spatial variation among lakes within landscapes: ecological … · 2006-09-01 · viewed as an...

Spatial Variation among Lakeswithin Landscapes: Ecological

Organization along Lake ChainsPatricia A. Soranno,1* Katherine E. Webster,1,2 Joan L. Riera,1

Timothy K. Kratz,1 Jill S. Baron,3 Paul A. Bukaveckas,4 George W. Kling,5David S. White,6 Nel Caine,7 Richard C. Lathrop,1,2 and Peter R. Leavitt8

1University of Wisconsin, Center for Limnology, 680 N. Park Street, Madison, Wisconsin 53706, USA; 2Wisconsin Department ofNatural Resources, 1350 Femrite Drive, Monona, Wisconsin 53716, USA; 3US Geological Survey Biological Resources Division and

Natural Resources Ecology Laboratory, Colorado State University, Fort Collins, Colorado 80523, USA; 4Department of Biology,University of Louisville, Louisville, Kentucky 40292, USA; 5Department of Biology, University of Michigan, Ann Arbor, Michigan48109, USA; 6Hancock Biological Station, Murray State University, Murray, Kentucky 42071, USA; 7Department of Geography,

and Institute for Arctic and Alpine Research, Campus Box 450, University of Colorado, Boulder, Colorado 80309, USA;8Limnology Laboratory, Department of Biology, University of Regina, Regina, Saskatchewan S4S 0A2, Canada

ABSTRACTAlthough limnologists have long been interested inregional patterns in lake attributes, only recentlyhave they considered lakes connected and orga-nized across the landscape, rather than as spatiallyindependent entities. Here we explore the spatialorganization of lake districts through the concept oflandscape position, a concept that considers lakeslongitudinally along gradients of geomorphologyand hydrology. We analyzed long-term chemicaland biological data from nine lake chains (lakes in aseries connected through surface or groundwaterflow) from seven lake districts of diverse hydrologicand geomorphic settings across North America.Spatial patterns in lake variables driven by land-scape position were surprisingly common acrosslake districts and across a wide range of variables.On the other hand, temporal patterns of lake vari-ables, quantified using synchrony, the degree towhich pairs of lakes exhibit similar dynamics through

time, related to landscape position only for lakechains with lake water residence times that spanneda wide range and were generally long (close to orgreater than 1 year). Highest synchrony of lakeswithin a lake chain occurred when lakes had shortwater residence times. Our results from both thespatial and temporal analyses suggest that certainfeatures of the landscape position concept are robustenough to span a wide range of seemingly disparatelake types. The strong spatial patterns observed inthis analysis, and some unexplained patterns, sug-gest the need to further study these scales and tocontinue to view lake ecosystems spatially, longitu-dinally, and broadly across the landscape.

Key words: landscape position; lake variability;lake districts; synchrony; coherence; north temper-ate lakes; lake chains; lake order; lake number;water residence time.

INTRODUCTION

Spatial variation within and among lakes has beenrecognized as playing an important role in structur-

ing lake ecosystems at a variety of scales. Thenatural boundary of the lake shoreline has longfocused the attention of limnologists on the lake asthe unit of study (Forbes 1887). Understandingspatial patterns within individual lakes occupiedearly limnologists since they first lowered thermom-eters and water samplers into deep stratified lakes(Forel 1892; Birge and Juday 1911). The observedvertical gradients in light, temperature, oxygen, andnutrients were viewed as major drivers of ecological

Received 2 February 1999; accepted 14 June 1999.*Corresponding author’s current address: Department of Fisheries and Wild-life, 10A Natural Resources Building, Michigan State University, EastLansing, Michigan 48824, USA.e-mail: [email protected]

Ecosystems (1999) 2: 395–410 ECOSYSTEMSr 1999 Springer-Verlag

395

processes. A more systemwide view evolved withthe recognition of the key role played by interac-tions between lakes and their watersheds and air-sheds (Likens 1984; Likens 1985; Schindler andothers 1990; Kling and others 1991). Lakes wereviewed as an integral component of the largeraquatic-terrestrial ecosystem with catchment char-acteristics, such as geology, land use/cover, andtopography used to explain interlake variation inthe chemical and trophic status of lakes (Dillon andKirchner 1975; Omernik 1976; Driscoll and vanDreason 1992; Gibson and others 1995).

Limnologists now recognize that interlake variabil-ity is driven by factors acting at both regional (forexample, climate, atmospheric deposition, and re-gional landscape characteristics) and local scales (forexample, lake morphometry, foodwebs, land use/cover, geomorphic setting, and hydrology). How-ever, there is still a significant amount of unex-plained variation among lakes of similar featuresand occupying seemingly identical physical settings.General organizing principles that explain interac-tions among lakes within a region in a spatiallyexplicit manner may help explain some of thisvariation. Until recently, no such principles haveexisted for lakes.

In contrast to lakes, streams have been viewedspatially within a landscape for some time. Thelongitudinal view of streams from headwaters toriver mouth has allowed stream ecologists to usegeomorphic and hydrologic gradients as a fundamen-tal physical template for streams at large spatialscales. The River Continuum Concept (Vannote andothers 1980; Minshall and others 1985) and morerecent constructs have postulated differences inland—water linkages along important spatial gradi-ents (Frissell and others 1986; Junk and others1989; Wiley and others 1990). Streams have beenviewed as open systems, strongly linked to surround-ing landscapes through material exchange, withexplicit linkages between up- and downstream seg-ments (Hynes 1975; Ward and Stanford 1983; Fris-sell and others 1986; Fisher and Grimm 1991).

Recently, lake ecologists have begun to recognizespatial patterning among lakes at larger scales byusing the idea of landscape position (Kratz andothers 1997; Magnuson and others forthcoming;Riera and others forthcoming; Webster and othersforthcoming). The landscape position concept wasdeveloped in northern Wisconsin, a region wheregroundwater flowpaths generate strong spatial pat-terns in lakes across the landscape (Kratz and others1997). In fact, distinctions between high- and low-gradient lakes were first identified in the early 1900sby E.A. Birge and C. Juday, who contrasted features

of seepage and drainage lakes, defined by absence orpresence, respectively, of surface water inlets andoutlets (Juday and Meloche 1943). Neighboringlakes sharing a common climate, geologic setting,and regional species pool differ systematically inmany of their features as a function of their hydrol-ogy and geomorphology (Winter 1977; Eilers andothers 1983; Rochelle and others 1989; Kratz andothers 1997; Riera and others forthcoming; Websterand others forthcoming; Magnuson and others forth-coming). Position of lakes along gradients such asthese is defined in part by the strength of theinteraction between lakes and the landscape. Higherinputs of landscape-derived materials occur in lakeslower in the landscape because water entering theselower lakes has been in contact with soil longer(Drever 1982).

Here, we explore this landscape position conceptfor lakes by analyzing patterns along lake chains(lakes in a series connected through surface orgroundwater flow). We examine lakes in a series asa simple quantification of landscape position to tryto capture the spatially explicit linkages betweenlakes and the landscape. We ask the following twoquestions: (a) Are there predictable spatial patternsin chemical, algal, and water clarity variables withinlake chains?; and (b) Do spatial patterns of variableswithin a lake chain influence the temporal dynam-ics of lakes? To answer the above questions, weexamined spatial patterns of lake variables alongnine different lake chains. To test the generality ofthe concept, we considered lake chains from a widerange of landscape types (alpine, prairie, tundra,and glacial plains) and hydrologic settings (ground-water-dominated seepage lakes, surface drainedlakes, and large main stem and tributary reservoirs).

