Effects of nutrients and physical lake characteristics on bacterial and phytoplankton production: A meta-analysis

Effects of nutrients and physical lake characteristics on bacterial and phytoplankton

production: A meta-analysis

C. L. Faithfull,a,* A.-K. Bergstrom,a and T. Vredeb

aDepartment of Ecology and Environmental Science, Umea University, Umea, SwedenbDepartment of Aquatic Sciences and Assessment, Swedish University of Agricultural Sciences, Uppsala, Sweden

Abstract

We performed a meta-analysis comprising field (300 studies) and experimental data (249 studies) from a widerange of lake trophic states and locations. We examined the effects of nitrogen (N), phosphorus (P), carbon(dissolved organic matter [DOM]), temperature, latitude, and lake morphometry on the absolute and relative ratesof phytoplankton primary production (PPr) and secondary bacterial production (BP). Areal and volumetric rates ofPPr, BP, and BP : PPr were compared, and we analyzed differences between experimental and natural systems. Bothfield studies and experimental results showed agreement with regard to N and P as predictors of volumetric PPr andBP, respectively, despite the large variation in study duration, size, and nutrient addition rates in experimentalsystems. This indicates that bacteria and phytoplankton do not seem to be competing for the same nutrients. Arealmeasurements were more difficult to predict and were more dependent on physical lake characteristics thannutrients. Temperature was positively correlated with PPr, but not with BP. BP : PPr was stable across experimentsregardless of N, P, DOM, or glucose additions. In contrast, BP : PPr ratios varied greatly in the field data set andwere highest in systems with low total N and at high latitudes. This pattern was driven by reduced PPr, not BP;therefore, experimenters may need to manipulate PPr to change BP : PPr. Collectively, our results indicate thatincreased temperatures and N availability will lead to higher PPr and lower BP : PPr, potentially decreasing theimportance of energy mobilized through the microbial food web on a global scale.

One aspect of global climate change is the anthropogenicalteration of the global biogeochemical cycles of nitrogen(N), phosphorus (P), and carbon (C), which have increasedby c. 100%, c. 400%, and c. 13%, respectively, frompreindustrial levels (Falkowski et al. 2000). These changescan be expected to have a huge effect on freshwaterecosystems, as N, P, and C are essential elements that mostoften limit phytoplankton primary production (PPr) andsecondary bacterial production (BP) in the pelagic zone oflakes (Tranvik 1988; Elser et al. 2007). Another aspect ofglobal climate change is enhanced air and water temper-atures (IPCC 2007), and increased water temperature hasthe potential to affect growth and respiration rates ofbacteria and phytoplankton (Wetzel 2001; Berggren et al.2010). The relative and absolute rates of PPr and BP notonly represent the total basal energy source for the pelagicaquatic food web (Jones 1992), but also influenceecosystem function and biogeochemical cycles (del Giorgioand Peters 1994; Jansson et al. 2007). Hence, it is importantto describe and understand the elemental and physical lakecharacteristics that determine the production of bacteriaand phytoplankton and their relative proportions.

Historically, increasing P concentrations have beencorrelated with increasing phytoplankton biomass infreshwater ecosystems, as N-fixing bacteria or cyanobacte-ria were assumed to compensate for phytoplankton Nlimitation (Schindler 1977). However, this can differ alonga gradient from oligotrophic to eutrophic lakes (Downingand McCauley 1992), across lake size (mean depth andarea) (Thebault et al. 1999), and along gradients ofatmospheric N deposition (Elser et al. 2009). Additionally,

like all photosynthetic organisms, PPr is also driven bylight availability (Jones 1992; Kalff 2003), which is knownto be affected by dissolved organic matter (DOM) (Jones1992) and eutrophication, i.e., high biomass developmentof phytoplankton may cause self-shading with increasednutrient inflows (Wetzel 2001; Vadeboncoeur et al. 2003).

