Freshwater Biology (2001) 46; 1503-1517

Inter-biome comparison of factors controlling stream

metabolism

P. J. MULHOLLAND,* C. S. FELLOWS,t J. L. TANK4 N. B. GRIMM,§ J. R. WEBSTER,!S. K. HAMILTON,** E. MARTI,tt L. ASHKENAS,# W. B. BOWDEN,§§ W. K. DODDS,fIW. H. McDOWELL,*** M. J. PAULttt and B. J. PETERSONftt*Environmental Sciences Division, Oak Ridge National Laboratory, Oak Ridge, TN, U.SA.^Centre for Catchment and In-Stream Research, Faculty of Environmental Sciences, Griffith University, Nathan,Queensland, Australia^Department of Biological Sciences, University of Notre Dame, Notre Dame, IN, U.S A.^Department of Biology, Arizona State University, Tempe, AZ, U.SA.^Department of Biology, Virginia Polytechnic Institute and State University, Blacksburg, VA, U.SA.**Kellogg Biological Station, Michigan State University, Hickory Comers, MI, U.SA. .^Centre d'Estudis Avancats de Blanes (CSIC), Cami de Sta. Barbara s/n, Blanes, Girona, Spain•^Department of Fisheries and Wildlife, Oregon State University, Corvallis, OR, U.SA.§§Landcare Research, PO Box 69, Lincoln, New Zealand^Division of Biology, Kansas State University, Manhattan, KS, U.SA.

> ***Department of Natural Resources, James Hall, University of New Hampshire, Durham, NH, U.SA.tttlnsfitMfe of Ecology, University of Georgia, Athens, GA, U.SA.^Ecosystems Center, Marine Biological Laboratory, Woods Hok, MA, U.SA. '

SUMMARY

1. We studied whole-ecosystem metabolism in eight streams from several biomes in NorthAmerica to identify controls on the rate of stream metabolism over a large geographicrange. The streams studied had climates ranging from tropical to cool-temperate and fromhumid to arid and were all relatively uninfluenced by human disturbances.2. Rates of gross primary production (GPP), ecosystem respiration (R) and net ecosystemproduction (NEP) were determined using the open-system, two-station diurnal oxygenchange method.3. Three general patterns in metabolism were evident among streams: (1) relatively highGPP with positive NEP (i.e. net oxygen production) in early afternoon, (2) moderateprimary production with a distinct peak in GPP during daylight but negative NEP at alltimes and (3) little or no evidence of GPP during daylight and a relatively constant andnegative NEP over the entire day. ',4. Gross primary production was most strongly correlated with photosynthetically activeradiation (PAR). A multiple regression model that included log PAR and stream watersoluble reactive phosphorus (SRP) concentration explained 90% of the variation in log GPP.5. Ecosystem respiration was significantly correlated with SRP concentration and sizeof the transient storage zone and, together, these factors,explained 73% of the variationin R. The rate of R was poorly correlated with the rate of GPP.6. Net ecosystem production was significantly correlated only with PAR, with 53% of thevariation in log NEP explained by log PAR. Only Sycamore Creek, a desert stream inArizona, had positive NEP (GPP: R > I), supporting the idea that streams are generally netsinks rather than net sources of organic matter.

Correspondence: P. J. Mulholland, Environmental Sciences Division, Oak Ridge National Laboratory, PO Box 2008, Oak Ridge,TN 37 831, U.S.A. E-mail: mulhollandpj@ornl.gov

© 2001 BlackweU Science Ltd ' . 1503

1504 P.J. Mulhplland et al.

7. Our results suggest that light, phosphorus concentration and channel hydrauh'csare important controls on the rate of ecosystem metabolism in streams over veryextensive geographic areas.

Keywords: inter-biome, metabolism, primary production, respiration, stream

Introduction

Primary production and respiration are importantdeterminants of ecosystem biomass and trophic struc-ture as well as important drivers of nutrient cyclingand other ecosystem processes. Primary productionrepresents the organic matter supply produced withinthe ecosystem whereas respiration provides an indi-cation of total consumption of organic matter suppliedby sources both within (autochthonous) and outside(allochthonous) the ecosystem. Broad comparison ofthe patterns of gross primary production (GPP) andrespiration exhibited by ecosystems in differentbiomes is an important approach to the determinationof fundamental controls on these processes (Cole,Lovett & Findlay, 1991; Schlesinger, 1997).

Streams present unique challenges for the meas-urement of GPP and total respiration (R). For exam-ple, the use of chambers is problematic because ofdifficulties in incorporating realistic flow regimes andhabitat complexity (Bott, 1996). These problems areparticularly challenging in small streams with highspatial heterogeneity in water velocity and sedimenttypes. Open-system oxygen change methods weredeveloped to measure the whole-ecosystem rate ofGPP and respiration in running waters (Odum,w1956;Hoskin, 1959; Hall, 1972). The open-system methodscircumvent many of the problems associated withhigh spatial variability but require good estimates ofair-water oxygen exchange rates. These methods havebeen used primarily in streams with a relatively highrate of primary productivity, where diurnal changesin dissolved oxygen are relatively large/Recentrefinements to the two-station diurnal oxygen changemethod by Marzolf, Mulholland & Steinman (1994)and Young & Huryn (1998) have improved theperformance of the open-system approach in small,relatively unproductive streams.

This study examines broad-scale controls on therate of stream metabolism across biomes using thecomparative ecosystem approach. We report metabo-lism measurements made using the open-systemmethod in streams in several different biomes of

North America. These measurements were made aspart of the lotic intersite nitrogen experiment (LJNX),a study of nitrogen uptake and cycling using tracer15N additions (Peterson et al, 2001).

Study sites .

We measured metabolism in eight first, second andthird order streams across a number of differentbiomes of North America (Fig. 1) spanning a widerange of physical, chemical and biological conditions(Table 1). All of the streams were relatively undis-turbed by current human activities (e.g. dissolved inor-ganic nitrogen (DIN) concentrations < 0.15 mg IT1).The metabolism measurements were made on onedate during the 6-week tracer 15N addition in eachstream. The period chosen for the LINX study in eachstream was designed to represent one of relativelyhigh rates of N uptake and cycling. Thus, our meta-bolism measurements presumably represent periodsof relatively high biological activity. Further, themetabolism measurements were conducted on a datewith generally clear to partly dear weather conditionsand, consequently, probably represent values that arehigher than the annual mean for most of the streams.

Methods

Whole-stream rates of GPP and R were determinedusing the upstream-downstream diurnal dissolved

:Bear

Fig. 1 Map showing location of streams used in study.