General PredictionsPredicting and understanding lake concentrationsof a wide range of chemical variables has been thesubject of much research over the past 30 y (seeReckhow and Chapra 1983). In particular, thesteady state mass-balance model has been usedsuccessfully to understand dynamics of nitrogen(Molot and Dillon 1993; Windolf and others 1996),dissolved organic carbon (Dillon and Molot 1997;Schindler and others 1997), and most notably phos-phorus (Vollenweider 1969; Reckhow and Chapra1983; Dillon and Molot 1996). The model states thatthe steady state standing stock of a constituent (overa given time period) will equal total loadings plus orminus transports, plus or minus reactions (Reckhowand Chapra 1983). Unfortunately, this model typi-cally has been applied only to individual lakes orgroups of unconnected lakes. We are interested in

396 P. A. Soranno and others

how it can be applied to lakes in a landscapecontext. The mass-balance model provides themechanisms for determining how landscape posi-tion can influence the concentrations of importantlake constituents along a landscape gradient. Be-cause building a complete mass-balance model formultiple lakes along nine different lake chains is notpossible in our study due to data limitations, we arenot able to fully answer our questions by using themass-balance model. However, to make predictionsabout how landscape position influences lake soluteconcentrations, we can examine patterns and inferlikely mechanisms based on the following assump-tions: (a) loading increases as the watershed to lakearea ratio (WS:LK) increases; (b) transport andretention within a lake depends on water residencetime (WRT); and (c) patterns for individual variablesdepend on retention strength and reactivity.

We are interested in whether concentrations oflake constituents vary predictably down lake chains.For example, higher solute concentrations in lakesfurther down the lake chain mean that either moresolutes are entering the lake, less solutes are leavingthe lake, or less solutes are being transformedwithin the lake. We assume that landscape positionaffects only the first case. We make the followingpredictions based on the above assumptions. First,for relatively nonreactive solutes (for example, thosesubject to weathering processes, such as calciumand alkalinity), biological or chemical reactionsaffecting them in the lake itself will be minor, andthus inputs should equal outputs. However, becausewe expect weathering to increase down the lakechain due to increased contact time of incomingwater with soil, total loading, and standing stocks ofthese variables should increase down the lake chain.And, because these nonreactive solutes will un-dergo little change once in the lake, WRT shouldhave no effect on lake concentrations. For morereactive solutes or particulates (such as other dis-solved ions, dissolved and total nutrients, and algalvariables), reactions within the lake are most likelya large term in the equation, and so lake concentra-tions should be influenced by WRT. WRT willinfluence the length of time that these reactionshave to proceed, and that may lead to eitherincreases or decreases in constituents down the lakechain depending on the variable. For example, it hasbeen shown that lakes often retain a significantportion of the phosphorus that is transported tothem (Reckhow and Chapra 1983). Thus, we expectthat lakes will act as retention basins and nutrienttransport down the chain will decrease, especially inlakes with long WRT.

Landscape position also may influence the tempo-ral dynamics of lake constituents. We expect that if

large spatial gradients exist in solute concentrationsalong the lake chain, then interannual dynamicsamong lakes may be different, and lakes that arecloser together along the chain may be more similarthan lakes farther apart. In lake chains where no orfew spatial gradients exist in lake concentrations,similar drivers acting on similar time scales shouldoverride spatial drivers, and lakes should vary syn-chronously through time.

We expect that specific landscape characteristicsof each lake district will determine the presence andthe strength of the above spatial and temporalpatterns along lakes chains. For example, the geo-logic setting, and in particular the characteristics ofthe soil and the presence of glacial till, will deter-mine the rate of weathering in the surroundinglandscape and the availability of materials to betransported from land to water. Also, the hydrologicsetting of the chain will determine the WRT. Ingeneral, the slower flowpaths of groundwater-fedlakes cause them to have longer WRT than surfacedrainage lakes of similar size and volume. Amongsurface drainage lakes, the shortest water residencetimes typically are observed in reservoir chains.

In summary, we expect that for nonreactivevariables all lake chains regardless of WRT will showincreasing concentrations down the lake chain. Forreactive variables, concentrations should either in-crease or decrease down the chain depending on thevariable and on lake-specific reaction rates. Thus thepredictability of changes in reactive variables acrossspace should not be generalizable, but patterns ofbehavior of specific variables among different lakechains may emerge. Our goal here is not to exhaus-tively explain patterns along individual lake chains,but rather, to identify general patterns to arrive at amore synthetic understanding of lake variationalong spatial gradients within a wide range of lakedistricts. We test the general utility of the landscapeposition concept by considering lake chains in dis-tricts of diverse hydrologic and geomorphic settings.In short, we address when and where the spatialarrangement of lakes within a landscape matters tolake functioning.

LAKE DISTRICTS

Nine lake chains from seven lake districts in NorthAmerica were analyzed (Figure 1; see Table 1 forreferences). We analyzed data from lake chains intwo lake districts that are part of the North Temper-ate Lakes Long Term Ecological Research (LTER)program, one in the Northern Highlands District ofnorthern Wisconsin (WIS-N), and one near Madi-son in southern Wisconsin (WIS-S); the Arctic LTER

Spatial Variation among Lakes 397

lake chain in Alaska (ARC); two lake chains of theNorth Branch of the Moose River in the AdirondackRegion Lake District (ADK-A and ADK-B); two lakechains in the Rocky Mountain Front Range regionof Colorado [Green Lakes Valley chain (GRN) andLoch Vale Watershed lake chain (VALE)]; a singleprairie lake chain in the Qu’Appelle Valley RiverBasin in Saskatchewan that includes one headwaterreservoir and three natural lakes (QUA); and a lakechain of the tributary and main-stem reservoirseries of the Tennessee River (TVA). In addition tothe TVA lake chain, dams are present on the twolower lakes in the GRN chain and in the second lakein the QUA chain.

The lake districts represent a wide range of geo-logic conditions, land use/cover, and hydrology(Table 1). Geologic settings range from glaciatedregions with thick glacial till over sedimentarydeposits to exposed Precambrian granite with littleor no till. Land cover in two districts is predomi-nantly second growth forest (WIS-N, ADK-A, andADK-B), the Rocky Mountain district is a combina-tion of forest and alpine habitats (GRN and VALE),

one district is arctic tundra with only an approxi-mately 0.5-m thaw depth in the summer months(ARC), and in three districts land use is dominatedby agriculture with some urban components (WIS-S,QUA, and TVA). Hydrology of lake chains is equallydiverse, including groundwater-dominated lakes(WIS-N), surface-water–dominated chains of allnatural lakes (WIS-S, ARC, ADK-A, ADK-B, GRN,and VALE), a combination of natural lakes andreservoirs (QUA), and all reservoirs (TVA). The lakedistricts with reservoirs appear to be relatively simi-lar hydrologically to the other chains given thatWRT in the reservoirs are well within the range ofWRTs found in the other chains (see Figure 2).

Although all lake chains have direct linear connec-tions through either surface streams or the regionalgroundwater flow system, midchain lakes were notsampled in some chains. In GRN, lake 3 (of a total offive) was not included, but when this lake wassystematically sampled for 1 y in 1985, its character-istics were found to be intermediate between theadjacent lakes in the series (N. Caine, unpublisheddata). In the QUA chain, only lakes 1, 2, 7, and 8were sampled. The number of lakes per chainranges from 3 (ADK-B and RLCH) to 13 (TVA; Table1). The two Adirondack lake chains share the samefinal lake in the chain. The total distance fromheadwater lake to the last downstream lake rangesfrom 2 km to more than 1000 km.

