observed for one of the streams). These values arevery similar to other rural und upland areas of theUK where Be rich bedrock and industrialyurbansources are absent(Edmunds and Trafford, 1993;Neal, 2003). The Be concentrations in the upperSevern are typically approximately a fiftieth ofthose in an acidified stream in the Czech Republic(range of 2–24mg l ) draining granite with highy1

fluoride concentrations of mean 490mg-F ly1

(Vesely et al., 2002): the mean fluoride concentra-tion for the Upper Severn is 50mg-F l (Neal ety1

al., 1997a).

Beryllium concentrations for the upper Severnstreams increased with decreasing pH and increas-ing Al in relation to increased inputs of water fromthe acidic soil zone where both Al and Be can bemobilised from the soil matrix. The lowest levelsof Be occurred for the upper Hafren and it is herewhere acidity is the lowest and there is a reducedleaching of Al (and hence Be). However, the Beconcentrations increased towards the end of thestudy period starting late in November 1995. Thisincrease was broadly at the time when anomalouslyhigh Be concentrations were encountered within


Fig. 5. Plots of beryllium against aluminium, pH, alkalinity and flow for the Quarry borehole. Note that the lowest quotable valuefor individual data points is 0.05mg l . This corresponds to an area"0.05mg l about thex-axis.y1 y1

lowland UK rivers. However, on closer inspection,the upper Severn data do not tie in with thelowland observations as

● The Be increase occurred later in the year inthe upper River Severn.

● The Be increase did not correspond to very lowflows.

● The Be increase occurred under acidic condi-tions (pH 5.1–5.6).

● The Be increase did not simply return to back-ground levels and remained there: there was a

subsequent increase to equivalent levels in1997–1998.

● The Be increase in the lowland rivers was linkedto urbanyindustrial sources while the upperSevern area is a considerable distance fromsuch sources.Rather, it seems that towards the end of the

study period there were increased atmosphericinputs of Be that increased the levels in the streamin the upper Severn area.

With regards to the groundwaters, Be concentra-tions are higher than in rainfall and stream waters,


but they are still low averaging only 0.14 with amaximum concentration of 1.56mg l . There arey1

increases in Be concentrations towards the end ofthe study period(1995–1998) in line with thoseobserved in the streams although there is nocorresponding relationship with pH and alkalinity.These Be concentrations compare with values typ-ically less than 0.04mg l for shallow drilledy1

wells in England(Edmunds and Trafford, 1993)but the values are two to four orders of magnitudelower than for acid waters associated with granitesand deeper groundwaters enriched in fluoride(Vesely et al., 2002).

While higher Be concentrations occur in ground-water compared to stream water for the upperSevern, this would not be expected due to thehigher pHs involved and the reduced potential forsolubilizationyprecipitation. However, set againstthis, Be has the potential to occur in micro-particulate form and to pass through filter papersgiving the impression of higher truly dissolvedconcentrations than there actually is. This featuremay account, in part, for the relatively high Aland Be concentrations for two sample points inupper Severn groundwaters: both Be and Al havevery similar hydrogeochemistry in that they canbe present in micro particulate form, they canprecipitate under pH neutral conditions and Be canco-precipitate with Al.

8. Conclusion

The hydrochemical study for the upper RiverSevern reveals Be concentrations that are relativelylow in rainfall, runoff and groundwater—as wouldbe anticipated for an area with low pollutant inputsand low Be content within the bedrock. As theupper Severn system is acidic and acid sensitive,Be shows increased concentrations during highflows when conditions are at their most acidic andwhere the potential for Be mobilisation from thesoil is at its greatest. Aluminium concentrations inthe stream respond not only to hydrology, but therealso seems to be an atmospheric component thathas increased during the latter part of the study.Nonetheless, the concentrations of Be within theatmospheric input, stream runoff and groundwater

are sufficiently low not to be of environmentalconcern.

The findings presented here and in an earlierstudy by Neal et al.(1992) provide a detailedaccount of Be transfer through the upper Severncatchment and this resource represents an upland‘endmember case’ to set against data for otherparts of the UK and elsewhere. The study indicatesthat Be mobilisation cannot simply be related toacid mobilisation and micro-particulate transportmay also play an important role under less acidicconditions such as those occurring within ground-water areas in the uplands and the lowland rivers.The role of micro-particulate in element flux trans-fers through the aquatic environment requires fur-ther study not just for Be but for many transitionmetals as well(Neal et al., 1997d). Longer termmonitoring is required to see if the increases in Beconcentration in rainfall and runoff are maintained.

