WR S OD Q N WR Q D Q G LQ Y H UWH E UD WH D V V H P E OD … · (sensu Lake 2000), often results in distinctive and persistent changes in biological assemblages (Økland and Økland

Tracing recovery under changing climate: response ofphytoplankton and invertebrate assemblages to

decreased acidification

Richard K. Johnson1AND David G. Angeler2

Department of Aquatic Sciences and Assessment, Swedish University of Agricultural Sciences, P.O. Box7050, SE-750 07, Uppsala, Sweden

Abstract. Phytoplankton and littoral invertebrate assemblages in 4 boreal lakes recovering fromacidification and 4 minimally disturbed reference lakes studied over 2 decades were used to determinethe pathways and trajectories of change under the influence of climatic variability. Assemblagecomposition (species presence–absence data) but not dominance patterns (invertebrate abundance/phytoplankton biovolume) of acidified lakes became more similar to those of reference lakes (distancedecreased with time), indicating that detection of recovery varies as a function of chosen metrics. Acidifiedlakes had more pronounced shifts in assemblage composition than did reference lakes. The most markeddifferences were noted for phytoplankton assemblages. Assemblages in acidified lakes had mean between-year Euclidean distances almost 23 greater than those of assemblages in reference lakes. Trends in waterchemistry showed unequivocal recovery, but responses of phytoplankton and invertebrate assemblages,measured as between-year shifts in assemblage composition, were correlated with interannual variabilityin climate (e.g., North Atlantic Oscillation, water temperature) in addition to decreased acidity. The findingthat recovery pathways and trajectories of individual acidified lakes and the environmental driversexplaining these changes differed among assemblages shows that biological recovery is complex and theinfluence of climatic variability is poorly understood.

Key words: acidification, recovery, phytoplankton, benthic invertebrates, monitoring, boreal lakes,climate.

Acidification of inland waters has been recognizedas a major environmental problem for .3 decades.Awareness that emissions from burning of fossil fuelswere causing biodiversity loss in lakes and streamsled to international action plans to protect and restorenatural resources (Stoddard et al. 1999). Ensuingreductions in the emissions of sulfur dioxide andNOx compounds resulted in improved air quality,and subsequent increases in surface-water pH andalkalinity have been attributed to the decreaseddeposition of these acidifying compounds (e.g.,Stoddard et al. 1999, Skjelkvale et al. 2005). However,whereas a growing body of literature supportschemical recovery of surface waters, empirical evi-dence of biological recovery is scarce and findings areequivocal (Skjelkvale et al. 2003, Stendera andJohnson 2008, Ormerod and Durance 2009). Further-more, because biological recovery is often the ultimate

goal of legislative action, failure to achieve biologicalobjectives means that acidification is still considered aforemost problem affecting the biodiversity of inlandsurface waters in northern Europe (Johnson et al.2003) and elsewhere (e.g., Monteith et al. 2005,Kowalik et al. 2007, Burns et al. 2008).

Acidification, a prolonged press type of disturbance(sensu Lake 2000), often results in distinctive andpersistent changes in biological assemblages (Øklandand Økland 1986). Response to disturbance often isstabilized by ecological feedback mechanisms result-ing in resilience (e.g., Blindow et al. 2002), andgrowing evidence indicates that perturbed systemsresist returning to the predisturbed state. For exam-ple, acidified sites can be maintained in the impairedstate by inadequate physicochemical properties (e.g.,acid episodes can retard recovery, Kowalik et al. 2007)and biological interactions (Ledger and Hildrew 2005,Monteith et al. 2005). In addition to resistance tochange, ecosystems can undergo transitions betweenstates (regime shifts) induced by interactions between

1 E-mail addresses: [email protected] [email protected]

J. N. Am. Benthol. Soc., 2010, 29(4):1472–1490’ 2010 by The North American Benthological SocietyDOI: 10.1899/09-171.1Published online: 26 October 2010

1472

internal ecosystem processes and external fluctuations(Scheffer and Carpenter 2003). Scheffer and Carpenter(2003) used the expression catastrophic regime shifts forirreversible ecosystem changes. Hence, a number ofinternal and external factors, such as habitat quality,connectivity, dispersal abilities, and species interac-tions, can individually or collectively affect the ratesand trajectories of biological recovery (Yan et al. 2003).

Climatic variability also influences the recovery ofstream invertebrate assemblages. Wet winters in-crease acidity in Welsh (UK) moorland and foreststreams and reset recovery (Ormerod and Durance2009). Positive (i.e., warmer, wetter) North AtlanticOscillation phases were accompanied by reducedinterannual stability (=similarity) in invertebrateassemblages (Durance and Ormerod 2007). Circum-neutral streams were more sensitive to climaticvariability than were acidified streams, a resultsuggesting that acidification stress could override orinteract complexly with the stress arising from globalwarming (Christensen et al. 2006, Durance andOrmerod 2007). Thus, understanding the factorsimportant for ecosystem recovery is difficult. Long-term data can be valuable for disentangling many ofthe factors driving or impeding recovery, and thesetypes of data are of particular interest for determiningthe pathways and trajectories of systems undergoingrecovery under the constraints set by climatic vari-ability.

We used water-chemistry, phytoplankton, andlittoral benthic invertebrate assemblage data from 4minimally disturbed (reference) and 4 acidified lakesstudied during the last 2 decades to track recoverypathways from acidification. We tested 4 specifichypotheses: 1) acidified lakes show chemical andbiological recovery as indicated by increased pH anddiversity (e.g., biological assemblages of acidifiedlakes are more dissimilar to those of reference sitesat the start than at the end of the study period, anddissimilarity decreases with time), 2) trajectories ofbiological recovery vary between assemblages, 3)long-term assemblage trends, indicated by between-year shifts in assemblage composition (measured asEuclidean distance), are related to decreased acidity,and 4) climatic variability subsumes the recoverysignal from acidification.

Methods

Study area

In the late 1980s, Sweden initiated long-termmonitoring of multiple habitats and trophic levels oflakes to follow the effects of acidification and recoveryof regionally representative lake ecosystems (Johnson

1999). Two lake categories were monitored: 1)minimally disturbed reference lakes (sensu Stoddardet al. 2006) to determine natural, interannual changesand 2) acidified lakes (defined as exceedence ofcritical load of S . 0) to assess the degree ofdegradation and subsequent recovery.

Samples for analysis of water chemistry andphytoplankton and littoral invertebrate assemblageshave been collected during the last 21 y (1988–2008)from 8 lakes in the boreonemoral or mixed forestecoregion of southern Sweden (Fig. 1). Four lakeswith a mean annual minimum pH . 6.2 wereconsidered minimally disturbed reference lakes, and4 lakes with mean annual maximum pH , 5.8 werejudged as being acidified based on estimates ofexchange of preindustrial alkalinity for anthropogenicSO4

22 (Persson 1996). All 8 lakes are relatively small(mean lake surface area = 0.47 6 0.52 km2), shallow(mean depth = 5.8 6 2.5 m), and nutrient poor (meanannual PO4-P concentration ,6.5 mg/L) (Appendix 1).Catchment land use/cover also is similar betweenreference and acidified lakes (% forest: mean = 66%,range = 54–85%; % mire mean = 11%, range = 4–20%). More information on these lakes is availablefrom: http://www.ma.slu.se.

Sampling

Surface-water samples (0.5 m) were collected 6 to 8times a year (monthly during the open-water phaseand once in February) at a mid-lake station in eachlake. Water was collected with a PlexiglasH samplerand kept cool during transport to the laboratory.Samples were analyzed for variables indicative ofacidity (pH and SO4

22 concentration), nutrients (PO4-P), and water color (absorbance at 420 nm of filteredwater). All physicochemical analyses were done at theDepartment of Aquatic Sciences and Assessmentfollowing international (ISO) or European (EN)standards when available (Wilander et al. 2003).Annual mean values were used in statistical analysesto down-weight seasonal differences.

Phytoplankton was sampled in August of each yearby taking a water sample from the epilimnion (0–4 m)with a Plexiglas tube sampler (diameter = 3 cm). Inlakes with a surface area .1 km2 a single mid-lake sitewas used for sampling. In lakes with a surface area,1 km2, 5 random epilimnetic water samples weretaken and mixed to form a composite sample fromwhich a subsample was taken and preserved withLugol’s iodine solution (2 g potassium iodide and 1 giodide in 100 mL distilled water) supplemented withacetic acid. Phytoplankton counts were made using aninverted light microscope and the modified Utermohl

2010] RECOVERY PATHWAYS IN BOREAL LAKES 1473

technique commonly used in the Nordic countries(Olrik et al. 1989). Taxa were identified to the lowesttaxonomic unit possible (usually species), and spe-cies-specific biovolume measures were calculatedfrom geometric formulae. In 1992, enumeration ofphytoplankton changed from counts of predominat-ing taxa to counts of all species in a sample. Therefore,

phytoplankton counts made before 1992 are notincluded here.