METHODS

Lake DatabaseLake data were assembled from either publishedlong-term studies of the lakes (WIS-N, ARC, ADK-A,ADK-B, RGRN, VALE, and QUA) or a combinationof unpublished and published data (WIS-S and TVA;see Table 1 for citations). Data records generallyextend from the early 1980s through the mid-1990sexcept for ARC, QUA, and TVA, which were sampledfrom 1991 to 1997, 1994 to 1997, and 1990 to 1994,respectively (Table 1). Analyses were based onannual open-water (nonwinter in the case of TVA)epilimnetic means, except for chlorophyll a in WIS-Nfor which we used the June through August mean.Samples were collected every 2–4 weeks for allvariables, except in WIS-N and WIS-S where cationswere sampled quarterly (excluding winter). Sampleswere collected midlake from all lakes except inADK-A, ADK-B, GRN, and VALE where outletsamples were collected.

Variables were divided into three main groups: (a)weathering products (conductivity, pH, alkalinity,and Ca); (b) dissolved nutrients and other majorions (SO4, NO3, NH4, and SiO2, measured as dis-

Figure 1. Location of the nine lake chains in NorthAmerica: the LTER lake chain in Northern HighlandsDistrict of northern Wisconsin (WIS-N), and of southernWisconsin (WIS-S); the Arctic Lakes LTER lake chain(ARC); two lake chains in the Adirondack Region lakedistrict (ADK-A and ADK-B); two lake chains in theRocky Mountain Front Range region of Colorado, theGreen Lakes valley chain (GRN), and the Loch Valewatershed lake chain (VALE); a single lake chain in theQu’Appelle Valley River Basin in Saskatchewan, Canada(QUA); and a single lake chain of a tributary and main-stem reservoir series of the Tennessee River (TVA).


Table 1. Characteristics of Lake Districts and Study References

Lake DistrictState orProvince

Lakes inLake Chaina Hydrology Geology

Land Use orLand Cover

ElevationChangeb

(m)

ChainLength(distance,km)c

ChainLength(no.lakes) Years

Northern HighlandLakesDistrict–LTERd

(WIS-N)

Wisconsin Crystal, Big Muskellunge,Sparkling, Allequash,Trout

Groundwater Sandy outwash plains of LaurentianShield sand or coarse till

,40 m of noncalcareous sandy tills andoutwash

Forested ,9 ,5 5 1982–1995

Southern WisconsinLakese (WIS-S)

Wisconsin Mendota, Monona,Waubesa, Kegonsa

Surface water Unconsolidated glacial depositsunderlain by sedimentary rocks

,50–100 m glacial till

Agricultural andurban

2 ,18 4 1982–1994

Arctic—LTERf (ARC) Alaska I-1, I-3, I-4, I-5, I-6, I-7,Toolik Lake

Surface water Glacial outwash surrounded by glacialsurfaces of ,10,000 to 100,000 y inage

Arctic tundra (cov-ered in 0.3–1.2 mof peat). Thawdepth ,0.5 m insummer

66 ,8.5 7 1991–1997

Adirondacks Regionlakesg (ADK-A,ADK-B)

New York A: Constable, Big Moose,Dart’s, Rondaxe

B: Cascade, Moss, Rondaxe

Surface water Granitic gneiss with metasedimentaryrock and heterogeneous soils

,3 m thin till

Forested A: 57B: 29

,11,8

A: 4B: 3

1983–1993

Rocky Mountainlakesh (GRN,VALE)

Colorado GRN: Green Lake 5, GreenLake 4, Green Lake 2,Lake Albion

VALE: Sky Pond, Glass Lake,The Loch

Surface water Precambrian granite, schist, and gneissNone or thin soils

GRN: subalpineVALE: alpine

GRN: 275VALE: 274

GRN: ,5VALE: ,2

GRN: 4VALE: 3

1984–1996

Qu’Appelle Valleyreservoirs andlakesi (QUA)

Saskatchewan Diefenbaker, Buffalo Pound,Katepwa, Crooked Lake

Surface waterlakes andreservoirs

Northern great plains prairies. Agricultural 100 ,315 4 1994–1997

Tennessee RiverValley Authorityreservoirsj (TVA)

Tennessee, Alabama,andKentucky

Upper tributary: Wautauga,Boone, Ft. Patrick Henry,Cherokee

Middle mainstem: FortLoudon, Watts Bar,Chickamauga, Nickajack

Lower mainstem:Guntersville, Wheeler,Wilson, PickwickLanding, Kentucky

Reservoirs Upper: Ridge and valley province,underlain by folded and faultedpaleozoic limestones, shale, andsanstone (thin soils, no till)

Middle: Cumberland section ofAppalachian Plateau, underlain byMississippian limestones andsandstone (thin soils, no till)

Lower: interior Low Province underlainby Mississippiam, Ordovician, andDevonian limestones

Soils weathered limestones

Agricultural withsome forestedareas

411 ,1000 13 1990–1994

aIn order from upstream to downstream.bChange in elevation from uppermost lake to the final lake in the chain.cDistance from headwater lake to end of the lake chain along the lake/stream network (that is, not straight-line distance).dKratz and others 1997, 1998; Magnuson and others 1990; Riera and others forthcoming; Webster and others forthcoming.eLathrop 1992; Lathrop and others 1992; R. Lathrop, Wisconsin Department of Natural Resources, unpublished data.fKling and others forthcoming.gDriscoll and others 1987, 1995; Driscoll and van Dreason 1993; Goldstein and others 1987.hBaron and Caine forthcoming; Baron and Bricker 1987; Baron 1992; Caine 1995; Caine and Thurman 1990; Campbell and others 1995; Mast and others 1990.iDixit and others forthcoming; Hall and others forthcoming; Hall and others 1999.jUnpublished data.

Spatial

Variatio

nam

on

gLak

es399

solved reactive silica); and (c) total nutrients andwater clarity (total P, total N, chlorophyll a, andSecchi depth). Specific methods and sampling de-tails are found in the data references listed inTable 1.

AnalysisEach lake in a chain was assigned a sequentialnumber starting with 1 for the lake highest in thelandscape (headwater lake). Spatial patterns alongeach lake chain were then examined using box plotsof the average of annual solute concentrationsversus lake chain number. These box plots show therange of annual values observed for each lakethroughout the period of record.

Temporal patterns in lake dynamics within indi-vidual lake chains were assessed by calculatingsynchrony. Synchrony, also called temporal coher-ence, is a measure of the similarity in interannualconcentrations between a lake pair (Magnuson and

others 1990; Kratz and others 1998; Baines andothers 1999). Synchrony for each variable wascalculated as the Pearson product-moment correla-tion coefficient (r) between time series of annualdata for every lake pair within a chain. QUA andTVA chains, which only had 4 y of data, wereexcluded from the analysis.