© 2001 Blackwell Science Ltd, Freshwater Biology, 46,1503-1517

Table 1 list of streams and physical, chemical and biological characteristics at or near the time of metabolism measurements. Although Walker Branch is a forest streamwith a closed riparian canopy, it was studied during a period of the year when leaves were absent from the deciduous riparian vegetation. Streams are listed from lowest to highestin latitude

Stream

Quebrada Bisley, Puerto RicoSycamore Creek, ArizonaWalker Branch, TennesseeGallina Creek, New MexicoSouth Kings Creek, KansasEagle Creek, MichiganMack Creek, OregonBear Brook, New Hamphsire

ID

QBPRSCAZWBTNGCNMKCKSECMIMCORBBNH

Date ofmeasure-ment

20 January 9827 May 9710 April 9727 August 9722 April 989 July 98

30Jul9827 June 97

Ripariancanopy

ClosedOpenClosedSemi-closedOpenClosedSemi-closedClosed

Channelgradient(mm-1)

0.080.0030.0350.170.0250.00250.090.13

Wettedwidth(m)

4.75.83.11.32.45.05.12.1

Discharge(Ls-1)

17317.83.1

10.4198602.3

Mean watervelocity(cm s'1)

4.913.85.85.74.3

249.92.2

Meanwaterdepth (m)

0.120.040.040.070.100.190.110.13

PAR(mol m"2 day"1)

0.35012.66.8

38183.82.2

Mean watertemperature

21.92311.912.612.12513.813.8

Table 1 (Continued)

Stream

Quebrada Bisley, Puerto RicoSycamore Creek, ArizonaWalker Branch, TennesseeGallina Creek, New MexicoSouth Kings Creek, KansasEagle Creek, MichiganMack Creek, OregonBear Brook, New Hampshire

AS : A ratio, A*(m2)

0.38, 0.090.59, 0.180.17, 0.020.06, 0.0040.16, 0.040.25, 0.220.29, 0.340.19, 0.04

SRP concentration

14143833

134

DIN concentration

132152395

336159

Autotroph biomass(g AFDM m-2), algal (%)

1(100)178 (100)41(3)42 (2)10 (98)0.7 (100)

18(4)11(8)

Detritus standing crop(g AFDM m-2), leaves (%)

38(16)20(0)

385 (20)120 (6)215 (<1)394 (< 1)113 (< 1)53 (8)

1506 P.J.Mulhollandetai.

oxygen change technique (Marzolf et a/., 1994) withthe modification suggested by Young & Huryn (1998)for calculating the air-water exchange rate of oxygen.Measurements of dissolved oxygen concentrationand water temperature (Orbisphere Model 2607(Orbisphere Laboratories, Geneva, Switzerland) dis-solved oxygen analyzer or YSI 600 (YSI, yellowSprings, OH, U.S.A.) water quality monitor) weremade at 1-min intervals and averages recorded at5-min intervals over a 24-h period at two stations ineach stream. In one stream (Sycamore Creek, Arizona)dissolved oxygen concentration and water tempera-ture were measured at hourly intervals by Winklertitration (APHA, 1992). The distance between stationsranged from 35 to 300 m and, depending on watervelocity, water travel time ranged from 9 to 40 minamong streams. Exchange of oxygen with the atmo-sphere was calculated based on the average oxygensaturation deficit or excess within the study reach andthe reaeration rate determined from the decline indissolved propane concentration during steady-stateinjection of propane and a conservative tracer (toaccount for dilution of propane caused by ground-water inflow) performed within 1 day of the oxygenmeasurements. The reaeration rate of propane wasconverted to oxygen using a factor of 1.39 (Raihbunet al., 1978). The net rate of oxygen change as a resultof metabolism (equivalent to net ecosystem produc-tion, NEP) was then determined at 5-min intervalsfrom the change in mass flux of dissolved oxygenbetween stations corrected for air-water exchange ofoxygen within the reach. The daily rate of R wascalculated by summing the net oxygen change ratemeasured during the night and the daytime rate of Rdetermined by extrapolating between the net oxygenchange rate during the 1-h predawn and postduskperiods. The daily rate of GPP was determined bysumming the differences between measured net oxy-gen change rate and the extrapolated value of Rduring the daylight period. All metabolism rates wereconverted to rates per unit area by dividing by thearea of stream bottom between the two stationsdetermined from the measurement of wetted channelwidth at 1-m intervals over each reach.

Groundwater inputs to the study reach, having adissolved oxygen concentration lower than the streamwater, contribute to errors in the measured rate of R inthe whole-stream dissolved oxygen balance method(McCutchan, Lewis & Saunders, 1998). Measurable

groundwater inputs occurred in five of the eightstream reaches studied (ranging from 3 to 17% ofdischarge), based on dilution of the injected conser-vative tracer. For these streams, we corrected R usingmeasurements of dissolved oxygen concentration ingroundwater seeps made at the same time of the yearas the metabolism measurements. Differences inaverage dissolved oxygen concentration between sur-face water and groundwater seeps were multiplied bystream discharge rate to compute the groundwaterseepage contribution to R in each stream. The largestcorrection was for South Kings Creek, Kansas, whichamounted to a 33% reduction in R (from 3.6 to2.4 gO2 m~2 day"1); corrections to R for the other fourstreams with groundwater inputs, were < 15%. Thedaily rate of NEP was calculated as the differencebetween the daily rate of GPP and groundwater-corrected R.

To permit a comparison of the reaeration rate deter-mined experimentally using propane injections withthat determined from physical characteristics of thestream channel, the reaeration rate was also estimatedusing the energy dissipation model (Tsivoglou & Neal,1976) as follows:

= K' x (AH/AX) x V (1)

where fc2o is the oxygen reaeration rate at 20 °C (day a),K' is an empirical constant equivalent to28.3 x 103 s m"1 day"1 for streams with dischargevalues < 280 L s"1, AH/ AX is the channel slope(m m"1) and V is water velocity (m s"1). The estimatedrate was compared with that determined directlyfrom propane injections and corrected to oxygen asdescribed above and to 20 °C according to:

where fcj- is the reaeration rate measured at tempera-ture T (Elmore & West, 1961). Although there are anumber of physically based methods for estimatingreaeration rate (Genereux & Hemond, 1992), we choseto compare our experimentally derived values withthe energy dissipation method because the latter hasbeen recommended for use in open-system methodsfor determining stream metabolism (APHA, 1992).

We measured a number of physical, chemical andbiological characteristics in each stream to identifypossible causal relationships with stream metabolism.We determined average stream discharge and water

© 2001 Blackwell Science Ltd, Freshwater Biology, 46, 1503-1517

velocity from conservative tracer additions performedwithin 1 day of the oxygen measurements. We mon-itored photosynthetically active radiation (PAR)within 20 cm of the stream water level at onestreamside location in the experimental reach duringthe period of oxygen measurements using a quantumsensor (LiCor 190SA; LI-COR, Lincoln, MB, U.S.A.)and data logger (LiCor 1000). We characterizedchannel hydraulic conditions by applying a transientstorage model to data from conservative tracer injec-

\ tions performed 2-3 weeks prior to the metabolismmeasurements under similar flow conditions in eachstream (Stream Solute Workshop, 1990; Webster &Ehrman, 1996). We pumped a sodium chloride (Nad)or sodium bromide (NaBr) solution into the streamuntil a steady-state was reached ^across a 100-300-mstream reach (2-6 h). We monitored either Cl~ con-centration (Orion model 9417 (Orion Instruments,Beverley, MA, U.S.A.) B ion specific electrode), Br~concentration (ion chromatography) or specific con-ductance (YSI Model 30) at intervals ranging from oneto several minutes at the downstream station duringand for several hours after the injection. We then fit atwo-compartment transient storage zone model (Ben-cala & Walters, 1983; Hart, 1995) to the ion or specificconductance data to determine the rate of exchangebetween flowing and stationary water zones withinthe stream channel. From the model output, wecomputed the cross-sectional area of the stationarytransient storage zone (AJ, the ratio of the cross-sectional areas of the transient storage zone and thesurface flowing zone (As : A), the average traveldistance of a water molecule prior to uptake into atransient storage zone (water uptake distance) and ahydraulic retention factor. The hydrologic retentionfactor is the ratio of the water residence time in thetransient storage zone to the water uptake distance(Morrice et al, 1997).