Benthic invertebrates were collected from wind-exposed, vegetation-free littoral habitats in lateautumn (October–November) each year. Five repli-cate samples were taken by standardized kicksampling with a hand net (0.5-mm mesh size). Each

FIG. 1. Map of Sweden showing the 8 study lakes and 3 main ecoregions: 14 (Central Plains), 22 (Scandinavian Shield), and 20(Boreal Highlands) according to Illies (1966).

1474 R. K. JOHNSON AND D. G. ANGELER [Volume 29

sample was taken by disturbing the bottom substra-tum for 20 s along a 1-m-long stretch of the littoralregion at a depth of ,0.5 m. Samples were preservedin 70% ethanol in the field and processed in thelaboratory by sorting against a white backgroundwith 103 magnification. Invertebrates were identifiedto the lowest taxonomic unit possible (generally tospecies level) and counted with dissecting and lightmicroscopes.

Climatic data

The effect of interannual variability in climate onphytoplankton and invertebrate assemblages wastested using the North Atlantic Oscillation (NAO)winter index (NAOwinter; December–March, providedby Climate Analysis Section, National Center forAtmospheric Research, Boulder, Colorado; http://www.cgd.ucar.edu/cas/jhurrell/nao.stat.winter.html;Hurrell 1995). The NAOwinter index is calculated fromthe difference in sea surface pressure between theAzores and Iceland. Positive values are associatedwith mild, wet winters in northwestern Europe andnegative values are associated with cold, dry winters(Hurrell 1995). During the period 1988 to 1995, theNAOwinter index averaged +2.97 (min: +0.72 for 1988,max: +5.08 for 1989), compared to +0.32 for 1996 to2008 (min = 23.78 for 1996, max = +2.80 for 2007).For southern Sweden, the NAOwinter index shows aswitch from warm to cold winters between 1995 and1996 (NAOwinter decreased from +3.96 to 23.78,respectively).

Data analysis

Biological response variables.—Four measures ofdiversity and abundance, derived from the species3 site data sets, were calculated using the PASTsoftware (PAST - PAlaeontological STatistics, version1.73; available from http://folk.uio.no/ohammer/past/index.html) to determine if phytoplankton andinvertebrate assemblages are recovering from acidifi-cation: 1) total number of individuals (invertebrates)or biovolume (phytoplankton) in a sample, 2) totalnumber of taxa in a sample (taxon richness), 3)Simpson diversity (Simpson 1949), and 4) evenness,calculated as the logarithm of Shannon diversity (H;Shannon and Weaver 1949) divided by the number oftaxa (eH/S). In addition, 2 measures of assemblagecomposition were calculated using nonmetric multi-dimensional scaling (NMDS) on presence–absenceand abundance (invertebrate) or biovolume (phyto-plankton) data in PAST 1.73. NMDS was carried outon a similarity matrix built on presence–absence data(Sørensen similarity) and a dissimilarity matrix built

on log(x)-transformed biovolume (phytoplankton)and abundance (invertebrate) (Bray–Curtis dissimi-larity). Measures of both similarity/dissimilarity arerecommended because they ignore joint absences, acombination that predominates in site 3 speciesmatrices (Faith et al. 1987). Moreover, presence–absence data gives more weight to species with areduced distribution, whereas abundance/biovolumedata is more recommended for use where many taxaare widely distributed (Rabeni et al. 2002) (i.e.,presence–absence data distinguished differences inspecies composition without considering the abun-dance/biovolume of each species).

Functional groups also were determined. Phyto-plankton was divided into autotrophic, mixotrophic,and heterotrophic biomass groups based on theclassification scheme of Jansson et al. (1996). Bacillar-iophyceae, Conjugatophyceae, Cryptophyceae, Cyano-phyceae, Loxophyceae, Prasinophyceae, Xanthophy-ceae, and Chlorophyceae were considered to befunctionally autotrophic, except the green algae Poly-toma and Polytomella (mixotrophic taxa). Chrysophy-ceae, Craspedophyceae, Dinophyceae, Euglenophyceae,Haptophyceae, and Raphidophyceae were regarded asmixotrophic. Taxa were considered heterotrophic whenthey contained no photosynthetic apparatus (e.g.,euglenozoan flagellates). Functional feeding groupsfor invertebrates (gatherers/collectors, parasites, xy-lophagous taxa, predators, miners, active and passivefilter feeders, shredders and grazers/scrapers) wereassigned with Asterics 3.1.1. (IRV Software, Vienna,Austria; http://www.fliessgewaesserbewertung.de/en/download/berechnung/).

Analysis of trends

Water quality and temperature.—Regression was usedto test the conjecture that water-chemistry variablesare responding to decreased emissions of acidifyingcompounds (restoration from acidification). Prelimi-nary analyses showed that several chemical variableswere autocorrelated with time (year of sampling).Therefore, relationships between response variablesand time were tested using Kendall’s t rank correla-tion coefficient (Kendall 1938), a nonparametric test ofconcordance (agreement between 2 rankings). Thesame test was used to evaluate the temporal trends ofwater temperature in the lakes. The Wilcoxon signed-rank test was used to detect differences in physico-chemical variables between acidified and referencelakes (Wilcoxon 1945). Pearson product–momentcorrelation analyses were used to determine therelationship between the NAOwinter index and thewater-quality variables.


Biological variables.—Structural and functional met-rics of phytoplankton and invertebrates were log(x)-transformed when necessary to improve normalityand homoscedasticity before analyses with a mixed-model, repeated-measures analysis of variance (AN-OVA) in Statistica (version 5; Statsoft, Inc., Tulsa,Oklahoma). The model contained treatment (acidifiedlakes vs reference lakes; fixed factor), year (samplingyears between 1988 and 2008; fixed factor), and lakesnested within treatment (lakes[treatment]; randomfactor). Variance components were calculated for eachterm in the model to provide information beyond thesignificance level about the variation contributed byeach term. The statistical design for ANOVA showingfactors, multipliers, and formulas to determine esti-mated mean squares, variance components, and F-ratio tests are shown in Appendix 2.

The key term in this ANOVA was the interactiontreatment 3 year which allowed us to test whether allacidified lakes converged with control lakes over thestudy period, a result that would indicate consistentrecovery patterns. To avoid detection of false posi-tives, significance levels were corrected using theprocedure by Holm (1979), which is less conservativethan the sequential Bonferroni correction.

Interannual changes in assemblage composition

NMDS was used to condense the species 3 sitedata sets into 2 measures (1st and 2nd axis scores)representing lake-specific between-year and among-lake similarities/dissimilarities in assemblage com-position. NMDS axis scores for phytoplankton andinvertebrate assemblages were calculated using Sør-ensen similarity and Bray–Curtis dissimilarity asdescribed above. A distance matrix based on theindividual lake–year NMDS results was calculated asthe Euclidean distance (![(xji – xi1)2 + (xi2 – xj2)2])between the NMDS axis 1 and 2 scores. Between-yearsimilarity/dissimilarity for pairs of successive yearsfor each lake were extracted from the Euclideandistance matrix and used to quantify the degree ofdissimilarity of assemblage composition betweensuccessive sampling years for the individual lakesand for the lake groups (acidified and referencelakes).

Recovery—movement towards reference conditions

The mean coordinates of individual lake–yearcombinations of the reference sites was used as ameasure of the reference condition and to test theassumption that biological assemblages of acidifiedlakes became less dissimilar to those of reference lakeswith time. The reference condition for phytoplankton

and invertebrate assemblages was estimated as themean of the NMDS axis 1 and 2 scores for all 4reference lakes. The Euclidean distance betweenindividual lake–year combinations of acidified andreference lakes and the mean of the reference lakeswas extracted from a distance matrix of NMDS axis 1and 2 scores calculated with Sørensen similarity andBray–Curtis dissimilarity as described above. AWilcoxon signed-rank test was used to test ifbetween-year distance was lower for acidified thanreference lakes. Kendall concordance of Euclideandistance of lake–year combinations to the referencepopulation was used to determine if distance de-creased in acidified lakes over time.

Environmental correlates of between-year shifts inassemblage composition

The Euclidean distance between individual lake–year combinations of acidified and reference lakeswas extracted from a distance matrix of NMDS axis 1and 2 scores calculated with Sørensen similarity andBray–Curtis dissimilarity as described above. Spear-man rank correlation analysis was used to determinethe associations between shifts (between-year Euclid-ean distance) in phytoplankton and littoral inverte-brate assemblages and water-chemistry and theNAOwinter index across lakes.