We first examined synchrony within individuallake chains to examine the importance of landscapeposition and to determine if nearby pairs were moresynchronous than spatially separated pairs. We re-gressed lake chain number difference (the absolutedifference between lake chain number for the lakepair) against lake pair synchrony for each lakechain. We then examined average synchrony amonglake chains. For each variable group, we regressedthe average synchrony for each variable within alake chain versus the median WRT and WS:LK foreach lake chain. Only weathering and dissolvednutrient/ion groups were used for this analysis

Figure 2. Individual charac-teristics for lake and water-shed morphometry and wa-ter residence time. WS:LK isthe watershed area to lakearea ratio (unitless); cumu-lative WS is the cumulativewatershed area (%) of lakesalong each lake chain (thuslower lakes in the chain in-clude watershed areas ofupper lakes); and WRT iswater residence time (yr).


because data for total P, total N, chlorophyll a, andSecchi depth were missing from many sites.

RESULTS

Lake Chain Spatial Patterns

To more fully understand the landscape position oflakes, we needed to first identify whether individuallake morphometry and WRT change predictablywith lake number. We found few consistent pat-terns in individual lake characteristics along lakechains (Figure 2). There was a weak tendency insome lake chains for mean depth to decrease andlake area to increase with lake chain number.However, the strongest pattern observed was anincrease in WS:LK, which supports our assumptionthat loading increases down the lake chain. We alsosaw a general decrease in WRT down the lake chain,although these trends were not present in all lakechains. In chains where WRT did not follow thistrend, the absolute range in WRT among lakes inthe chain was very small (ADK-A, ADK-B, GRN,and VALE). Cumulative watershed area obviouslyincreased along each lake chain because lake chainwatersheds are hierarchical and lower lake water-sheds include basins from upper lakes. However,watershed area shows three patterns of accumula-tion (Figure 2). In the first case, there is a gradual,proportional increase in cumulative watershed areaalong the lake chain (ADK-A, GRN, and TVA). Inthe second case, the watershed of the last lake in thechain is disproportionally large relative to the lowerlakes (WIS-N, ADK-B, and VALE). In the third case,the watershed of the first lake in the chain isdisproportionally large relative to the lower lakes(WIS-S and QUA). These patterns and other lack ofpatterns in morphometry are necessary to interpretthe role of lake number in influencing lake vari-ables.

Weathering variables, alkalinity, conductivity, andcalcium generally increased along almost all lakechains, consistent with our expectation that loadingfrom the landscape would be higher in lakes lowerin lake chains (Figure 3). However, these patternswere less developed where lakes were situated incalcium-rich tills (WIS-S) or areas with high localheterogeneity in geologic substratum (ADK-A andADK-B; Driscoll and van Dreason 1993).

Although silica is a dissolved nutrient subject tobiological uptake and driven by in-lake processes,spatial patterns were more suggestive of a nonreac-tive weathering variable than a reactive element(Figure 4). In three lake chains, WIS-N, WIS-S, andADK-A, dissolved silica concentration increased with

increasing lake chain number, and in ADK-B, theonly chain where other weathering variables actu-ally decreased due to geologic heterogeneity, silicaalso decreased. In WIS-S, silica was higher in thelower two lakes, which are shallow and likelysubject to recycling of biogenic silica from thesediments.

The remaining dissolved nutrients and reactivecompounds showed fewer patterns with landscapeposition and appeared to be sensitive to lake chain–specific or local conditions affecting internal process-ing rates. However, there were a few commonalitiesamong lake chains (Figure 4). For example, becausesulfate is of both atmospheric and geologic origin,we expected strong spatial patterns only wheresulfur-bearing minerals are present in the surround-ing landscape. Sulfate does slightly increase withlake chain number in two districts that do havesulfur-bearing minerals (WIS-S and GRN). Spatialpatterns in nitrate on the other hand were morevariable and not predictable from landscape positionalone and are not explainable with our data here.For example, patterns observed in WIS-S and QUAlake chains for nitrate may have been influenced bya combination of lake morphometry and land-useloading rates. A decrease in nitrate occurs in bothchains after the first lake. In both lake chains, thefirst lake’s watershed is not only large (making upapproximately 60% of the total chain watershedarea; Figure 2) but is dominated by agricultural landuse, which can increase nitrate inputs. In contrast,the lower lakes’ watersheds add little additionalwatershed area and, in the case of WIS-S, have lessproportional areas under agricultural production.

In contrast to our expectations based on mass-balance assumptions, concentrations of total nutri-ents [TP and TN (not shown)] and measures of algalbiomass (chlorophyll) increased with increasing lakenumber in two of the surface-drained lake chainsbut not the groundwater-dominated lake chain(Figure 5). Patterns for surface-drained chains TVAand ARC showed complex patterns that are notexplainable given our data. Secchi depth showedthe most consistent pattern, with decreases along alllake chains except ARC. As a measure of waterclarity, Secchi depth is the net result of chlorophyllconcentration, colored dissolved organic carbon(DOC), and suspended sediments—all of whichappear to be related to lake number.

Lake Chain Temporal Dynamics

Across all lake chains, weathering variables werethe most synchronous among lake pairs within achain, followed by the dissolved nutrients and ions,


and then total nutrients and algal variables (Table 2,Total row). This pattern fits well with results fromother studies of synchrony (Magnuson and others1990; Kratz and others 1998; Webster and othersforthcoming) that have shown nonreactive ions tobe the most synchronous and biological variables tobe the least synchronous.

At the scale of individual lake chains, our expecta-tions that lakes with strong spatial gradients shouldhave low synchrony was not proven. Landscape

position proved to be important in determiningsynchrony between lake pairs only for some vari-ables related to weathering (Table 3). Here, regres-sions of lake pair synchrony against the difference intheir lake chain number (suggesting that lakescloser in the chain will be more temporally similar)were significant for only three lake chains, WIS-N,WIS-S, and ARC. These three lake chains have thelongest average WRT, close to or exceeding 1 y(Figure 2). In addition, all of these lake chains had

Figure 3. Box plots of an-nual averages of three of thevariables used as a measureof weathering (ALK, alkalin-ity; COND, conductivity;Ca, calcium concentration)for each lake versus lakechain number. Note that forthe GRN and QUA lakechains, some intermediatelakes were not sampled sothat the maximum lakechain number is larger thanthe maximum number oflakes plotted (see text andFigure 2).


the largest range of WRT among lakes within achain. Whether these large differences in WRTamong the lake chains, or some other feature oflandscape position cause the differences in syn-chrony at these sites remains to be seen. In theremaining sites, however, the high average syn-chrony between lake pairs (Table 2) indicates thatsynchrony across these lake chains was unrelated tolandscape position.

At the scale of multiple lake chains, if we compareaverage lake chain synchrony (Table 2, Total col-umn), there appears to be large differences amonglake chains. To examine whether hydrologic charac-teristics of the lake chains influenced their temporalpatterns, we regressed synchrony for each variablewithin the group (except total nutrients and algae)against the median log-transformed WRT and themedian WS:LK (Figure 6) for each lake chain. We

Figure 4. Box plots of an-nual averages of three dis-solved nutrient and ion vari-ables (SO4, sulfate; NO3-N,nitrate; DSi, dissolved silica)for each lake chain versuslake chain number. Notethat for the GRN and QUAlake chains, some inter-mediate lakes were notsampled so that the maxi-mum lake chain number islarger than the maximumnumber of lakes plotted (seetext and Figure 2).


found that synchrony of both groups of variables(nonreactive solutes and dissolved nutrients/ions)decreased with increasing WRT and increased withincreasing WS:LK. However, only the relationshipswith the dissolved nutrient/ion variables were sig-nificant, suggesting that in longer WRT lake chains,longer residence times lead to higher processing ofthese variables, and more differences among lakeswithin a chain.