We measured concentrations of ammonium, nitrateand soluble reactive phosphorus (SRP) on three to fivedates within 3 weeks of the metabolism measure-ments using standard colorimetric methods (APHA,1992). We calculated DIN as the sum of ammoniumand nitrate concentration.

We measured the standing crop of detrital benthicorganic matter (BOM) 2-3 weeks prior to the meta-bolism measurements according to methods des-cribed by Mulholland etal. (2000). We placed anopen-ended metal cylinder (0.07 m2) into the stream

© 2001 BlackweU Science Ltd, freshwater Biology, 46,1503-1517

Inter-biome comparison of stream metabolism 1507

bottom at 10 locations and collected coarse particulateorganic matter (CPOM, > 1 mm diameter) and separ-ated it into leaves and wood. To estimate fineparticulate organic matter (FPOM), we vigorouslyagitated the sediments within the cylinder to a depthof about 10 cm, pumped the slurry through a 1-mmscreen into a container of known volume and sub-sampled the pumped slurry. Material was returned tothe laboratory, dried (60 °C), weighed, combusted(500 °C) and reweighed to determine ash-free drymass (AFDM) per unit area sampled. We calculatedtotal benthic detritus as the sum of CPOM (leaves andwood) and FPOM.

We measured epilithon standing crop by collectingrocks randomly from five to six locations in thestream, scraping the rock surfaces and washing thematerial into a container with stream water. We thenfiltered this slurry (Whatman GFF; Whatman, Maid-stone, U.K.), extracted the filters in 90% acetoneovernight and analysed spectrophotometrically forchlorophyll a, using the method of Lorenzen (1967).We measured the area of each rock scraped todetermine chlorophyll a per unit area. We estimatedthe biomass of the algal component of the epilithon as100 x chlorophyll a mass per unit area of rock surface(Reynolds, 1984). We determined the area! coverage offilamentous algae and bryophytes by establishingtransects across the stream every 5 m along the studyreach and determining presence/absence every10-20 cm across the transects. We estimated biomassby scraping or coring material from known areas ofsubstratum with 100% coverage of filamentous algaeor bryophytes and determining AFDM as the differ-ence between dry mass (60 °C) and ash mass (500 °C).We calculated the biomass of filamentous algae andbryophytes as the product of average per cent coverand biomass per unit area in areas of 100% cover. Wecalculated total autotroph biomass as the sum ofepilithon, filamentous algae and bryophyte biomass.

Statistical analysis

We analysed the data using bivariate correlation andstepwise multiple regression (SAS, 1985). Correlationanalysis was used to identify relationships betweensingle factors and metabolism rates, whereas multipleregression was used to determine whether predictiverelationships for the rate of metabolism could bedeveloped using more than one environmental factor.

1508 P.J. Mulkolland et al.

Variation among values for several of the variableswas considerable (> 20-fold) and data were normal-ized by log-transformation of these values. For step-wise multiple regression, we used P = 0.05 as thecriterion for entry into the model, and an analysis ofcollinearity was performed for all variables enteringthe regression model.

Results

Diurnal profiles of metabolism rate varied consider-ably among streams. Three general patterns wereevident. Streams with little canopy cover and highPAR, such as South Kings Creek, had a relatively highrate of GPP and a positive rate of NEP, peaking in theafternoon (Fig. 2a). Sites with somewhat greatershading and lower PAR, such as Walker Branch,Tennessee, in early spring, had an intermediate rate ofGPP but NEP rate remained negative throughout theday (Fig. 2b). Heavily shaded sites with low PAR,such as Quebrada Bisley, Puerto Rico, showed noevidence of GPP and the rate of NEP was highlynegative and somewhat variable with no clear diurnalpattern (Fig. 2c).

The daily rate of metabolism also varied consider-ably among streams. The GPP ranged from < 0.1-15 gO2 m~2 day"1, with Quebrada Bisley having thelowest and Sycamore Creek having the highest rate(Fig. 3a). There was substantially lower variation in Rthan GPP, with values ranging from 2.4 gOa m~2 day"1

in South Kings Creek to 11 gO2 m~2 day"1 in MackCreek, Oregon (Fig. 3b). The NEP was negative andP : R ratios were < 1 for all streams except SycamoreCreek (Fig. 3c). Six of the eight streams were stronglyheterotrophic, with P : R ratios < 0.25.

Relationships between the instantaneous rate ofGPP and PAR were variable among streams (Fig. 4).For South Kings Creek and Walker Branch, lightsaturation of GPP appeared to occur at PAR values> 200-500 nmol m2 s"1. However, there was no evi-dence of light saturation of GPP in Sycamore Creek. Inthe streams with the highest light levels and greatestalgal biomass (Sycamore and South Kings Creeks),GPP was consistently higher in the afternoon than inthe morning under the same PAR. This might reflect adelay in oxygen diffusion from within the algal mat tothe overlying water in streams with thick algal mats.Relationships between GPP and PAR were morevariable in Mack Creek, Gallina Creek and Bear

tE

I

o13

00:00:00 04:00:00 08:00:00 1200:00 16:00:00 20:00:00 00:00:0022 April 1998

030:00 4:00:00 8:00:00 12:00:00 16:00:00 20:00:00 0:00:0010 April 1997

"CD

0:00:00 4*0:00 8:0030 12:00:00 163030 20:00:00 0303020 January 1998

Fig. 2 Diurnal patterns of net ecosystem production (net oxygenchange corrected for reaeration) for South Kings Creek, Kansas(a), Walker Branch, Tennessee (b), and Quebrada Bisley, PuertoRico (c). The line extending from the predawn to postsunsetperiod hi each plot is the extrapolated respiration rate duringdaylight Sycamore Creek, Arizona, also fit the pattern shown in(a). Other streams fitting the pattern in (b) were: Eagle Creek,Michigan, and Mack Creek, Oregon, and to a lesser extent,Gallina Creek, New Mexico. Bear Brook, New Hampshire, alsofits the pattern in (c).

Brook, although GPP appeared to become light-saturated at relatively low PAR in these streams.