Results

Trends in surface-water chemistry

Acidified lakes had lower pH than reference lakes(Wilcoxon signed-rank test, p , 0.001). Mean annualpH was 5.3 6 0.41 for the 4 acidified compared to 6.66 0.17 for the 4 reference lakes (Appendix 1). Meansurface-water pH increased significantly with time in5 of the 8 lakes during the 21-y period (Appendix 1).However, acidic episodes occurred during the recov-ery period. Annual minimum pH was ,5.0 (mean: 4.46 0.41) in the acidified lakes compared to .6.0 (mean:6.2 6 0.17) for the reference lakes. One lake(Harsvatten) has had pH , 5.0 during the 21-y ofstudy, but the other 3 acidified lakes also haveexperienced acidic episodes. Annual mean pH , 5.5was recorded in 8 (Ovre Skarsjon) to 15 (Rotehogst-jarnen) of the 21 study years. Significant decreases inSO4 concentration were noted for all 8 of the lakes.Likewise, increasing trends in lake water color werenoted for all 8 of the lakes, with acidified lakes havinghigher water color. No temporal trends or differenceswere noted between acidified and reference lakes forPO4-P concentration.


Surface water temperatures increased significantlyonly in the reference lakes Fracksjon and StoraSkarsjon (Appendix 1). A common pattern arisingfrom the Pearson correlation analysis was that trendsin water temperature were independent of theNAOwinter index (p . 0.05). However, correlations ofthe NAO with physicochemical variables were lakespecific, and no common patterns were revealed foreither reference or acidified lakes (Appendix 3).

Patterns of biological recovery

Recovery patterns based on both structural andfunctional metrics of phytoplankton (Fig. 2A–I) andinvertebrates (Fig. 3A–J) were highly variable amongacidified lakes.

Phytoplankton.—Species evenness (Fig. 2D) and bio-volume-based assemblage composition (NMDS 1,Bray–Curtis; Fig. 2F) showed the most consistentrecovery for Brunnsjon and Ovre Skarsjon. Thesemetrics were either within or close to the 95%

confidence interval reflecting target conditions inneutral lakes. In contrast, phytoplankton in lakeRotehogstjarnen was most resistant to recovery.Species richness (Fig. 2A), biovolume (Fig. 2B), Simp-son diversity (Fig. 2C), and species composition(NMDS, Sørensen; Fig. 2E) showed different degreesof variability approaching reference conditions. Re-covery trends of functional groups were highlyvariable and were characterized by periods wherethese groups showed intermittent approaches toreference conditions (Fig. 2G–I). As a result of thisstrong variability between and idiosyncratic respons-es within lakes, most comparisons in the ANOVAwere not significant, except the significant temporalvariation detected for biovolume-based assemblagecomposition (NMDS 1, Bray Curtis; Table 1). Noneof the metrics had a significant treatment 3 yearinteraction, i.e., no consistent recovery occurredacross all acidified lakes (Table 1).

Invertebrates.—Overall, the invertebrates showed abetter recovery response than the phytoplankton.Most structural metrics indicated that acidified lakeswere within or close to the 95% confidence intervalrepresenting reference conditions (Fig. 3A–D). Brunn-sjon and Ovre Skarsjon seemed to have a betterrecovery response, whereas Harsvatten had a poorerresponse. Both composition- and abundance-basedassemblage similarity showed variable degrees ofapproaching assemblage structure in reference lakes(Fig. 3E, F). Recovery responses among functionalfeeding groups were variable among lakes and overtime. Predators generally showed the weakest re-sponse, and only Brunnsjon or Rotehogstjarnen were

consistently within or close to reference conditions(Fig. 3J). These patterns also were reflected by thecollectors, although Harsvatten and Ovre Skarsjonwere close to reference conditions (Fig. 3I). Shreddersand grazers showed the most consistent recoverypatterns, reflected by an approach to referenceconditions during the 2nd ½ of the study period(significant year term in the ANOVA; Fig. 3G, H,Table 2). Only shredders showed a consistent recov-ery across acidified lakes, indicated by the significantinteraction term, which explained 82% of the varia-tion in the model.

Between-year shifts in assemblage composition

Mean between-year shifts (Euclidean distance) ofassemblage composition supported the idea thatchanges were stronger in acidified than in referencelakes (Appendix 4). Changes in phytoplankton andlittoral invertebrate assemblage composition weresignificant when measured with Sørensen similarity(presence–absence data) (Wilcoxon signed-rank test, p, 0.05), but differences were not significant whenbiovolume or abundance data were used (Bray–Curtisdissimilarity). The strongest differences were notedfor phytoplankton assemblages. Acidified lakes had amean Euclidean distance of almost 23 that ofreference lakes (acidified: 0.037 6 0.039, reference:0.019 6 0.016; Sørensen similarity). Between-yearshifts in littoral invertebrate assemblages were 0.0356 0.024 for acidified and 0.027 6 0.025 for referencelakes (Sørensen similarity).

Distance to reference conditions

Distance to the mean of reference lakes was onaverage 2.83 (invertebrates, Bray–Curtis) to 4.03

(phytoplankton, Sørensen) greater for individualacidified lakes than for individual reference lakes.On average, phytoplankton assemblages in acidifiedlakes were 0.134 6 0.044 (Sørensen similarity) to 0.1136 0.038 (Bray–Curtis dissimilarity) Euclidean distanceunits from the mean of the 4 reference lakes (NMDSaxis 1 and axis 2 scores) (Table 3). Similar meandistances were found for littoral invertebrate assem-blages (0.100 6 0.054 for Sørensen and 0.101 6 0.060for Bray–Curtis). Harsvatten was the furthest from thereference lakes (Euclidean distance .0.152). Between-year distance of individual lakes to mean coordinatesof the 4 reference lakes decreased significantly withtime in all 4 acidified lakes, indicating recovery interms of assemblage composition (Kendall concor-dance, p , 0.05) (Fig. 4A–H). However, interpretationof recovery was dependent on the taxonomic groupevaluated. For example, phytoplankton assemblages


FIG. 2. Temporal trends of phytoplankton taxon richness (A), biovolume (B), Simpson Index (C), evenness (D), nonmetricmultidimensional scaling (NMDS) axis 1 (Sørensen’s similarity) (E), NMDS axis 1 (Bray–Curtis dissimilarity) (F), proportionalbiomass of autotrophs (G), heterotrophs (H), and mixotrophs (I) in reference lakes (represented by the mean and 95% confidenceinterval of 4 lakes: Allgjuttern, Fracksjon, Stora Envattern, Stora Skarsjon) and individual acidified lakes (Brunnsjon, Harsvatten,Ovre Skarsjon, Rotehogstjarnen) between 1992 and 2008.


FIG. 3. Temporal trends of littoral invertebrate taxon richness (A), abundance (B), Simpson Index (C), evenness (D), nonmetricmultidimensional scaling (NMDS) axis 1 (Sørensen’s similarity) (E), NMDS axis 1 (Bray–Curtis dissimilarity) (F), and %

abundance of grazers (G), shredders (H), collectors (I), and predators (J) in reference lakes (represented by the mean and 95%

confidence interval of 4 lakes: Allgjuttern, Fracksjon, Stora Envattern, Stora Skarsjon), and individual acidified lakes (Brunnsjon,Harsvatten, Ovre Skarsjon, Rotehogstjarnen) between 1988 and 2008.


TA

BL

E1.

An

aly

sis

of

var

ian

cere

sult

ssh

ow

ing

mea

nsq

uar

es(M

S),

sou

rces

inv

aria

tio

n(F

rati

os)

,an

dv

aria

nce

com

po

nen

ts(V

C;

%)

for

stru

ctu

ral

and

fun

c-ti

on

alm

etri

cso

fp

hy

top

lan

kto

nas

sem

bla

ges

.S

ign

ific

ant

term

sar

eh

igh

lig

hte

din

bo

ld.

Th

ere

mai

nin

gte

rms

wer

en

ot

sig

nif

ican

taf

ter

Ho

lmco

rrec

tio

no

fp

lev

els.

NM

DS

=n

on

met

ric

mu

ltid

imen

sio

nal

scal

ing

.*

=p

,0.

05,

**=

p,

0.01

,**

*=

p,

0.00

1.

Var

iab

le

Bet

wee

nsa

mp

lev

aria

tio

nW

ith

insa

mp

lev

aria

tio

n–

rep

eate

dm

easu

res

Tre

atm

ent

(df

1,6)

Lak

e(tr

eatm

ent)

a

(df

6)Y

ear

(df

16,9

6)T

reat

men

t3

yea

r(d

f16

,96)

Yea

r3

Lak

e(tr

eatm

ent)

b

(df

96)

MS

FV

C(%

)M

SM

SF

VC

(%)

MS

FV

C(%

)M

S

Tax

on

rich

nes

s12

3691

80.