DISCUSSION

We have gained new insight into the spatial andtemporal patterns of lakes across the landscapewithin and among lake districts. At the scale of thelake chain itself, many spatial patterns in lakevariables appear to be driven by landscape positionas measured simply by lake chain number, buttemporal patterns are not, except in a few cases. At

Figure 5. Box plots of an-nual averages for total nutri-ents and water clarity vari-ables [total phosphorusconcentration (TP), chloro-phyll a concentration (Chl),and Secchi depth]. Note thatfor the QUA lake chain,some intermediate lakeswere not sampled so thatthe maximum lake chainnumber is larger than themaximum number of lakesplotted (see text and Figure2). Data are not available forADK-A, ADK-B, GRN, andVALE.

Table 2. Average Synchrony (r) Values for Each Lake Chain

District

Weathering Variables Dissolved Nutrients and Ions Total Nutrients and Water Clarity

TotalAlk Cond PH Ca SO4 NO3-N NH4-N DSi TP TN Chl Secc

WIS-N 0.835 0.685 0.805 0.811 0.185 0.077 0.585 0.090 0.139 0.573 0.469 0.227 0.457WIS-S 0.094 0.717 0.545 0.758 0.638 0.120 0.331 0.563 0.674 0.311 0.493 0.292 0.461ARC 0.244 0.536 0.595 0.065 0.435 0.691 0.863 0.472 0.488ADK-A 0.618 0.921 0.399 0.627 0.946 0.926 0.492 0.658 0.698ADK-B 0.283 0.698 0.335 20.032 0.865 0.902 0.325 0.621 0.500GRN 0.875 0.801 0.814 0.850 0.577 0.280 0.277 0.639VALE 0.602 0.866 0.798 0.906 0.961 0.881 0.447 0.886 0.793Total 0.507 0.746 0.613 0.569 0.658 0.554 0.507 0.516 0.407 0.442 0.478 0.260

Alk, alkalinity; Cond, conductivity; DSi, dissolved reactive silica; TP, total phosphorus; TN, total nitrogen; Chl, chlorophyll a concentration, and Secc, Secchi depth. Lake chainacronyms are same as for Table 1. Blank cells are missing data.


the scale of comparisons across different lake dis-tricts, however, lake chain temporal dynamics mea-sured by average lake chain synchrony is influencedby hydrologic differences among lake chains, asmeasured by average WRT. We explore the signifi-cance of these findings below.

Lake Chain Spatial PatternsThe patterns observed in nine diverse lake chainssupport our contention that lake districts are spa-tially organized by position within a flow system.However, we were concerned that lake morphom-

etry would covary with lake number, thus confound-ing the effects of landscape position. Indeed, whenexamining lake landscape position across hundredsof lakes within a single lake district, Riera and others(forthcoming) found lake morphometry to varygradually with landscape position. In our analysis,however, lake morphometry did not change predict-ably with lake number, thus morphometric patternsgenerally were not responsible for the observedspatial patterns. It is important to note, however,that this general lack of a clear pattern may be dueto the limitations of our data, because we focused on

Table 3. Regressions of Lake Pair Synchrony versus Differencein the Lake Number (of the Lake Pair) per Chain

LakeChain

Weathering Variables Dissolved Nutrients and Ions Total Nutrients and Water ClarityTotal(%)Alk Cond pH Ca SO4 NO3-N NH4-N DSi TP TN Chl Secc

WIS-N a ns 1 a ns ns 1 ns ns ns ns ns 17WIS-S ns b ns 1 ns ns ns ns 1 ns ns ns 8ARC ns ns ns c ns 1 1 ns 13ADK-A 1 1 ns 1 1 1 ns 1 0ADK-B ns 1 ns ns 1 1 ns 1 0GRN 1 1 1 1 1 ns ns 0VALE 1 1 1 1 1 1 1 1 0Total (%) 14 14 0 29 0 0 0 0 0 0 0 0

ns, nonsignificant. 1 indicates that there is no relationship between lake number difference and average synchrony, but that synchrony for all lake pairs is high, and the majorityof lake pair coherence values are .0.50. Negative synchrony values were treated as if synchrony were 0 and were not used to develop the regression equations. Percentages in theTOT columns represent the percent significant regressions (P # 0.10) with lake chain number difference. Alk, alkalinity; Cond, conductivity; DSi, dissolved reactive silica; TP, totalphosphorus; TN, total nitrogen; Chl, chlorophyll a concentration; Secc, Secchi depth. Lake chain acronyms are same as for Table 1. Blank cells are missing data.aP # 0.01.bP # 0.10.cP # 0.001.

Figure 6. Average syn-chrony (r) of each variablewithin a group versus log-transformed median lakechain water residence time[Ln (WRT), yr] and versusmedian lake chain water-shed to lake area ratio(WS:LK, unitless). In thefirst panel, the four weath-ering variables for each lakechain are plotted. In the sec-ond panel, the four dis-solved nutrient and ion vari-ables are plotted.


only one lake chain per lake district (except for twosites), and often only a portion of the complete lakechain.

We found a surprisingly large number of commonpatterns in a wide range of variables and found thatdifferences in hydrologic setting (WRT in particular)among lake districts had fewer effects on spatialpatterns in lake variables than we expected. Ingeneral, nonreactive variables, such as those subjectto weathering, increased with increasing lake num-ber, and patterns were not related to WRT. Excep-tions occurred either where geologic heterogeneitywas large relative to lake chain length and had anoverriding influence on interlake variability (ADK-Aand ADK-B) or where products of weathering werealready high (WIS-S). For the more reactive solutes,such as dissolved nutrients and ions, we foundfewer patterns that could be explained with ourdata. However, we found surprisingly commonpatterns for total nutrient (TP and TN) and algal(chlorophyll) variables, which increased with lakenumber, with the exception of the lake chain(WIS-N) where groundwater flow was the domi-nant hydrologic flowpath between lakes, and thearctic lake chain. We expected fairly reactive vari-ables such as these to either show no pattern or todecrease with increasing lake number due to thehigh retention of nutrients that is known to occur inlakes (Rechkow and Chapra 1983). Finally, waterclarity was strongly related to lake number, reflect-ing changes in chlorophyll, turbidity, or DOC, all ofwhich appear to be associated with landscape posi-tion. The generality of these spatial patterns shouldhelp account for some of the unexplained amonglake variability that is observed within lake districts.In addition, these results suggest that certain fea-tures of the landscape position concept are robustenough to span a wide range of seemingly disparatelake types.

Although the presence of spatial patterns wassimilar across lake chains, we did see differences inthe strength of the patterns related to hydrogeomor-phic setting and land use. Across lake chains, spatialpatterns were strongest where there was a gradientin weatherable soils (GRN), strong gradients acrossgeologic boundaries (ADK-B, although the trendswere opposite from others), or long, slow flowpathsthrough groundwater (WIS-N). For example, thesteep elevational gradient of GRN is associated withchanges in geologic and soil conditions along thechain that lead to strong gradients in weathering,dissolved nutrient, and ion levels. Both QUA andWIS-S chains have land-use patterns that may leadto some of the observed patterns, especially indissolved and total nutrients, chlorophyll, and Sec-

chi depth. Finally, in TVA, it is not surprising thatthere were strong spatial patterns across such a longchain of reservoirs that are tightly linked hydrologi-cally, although we cannot explain the patterns givenour data. In contrast, spatial patterns were weakestwhere there was little weatherable material on thelandscape, resulting in very dilute lakes (ADK-Aand VALE). These results suggest that although it isimportant to recognize the tight coupling of aquaticecosystems with their surrounding landscape, thestrength of this interaction varies along geologic andhydrologic gradients among lakes districts.