Many diverse studies have investigated the factorsregulating BP in different aquatic environments (Tranvik1988; Kirchman 1994), but relatively few studies haveattempted to make generalizations as to what limits BP infreshwater ecosystems (Cole et al. 1988; Nurnberg and Shaw1998). Still, it is generally acknowledged that when there is noexternal (allochthonous) source of C, BP can be regulated bythe amount of autochthonously produced C as a by-productof photosynthesis (Kirchman 1994). Bacteria are often Plimited in natural systems and, because of the high affinityand cell P content of bacteria, they are good competitors forP compared to phytoplankton (Vadstein 2000). Inputs ofallochthonous DOM, especially to unproductive lake eco-systems, may relieve bacterial C limitation and allow bacteriato outcompete phytoplankton for P, potentially resulting in asystem in which BP exceeds PPr (Jansson 1998). However,this pattern is dependent on competition between BP and PPrfor nutrients, and may be masked when examining lakesalong a wide trophic gradient (Jones 1992).

Another factor influencing pelagic production is lakelocation, i.e., lower latitudes are generally considered moreP limited because of soil age and high N fixation (Schle-singer and Andrews 2000), but land use and atmospheric Ndeposition may disguise this pattern (Lebauer and Treseder2008). Latitude is also correlated with temperature, and inlaboratory studies BP has been found to increase exponen-tially with temperature (Berggren et al. 2010). PPr can also*Corresponding author: [email protected]

Limnol. Oceanogr., 56(5), 2011, 1703–1713

E 2011, by the American Society of Limnology and Oceanography, Inc.

doi:10.4319/lo.2011.56.5.1703

1703

show a positive response to temperature, although this hasnot been as well documented as for bacteria (Vrede et al.1999). The response to temperature may, thus, be greaterfor BP than PPr; therefore, increasing temperatures mayincrease the BP : PPr ratio (Muren et al. 2005; Hoppe et al.2008). Physical attributes of lake ecosystems such as meandepth and lake area may also affect PPr and BP, i.e., large,shallow lakes tend to have higher sediment resuspensionand P release from sediments than deep lakes (Kalff 2003).Lake size, mean depth, and water color also interact toaffect euphotic and epilimnion depths and, hence, the lightclimate (Jones 1992).

Although field studies allow us to examine what factorscontrol BP and PPr on a large scale, experimental studies ofN, P, and DOM additions enable researchers to examine themechanisms behind nutrient limitation of BP and PPr in acontrolled environment. However, the applicability of scalingup conclusions from experimental studies to natural ecosys-tems is questionable (Schindler 1998) and may depend on thesize and duration of the experiment (Petersen et al. 1999).Ideally, to truly understand the regulation of BP and PPr,survey and experimental approaches should be combined.

Here we combine results from field observations andexperimental studies to examine which elements (N, P, andC) and physical lake characteristics (temperature, latitude,and lake morphometry) determine the relative and absoluterates of areal and epilimnetic volumetric BP and PPr, andthus, the total basal energy mobilization within freshwaterpelagic systems. We compare areal and volumetric rates ofBP and PPr in field studies because volumetric measure-ments may overestimate PPr relative to BP, as PPr does notoccur below the euphotic zone, whereas BP occursthroughout the whole water column (Ochs et al. 1995).The data we collected were used to answer the following

questions: Which major elements (N, P, and C) and/orphysical factors (temperature, latitude, and lake morphom-etry) regulate the relative and absolute rates of BP and PPrin freshwater ecosystems? What factors determine if thesystem is dominated by BP or PPr? Are there differencesbetween the factors that determine volumetric and arealproduction rates? How do the results of field andexperimental studies compare, and if they differ, why?

Methods

Field studies—Relevant studies were found by searchingtitles and abstracts of publications found on the Web ofScienceE and Google ScholarE search engines usingcombinations of the keywords bacteria*, phyto*, primary,production, lake*, nutrient*, and carbon (* indicates allpossible word endings). Additional sources were found bysearching citation lists. Some data were obtained viapersonal communication (P. Kankaala, J. Ask, A. Wenzelunpubl.). Studies were included when (1) BP and/or PPrvalues were reported and measured at the same time andplace or the same season; (2) the method of measurementwas reported; and (3) data for total N (TN), total P (TP),DOM or C (from here on both types are referred to asDOM), latitude and longitude, and lake area and meandepth were reported for approximately the same period asBP and/or PPr. We restricted our analysis to inland waterbodies (Fig. 1).