2510

049

0328

610

9748

1.28

2415

3752

1.79

7685

895

Bio

vo

lum

e4.

261.

9210

02.

230.

061.

934

0.07

2.01

960.

03S

imp

son

ind

ex0.

822.

7010

00.

300.

021.

355

0.02

1.40

950.

01E

ven

nes

s1.

571.

4810

01.

060.

041.

4917

0.05

1.84

830.

02N

MD

S1

(Sø

ren

sen

)14

.78

1.91

100

7.75

0.11

0.70

180.

120.

7982

0.15

NM

DS

1(B

ray

–Cu

rtis

)0.

018

0.18

100

0.10

50

.00

22

.61

***

76

0.00

10.

6924

0.00

1A

uto

tro

ph

bio

mas

s2.

102

9.42

100

0.22

30.

031.

1083

0.01

0.55

270.

02H

eter

otr

op

hb

iom

ass

0.00

40.

0410

00.

105

0.00

71.

7722

0.00

61.

4978

0.00

4M

ixo

tro

ph

bio

mas

s1.

605.

8710

00.

270.

021.

1238

0.01

0.52

620.

02

aN

ore

sid

ual

erro

rco

uld

be

calc

ula

ted

for

the

bet

wee

n-s

amp

lev

aria

tio

nin

the

mo

del

.T

his

term

was

use

das

the

erro

rte

rmfo

rth

esp

atia

lv

aria

tio

n;

no

F-r

atio

s,si

gn

ific

ance

val

ues

,an

dv

aria

nce

com

po

nen

tsar

eav

aila

ble

bN

ore

sid

ual

erro

rco

uld

be

calc

ula

ted

for

the

wit

hin

-sam

ple

var

iati

on

inth

em

od

el.T

his

term

was

use

das

the

erro

rte

rmfo

rth

ete

mp

ora

lv

aria

tio

n;n

oF

-rat

ios,

sig

nif

ican

cev

alu

es,

and

var

ian

ceco

mp

on

ents

are

avai

lab

le


TA

BL

E2.

An

aly

sis

of

var

ian

cere

sult

ssh

ow

ing

mea

nsq

uar

es(M

S),

sou

rces

inv

aria

tio

n(F

rati

os)

,an

dv

aria

nce

com

po

nen

ts(V

C;

%)

for

stru

ctu

ral

and

fun

c-ti

on

alm

etri

cso

fli

tto

ral

inv

erte

bra

teas

sem

bla

ges

.S

ign

ific

ant

term

sar

eh

igh

lig

hte

din

bo

ld.

Th

ere

mai

nin

gte

rms

wer

en

ot

sig

nif

ican

taf

ter

Ho

lmco

rrec

tio

no

fp

lev

els.

NM

DS

=n

on

met

ric

mu

ltid

imen

sio

nal

scal

ing

.*

=p

,0.

05,

**=

p,

0.01

,**

*=

p,

0.00

1.

Var

iab

le

Bet

wee

nsa

mp

lev

aria

tio

nW

ith

insa

mp

lev

aria

tio

n–

rep

eate

dm

easu

res

Tre

atm

ent

(df

1,6)

Lak

e(tr

eatm

ent)

a

(df

6)Y

ear

(df

19,1

14)

Tre

atm

ent

3y

ear

(df

19,1

14)

Yea

r3

lak

e(tr

eatm

ent)

b

(df

114)

MS

FV

C(%

)M

SM

SF

VC

(%)

MS

FV

C(%

)M

S

Tax

on

rich

nes

s58

084.

7510

012

222

89

5.1

6**

*9

759

.15

1.06

356

Ab

un

dan

ce0.

950.

6410

01.

480.

131.

672

0.13

1.70

980.

08S

imp

son

ind

ex0.

008

0.61

100

0.01

30.

003

1.25

910.

002

1.01

90.

002

Ev

enn

ess

0.22

1.06

100

0.21

0.02

0.95

430.

021.

0957

0.02

NM

DS

1(S

øre

nse

n)

0.25

76.

9610

00.

037

,0.

001

1.15

240.

001

1.39

76,

0.00

1N

MD

S1

(Bra

y–C

urt

is)

0.29

46.

0910

00.

048

0.00

11.

2011

0.00

11.

1689

,0.

001

Gra

zers

0.58

310

.19

100

0.05

70.

005

1.47

100.

005

1.38

900.

004

Min

ers

0.01

20.

9010

00.

014

0.00

11.

4594

0.00

10.

996

0.00

1S

hre

dd

ers

0.59

4.42

100

0.13

0.0

11

.96

*1

80

.01

2.7

5**

82

,0.

001

Gat

her

ers/

coll

ecto

rs0.

002

0.01

100

0.20

50

.01

72

.04

*9

20.

008

0.95

80.

008

Act

ive

filt

erfe

eder

s0.

007

0.16

100

0.04

60.

005

1.43

620.

003

0.81

380.

003

Pas

siv

efi

lter

feed

ers

,0.

001

0.02

100

0.01

40.

001

1.12

270.

002

1.43

730.

001

Pre

dat

ors

1.32

12.

7110

00.

488

0.01

11.

2643

0.00

40.

5157

0.00

9

aN

ore

sid

ual

erro

rco

uld

be

calc

ula

ted

for

the

bet

wee

n-s

amp

lev

aria

tio

nin

the

mo

del

.T

his

term

was

use

das

the

erro

rte

rmfo

rth

esp

atia

lv

aria

tio

n;

no

F-r

atio

s,si

gn

ific

ance

val

ues

,an

dv

aria

nce

com

po

nen

tsar

eav

aila

ble

bN

ore

sid

ual

erro

rco

uld

be

calc

ula

ted

for

the

wit

hin

-sam

ple

var

iati

on

inth

em

od

el.T

his

term

was

use

das

the

erro

rte

rmfo

rth

ete

mp

ora

lv

aria

tio

n;n

oF

-rat

ios,

sig

nif

ican

cev

alu

es,

and

var

ian

ceco

mp

on

ents

are

avai

lab

le


of Harsvatten and Rotehogstjarnen showed markedchanges over time (Harsvatten: Sørensen t = 20.441,p = 0.014, Bray–Curtis t = 20.706, p , 0.0001, Fig. 4F;Rotehogstjarnen: Sørensen t = 20.103, p = 0.564,Bray–Curtis t = 20.529, p = 0.003, Fig. 4H), whereasinvertebrate assemblage dissimilarity did not de-crease with time (Fig. 4B, D). In contrast, invertebrateassemblages of 2 lakes (Brunnsjon and Ovre Skarsjon)had significant negative slopes during the 21-y studyperiod (Brunnsjon: Sørensen t = 20.162, p = 0.305,Bray–Curtis t = 20.457, p = 0.004, Fig. 4A; OvreSkarsjon: Sørensen t = 20.514, p = 0.001, Bray–Curtist = 20.571, p = 0.0003, Fig. 4C).

Environmental drivers of between-year shifts inassemblage composition

Between-year distance of phytoplankton assem-blages was negatively correlated with lake surface-water temperature (20.336) and pH (20.318) (Sør-ensen similarity) and water color (20.307; Bray–Curtis dissimilarity) (Table 4). Between-year distanceof invertebrate assemblages was negatively correlatedwith temperature (20.249), pH (20.281) (Sørensensimilarity), and N concentration (20.167; Bray–Curtisdissimilarity), but positively related to NAOwinter

index (0.249; Sørensen similarity).

Discussion

Decreasing trends in surface-water acidity areconsistent with results from several previous studiesshowing chemical recovery from acidification across

Europe and North America (Stoddard et al. 1999,Eshleman et al. 2008). However, despite widespreadevidence of chemical recovery, support for biologicalrecovery has been patchy, and the environmentalvariables driving or delaying recovery have beendifficult to ascertain. These results imply thatresponse of acid-sensitive species (and assemblages)to decreased acidity is more complex in the recoveryphase than in the impairment phase. A number ofhypotheses have been raised to explain discrepanciesbetween the impairment and recovery phases ofacidification (Yan et al. 2003, Ledger and Hildrew2005, Monteith et al. 2005). Briefly, recovery can beprevented or impeded by impoverished water quality(Lepori et al. 2003, Kowalik et al. 2007), bioticinteractions (Ledger and Hildrew 2005), or scarcityof acid-sensitive colonists (Monteith et al. 2005).Choice of response variable also can affect ability todetect recovery. Clear shifts in lake assemblages ofphytoplankton (Findlay 2003), zooplankton (Frost etal. 2006), and epilithic algae (Battarbee et al. 1988, butsee Vinebrooke et al. 2003) have been related todecreased acidity, whereas results for other groups,such as fish and benthic invertebrate assemblages,have been more equivocal (Burns et al. 2008, Stenderaand Johnson 2008). Our study further highlights thecomplexity of biological recovery responses to acid-ification. This complexity is reflected in stronglyidiosyncratic changes over time as a function of theresponse variable chosen (functional vs structural,univariate vs multivariate) to track change for bothassemblages.