The daily rate of GPP was significantly correlatedwith daily PAR (Fig. 5a). The correlation betweenGPP and other physical and chemical characteristics(water temperature, discharge, water velocity, DINconcentration and SRP concentration) was not signi-ficant (P > 0.05). Gross primary production wasmarginally correlated with total algal biomass (epil-ithon plus filamentous algae, r = 0.66, P = 0.077).Multiple regression analysis indicated that 90% ofthe variation in log GPP could be explained by amodel that included log PAR and SRP concentration(Table 2).

© 2001 Blackwell Science Ltd, Freshwater Biology, 46,1503-1517

w. 8 -

I <J

cT °S -4Ha.LU „ ,^ —8 -

1.8

0.01 0.2 0.06 0.75 0.12 0.17 0.03

£L§ i

g im01

Fig. 3 Daily rates of gross primary productivity (GPP), totalrespiration (R) and net ecosystem production (NEP). Streams arelisted using the codes in Table 1 and are ordered from lowest(left) to highest latitude. Values on bars in (c) are GPP : R ratios.

The daily rate of R was significantly correlatedwith SRP concentration and the size of the transientstorage zone, AS (Fig. 5b & c). Correlations betweenR and several other physical and chemical charac-teristics (water temperature, discharge, water velo-city, DBST concentration, AS : A ratio, water uptakedistance and hydraulic retention factor) were notsignificant (P > 0.05). In addition, R was not signifi-cantly correlated with total detritus standing crop(P > 0.05). Multiple regression analysis indicated thatonly SRP was a significant predictor of R using amodel entry criterion of P = 0.05 (Table 2). Relaxa-tion of the model entry criterion to P = 0.15 resultedin the addition of AS as a significant predictor in themultiple regression analysis, with 73% of the vari-ation in R explained by the model containing bothvariables (Table 2).

Inter-biome comparison of stream metabolism 1509

The daily rate of NEP was significantly correlatedonly with PAR (Fig. 5d). Multiple regression analysisindicated mat 53% of the variation in log(NEP + 10)could be explained by a model that included log PAR,with no other variables significantly improving themodel (Table 2).

The reaeration rate measured directly from thepropane injections was generally higher than esti-mates from the energy dissipation method (Fig. 6a).The differences in the reaeration rate between meth-ods appeared to be related to average water depth byan exponentially declining function (Fig. 6b). Foraverage water depth > 6 cm, the reaeration ratemeasured using direct propane injections was < 25%,higher than that calculated using channel physicalfeatures for five of seven data points.

Discussion

Methodology

Estimates of whole-ecosystem rates of GPP and R instreams have been made for many years by measuringdiurnal changes in dissolved oxygen in open systems(e.g. Odum, 1956; Hoskin, 1959; Hall, 1972; Meyer &Edwards, 1990; Young & Huryn, 1996; Uehlinger &Naegeli, 1998), although most previous studies havebeen of unshaded, relatively productive streams.Refinements to the two-station diurnal oxygen changemethod by Marzolf et al. (1994) and Young & Huryn(1998) have improved the performance! of this open-system approach in small streams with relatively lowrates of GPP. Most of the streams studied here were ofthis type.

Open-system methods offer some advantages overchamber methods for determining the rate of streammetabolism because they do not suffer from enclosureartefacts (e.g. nutrient and oxygen depletion), diffi-culties in transferring all components of the streamecosystem in correct proportions (e.g. fine sediments,hyporheic sediments) and scaling problems (e.g.accounting for spatial variability) associated withchamber measurements. Because of these problems,open-system methods should result in higher esti-mates of whole-ecosystem metabolism, particularlyrespiration, than chamber studies. Comparison of ourresults with those from several previous studiesseems to confirm this. Respiration rates determinedfrom chamber measurements reported by Bott et al.

© 2001 Blackwell Science Ltd, freshwater Biology, 46,1503-1517

1510 P.J. Mulhplland et al.

cE

E

0°^^O)E,

Q_Q.

TcE

\CM

OO)

Q.Q.ff\(D

"pcE

CM1

CMOO)E,D_a.CD

ou -

40 -

30-

20 -

10 -

0 _

a SycamoreCreek ^

• o

0

• o o

.•0 °T i l l

6 -

5 -

4 -

3-

-

1 -

b South KingsCreek

• ••Wjflft•••*25R(Tij^^^f^^^^ f^^^Sft^^^~ rifflo

rofssfeto. Q-*

Ifir

~^f 1 1 1 10 500 1000 1500 2000 o 500 1000 1500 2000

6 -

5 -

4 -

3 -

2 -

1 -

c Walker Branch

r ' 9

**

0 _

6 -

5 -

4 -

3 -

2 -

1 -

d Gallina Creek

0

*• ' '*• *A *• *^ I I I

0 500 1000 1500 2000 0 500 1000 1500 2000

6 '"

5-

4 -

3 -

2 -

1 -

• e Mack Creek

1 • ^»»

J •

. PbfIfT i l l

w

5 -

4 -

3 -

2 -

1 -

0 .

f Bear Brook

Er*i>* °T" I I I

0 500 1000 1500 2000 0 500 1000 1500 2000

PARdimoIrrfV1) PAR (umol m'V1)

'V...

Fig. 4 Relationships between rate of grossprimary productivity (GPP) and photo-synthetically active radiation (PAR) for sixof the eight streams. Eagle Creek is notshown because only daily PAR value wasrecorded, and Quebrada Bisley is notshown because GPP remained very lowthroughout the day. Open circles denotemeasurements made in the morning anddosed circles-measurements made in theafternoon.

(1985) for small streams in Pennsylvania, Michiganand Oregon (0.6-2.1 gOa m~2 day"1) are considerablylower than respiration rates measured in our study(2.4-11 gO2 rrf

2 day"1). Respiration rates for streamsin the Hubbard Brook Experimental Forest, NewHampshire, reported by Hedin (1990) for summerand autumn (0.1-0.8 gC>2 m~2 day"1), were consider-ably lower than our value of R in summer for BearBrook (6.7 gO2 m~2 day"1), also located in the Hub-bard Brook Experimental Forest. In a previous study,in Walker Branch, Marzolf et al. (1994) comparedopen-system measurements with chamber measure-ments and found that chamber measurements under-

estimated GPP by about 20% and R by about 300%.Chamber measurements of R for Sycamore Creek (2-5 gO2 m~2 day"1) by Grimm (1987) also were lowerthan the respiration rate measured in our study forthe same stream (8.3 gO2 m~2 day"1). Comparingchamber and open-system measurements on the samedate in Sycamore Creek, Grimm & Fisher (1984) arguedthat chamber measurements were lower because theydo not include the hyporheic component of respiration.Webster, Wallace & Benfield (1995) summarizedstream metabolism measurements made in over 30streams in the eastern U.S.A. and found that estimatesof mean primary production and respiration rate were

© 2001 BlackweU Science Ltd, Freshwater Biology, 46,1503-1517

Inter-biome comparison of stream metabolism 1511

10 -

CO•o

I ,.CM

O

Q.Q_CD 0.1 J

a SCAZ •

MCOR KCKS

WBTN *

ECMIGCNM*

BBNH «

QBPR r =0.85• P= 0.008

i i i1 10 10(

12-

^ 10-

co 0•0 8 -

E 6-CM

0 4-

n

b MCOR

SCAZBBNH f

ECM^ft 0 QBRR

^ GCNMWBTN

KCKS r =0.75KCKS P = 0.033

PAR (mol rrf2 day"1)

0 3 6 9 1 2 1 5

SRP concentration (|j,g LT1)

CD•0

O0)

12 -

8 -

4 -

2 -

0 -0

C MCOR• '

GCNM

/ »SCAZ/ « QBPR

• BBNH »ECMI

WBTN

KCKS r = °71

P = 0.049i i i

.0 0.1 0.2 0.3 0.