We collected summer epilimnion values of volumetric BP(mg C m23 h21) and PPr (mg C m23 h21) and areal BP (mgC m22 h21) and PPr (mg C m22 h21) for temperate lakes,and dry and rainy season data for tropical lakes. 92.4% ofPPr measurements were taken using the 14C method, 6.6%using oxygen measurements, and 1% of sources did not

Fig. 1. World map showing locations of experimental studies and field studies obtained from the literature or bypersonal communication.

1704 Faithfull et al.

report PPr method. For BP measurements, 63.8% weremade using leucine isotope, 35% with thymidine isotope,and 1.2% using changes in bacterial abundance. Areal dataeither were reported as areal measurements or werecalculated from volumetric rates using lake morphometryand euphotic depth (Zeu). From the compiled data wecalculated the BP : PPr ratio (BP : PPr 5 log10(BP/PPr). BP,PPr, and BP : PPr were used as response variables. Theexplanatory variables were TN (mg N L21), TP (mg P L21),and DOM (mg C L21) concentration, latitude, and lakeshape (lake area [km2]/lake mean depth [m]). A large lakeshape indicates a large shallow lake and a small lake shaperepresents a small deep lake. All data were log10transformed except for latitude. Latitude values used wereabsolute, without an indication of north or south. Wecalculated average epilimnetic production and nutrientconcentration values, which were weighted by the squareroot of the number of data points that contributed to themean. When mean values were not available, values forsingle sampling dates were used with a weight of 1. Meanwater temperature (uC) for the period of the productionmeasurements, Zeu, and trophic state, according to theoriginal authors’ definition, were recorded as descriptivevariables. Where Zeu was not given, but Secchi depth (SD)or the vertical light extinction coefficient (Kd) wasavailable, we calculated Zeu from SD according toLaPerriere and Edmundson (2000): ln(Zeu) 5 1.19 +

0.92ln(SD), and Zeu from the Kd: ln(Zeu) 5 1.19 +

0.92ln((Kd–0.03)/1.46).All BP and PPr rates were converted to hourly rates in

order to avoid a positive correlation with PPr and latitudedue to day length. PPr daily rates were converted to hourlyrates using an online photoperiod calculator (http://www.saunalahti.fi/,benefon/sol.html) and the lake coordinatesand excluding civil twilight. There were no differences inthe relationships between explanatory variables and arealor volumetric BP : PPr ratios when using either daily orhourly measurements, so we present the hourly resultsbelow. Volumetric PPr measurements were standardizedfor Zeu: PPr 3 Zeu/mean depth (or 1 if Zeu exceeded meandepth). This was done to standardize for the area of thelake where photosynthesis could take place. Because of thedata selection requirements, we included 300 lake years(temporally and spatially distinct data collection episodes)conducted in 220 different lakes (Fig. 1), which variedgreatly in trophic state; TN ranged between 41 and 4680 mgN L21, the range of TP was 1–700 mg P L21, and DOM was0.08–41 mg L21. Mean depth was 0.2–688 m and lake areavaried between 1.4 m2 and 31,500 km2. For a list of articlesincluded see Web Appendix (www.aslo.org/lo/toc/vol_56/issue_5/1703a.html).

Volumetric BP and PPr were correlated, as were theexplanatory variables to various degrees (Fig. 2). In orderto deal with the problem of multicollinearity for both ourexplanatory and response variables we used both multipleregression and hierarchical partitioning methods to exam-ine relationships between the variables. As temperaturedata were only available for a smaller subset of the lakedata (n5 36–55) we conducted multiple regression analysesboth with and without temperature. Nonsignificant terms

were removed stepwise from the model using backwardselection. The final model was chosen based on the simplestmodel with the lowest Akaike information criterion(Crawley 2002).

Hierarchical variance partitioning jointly considers allpossible models in a multiple regression data set, whichallows the identification of variables that are independentlycorrelated with a response variable even when multi-collinearity is present (Murray and Conner 2009). Thehierarchical partitioning method does not cope withmissing values, so only data where all explanatory variableswere available were used. This resulted in a subsample of117 lake years (data for each year and each lake wereconsidered as independent data points), including 78different water bodies: Scandinavian lakes (23 lakes),North American temperate lakes (27 lakes), Texas sub-tropical reservoirs (9 lakes), Brazilian tropical lakes (16lakes) and Polish lakes (3 lakes). The range in nutrientconcentrations was TN, 62–4682 mg N L21; TP, 2.9–313 mgP L21; DOM, 0.5–36 mg C L21; lake area, 14 m2–93 km2;mean depth, 0.2–23 m; latitude, 10.35–74.50u. The Rpackage hier.part (Walsh and MacNally 2008) was usedfor the hierarchal partitioning analysis.