TABLE 3. Mean (61 SD) Euclidean distance of individual lake–year combinations to mean coordinates of nonmetricmultidimensional scaling (NMDS) axis 1 and axis 2 scores of 4 reference lakes. Euclidean distance was calculated using NMDS onSørensen similarity and Bray–Curtis dissimilarity for phytoplankton (1992–2008) and invertebrate (1988–2008) assemblages of 4reference and 4 acidified lakes. Different letters in parentheses show significant differences (p , 0.05) between reference andacidified lakes using a Wilcoxon signed-rank test.

Lake

Phytoplankton Invertebrates

Sørensen similarityBray–Curtisdissimilarity Sørensen similarity

Bray–Curtisdissimilarity

Reference

Allgjuttern 0.041 6 0.023 0.045 6 0.011 0.026 6 0.020 0.044 6 0.026Fracksjon 0.029 6 0.019 0.040 6 0.020 0.028 6 0.036 0.033 6 0.033Stora Envattern 0.033 6 0.024 0.019 6 0.011 0.036 6 0.020 0.034 6 0.015Stora Skarsjon 0.031 6 0.024 0.046 6 0.025 0.024 6 0.018 0.031 6 0.015Mean 6 1 SD 0.034 6 0.005(a) 0.038 6 0.013(a) 0.029 6 0.005(a) 0.036 6 0.006(a)

Acidified

Brunnsjon 0.119 6 0.023 0.086 6 0.012 0.058 6 0.028 0.057 6 0.020Harsvatten 0.199 6 0.046 0.152 6 0.050 0.177 6 0.029 0.189 6 0.048Ovre Skarsjon 0.107 6 0.022 0.075 6 0.007 0.097 6 0.023 0.069 6 0.019Rotehogstjarnen 0.109 6 0.029 0.140 6 0.016 0.069 6 0.019 0.087 6 0.017Mean 6 1 SD 0.134 6 0.044(b) 0.113 6 0.038(b) 0.100 6 0.054(b) 0.101 6 0.060(a)


FIG. 4. Plots of Euclidean distance of acidified lakes Brunnsjon (A, E), Harsvatten (B, F), Ovre Skarsjon (C, G), andRotehogstjarnen (D, H) to the mean of the reference condition of 4 minimally disturbed lakes against year. Euclidean distance wasbased on nonmetric multidimensional scaling (NMDS) axes 1 and 2 coordinates for Bray–Curtis dissimilarity and Sørensensimilarity of presence–absence of invertebrate (A, B, C, D) and phytoplankton (E, F, G, H) assemblages. Statistics are shown forsignificant (p , 0.05) regressions.


Quantitative and qualitative aspects of recovery

Our results were based on a complementaryanalysis approach that allowed identification oftrends of convergence of acidified lakes with refer-ence lakes (quantitative recovery aspect) and oftemporal variability shown by individual lakesduring recovery (qualitative aspect). Our studyshowed that phytoplankton and invertebrate assem-blage composition (Sørensen similarity) tended toconverge quantitatively with those in reference lakes.The rate of recovery varied among lakes (indicated bythe negative slopes in Fig. 4), but the result suggeststhat acidified lakes tend to regain sets of speciespresent in reference lakes. This result supports thefindings of other studies that have documentedrecovery of phytoplankton (Findlay 2003, Stenderaand Johnson 2008) and invertebrate assemblages fromacidification (Raddum et al. 2001, Lento et al. 2008,Stendera and Johnson 2008). However, convergencewith reference conditions was equivocal with abun-dance-based measures of similarity (Bray–Curtisdissimilarity). Invertebrate assemblages in Harsvattenand Rotehogstjarnen and phytoplankton assemblagesin Ovre Skarsjon tended to diverge from referenceconditions during recovery. This result suggests thatrecovery patterns of assemblage composition anddominance patterns are not harmonic and thatinternal abiotic and biotic lake characteristics (Yan etal. 2003) individually or collectively influence popu-lation frequencies during recovery.

Recovery of phytoplankton and invertebrate as-semblage composition (Sørensen similarity) ratherthan abundance-based assemblage structure (Bray–Curtis dissimilarity) underlines the importance ofconsidering species dominance patterns for under-standing ecological patterns and processes (Hille-brand et al. 2008). Species distributions tend tobecome more homogenized regionally as a result of

recolonization of acid-sensitive biota, a pattern con-sistent with landscape-level recovery from acidifica-tion (DGA, RKJ, and Willem Goedkoop, SwedishUniversity of Agricultural Sciences, unpublisheddata). However, local lake conditions contribute tospecies sorting and mediate the degree to whichrecolonizing species can become numerically domi-nant in the assemblages. Consequently, the recoverysignal from patterns of species distribution can beobscured when assessing evenness, which can bemediated through biological interactions (Ledger andHildrew 2005). Thus, our results provide mixedsupport of our 1st hypothesis (decreased dissimilaritybetween acidified and reference lakes over time).Compositional similarity data support it, whereasabundance-based dissimilarity data do not. However,the results do support our 2nd hypothesis (recoverytrajectories differ between assemblages as a functionof the response variable used to detect change). Thisfinding agrees with those of earlier studies (e.g.,Stendera and Johnson 2008) and shows that inferenceabout biological recovery can be biased if results arebased on single-assemblage studies. Comparativeassessment approaches using multiple assemblagesare needed to evaluate integral ecological responses toenvironmental stress (Johnson and Hering 2009).

We acknowledge that quantitative assessment ofassemblage recovery, expressed as the Euclideandistances of acidified lakes to the mean of referencelakes between sampling years is a static approach thatunderestimates temporal variability in the referencelakes. Reference lakes also can undergo environmen-tal change over time, particularly when broad-scaleenvironmental drivers (e.g., climate) cause coherentchanges across aquatic ecosystems at macroecologicalscales (Li et al. 2006, Kent et al. 2007, Fisher et al.2008). Accordingly, defining convergence of impactedsites on a temporal mean of reference sites couldeither over- or underestimate recovery when the

TABLE 4. Spearman correlation (r) of between-year Euclidean distance of phytoplankton and littoral invertebrate assemblagesin 4 acidified and 4 reference lakes and selected water chemistry variables (annual means between 1988 and 2008) and the NorthAtlantic Oscillation winter index (NAOwinter). Sørensen similarity and Bray–Curtis dissimilarity were used in a nonmetricmultidimensional scaling ordination. * = p , 0.05, ** = p , 0.01, *** = p , 0.001

Variable


Sørensen similarity Bray–Curtis dissimilarity Sørensen similarity Bray–Curtis dissimilarity

Temperature 20.336*** 0.094 20.249** 20.027pH 20.318*** 20.076 20.281*** 20.112NO2 + NO3 (mg/L) 0.044 0.167 0.109 20.167*PO4-P (mg/L) 0.035 0.015 0.061 20.015Water color 0.057 20.307*** 20.069 0.065NAOwinter 0.115 0.073 0.249** 20.055


reference sites also show directional temporal change.Assemblage structure in our reference lakes showedno marked directional change despite variability overthe study period (Figs 2A–I, 3A–J). Thus, our quan-titative recovery estimates can be viewed withconfidence.

Our results provide quantitative evidence of recov-ery and evidence for qualitative differences betweenlakes and assemblages. Thus, they highlight theimportance of addressing short- and long-termtemporal dynamics when assessing recovery. Acidi-fied lakes had greater temporal variability thanreference lakes, as has been observed in other stressedenvironments (Angeler and Moreno 2007), but vari-ability in temporal patterns was assemblage and lakespecific. As a result, a consistent convergence ofstructural and functional metrics in all acidified lakeswith reference conditions was not observed. Instead,each acidified lake showed specific periods ofapproximation to and convergence with referenceconditions, punctuated by periods during whichstressed lakes deviated from these desired states. Thispronounced variability can be partially explained bythe environmental variables that were correlated withthis temporal variability.