As(m2)

7 20 -IoT

^ 10 ME.

o- 5 H

O 2-\

Q.HI

QBPR

SCAZ

KCKS

WBTN:

GCNM• •BBNH

ECMI

MCOR r = 0.73P= 0.041

1 10 . 100

PAR (mol m-2 day'1)

Fig. 5 Significant correlations between gross primary productivity (GPP), total respiration (R) and net ecosystem production (NEP)and various physical and chemical characteristics.

significantly higher in studies where open-systemmethods were used than those using chamber methods.As they have pointed out, however, their comparisonwas confounded by the fact that open-system methodswere generally used where a higher rate of metabolismmight be expected Garger, more nutrient-rich streams).Our range in respiration rates was similar to thatreported for New Zealand streams (1-8 gO2 m~2 day"1)by Young & Huryn (1999) and that reported over a 2-year period for a Swiss river (1-13 gO2 m~2 day"1) byUehlinger & Naegeli (1998), also using open-systemoxygen change methods.

© 2001 Blackwell Science Ltd, Freshwater Biology, 46,1503-1517

Limitations of the open-system method includethe need for accurate measurements of dissolvedoxygen concentration and saturation deficit as well asgood estimates of reaeration rates. Accurate dis-solved oxygen measurements require high-precisioninstruments, accurate field calibration and frequentchecks, on calibration. McCutchan et al. (1998), in adetailed analysis of uncertainties associated withopen-system methods, demonstrated that estimatesof R are subject to greater uncertainty than estimatesof GPP7 particularly in high gradient streams. Reaera-tion rates were > 100 day"1 in five of our eight

1512 P.J.Mulhqlhndetai.

Table 2 Results of stepwise multiple regression analysis forrates of gross primary production (GPP), respiration (R) and netecosystem production (NEP) (n = 8 for each regression)

350

Dependentvariable

log GPP

R

R (P = 0.15)

log(NEP + 10)

Independentvariable

Interceptlog PARSRPFull model

InterceptSRP

InterceptSRPA.Full model

Interceptlog PAR

Parameterestimate (SB)

-1.737 (0.349)0.994 (0.147)1.027 (0.338)

4.104 (1.175)0.356 (0.129)

3.775 (1.031)0.255 (0.125)9.572 (5.463)

0.298 (0.164)0.381 (0.150)

,»

0.7200.1810.901

0.561

0.5600.1670.73

0.529

Prob>F

0.00420.00110.02880.003

0.0130.033

0.01460.09660.14010.0387

0.11950.0437

Criterion for entry into the model was P — 0.05, except forR where results for a relaxed entry criterion (P = 0.15)are also given.

streams, a level that could result in uncertainties inmetabolism rates of > 30% according to McCutchanet al. (1998).

Reaeration rate is often the most problematiccomponent of open system methods. For directmeasurements, using injections of volatile gas trac-ers, achieving complete mixing can be difficult,particularly in larger streams. Several predictiveequations have been developed to estimate reaer-ation rate from more readily determined physicalcharacteristics of streams (channel slope, watervelocity and depth); however, these indirect meth-ods for determining oxygen reaeration rate weregenerally developed for larger rivers with lessturbulent flow. Genereux & Hemond (1992) com-pared a number of these indirect estimates withdirect measurements of reaeration made using pro-pane injections in Walker Branch and found pooragreement, with most of the indirect methodsunderestimating reaeration rate by 25% or more.Young & Huryn (1999) made a similar comparisonbetween reaeration rates determined by propaneinjections and those estimated by indirect methodsfor streams in New Zealand and found that theindirect methods substantially underestimated re-aeration rates, particularly for rates > 50 day"1 asdetermined by the direct propane method. Ourcomparison of direct and indirect measurementsindicates that the underestimation of reaeration rate

300 -

250-

200-

s

.o£CD|g 150 H

O 100 -\T3

I_OCOo

50-

00 50 100 150 200 250 300 350

"1Measured O2 reaeration rate (day")

14

12

ISV •*-»

ll«- "S!rj jo

M

3-

2-

1 -- __•••_•_

2 4 6 8 10 12 14 16 18 20

Average water depth (cm)

Fig. 6 Comparison of reaeration rates measured using propaneinjections with rates calculated using the energy dissipationmodel of Tsivoglou & Neal (1976) (a) and measured : calculatedrate ratios as a function of average water depth (b). Included inthe plot are data from measurements on three different dates inWalker Branch (Mulholland et al., 2000) and data from meas-urements in Upper Ball Creek, North Carolina.

by the energy dissipation method declines with anincrease in average water depth. In streams with awater depth > 6 cm, the energy dissipation methodmay provide acceptable estimates of reaeration ratefor use in open-channel measurements of streammetabolism in many cases.

Gross primary production rate

Our results indicate that light (PAR) is the dominantcontrol on stream GPP. Other multi-stream compar-ison studies have also shown that available light, as

© 2001 Blackwell Science Ltd, Freshwater Biology, 46,1503-1517

indicated by canopy cover, is a strong determinant ofprimary production rate (Naiman, 1983; Bott etal.,1985; Webster at al, 1995; Young & Huryn, 1999). Inthese studies light availability was primarily afunction of stream size or land use, with largerstreams or those in grazed pasture having more opencanopies (higher light) and higher rates of primaryproductivity. The streams we studied were all relat-ively small (average width ranging from 0.8 to 5.8 m)and high PAR values were primarily the result of thelower density of riparian trees found in more aridclimates (e.g. South Kings Creek and SycamoreCreek). In addition, PAR was moderately high inWalker Branch because the period of measurementwas prior to spring leaf emergence in the surround-ing deciduous forest. In a previous paper, Mulhol-land et al. (2000) showed that PAR values declinedby about 85% between early April and the middle ofMay as the leaves emerged and shaded the stream. A75% reduction in GPP and a sharp decline in the rateof nitrate uptake accompanied the reduction in PAR,demonstrating that the phenology of riparian veget-ation is an important determinant of light availabilityand, consequently, of GPP. Our study did notaddress such seasonal changes in GPP in the otherstreams.