Lake Chain Temporal PatternsSynchrony is a measure of the degree to which lakepairs within a lake district behave similarly throughtime (Magnuson and others 1990). High synchronyis generated when lakes respond similarly to acommon driver. However, for synchrony to bedetected, lakes must vary through time at thetemporal scale of the synchrony measure (in ourcase, annual). In other words, if a particular constitu-ent does not change through time, the correlationcoefficient (our measure of synchrony) between thetwo lakes will be low. For lakes, climate is one of themore important drivers that lead to high synchrony(Magnuson and others 1990; Webster and othersforthcoming). Thus, in this analysis, lake chainswith high synchrony are likely being strongly drivenby a climatic driver that varies on an annual scale.For example, snowmelt is recognized as being oneof the major drivers determining interannual andseasonal dynamics in two of the lake districts, theRocky Mountain lakes and the Adirondack Regionlakes (Driscoll and others 1987; Baron 1992; Camp-bell and others 1995). Although other strong cli-matic or regional drivers of lakes have been identi-fied from other studies (that is, droughts, El Nino,atmospheric deposition), their role in determiningsynchrony is only beginning to be explored (Dillonand others 1997; Webster and others forthcoming).

Because of the strong spatial patterns observed inmost lake chains, we expected synchrony withinlake chains to also be related to landscape position.This was not the case, and synchrony was related tolandscape position only for some weathering vari-ables and only in those lake chains with averageWRTs close to 1 y and with large ranges among thelakes. Surprisingly, for many lake chains and vari-ables, average synchrony among most lakes in thechain was high. It is important to note that anymeasure of synchrony is going to be strongly scaledependent. Because we have measured synchronyat the annual scale, our conclusions are most appro-priate at that scale. It is likely that at other temporal


scales, other factors than the ones identified herecould influence synchrony (Baines and others 1999).Further analysis of synchrony calculated at addi-tional temporal scales is clearly needed.

A Landscape Perspective for LakesOur analysis suggests that a landscape perspectivefor lakes based on landscape position is surprisinglyrobust. Although these ideas initially evolved in thegroundwater dominated setting of northern Wiscon-sin, we found that the concept generally was appli-cable to a variety of hydrogeomorphic settings, fromlow-till subalpine lakes to midcontinent reservoirs.Our results indicate that lake ecosystems should beviewed in a spatial context across the landscape. Wehave addressed the spatial organization of lakedistricts through a landscape view that is longitudi-nal and that uses surface and subsurface flowpathsas a gradient that drives lake variability in additionto the known drivers of lake variability that includegeology (till depth and spatial heterogeneity), landuse, and hydrology (water residence times andsurface vs groundwater flow). Our discovery ofcommon patterns in such disparate settings suggeststhis is a fruitful area for future research.

The strength of using lake number as a metric forlandscape position is its simplicity and its ability tocapture variability of a whole suite of lake character-istics. However, this strength is also a weakness andlimits its use in some key ways. First, the landscapeposition concept could benefit from considering amore realistic model of lake networks as dendriticrather than longitudinal. The importance of thisdistinction recently has been recognized for bothstreams (Fisher 1997) and lakes (Riera and othersforthcoming). The lake order scheme used by Rieraand others (forthcoming), analogous to stream or-der, is one way to improve on lake number becauseit incorporates inputs from multiple stream and lakesources. Second, we have assumed that patternsalong lake chains are gradual, an assumption thatclearly was not universal. Discontinuities along lakechains may be caused by lakes that ‘‘reset’’ thespatial template and represent important sinks orsources of material to downchain lakes. Third, muchwork remains to be done to map biological commu-nities and ecosystem-level processes onto thesespatial templates of hydrology, water flow, mor-phometry, and chemical and nutrient composition.Applications of the principles of biogeography clearlycould benefit from this landscape approach (Magnu-son and others 1998; Lewis and Magnuson forth-coming). Finally, we have limited our analysis ofsynchrony to the annual time scale. Resolution ofthe data at different time scales better suited to the

water residence time of the faster flushing systemsshould provide interesting insights. Although ourresults presented here are promising, further refine-ments of the landscape position idea are clearlyneeded.

Although limnologists have long been interestedin regional patterns in lake attributes, only recentlyhave they considered lakes as connected and orga-nized across the landscape, rather than as spatiallyindependent entities. In this way, the concept oflake landscape position pays obvious homage to theRiver Continuum Concept (Vannote and others1980). But, rather than borrow the mechanisms anddetailed predictions from this concept, we borrowprimarily the perspective of examining spatial rela-tionships among aquatic ecosystems and their sur-rounding landscape. Hierarchically ordered systemsof streams and lakes result from spatial processesthat occur at broad landscape scales. Both conceptsdescribe the position of lakes or streams along theflowpath of water moving from upland to lowlandreaches within medium to large watersheds. Also,both address the complex interactions betweeninternal and external factors influencing aquaticprocesses, which change in relative strength fromupstream to downstream. One of the major differ-ences between these two concepts may lie in thevastly different WRTs of lakes and streams, whichare quantified and perceived of differently by lakeand stream ecologists. For example, a nutrientresidence time in lakes may be analogous to anutrient spiral length in streams (Newbold andothers 1982). Other analogies between differentconcepts for lakes and streams may be fruitful toexplore as well. Interestingly, a reservoir lake chainmay represent some intermediate between thesetwo concepts and suggests that comparative studiesand paradigms that span the distinctions betweenlake, reservoir, and stream ecosystems may bepossible (for example, see Soballe and Kimmel1987).

Future Directions for Spatial Analyses ofLake EcosystemsOur analysis promotes the perspective of lake ecosys-tems imbedded in a landscape matrix interactingwith each other and with other terrestrial andaquatic ecosystems. In this context, the landscapeposition idea not only can but should incorporateother aquatic ecosystems into its framework. Wehave treated streams and groundwater connectionsas functioning essentially as nonreactive materialconduits (but see Kling and others forthcoming),which is clearly an oversimplification of the actualprocesses occurring. At larger spatial scales, it is


obvious that lakes, streams, and wetlands often areconnected spatially and functionally, and, in fact,the distinctions between them sometimes can evenbe difficult to discern. Unfortunately, with only afew exceptions, these aquatic ecosystems currentlyare being studied as isolated entities as reflected inthe separate journals, conferences, and scientificsocieties that each subdiscipline uses (Lewis andothers 1995; Gorham 1996). And, in the few caseswhere more than one ecosystem type has beenstudied (usually only two), it is usually to examinethe effect of one on the other (Gorham 1996),rather than to integrate both into a single concep-tual framework.

There are two major needs in the immediatefuture, one concerns data and the other concernsconcepts. The data needs are to quantify the majorprocesses and fluxes of material between all compo-nents that include lakes, streams, wetlands, andland, all put in a mass-balance context. In our study,we have examined the question of spatial processesamong lakes only using the mass-balance modelconceptually. Unfortunately, we could not quantifythe major fluxes or compartments to solve themass-balance equations, and we only were able toinfer process from pattern. We are aware of onlyone study that has adequate data to at least examinethe processing that occurs between lakes and con-necting streams (Kling and others forthcoming),which is an important next step. Having the aboveinformation, we should be able to develop newconcepts that address how the ‘‘patterns’’ of freshwa-ter ecosystems (for example, lake features, connec-tions to other aquatic ecosystems, corridors, etc.)influence the ‘‘processes’’ in them (primary produc-tion, nutrient cycling, community dynamics, etc.)across a wide range of spatial scales. There are manytools, approaches, and perspectives available fromthe discipline of landscape ecology that may helpaquatic ecologists unify the study of aquatic ecosys-tems, at least at the landscape scale (Turner 1989;Fisher 1994; Magnuson and Kratz and others). Thisneed for a comprehensive perspective is critical notonly to integrate the science of freshwater ecologybut also to aid in the management, conservation,and restoration of aquatic ecosystems.