Environmental correlates of recovery

Our results provide mixed support for our 3rd

hypothesis (biological recovery, measured as decreas-es in between-year Euclidean distance, is a result ofdecreased acidity). The effects of decreased aciditywere evident in species composition of assemblages(Sørensen similarity) but not in species abundances(Bray–Curtis dissimilarity) in both phytoplanktonand invertebrate assemblages. Stendera and Johnson(2008) showed that changes in phytoplankton taxonrichness in these boreal lakes were related not only toincreased pH but also to other environmentalvariables, such as increased water color (a proxy fordissolved organic C [DOC]) and nutrient concentra-tions. Our study showed that high between-yearshifts in phytoplankton assemblage composition werenegatively correlated with temperature, water color,and pH. Increased pH should permit colonization ofacid-sensitive species, but concurrent increases inwater color (DOC concentrations) and temperaturealso could have influenced assemblage structure(Wetzel 1995). Interannual shifts in invertebrateassemblage composition also were negatively corre-lated with temperature and pH. In addition, NO2 +NO3 concentration and the NAOwinter index weresignificantly correlated with temporal assemblageturnover.

The imprint of climate on recovery

The NAO is an important regional-level driver ofclimate in southern Sweden (Weyhenmeyer 2004).Positive NAO values signify higher precipitation and,thus, more variable hydraulic and chemical condi-tions. Invertebrates seemed to respond more readilyto the NAOwinter index than did phytoplanktonassemblages. Life-history characteristics mediate thestrength of biological responses to climate forcing(Adrian et al. 2006). Much shorter generation times ofphytoplankton relative to invertebrates could uncou-ple phytoplankton responses from climate signals andcould make these responses visible only via indirector integrated effects (Ottersen et al. 2001). Theseeffects could be manifested through responses toabiotic or biotic variables that are more stronglyaffected by the NAO. Rusak et al. (2008) showedclimate imprints on zooplankton dynamics, and theseeffects could have at least some influence onphytoplankton. We acknowledge that using singleannual summer samples in our analyses could havecaused us to miss potential NAO effects on springphytoplankton development (Weyhenmeyer et al.1999), but our results suggest that this potentialNAO effect does not influence phytoplankton duringlater successional stages. This conclusion is inagreement with the results of a recent meta-analysisof data from European lakes, which showed nocorrelations between phytoplankton assemblages inall seasons with the NAO (Blenckner et al. 2007).

Between-year Euclidean distance of invertebrateassemblage composition was correlated positivelywith the NAOwinter index across lakes. Durance andOrmerod (2007) found that the imprints of the NAOon stream invertebrate assemblages were particularlystrong in circumneutral streams. They suggested thatacidification stress in streams could override a climatesignal, possibly because of the complex interaction ofthe natural and anthropogenic disturbance regimes towhich stream organisms are exposed under theconstraints set by climatic conditions. The differencebetween the study of Durance and Ormerod (2007)and ours could be because lake ecosystems are muchless affected by changes in surface-water hydrologythan are stream ecosystems, and hence, streaminvertebrate assemblages respond more readily toclimatic variability regardless of acidification status.

The significant correlations of both assemblageswith temperature also highlight the importance ofclimatic variables in shaping among-year shifts inassemblage composition. Correlation of between-yearEuclidean distance of invertebrate assemblages andNAOwinter index values were significant for acidified


lakes, and greater between-year distance of inverte-brate assemblages was associated with warmerwinters (i.e., positive NAO values). Several studieshave shown how interannual variability in NAOaffects lake assemblages by affecting temperature, icecover, and timing of spring algae blooms (Weyhen-meyer et al. 1999). In the boreal lakes of our study, theeffect of broad-scale climatic variability might havecontext-dependent effects on water quality which, inturn, can mediate shifts in invertebrate and phyto-plankton assemblages. Thus, recovery might be offsetby effects of interannual, climate-driven changes onlocal lake variables (but see Evans et al. 2006). Thissupports our 4th hypothesis (climatic effects subsumethe signal of biological recovery). In addition toclimate-mediated imprints on recovery, a furtherplausible explanation for the marked interannualvariability in assemblage structure is that acidicepisodes can inflict recurrent stress on lake assem-blages. This conjecture is supported by data showingthat annual mean pH was ,5.5 during several of the21 y of study in the acidified lakes.

Predisturbance conditions

An apparent weakness of our study is that predis-turbance measures of water chemistry and phyto-plankton and littoral invertebrate assemblages are notavailable. However, palaeoreconstruction of prein-dustrial pH indicates that all 4 lakes were relativelywell buffered prior to acidification (diatom-inferredpH ranged from 6.0 [Rotehogstjarnen] to 6.7 [OvreSkarsjon]) (Guhren et al. 2003, Erlandsson et al. 2008).These preindustrial estimates of pH are close to therange of present-day measures of pH in the 4reference lakes (mean pH ranged from 6.4 to 6.8)and indicate that water quality of the 4 reference lakesis a reasonable estimate of the predisturbed state.Furthermore, the finding that phytoplankton andlittoral invertebrate assemblage composition are mov-ing toward the reference population lends support tothe conjecture that the acidified lakes show signs ofapproaching their predisturbed abiotic state.

Conclusion

Our results are of general interest concerningrecovery of ecosystems from disturbance, but areparticularly relevant to lake management. Phyto-plankton and littoral invertebrate assemblages re-sponded differently to improved water quality amongacidified lakes. Thus, investigators should considerusing multiple groups of organisms when a prioriknowledge of response (recovery) signatures is poor.Use of phytoplankton or invertebrate assemblages

alone would have led us to a biased conclusionregarding biological recovery. Moreover, the choice ofthe response variable influenced our ability to detectrecovery. Assemblage composition (NMDS ordina-tion) was powerful for detecting changes, but thedissimilarity measure used (whether based on pres-ence–absence or biovolume/abundance) influencedthe results. One of our most interesting findings wasthe congruence of trends in water chemistry andbiology in acidified and reference lakes. Our resultsshow the influence of factors other than decreasedacidity, such as the underlying importance of climate,as drivers of recovery trajectories and pathways.

Acknowledgements

We thank the many people involved in collectingand processing the physicochemical, phytoplankton,and invertebrate samples. In particular, we thankBjorn Wiklund and Putte Olsson for sorting and LarsEriksson for identifying the invertebrates, and EvaHerlitz, Ann-Marie Wiederholm, and Isabel Quintanafor identifying phytoplankton. Financial support forthis project was partly provided by the EU fundedprojects EURO-LIMPACS (contract no. GOEC-CT-2003-5055), WISER (contract no. 226273), and RE-FRESH (contract no. 244121). The Swedish Environ-mental Protection Agency is acknowledged formaking these data available. We thank S. Ormerod,M. Barbour, and an anonymous referee for helpfulcomments on a previous version of the manuscript.

Literature Cited

ADRIAN, R., S. WILHELM, AND D. GERTEN. 2006. Life-historytraits of lake plankton species may govern theirphenology response to climate warming. Global ChangeBiology 12:652–661.

ANGELER, D. G., AND J. M. MORENO. 2007. Zooplanktoncommunity resilience after press-type anthropogenicstress in temporary ponds. Ecological Applications 17:1105–1115.

BATTARBEE, R. W., R. J. FLOWER, A. C. STEVENSON, V. J. JONES, R.HARRIMAN, AND P. G. APPLEBY. 1988. Diatom and chemicalevidence for reversibility of acidification of Scottishlochs. Nature 322:530–532.

BLENCKNER, T., R. ADRIAN, D. M. LIVINGSTONE, E. JENNINGS, G.A. WEYHENMEYER, D. G. GEORGE, T. JANKOWSKI, M.JARVINEN, C. N. AONGHUSA, T. NOGES, D. STRAILE, AND K.TEUBNER. 2007. Large-scale signatures in lakes acrossEurope: a meta-analysis. Global Change Biology 13:1314–1326.

BLINDOW, I., A. HARGEBY, AND G. ANDERSSON. 2002. Seasonalchanges of mechanisms maintaining clear water in ashallow lake with abundant Chara vegetation. AquaticBotany 72:315–334.


BURNS, D. A., K. RIVA-MURRAY, R. W. BODE, AND S. PASSY. 2008.Changes in stream chemistry and biology in response toreduced levels of acid deposition during 1987–2003 inthe Neversink River Basin, Catskill Mountains. Ecolog-ical Indicators 8:191–203.

CHRISTENSEN, M. R., M. D. GRAHAM, R. D. VINEBROOKE, D. L.FINDLAY, AND M. A. TURNER. 2006. Multiple anthropo-genic stressors cause ecological surprises in boreal lakes.Global Change Biology 12:2316–2322.

DURANCE, I., AND S. J. ORMEROD. 2007. Climate change effectson upland stream macroinvertebrates over a 25-yearperiod. Global Change Biology 13:942–957.