Although we observed a saturating effect of light onthe instantaneous rate of GPP in most streams overthe course of the day, we did not observe lightsaturation of daily GPP across streams. With theexception of Sycamore Creek, which showed noevidence of light saturation, our GPP-PAR relation-ships for individual streams were generally consistentwith results from chamber studies of stream periphy-ton showing light saturation of photosynthesis at anirradiances above 20Q-400 nmol m~2 s"1 (Hill & Bos-ton, 1991; Hill, Ryon & Schilling, 1995). Our resultswere also consistent with those of Young & Huryn(1996), who measured whole-system GPP using theopen-system oxygen change method in New Zealandstreams, and commonly found evidence of lightsaturation at PAR of 250-500 junol m~2 s"1. However,several other stream studies using the open-systemoxygen change method have shown no evidence ofsaturation at high irradiance (Duffer & Dorris, 1966;Kelly, Hornberger & Cosby, 1974; Hornberger, Kelly& Fuller, 1976). Uehlinger, Konig & Reicher (2000)monitored GPP at 3-day intervals in a small, mostlyunshaded Swiss river using the open-system method

© 2001 Blackwell Science Ltd, Freshwater Biology, 46,1503-1517

Inter-Nome comparison of stream metabolism 1513

and reported that light saturation was restricted to thewinter months only. It is not surprising that lightsaturation of GPP would be observed less frequentlywhen open-system, whole-stream measurements ofGPP are used than in chamber studies because thelatter include only one component of the streamautotroph community. The lack of evidence for lightsaturation of daily rates of GPP observed acrossstreams in our study also suggests long-term adapta-tion to high light levels. Thus, it would appear that,although individual communities at specific sites cansometimes become light saturated, patterns of whole-ecosystem GPP across biomes may rarely show lightsaturation effects because different types of autotrophcommunities develop and adapt to use the greateravailable light resource.

Nutrient concentration appeared to be a secondarydeterminant of GPP in our streams, as indicated bythe inclusion of SRP concentration in the best-fit,multiple regression model. The streams in our studyhad relatively low SRP concentration (3-14 ng LT1),potentially sufficient to limit primary production(Bothwell, 1989). A number of studies have demon-strated nutrient limitation of algal biomass accrual inoligotrophic streams (e.g. Elwood et al., 1981; Petersonetal, 1983; Grimm & Fisher, 1986; Hill & Knight,1988; Rosemond, Mulholland & Elwood, 1993). Theeffect of SRP on GPP in our study was highlyinfluenced by one stream (Sycamore Creek) withhigh values of GPP and SRP. When Sycamore Creekwas removed from the multiple regression analysis,SRP no longer entered the model as a significantpredictor. Because of the low number of streams inour study, the power of these tests is rather low andthe importance of SRP as a predictor of GPP at broadspatial scales is unclear.

Respiration rate

Our results suggest that phosphorus concentrationsand channel hydraulic conditions control the whole-stream rate of respiration over large geographic areas.The effect of nutrients on heterotrophic microbialprocesses hi streams is relatively well documentedand the rate of leaf decomposition increases withnutrient enrichment (Elwood et al., 1981; Meyer &Johnson, 1983; Suberkropp & Chauvet, 1995). Theproduction and respiration rates of bacteria and fungicolonizing leaf detritus have also been shown to be

1514 P.J. Mulholland et al.

nutrient limited in some streams (Tank & Webster,1998; Grattan & Suberkropp, 2001). Our study sug-gests that whole-ecosystem respiration rate in streamsis also influenced by nutrients and that nutrientlimitation may be an important large-scale controlon heterotrophic metabolism in streams.

Channel hydraulic conditions, notably the extent ofthe hyporheic zone, have been shown to have strongeffects on respiration rate in streams. Other studies intwo of our streams, Sycamore Creek (Grimm & Fisher,1984) and Gallina Creek, New Mexico (Fellows et al.2001) have shown that > 50% and about 85% ofwhole-ecosystem R, respectively, was the result ofhyporheic respiration. Similarly, Naegeli & Uehlinger(1997) reported that hyporheic respiration contributedabout 85% of the total ecosystem respiration rate in agravel-bed river in Switzerland. In a comparativestudy of metabolism and phosphorus uptake in twosmall forested streams, with similar temperature,nutrient concentration and organic matter input buta 15-fold difference in the size of hyporheic zone,Mulholland et al. (1997) showed that the stream withthe larger hyporheic zone had 2.5-fold higher R and Puptake rates. Our study indicates that the size oftransient storage zones (as defined by AS) was asecondary predictor of R, accounting for an additional17% of the variation in R beyond that resulting fromvariation in SRP concentration. Although backwaterzones along the margins of channels may have

. accounted for some of the transient storage zoneareas determined in our streams, hyporheic sedimentsprobably also accounted for a substantial portion ofA,, particularly in the streams with relatively large AS-Presumably, the effect of Ag on R was the result ofgreater storage of organic matter and increasedsurface area for heterotrophic microbes in streamswith larger AS values. However, the correlationbetween benthic detritus standing crop and R hi ourstudy was not strong (r = -0.41, P = 0.25) and oppos-ite in sign to that expected. In addition, Webster et al.(1995) reported that evidence for a positive effect ofBOM storage on respiration rate in eastern U.S.A.streams is weak Perhaps most techniques for meas-uring BOM standing crop (including ours) do notaccount for the deeper storage of material hi streamswith higher AS.

The variation hi R among our streams was not theresult of variation in concurrent autochthonous pro-duction. The correlation between R and GPP was poor

(i2 = 0.05) and the fivefold variation in R was smallcompared with the 150-fold variation in GPP amongstreams. These results probably reflect the fact thatstream respiration is fuelled by both autochthonousand allochthonous sources of organic matter contri-buted over extended periods of time.

We were surprised by the lack of evidence for aneffect of water temperature on R hi our study. In part,this may have been the result of the relatively smallrange in temperature (12-25 °C) compared with the10-30-fold range in nutrient concentration and size oftransient storage zones among streams (Table 1).Several other studies have suggested a modest effectof water temperature on respiration rate in streams.Bott et al. (1985) found that temperature was the bestsingle predictor of R in a study of streams hi fourdifferent biomes in the U.S.A., explaining 33% ofthe variation hi R for all streams. Unlike our study,however, Bott et al. (1985) made measurementsduring all seasons in each stream, which may haveincreased the likelihood of showing an effect oftemperature. Sinsabaugh (1997) found that meanannual temperature explained 38% of the variationhi mean annual respiration rate hi a comparativestudy of 22 streams, hi an intensive study of metabo-lism in a Swiss river over an annual period, Uehlingeret al. (2000) found that R was significantly related towater temperature, although temperature explainedonly 22% of the variation in R. In our study, the effectof temperature on R may have been obscured byeffects of differences in organic matter supply andnutrient concentration.