ACKNOWLEDGEMENTS

This article is a result of a National Science Founda-tion (NSF)—sponsored Long Term Ecological Re-search Intersite comparison workshop ‘‘EcologicalVariability and Organization of Lake Districts’’ held5–7 February 1997 at the University of WisconsinTrout Lake Station. We thank all participants at theworkshop for dynamic and fruitful discussions that

led to this synthesis. In addition, thanks to J.J.Magnuson and S.B. Baines for additional discus-sions and insights on these ideas. Helpful reviews ofearlier drafts were provided by M. Bremigan, K.Cheruvelil, S. Nelson, O. Sarnelle, S. Walsh, andtwo anonymous reviewers. In addition, we thankthe many workers who have contributed to thecollection of these extensive databases, and wethank the following personnel in particular. PASthanks B.J. Benson and S.B. Baines with help withdata retrieval of WIS-N data. D.S.W. thanks K.Johnston for help with retrieval of TVA data. P.A.B.thanks C. Driscoll and W. Kretser (Adirondack LakesSurvey Corporation) for providing access to Adiron-dack long-term monitoring data. J.S.B. thanks E.Allstott and M. Bashkin for help with data retrievalof VALE data. G.W.K. acknowledges G. Kipphut,W.J. O’Brien, M. Miller, and NSF grants 9211775,9615949, and 9553064 for ARC data. Research byP.R.L. was supported by the Natural Sciences andEngineering Research Council of Canada, the De-partment of Fisheries and Oceans, and SaskWaterCorporation.

REFERENCES

Baines SB, Webster KE, Kratz TK, Carpenter SR, Magnuson JJ.Synchronous behavior of temperature, calcium and chloro-phyll in lakes of northern Wisconsin. Ecology. Forthcoming.

Baron J, editor. 1992. Biogeochemistry of a subalpine ecosystem:Loch Vale watershed. New York: Springer-Verlag. 247 pp.

Baron JS, Bricker OP. 1987. Hydrologic and chemical flux in LochVale Watershed, Rocky Mountain National Park. In: AverettRC, McKnight DM, editors. Chemical quality of water and thehydrologic cycle. Ann Arbor, MI: Lewis Publishers.

Baron JS, Caine N. The temporal coherence of two alpine lakebasins of the Colorado front range. USA Freshwater Biol.Forthcoming.

Birge EA, Juday C. 1911. The inland lakes of Wisconsin: thedissolved gases of the water and their biological significance.Wisconsin Geol Nat Hist Bull 22:

Campbell DH, Clow DW, Ingersoll GP, Mast MA, Spahr NE, TurkJT. 1995. Processes controlling the chemistry of two snowmelt-dominated streams in the Rocky Mountains. Water Res Res31:2811-21.

Caine N. 1995. Temporal trends in the quality of streamwater inan alpine environment: Green Lakes Valley, Colorado FrontRange, U.S.A. Geographiska Annaler 77A(4):207–20.

Caine N, Thurman EM. 1990. Temporal and spatial variations inthe solute content of an alpine stream, Colorado Front Range.Geomorphology 4:55–72.

Dillon PJ, Kirchner WB. 1975. The effects of geology and land useon the export of phosphorus from watersheds. Water Res9:135–48.

Dillon PJ, Molot LA. 1996. Long-term phosphorus budgets andan examination of a steady-state mass-balance model forcentral Ontario. Water Res 30(10):2273–80.

Dillon PJ, Molot LA. 1997. Dissolved organic and inorganiccarbon mass balances in central Ontario lakes. Biogeochemis-try 36(1):29–42.


Dillon PJ, Molot LA, Futter MN. 1997. A note on the effect of ElNino-related drought on the recovery of acidified lakes. Envi-ron Monitoring Assess 46:105–11.

Dixit AS, Hall RI, Leavitt PR, Smol JP, Quinlan R. Effects ofsequential depositional basins on lake response to urban andagricultural pollution: a paleolimnological analysis of theQu’Appelle Valley, Saskatchewan, Canada. Freshwater Biol.Forthcoming.

Drever JI. 1982. The geochemistry of natural waters. UpperSaddle River, NJ: Prentice Hall. 388 pp.

Driscoll CT, Postek KM, Kretser W, Raynal DJ. 1995. Long-termtrends in the chemistry of precipitation and lake water in theAdirondack region of New York, U.S.A. Water, Air Soil Pollut85:583–8.

Driscoll CT, Yatsko CP, Unangst FJ. 1987. Longitudinal andtemporal trends in the water chemistry of the North Branch ofthe Moose River. Biogeochemistry 3:37–61.

Driscoll CT, van Dreason R. 1993. Seasonal and long-termtemporal patterns in the chemistry of Adirondack lakes. Water,Air, Soil Pollut 67:319–44.

Eilers JM, Glass GE, Webster KE, Rogalla JA. 1983. Hydrologiccontrol of lake susceptibility to acidification. Can J Fish AquatSci 40:1896–904.

Fisher SG. 1994. Pattern, process and scale in freshwater ecosys-tems: some unifying thoughts. In: Hildrew A, Giller P, RafaelliD, editors. Pattern, process and scale in aquatic ecology.Blackwell Scientific.

Fisher SG. 1997. Creativity, idea generation, and the functionalmorphology of streams. J North Am Benthol Soc 16(2):305–18.

Fisher SG, Grimm NB. 1991. Streams and disturbance: arecross-ecosystem comparisons useful? In: Cole J, Lovett G,Findlay S, editors. Comparative analyses of ecosystems. NewYork: Springer-Verlag. 375 pp.

Frissell CA, Liss WJ, Warren CE, Hurley MD. 1986. A hierarchicalframework for stream habitat classification: viewing streams ina watershed context. Environ Manage 10(2):199–214.

Forbes SA. 1887. The lake as a microcosm. Bulletin ScienceAssociation of Peoria, IL. 1887:77–87.

Forel FA. 1892. Lac Leman: monographie limnologique. Lau-sanne: Rouge.

Gibson CE, Wu Y, Smith SJ, Wolfe-Murphy SA. 1995. Synopticlimnology of a diverse geological region: catchment and waterchemistry. Hydrobiologia 306(3):213–27.

Goldstein RA, Gherini SA, Driscoll CT, April R, Schofield CL,Chen CW. 1987. Lake-watershed acidification in the NorthBranch of the Moose River. Introduction. Biogeochemistry3:5–20.

Gorham E. 1996. Wetlands: an essential component of curriculain limnology. In: Freshwater ecosystems: revitalizing educa-tional programs in limnology. Washington, DC: National Re-search Council.

Hall RI, Leavitt PR, Quinlan R, Dixit AS, Smol JP. 1999. Effects ofagriculture, urbanization and climate on water quality in theNorthern Great Plains. Limnol Oceanogr 44:739–56.