ERLANDSSON, M., K. BISHOP, J. FOLSTER, M. GUHREN, T. KORS-

MAN, V. KRONNAS, AND F. MOLDAN. 2008. A comparison ofMAGIC and paleolimnological predictions of preindus-trial pH for 55 Swedish lakes. Environmental Scienceand Technology 42:43–48.

ESHLEMAN, K. N., K. M. KLINE, R. P. MORGAN, N. M. CASTRO,AND T. L. NEGLEY. 2008. Contemporary trends in theacid-base status of two acid-sensitive streams in westernMaryland. Environmental Science and Technology 42:56–61.

EVANS, C. D., B. REYNOLDS, A. JENKINS, R. C. HELLIWELL, C. J.CURTIS, C. L. GOODALE, R. C. FERRIER, B. A. EMMETT, M. G.PILKINGTON, S. J. M. CAPORN, J. A. CARROLL, D. NORRIS, J.DAVIES, AND M. C. COULL. 2006. Evidence that soil carbonpool determines susceptibility of semi-natural ecosys-tems to elevated nitrogen leaching. Ecosystems 9:453–462.

FAITH, D. P., P. MINCHIN, AND L. BELBIN. 1987. Compositionaldissimilarity as a robust measure of ecological distance.Vegetatio 69:57–68.

FINDLAY, D. L. 2003. Response of phytoplankton communi-ties to acidification and recovery in Killarney Park andthe Experimental Lakes Area, Ontario. Ambio 32:190–195.

FISHER, J. A. D., K. T. FRANK, B. PETRIE, W. C. LEGGETT, AND N.L. SHACKELL. 2008. Temporal dynamics within a con-temporary latitudinal diversity gradient. Ecology Let-ters 11:883–897.

FROST, T. M., J. M. FISCHER, J. L. KLUG, S. E. ARNOTT, AND P. K.MONTZ. 2006. Trajectories of zooplankton recovery in theLittle Rock Lake whole-lake acidification experiment.Ecological Applications 16:353–367.

GUHREN, M., P. ROSEN, T. KORSMAN, AND I. RENBERG. 2003.Tidigare pH i referenssjoar i Sverige, Department ofEcology and Environmental Science, Umea University,Umea, Sweden. (Available from: Department of Ecologyand Environmental Science, Umea University, Umea,Linnaeus vag 6, SE- 901 87 Umea, Sweden.)

HILLEBRAND, H., D. M. BENNETT, AND M. W. CADOTTE. 2008.Consequences of dominance: a review of evennesseffects on local and regional ecosystem processes.Ecology 89:1510–1520.

HOLM, S. 1979. A simple sequentially rejective test proce-dure. Scandinavian Journal of Statistics 6:65–70.

HURRELL, J. W. 1995. Decadal trends in the North AtlanticOscillation: regional temperatures and precipitation.Science 269:676–679.

ILLIES, J. 1966. Verbreitung der Susswasserfauna Europas.Verhandlungen der Internationalen Vereinigung furtheoretische und angewandte Limnologie 16:287–296.

JANSSON, M., P. BLOMQVIST, A. JONSSON, AND A. K. BERGSTROM.1996. Nutrient limitation of bacterioplankton, autotro-phic and mixotrophic phytoplankton, and heterotrophicnanoflagellates in Lake Ortrasket. Limnology andOceanography 41:1552–1559.

JOHNSON, R. K. 1999. Regional representativeness of Swedishreference lakes. Environmental Management 23:115–124.

JOHNSON, R. K., S. BELL, L. DAVIES, S. DECLERCK, E. LACZKO, M.TONDER, AND Y. UZUNOV. 2003. Wetlands. Pages 98–122 inJ. Young, P. Nowicki, D. Alard, K. Henle, R. K. Johnson,S. Matouch, J. Niemela, and A. Watt (editors). Conflictsbetween human activities and the conservation ofbiodiversity in agricultural landscapes, grasslands,forests, wetlands, and uplands in Europe. A report ofthe BIOFORUM project. Centre for Ecology andHydrology, Banchory, Scotland.

JOHNSON, R. K., AND D. HERING. 2009. Response taxonomicgroups in streams to gradients in resource and habitatcharacteristics. Journal of Applied Ecology 46:175–186.

KENDALL, M. 1938. A new measure of rank correlation.Biometrika 30:81–89.

KENT, A. D., A. C. YANNARELL, J. A. RUSAK, E. W. TRIPLETT, AND

K. D. MCMAHON. 2007. Synchrony in aquatic microbialcommunity dynamics. ISME Journal 1:38–47.

KOWALIK, R. A., D. M. COOPER, C. D. EVANS, AND S. J. ORMEROD.2007. Acidic episodes retard the biological recovery ofupland British streams from chronic acidification.Global Change Biology 13:2439–2452.

LAKE, P. S. 2000. Disturbance, patchiness, and diversity instreams. Journal of the North American BenthologicalSociety 19:573–592.

LEDGER, M. E., AND A. G. HILDREW. 2005. The ecology ofacidification and recovery: changes in herbivore–algalfood web linkages across a stream pH gradient.Environmental Pollution 137:103–118.

LENTO, J., P. J. DILLON, K. M. SOMERS, AND R. A. REID. 2008.Changes in littoral benthic macroinvertebrate commu-nities in relation to water chemistry in 17 PrecambrianShield lakes. Canadian Journal of Fisheries and AquaticSciences 65:906–918.

LEPORI, F., A. BARBIERI, AND S. J. ORMEROD. 2003. Effects ofepisodic acidification on macroinvertebrate assemblagesin Swiss Alpine streams. Freshwater Biology 48:1873–1885.

LI, W. K. W., W. G. HARRISON, AND E. J. H. HEAD. 2006.Coherent assembly of phytoplankton communities indiverse temperate ocean ecosystems. Proceedings of theRoyal Society of London Series B: Biological Sciences273:1953–1960.

MONTEITH, D. T., A. G. HILDREW, R. J. FLOWER, P. J. RAVEN, W.R. B. BEAUMONT, P. COLLEN, A. M. KREISER, E. M. SHILLAND,AND J. H. WINTERBOTTOM. 2005. Biological responses to thechemical recovery of acidified fresh waters in the UK.Environmental Pollution 137:83–101.


ØKLAND, J., AND K. A. ØKLAND. 1986. The effects of aciddeposition on benthic animals in lakes and streams.Experentia 42:471–486.

OLRIK, K. P., P. BLOMQVIST, P. BRETTUM, G. CRONBERG, AND P.ELORANTA. 1989. Methods for quantitative assessment ofphytoplankton in freshwaters, part I. Swedish Environ-mental Protection Agency, Stockholm, Sweden. (Avail-able from: Department of Aquatic Sciences and Assess-ment, Swedish University of Agricultural Sciences, POBox 70 50, SE-750 07 Uppsala, Sweden)

ORMEROD, S. J., AND I. DURANCE. 2009. Restoration andrecovery from acidification in upland Welsh streamsover 25 years. Journal of Applied Ecology 46:164–174.

OTTERSEN, G., B. PLANQUE, A. BELGRANO, E. POST, P. C. REID,AND N. C. STENSETH. 2001. Ecological effects of the NorthAtlantic Oscillation. Oecologia (Berlin) 128:1–14.

PERSSON, G. 1996. 26 svenska referenssjoar 1989–1993: enkemisk-biologisk statusbeskrivning. 141. Naturvards-verket, Stockholm, Sweden (in Swedish). (Availablefrom: Department of Aquatic Sciences and Assessment,Swedish University of Agricultural Sciences, PO Box 7050, SE-750 07 Uppsala, Sweden)

RABENI, C. F., K. E. DOISY, AND D. L. GALAT. 2002. Testing thebiological basis of a stream habitat classification usingbenthic invertebrates. Ecological Applications 12:782–796.

RADDUM, G. G., A. FJELLHEIM, AND B. L. SKJELKVALE. 2001.Improvements in water quality and aquatic ecosystemsdue to reductions in sulphur deposition in Norway.Water, Air, & Soil Pollution 130:87–98.

RUSAK, J. A., N. D. YAN, AND K. M. SOMERS. 2008. Regionalclimatic drivers of synchronous zooplankton dynamicsin north-temperate lakes. Canadian Journal of Fisheriesand Aquatic Sciences 65:878–889.

SCHEFFER, M., AND S. R. CARPENTER. 2003. Catastrophic regimeshifts in ecosystems: linking theory to observation.Trends in Ecology and Evolution 18:648–656.

SHANNON, C. E., AND W. WEAVER. 1949. The mathematicaltheory of communication. University of Illinois Press,Urbana, Illinois.

SIMPSON, E. H. 1949. Measurement of diversity. Nature 163:688.