Net ecosystem production and P : R ratio

Respiration dominated whole-stream metabolism himost of our streams. Only Sycamore Creek, withhigh light (PAR of 50 mol m~2 day"1), had a positiveNEP (P : R ratio > 1). Even hi South Kings Creek,with PAR of 38 mol m~2 day"1, R exceeded GPP onthe date we measured metabolism (P : R of 0.75),although earlier hi the spring NEP may have beenpositive. We observed that the large biomass ofperiphyton appeared to be undergoing partial sen-escence when the metabolism measurements weremade hi this stream. The highly negative NEP valuesand very low P : R ratios for most of our streamsemphasize the importance of allochthonous sourcesof carbon hi fuelling heterotrophic metabolism. This

© 2001 Blackwell Science Ltd, Freshwater Biology, 46,1503-1517

is not surprising because six of the eight streams (allthose with low P : R ratios) were in forests withdosed or semi-closed canopies. Even when lightsupply in the forested streams was moderately high,as for Walker Branch (12.6 mol m"2 day"1) and EagleCreek, Michigan (18 mol m~2 day'1), NEP and P : Rratios were quite low, presumably because of respir-ation associated with large allochthonous organicmatter inputs.

Our multiple regression results suggested that NEPwas controlled primarily by factors influencing pro-duction (PAR), probably because rates of GPP variedconsiderably more than rates of R among our streams.Others have also shown the positive influence of lighton NEP. Bott et d. (1985) reported that NEP increasedwith stream size as the canopy opened and lightincreased. They reported that light was the strongestpredictor of NEP, although it accounted for onlyabout 14% of the variation in NEP. Young & Huryn(1999) reported that NEP was positively correlatedwith incident light which, in turn, was related to landuse. In their intensive, 2-year study of metabolism in asixth order prealpine Swiss river, Uehlinger & Naegeli(1998) showed that hydrologic fluctuations stronglyinfluenced the balance between GPP and R. Bed-moving spates resulted in a sharp decline in P : Rratio because they had greater negative effects oh GPPthan on R. Although we selected baseflow periods toperform our measurements, variation in the length oftime since large storms among our streams may haveadded additional variation in rates of metabolism inour study.

The generality of our results is somewhat limitedbecause of the low number of streams in the studyand because measurements were made on only onedate in each stream. As a consequence, the statisticalpower of tests for the effect of various factors onrates of metabolism was generally low. In eachstream our results are a snapshot in time, represen-tative of a period of relatively high metabolism formost of the streams. However, there have been fewstudies that have examined stream metabolismacross large geographic areas using the samemethod. Our study included streams in climatesranging from tropical to cool temperate and fromhumid to arid. Further, our measurements are ofmetabolism of entire stream ecosystems as weused the open-system, two-station diurnal oxygenchange method. Thus, our findings provide a large-

Inter-biome comparison of stream metabolism 1515

scale and synthetic picture of the factors that controlmetabolism in streams, but they must await fur-ther test across a larger number of streams andseasons.

In conclusion, we show that inter-biome variation inthe whole-ecosystem rate of GPP, respiration andNEP in streams is related primarily to differences inthe availability of light and phosphorus. Variation inthe rate of ecosystem respiration also appears to berelated to differences in channel hydraulic character-istics, such as the size of the transient storage zone.While the effect of light on primary production instreams is relatively well documented, our resultssuggest that nutrient limitation and channel hydraul-ics may also be important in the control of streamecosystem metabolism across large geographic areas.

Acknowledgments

We thank Sherri Johnson, Melody Kemp, AmandaLopez, Jeff Meriam and Diane Sanzone for help withthe field measurements. Walter Hill and Yetta Jager,Environmental Sciences Division, Oak Ridge NationalLaboratory provided helpful reviews of earlier draftsof the manuscript. This research was supported by agrant from the Ecosystems Program, National ScienceFoundation, to Virginia Tech (DEB-9628860). Theresearch was conducted in part on the Oak RidgeNational Environmental Research Park, Environmen-tal Sciences Division, Office of Biological and Envi-ronmental Research, U.S. Department of Energyunder contract DE-AC05-OOOR2725 with Universityof Tennessee-Battelle LLC.

References

American Public Health Association (APHA) (1992)Standard Methods for the Examination of. Water andWastewater. American Public Health Association,Washington, DC.

Bencala K.E. & Walters R.A. (1983) Simulation of solutetransport in a mountain pool-and-riffle stream: atransient storage model. Water Resources Research, 19,718-724.

Bothwell M.L. (1989) Phosphorus-limited growthdynamics of lotic periphyton diatom communities:area! biomass and cellular growth responses. CanadianJournal of Fisheries and Aquatic Sciences, 46,1293-1301.

Bott T.L. (1996) Primary productivity and communityrespiration. In: Methods in Stream Ecology (Eds F.R.

© 2001 Blackwell Science Ltd, Freshwater Biology, 46,1503-1517

1516 P./. Mulhplland et al.

Hauer & G.A. Lamberti), pp. 533-556. Academic Press,San Diego, CA.

Bott T.L., Brock J.T., Dunn C.S., Naiman R.J., Ovink R.W.& Petersen R.C. (1985) Benthic community metabolismin four temperate stream systems: an inter-biomecomparison and evaluation of the river continuumconcept. Hydrobiologia, 123, 3-45.

Cole ]., Lovett G. & Findky S. (Eds) (1991) ComparativeAnalysis of Ecosystems. Springer-Verlag, New York

Duffer W.R. & Dorris T.C. (1966) Primary production in asouthern great plains stream. Limnology and Oceano-graphy, 11,143-151.

Elmore H.L. & West W.F. (1961) Effect of water tem-perature on stream reaeration. Journal of the SanitaryEngineering Division. Proceedings of the American Societyof Civil Engineering, 87 (SA6), 59-71.

Elwood J.W., Newbold J.D., Trimble A.F. & Stark R.W.(1981) The limiting role of phosphorus in a woodlandstream ecosystem: effects of P enrichment on leafdecomposition and primary producers. Ecology, 62,146-158.

Fellows C.S., Valett H.M. & Dahm CJM. (2001) Whole-stream metabolism in two montane streams: contribu-tion of the hyporheic zone. Limnology and Oceanography46, 523-531.

Genereux D.P. & Hemond H.F. (1992) Determination ofgas exchange rate constants for a small stream onWalker Branch Watershed, Tennessee. Water ResourcesResearch, 28,2365-2374.

Grattan R.M. & Suberkropp K (2001) Effects of nutrientenrichment on yellow poplar leaf decomposition andfungal activity in streams. Journal of the North AmericanBenthological Society, 20,33-43.

Grimm N.B. (1987) Nitrogen dynamics during successionin a desert stream. Ecology, 68,1157-1170.

Grimm N.B. & Fisher S.G. (1984) Exchange betweeninterstitial and surface water: implications for streammetabolism and nutrient cycling. Hydrobiologia, 111,219-228.

Grimm N.B. & Fisher S.G. (1986) Nitrogen limitation in aSonoran Desert stream. Journal of the North AmericanBenthological Society, 5,2-15.

Hall C.A.S. (1972) Migration and metabolism ina temperate stream ecosystem. Ecology, 53, 585-604.

Hart D.R. (1995) Parameter estimation and stochasticinterpretation of the transient storage model for solutetransport in streams. Water Resources Research, 31,323-328.

Hedin L.O. (1990) Factors controlling sediment commu-nity respiration in woodland stream ecosystems. Oikos,57, 94-105.