References

Edmunds WM, Smedley Pl. Groundwater geochemistry andhealth: an overview. Environ Geochem Health Geol Soc(Lond) Special Pub 1996;113:91–105.

Edmunds WM, Trafford JM. Beryllium in river baseflow,shallow groundwaters and major aquifers of the UK. ApplGeochem Suppl Issue 1993;2:223–233.

Haines, T Jagoe, CH, Matey, V, 1991. Gill histopathology andtoxicology of beryllium in acid water to perch(PercaFlaviatilis and P. Flavescens). 47th Northeast Fish andWildlife Conf, Portland Or., p 8.

Heath RC, Miller, LM, Perry, CM, Norton, SA, 1988. Beryl-lium (Be) in surface water sources, sinks, mobilisation andpotential toxicity. International conference on metals inlakes, Hamilton Ont., p 35.

Hudson JA, Crane SB, Blackie JR. The Plynlimon waterbalance 1969–1995: the impact of forest and moorlandvegetation on evaporation and streamflow in upland catch-ments. Earth Sys Sci 1997;1(3):409–427.

Kaste, JM, 1999. Dynamics of cosmogenic beryllium-7 andbedrock-derived beryllium-9 in forested ecosystems inMaine, USA. Unpub. M. Sci. Thesis, University of Maine,96 pp.

Kirchner JW, Dillon PJ, Lazerte BD. Separating hydrologicaland geochemical influences on runoff in spatially heteroge-neous catchments. Wat Resour Res 1993;29:3903–3916.

Neal C, Jeffery HA, Conway T, Ryland GP, Smith CJ, NealM, Norton SA. Beryllium concentrations in rainfall, stem-flow, throughflow, mist and stream waters for an uplandacidified area of mid-Wales. J Hydrol 1992;136:33–49.

Neal C, Wilkinson J, Neal M, Harrow M, Wickham H, HillL, Morfitt C. The hydrochemistry of the River Severn,


Plynlimon, mid-Wales. Hydrol Earth Syst Sci 1997;1(3):583–618.

Neal C, Robson AJ, Shand P, Edmunds WM, Dixon AJ,Buckley DK, Hill S, Harrow M, Neal M, Reynolds B. Theoccurrence of groundwater in the Lower Palaeozoic rocksof upland Central Wales. Hydrol Earth Syst Sci1997;1(1):3–18.

Neal C, Hill T, Alexander S, Reynolds B, Hill S, Dixon AJ,Harrow M, Neal M, Smith CJ. Stream water quality in acidsensitive upland areas, an example of potential water qualityremediation based on groundwater manipulation. HydrolEarth Syst Sci 1997;1(1):185–196.

Neal C, Robson AJ, Jeffery HA, Harrow M, Neal M, SmithCJ, Jarvie HP. Trace element inter-relationships for theHumber rivers: inferences for hydrological and chemicalcontrols. Sci Total Environ 1997;194y195:321–344.

Neal, C, 2003. Dissolved and particulate beryllium in easternUK surface waters. Sci Total Environ, 2003;314–316:185–208.

Neal, C, Reynolds, B, Neal, M, Hill, L, Wickham, H, Pugh,B, 2003, Nitrogen in rainfall, cloud water, throughfall,stemflow, stream water and groundwater for the Plynlimoncatchments of mid-Wales. Sci Total Environ, 2003;314–316:121–151.

Norton S, Kahl J, Ferandez I, Haines T, Rustad L, Nodvin S,Scofield J, Strickland T, Erickson H, Wiggington P, Lee J.The Bear Brook Watershed, Maine(BBWM), USA. EnvironMonit Asssessment 1999;55:7–51.

Reeves AL. Beryllium. In: Friberg L, editor. Handbook on thetoxicology of metals. Amsterdam: Elsevier, 1979. p. 329–343.

Vesely J, Majer V. The effect of pH and atmospheric depositionon concentrations of trace elements in acidified freshwaters:a statistical approach. Water Air Soil Poll 1996;88:227–246.

Vesely J, Norton, SA, Skrivan, P, Majer, V, Kram, P, Navratil,T, Kaste, JM, 2002. Environmental chemistry of Beryllium.Min Soc Am, in press.

dissolved beryllium in rainfall, stream and shallow groundwaters in the upper river severn...

Documents