SKJELKVALE, B. L., C. D. EVANS, T. LARSSEN, A. HINDAR, AND G.RADDUM. 2003. Recovery from acidification in Europeansurface waters: a view to the future. Ambio 32:170–175.

SKJELKVALE, B. L., J. L. STODDARD, D. S. JEFFRIES, K. TORSETH, T.HOGASEN, J. BOWMAN, J. MANNIO, D. T. MONTEITH, R.MOSELLO, M. ROGORA, D. RZYCHON, J. VESELY, J. WIETING, A.WILANDER, AND A. WORSZTYNOWICZ. 2005. Regional scale

evidence for improvements in surface water chemistry1990–2001. Environmental Pollution 137:165–176.

STENDERA, S., AND R. K. JOHNSON. 2008. Tracking recoverytrends of boreal lakes: use of multiple indicators andhabitats. Journal of the North American BenthologicalSociety 27:529–540.

STODDARD, J. L., D. S. JEFFRIES, A. LUKEWILLE, T. A. CLAIR, P. J.DILLON, C. T. DRISCOLL, M. FORSIUS, M. JOHANNESSEN, J. S.KAHL, J. H. KELLOGG, A. KEMP, J. MANNIO, D. T. MONTEITH,P. S. MURDOCH, S. PATRICK, A. REBSDORF, B. L. SKJELKVALE,M. P. STAINTON, T. TRAAEN, H. VAN DAM, K. E. WEBSTER, J.WIETING, AND A. WILANDER. 1999. Regional trends inaquatic recovery from acidification in North Americaand Europe. Nature 401:575–578.

STODDARD, J. L., P. LARSEN, C. P. HAWKINS, R. K. JOHNSON, AND

R. H. NORRIS. 2006. Setting expectations for the ecolog-ical condition of running waters: the concept ofreference condition. Ecological Applications 16:1267–1276.

VINEBROOKE, R. D., M. D. GRAHAM, D. L. FINDLAY, AND T. A.TURNER. 2003. Resilience of epilithic algal assemblages inatmospherically and experimentally acidified boreallakes. Ambio 32:196–202.

WETZEL, R. G. 1995. Death, detritus, and energy flowin aquatic ecosystems. Freshwater Biology 33:83–89.

WEYHENMEYER, G. A. 2004. Synchrony in relationshipsbetween the North Atlantic Oscillation and waterchemistry among Sweden’s largest lakes. Limnologyand Oceanography 49:1191–1201.

WEYHENMEYER, G. A., T. BLENCKNER, AND K. PETTERSSON. 1999.Changes of the plankton spring outburst related to theNorth Atlantic Oscillation. Limnology and Oceanogra-phy 44:1788–1792.

WILANDER, A., R. K. JOHNSON, AND W. GOEDKOOP. 2003.Riksinventering 2000. En synoptisk studie av vattenkemioch bottenfauna i svenska sjoar och vattendrag. Institu-tionen for Miljoanalys, Uppsala, Sweden (in Swedish).(Available from Department of Aquatic Sciences andAssessment, Swedish University of Agricultural Scienc-es, PO Box 70 50, SE-750 07 Uppsala, Sweden)

WILCOXON, F. 1945. Individual comparisons by rankingmethods. Biometrics 1:80–83.

YAN, N. D., B. LEUNG, B. KELLER, S. E. ARNOTT, J. M. GUNN, AND

G. G. RADDUM. 2003. Developing conceptual frameworksfor the recovery of aquatic biota from acidification.Ambio 32:165–169.

Received: 10 December 2009Accepted: 15 July 2010


Source of variation Multipliers

E(MS) Den. Variance componentBetween samples i r j –

1. Ti 0 n – – se2 + ns2

T 2 s2T = [MST – MSL(T)]/n

2. L(T)r(i) 1 n – – se2

Within samples—repeated measures i – j s

3. Yj a – 0 N sE2 + Ns2

YL(T) + aNs2Y 4 s2

Y = [MSY – MSTY]/aN4. T 3 Yij 0 – 0 N sE

2 + Ns2YL(T) 5 s2

TY = [MSTY – MSYL(T)]/N5. Y 3 L(T)r(i)j 1 – 1 N sE

2

APPENDIX 2. Statistical design for analysis of variance showing factors, degrees of freedom, multipliers, and formulas todetermine estimated mean squares (E[MS]), variance components, and F-ratio tests. Factors and levels: Treatment (T) has i =…alevels, Lake(treatment) (L[T]) was replicated n times (r =…n) and served as the error term for the spatial variation, Year (Y) has j=…b levels, all combinations of the temporal variation effect were replicated N times (s =…N), and the term Y 3 L(T) is the errorterm for the temporal variation. Two factors are orthogonal and 1 is nested: T and Y are fixed, and L nested in T (L[T]) is random.Multipliers and formulas to determine E(MS) and variance components, denominators for F-ratio tests (den.; numbers correspondto numbers in the sources of variation column), and degrees of freedom (df) are shown. se

2, error term for between-samplevariation; sE

2, error term for within-sample (repeated measures) variation. Models: between samples: Yir = m + Ti + L(T)r(i), withinsamples: Yijr = m + Yj + TYij + YL(T)jr(i)

Lake

Acidity

NutrientsPO4-P (mg/L)

Water colorAbsorbance (420 nm)

TemperatureuCpH

Diatom-inferred pHa SO4

22(meq/L)

Reference

Mean 6 1 SD 6.6 6 0.17(a) 0.171 6 0.043(a) 2.25 6 0.93(a) 0.069 6 0.032(a) 13.5 6 1.18(a)

Allgjuttern 0.467** – 20.848*** 0.476**Fracksjon 0.581*** – 20.876*** 0.591*** 0.349*Stora Envattern – 20.867*** 20.369* 0.543***Stora Skarsjon 0.352* – 20.848*** 0.543** 0.444**

Acidified

Mean 6 1 SD 5.3 6 0.41(b) 0.155 6 0.078(b) 2.66 6 1.32(a) 0.187 6 0.152(b) 12.8 6 1.33(a)

Brunnsjon 6.1 20.619*** 0.448**Harsvatten 0.848*** 6.4 20.943*** 0.610***Ovre Skarsjon 0.676*** 6.7 20.950*** 0.514***Rotehogstjarnen 6.0 20.600*** 0.495**

a Data from Guhren et al. 2003 and Erlandsson et al. 2008

APPENDIX 1. Mean (61 SD) values of water-chemistry and temperature variables and Kendall t rank correlation coefficients ofwater-chemistry and temperature variables against year for 4 reference and 4 acidified lakes sampled from 1988 to 2008. Differentletters in parentheses shows significant differences (p , 0.05) between reference and acidified lakes. Only significant regressionsare shown. * = p , 0.05, ** = p , 0.01, *** = p , 0.001.



Sørensen similarity Bray–Curtis dissimilarity Sørensen similarity Bray–Curtis dissimilarity

Reference 0.019 6 0.016(a) 0.027 6 0.018(a) 0.027 6 0.025(a) 0.034 6 0.035(a)

Acidified 0.037 6 0.039(b) 0.029 6 0.033(a) 0.035 6 0.024(b) 0.035 6 0.034(a)

APPENDIX 4. Mean (61 SD) between-year shifts (Euclidean distance) of phytoplankton and invertebrate assemblages in 4reference and 4 acidified lakes. Euclidean distance was calculated using NMDS on Sørensen similarity and Bray–Curtisdissimilarity for phytoplankton (1992–2008) and invertebrate (1988–2008) assemblages. Different letters in parentheses showsignificant differences (p , 0.05) between reference and acidified lakes using a Wilcoxon signed-rank test.

APPENDIX 3. Correlation coefficients from Pearson correlation relating the North Atlantic Oscillation winter [NAOwinter] indexto water-quality variables. Correlations that were not significant (p . 0.05) are not shown.* = p , 0.05, ** = p , 0.01.

Variable

Reference lakes Acidified lakes

Allgjuttern FracksjonStora

SkarsjonStora

Envattern Brunnsjon Harsvatten RotehogstjarnenOvre

Skarsjon

Secchi depth 0.592**Dissolved O2 0.492*pH 20.463* 20.679**Alkalinity 20.459* 20.550* 0.553**Electrical

conductivity0.452*

CaMg 0.442*K 0.458*SO4

22

NO2 + NO3-N 0.471* 0.438* 0.471*NH4-N 20.534* 20.565** 20.589**Total P 0.471* 0.460* 0.544*Total organic C 0.621** 20.535**Water color 20.489* 20.455* 20.522* 20.448*


Documents

WR S OD Q N WR Q D Q G LQ Y H UWH E UD WH D V V H P E OD … · (sensu Lake 2000), often results in distinctive and persistent changes in biological assemblages (Økland and Økland