Hill W.R. & Boston H.L. (1991) Community developmentalters photosynthesis-irradiance relations in streamperiphyton. Limnology and Oceanography, 36,1375-1389.

Hill W.R. & Knight A.W. (1988) Nutrient and lightlimitation of algae in two northern California streams.Journal of Phycology, 24,125-132.

Hill W.R., Ryon M.G. & Schilling E.M. (1995) Lightlimitation in a stream ecosystem: responses by primaryproducers and consumers. Ecology, 76,1297-1309.

Hornberger G.M., Kelly M.G. & Fuller R.M. (1976) Therelationship between light and photosynthetic rate in ariver community and implications for water qualitymodelling. Water Resources Research, 12, 215-232.

Hoskin C.M. (1959) Studies of oxygen metabolism ofstreams of North Carolina. Publications of the Institute ofMarine Science, Texas, 6,186-192V

Kelly M.G., Hornberger G.M. & ""Cosby B.J. (1974)Continuous automated measurement of rates of pho-tosynthesis and respiration in an undisturbed rivercommunity. Limnology and Oceanography, 19,305-312.

Lorenzen C.J. (1967) Determination of chlorophyll andphae-pigments: spectrophotometric equations. Limnol-ogy and Oceanography, 12, 343-346.

Marzolf E.R., Mulholland P.J. & Steinman A.D. (1994)Improvements to the diurnal upstream-downstreamdissolved oxygen change technique for determiningwhole-stream metabolism in small streams. CanadianJournal of Fisheries and Aquatic Sciences, 51,1591-1599.

McCutchan J.H., Lewis W.M. .& Saunders J.F. (1998)Uncertainty in the estimation of stream metabolismfrom open-channel oxygen concentrations. Journal ofthe North American Benthological Society, 17,155-164.

Meyer J.L. & Edwards R.T. (1990) Ecosystem metabolismand turnover of organic carbon along a blackwaterriver continuum. Ecology, 71, 668-677.

Meyer J.L. & Johnson C. (1983) The influence of elevatednitrate concentration on rate of leaf decomposition in astream. Freshwater Biology, 13,177-183.

Morrice J.A., Valett H.M., Dahm C.M. & Campana M.E.(1997) Alluvial characteristics, groundwater-surfacewater exchange and hydrologic retention in headwaterstreams. Hydrological Processes, 11, 253-267.

Mulholland P.J., Marzolf E.R., Webster J.R., Hart D.R. &Hendricks S.P. (1997) Evidence that hyporheic zonesincrease heterotrophic metabolism and phosphorusuptake in forest streams. Limnology and Oceanography,42,443-451.

Mulholland P.J., Tank J.L., Sanzone D.M., WollheimW.M., Peterson B.J., Webster J.R. & Meyer J.L. (2000)Nitrogen cycling in a forest stream determined bya 15N tracer addition. Ecological Monographs, 70,471-493.

© 2001 Bkckwell Science Ltd, Freshwater Biology, 46,1503-1517

Naegeli M.W. & Uehlinger U. (1997) Contribution of thehyporheic zone to ecosystem metabolism in a preal-pine gravel-bed river. Journal of the North AmericanBenthological Society, 16, 794-804.

Naiman R.J. (1983) The annual pattern and spatialdistribution of aquatic oxygen metabolism in borealforest watersheds. Ecological Monographs, 53, 73-94.

Odum H.T. (1956) Primary production in flowing waters.Limnology and Oceanography, 2, 85-97.

Peterson B.J., Hobbie J.E., Corliss T.L. & Kriet K. (1983) Acontinuous-flow periphyton bioassay: tests of nutrientlimitation in a tundra stream. Limnology and Oceano-graphy, 28, 583-591.

Peterson B.J., Wollheim W.M., Mulholland P.J. et al.(2001) Control of nitrogen export from watersheds byheadwater streams. Science, 292, 86-90.

Rathbun R.E., Stephens D.W., Schultz DJ. & Tai D.Y.(1978) Laboratory studies of gas tracers for reaeration.Proceedings of the American Society of Civil Engineering,104,215-229.

1 Reynolds C.S. (1984) The Ecology of Freshwater Phytoplank-ton. Cambridge University Press, London.

Rosemond A.D., Mulholland P.J. & Elwood J.W. (1993)Top-down and bottom-up control of periphyton in awoodland stream: effects of and between nutrients andherbivores. Ecology, 74,1264-1280.

SAS (1985) SAS User's Guide Statistics. SAS Institute,Gary, NC

Schlesinger W.H. (1997) Biogeochemistry: An Analysis ofGlobal Change. Academic Press, San Diego, CA.

Sinsabaugh R.L. (1997) Large-scale trends for streambenthic respiration. Journal of the North AmericanBenthological Society, 16,119-122.

Stream Solute Workshop (1990) Concepts and methods forassessing solute dynamics in stream ecosystems. Journalof the North American Benthological Society, 9,95-119.

Suberkropp K. & Chauvet E. (1995) Regulation of leafbreakdown by fungi in streams: influences of waterchemistry. Ecology, 76,1433-1445.

Inter-biome comparison of stream metabolism 1517

Tank J.L. & Webster J.R. (1998) Interaction of substrateand nutrient availability on wood biofilm processes instreams. Ecology, 79,2168-2179.

Tsivoglou B.C. & Neal L.A. (1976) Tracer measurement ofreaeration. HI. Predicting the capacity of inlandstreams. Journal of the Water Pollution Control Federation,48,2669-2689.

Uehlinger U. & Naegeli WM. (1998) Ecosystem metabo-lism, disturbance, and stability in a prealpine gravelbed river. Journal of the North American BenthologicalSociety, 17,165-178.

Uehlinger U., Konig C. & Reichert P. (2000) Variabilityof photosynthesis-irradiance curves and ecosystemrespiration in a small river. Freshwater Biology, 44,493-507.

Webster J.R. & Ehrman T.P. (1996) Solute dynamics. In:Methods in Stream Ecology (Eds F.R. Hauer & G.A.Lamberti), pp. 145-160. Academic Press, San Diego,CA.

Webster J.R., Wallace J.B. & Benfield E.F. (1995) Organicprocesses in streams of the eastern United States.In: Ecosystems of the World 22: River and StreamEcosystems (Eds C.E. Gushing, K.W. Cummins &G.W. Minshall), pp. 117-187. Elsevier, Amsterdam.

Young R.G. & Huryn A.D. (1996) Interannual variation indischarge controls ecosystem metabolism along agrassland river continuum. Canadian Journal of Fisheriesand Aquatic Sciences, 53, 2199-2211.

Young R.G. & Huryn A.D. (1998) Comment: improve-ments to the diurnal upstream-downstream dis-solved oxygen change technique for determiningwhole-stream metabolism in small streams. CanadianJournal of Fisheries and Aquatk Sciences, 55, 1784-1785.

Young R.G. & Huryn A.D. (1999) Effects of land use onstream metabolism and organic matter turnover.Ecological Applications, 9,1359-1376.

(Manuscript accepted 10 March 2001)

© 2001 Blackwell Science Ltd, Freshwater Biology, 46,1503-1517

I