Hall RI, Leavitt PR, Dixit AS, Quinlan R, Smol JP. Limnologicalsuccession in reservoirs: a paleolimnological comparison oftwo methods of reservoir formation. Can J Fish Aquat Sci.Forthcoming.

Hynes HBN. 1975. The stream and its valley. Verh Int VereinLimnol 19:1–15.

Juday C, Meloche VW. 1943. Physical and chemical evidencerelating to the lake basin seal in certain areas of the Trout Lakeregion of Wisconsin. Trans Wis Acad Sci Arts Lett 35:157–74.

Junk WJ, Bayley PB, Sparks RE. 1989. The flood pulse concept inriver-floodplain systems. In: Dodge DP, editor. Proceedings ofthe International Large River Symposium. Can J Fish Aquat Sci106:110–27.

Kling GW, Kipphut GW, Miller MC. 1991. Arctic lakes and riversas gas conduits to the atmosphere: implications for tundracarbon budgets. Science 251:298–301.

Kling GW, Kipphut GW, Miller MM, O’Brien WJ. The influenceof spatial position, morphology, and downslope processing onthe coherence of lake properties. Freshwater Biol. Forthcom-ing.

Kratz TK, Webster KE, Bowser CJ, Magnuson JJ, Benson BJ.1997. The influence of landscape position on lakes in northernWisconsin. Freshwater Biol 37:209–17.

Kratz TK, Soranno PA, Baines SB, Benson BJ, Magnuson JJ, FrostTM, Lathrop RC. 1998. Interannual synchronous dynamics innorth temperate lakes in Wisconsin, USA. In: George DG andothers, editors. Management of lakes and reservoirs duringglobal climate change. The Netherlands: Kluwer AcademicPublishers.

Lathrop RC. 1992. Nutrient loadings, lake nutrients, and waterclarity. In: Kitchell JF, editor. Food web management: a casestudy of Lake Mendota. New York: Springer-Verlag. p 69–96.

Lathrop RC, Nehls SB, Brynildson CL, Plass KR. 1992. The fisheryof the Yahara Lakes. Technical Bulletin nr 181. Department ofNatural Resources, Madison, WI.

Lewis WM, Chisholm S, D’Elia C, Fee E, Hairston NG, Hobbie J,Likens GE, Threlkeld S, Wetzel R. 1995. Challenges for limnol-ogy in North America: an assessment of the discipline of the1990s. Am Soc Limnol Oceanogr Bull 4(2):1–20.

Lewis DB, Magnuson JJ. Landscape spatial patterns in freshwatersnail assemblages across Northern Highland catchments. Fresh-water Biol. Forthcoming.

Likens GE. 1984. Beyond the shoreline: a watershed-ecosystemapproach. Verh Int Verein Limnol 22:1–22.

Likens GE, editor. 1985. An ecosystem approach to aquaticecology: Mirror Lake and its environment. New York: Springer-Verlag. 516 pp.

Magnuson JJ, Kratz TK. Lakes in the landscape: Approaches toregional limnology. Verh Int Verein Limnol. Forthcoming.

Magnuson JJ, Benson BJ, Kratz TK. 1990. Temporal coherence inthe limnology of a suite of lakes in Wisconsin, U.S.A. Freshwa-ter Biol 23:145–59.

Magnuson JJ, Tonn WM, Banerjee A, Toivonen J, Sanchez O,Rask M. 1998. Isolation vs. extinction in the assembly of fishesin small northern lakes. Ecology 79(8):2941–56.

Mast MA, Drever JI, Baron J. 1990. Chemical weathering in theLoch Vale Watershed, Rocky Mountain National Park, Colo-rado. Water Resources Res 26:2971–8.

Minshall GW, Cummins KW, Peterson RC, Cushing CE, BrunsDA, Sedell JR, Vannote RL. 1985. Developments in streamecosystem theory. Can J Fish Aquat Sci 42:1045–55.

Molot LA, Dillon PJ. 1993. Nitrogen mass balances and denitrifi-cation rates in central Ontario lakes. Biogeochemistry 20(3):195–212.

Newbold JD, O’Neill RV, Elwood JW, VanWinkle W. 1982.Nutrient spiraling in streams: implications for nutrient limita-tion and invertebrate activity. Am Nat 120:628–52.


Omernik J. 1976. The influence of land use on stream nutrientlevels. Corvallis, OR: US Environmental Protection Agency.

Reckhow KH, Chapra SC. 1983. Engineering approaches for lakemanagement. Volume 1. Data analysis and empirical model-ing. Boston: Butterworth Publishers. 340 pp.

Riera JL, Magnuson JJ, Kratz TK, Webster KE. A geomorphictemplate for the analysis of lake districts applied to theNorthern Highland Lake District, Wisconsin, U.S.A. Freshwa-ter Biol. Forthcoming.

Rochelle BP, Liff CI, Campbell WG, Cassell DL, Church MR, NuszRA. 1989. Regional relationships between geomorphic/hydrologic parameters and surface water chemistry relative toacidic deposition. J Hydrology 112:103–20.

Schindler DW, Beaty KG, Fee EJ, Cruikshank DJ, DeBruyn ED,Findlay DL, Linsey GA, Shearer JA, Stainton MP, Turner MA.1990. Effects of climatic warming on lakes of the central borealforest. Science 250:967–70.

Schindler DW, Curtis PJ, Bayley SE, Parker BR, Beaty KG,Stainton MP. 1997. Climate-induced changes in the dissolvedorganic carbon budgets of boreal lakes. Biogeochemistry 36(1):9–28.

Soballe DM, Kimmel BL. 1987. A large-scale comparison offactors influencing phytoplankton abundance in rivers, lakes,and impoundments. Ecology 68(6):1943–54.

Turner MG, Gardner RH. 1989. Quantitative methods in land-scape ecology. New York: Springer-Verlag.

Vannote RL, Minshall GW, Cummins KW, Sedell JR, Cushing CE.1980. The river continuum concept. Can J Fish Aquat Sci37:130–7.

Vollenweider RA. 1969. Possibilities and limits of elementarymodels concerning the budget of substances in lakes. ArchHydrobiol 66(1):1–36.

Ward JV, Stanford JA. 1983. The serial discontinuity concept oflotic ecosystems. In: Fontaine TD, Bartell SM, editors. Dynam-ics of lotic ecosystems. Ann Arbor Science Publishers. 494 pp.

Webster KE, Kratz TK, Bowser CJ, Magnuson JJ, Rose WJ. 1996.The influence of landscape position on lake chemical responses todrought in northern Wisconsin. Limnol Oceanogr 41(5):977–84.

Webster KE, Soranno PA, Baines SB, Kratz TK, Bowser CJ, DillonPJ, Campbell P, Fee EJ, Hecky RE. Structuring features of lakedistricts: geomorphic and landscape controls on lake chemicalresponses to drought. Freshwater Biol. Forthcoming.

Wiley MJ, Osborne LL, Larimore RW. 1990. Longitudinal struc-ture of an agricultural prairie river system and its relationshipto current stream ecosystem theory. Can J Fish Aquat Sci47:373–83.

Windolf J, Jeppson E, Jensen JP, Kristensen P. 1996. Modelling ofseasonal variation in nitrogen retention and in-lake concentra-tion: a four-year mass balance study in 16 shallow Danishlakes. Biogeochemistry 33(1):25–44.

Winter TC. 1977. Classification of the hydrologic settings of lakesin the North Central United States. Water Resources Res13(4):753–67.


spatial variation among lakes within landscapes: ecological … · 2006-09-01 · viewed as an...

Documents