Experimental studies—Studies for the second part of ouranalysis were found using the same databases and keywordsas above, but with the inclusion of the keywords experi-ment*, mesocosm* and glucose*.We included freshwater (,5% salinity) mesocosm, microcosm, chemostat, and nutrientbioassay experiments. Our selection requirements were (1)the study must be experimental and means and variancesmust be available for both control and manipulatedtreatments; (2) the study must measure BP and/or PPr whilemanipulating one or all of P, N, glucose (G), or DOM or C(DOC or humic substances, hereon referred to as DOM);and (3) the study must have been published or performed(for unpublished studies) in the last 20 yr. Some data wereincluded from the currently unpublished work of the authorsand via personal communication (K. Dahlgren unpubl.).Study locations are shown in Fig. 1.

Data collected were the response variables volumetric BPand PPr (mg C m23 h21), and the rate of treatmentaddition of inorganic N, inorganic P, G, or DOM. Wherethese treatments were applied simultaneously the abbrevi-ations have been combined in the text, e.g., NP, DOMP,etc. Additional information about the study, such astemperature, volume of experimental units, the durationof the experiment, and method of BP and PPr measure-ment, was also recorded. The ratio of BP : PPr wascalculated as for the field studies (see above). Anindependent study was defined as a temporally andspatially distinct experiment with internally consistentcontrols. Therefore, data from a study that performedreplicate experiments in the same environment overdifferent time periods (i.e., separate experiments in spring,summer and autumn) were averaged, with the exclusion ofthe initial or ‘‘zero’’ measurement, as these measurementscould not be treated as independent. In total, 247independent studies were used in our analysis (see WebAppendix). Unfortunately, few studies reported TN, TP,

Reviews in L&O 1705

and DOM values for source water, so the ranges for TN(1.4–1500 mg N L21), TP (1.5–50 mg P L21), and DOM(0.15–37.4 mg C L21), may not be a good indication of thetrue trophic range of the data. Nutrient addition rangeswere N addition, 0.54–1400 mg N L21; P addition, 0.170–310 mg P L21; G addition, 0.6–18,600 mg C L21; and DOMaddition, 0.08–4640 mg L21. We used the standardized

mean difference (Hedges g) as our effect size metric and wecalculated the standard error of the effect size and inverseweighted variance of the effect size according to Lipsey andWilson (2001). Positive effect sizes indicate a positiveresponse to the treatment. Because of the prevailing smallsample sizes (n , 20) in many of the included studies wecorrected the effect size for small sample size according to

Fig. 2. Bivariate Pearson’s production moment correlation coefficients for areal and volumetric rates of BP, PPr, and explanatoryvariables from the field data set. Correlation coefficients are shown under each scatter plot. Significant correlations (p, 0.010) are shown inbold type. Lake shape is represented by ‘‘Shape’’ and temperature by ‘‘Temp.’’ All other abbreviations and units are as stated in the text.

1706 Faithfull et al.

Lipsey and Wilson (2001). All effect sizes were weightedaccording to the inverse weighted variance and the datawere analyzed for homogeneity (Lipsey and Wilson 2001).All of the treatment groups were found to be heteroge-neous, which is a common feature of ecological data sets,and in our case the studies come from many differentaquatic ecosystems, which may experience a broad range oflimiting nutrients. To allow for this heterogeneity, allfurther analysis was carried out assuming a random errorstructure, which produces slightly more conservativeconfidence intervals (Lipsey and Wilson 2001). In additionto the analysis of the treatment groups, we analyzed theeffects of covariables on effect size by using multipleregression for continuous covariables (i.e., volume ofexperimental unit, temperature, and study duration,latitude, rate of N, P, G, or DOM addition) or analysisof variance for factorial covariables (i.e., mixing, presenceor absence of grazers). All covariable data were log10transformed to maintain a normal distribution and allstatistical analyses were conducted in the R program (RDevelopment Core Team 2009).

Results

Areal PPr ranged between 0.0044 and 1620 mg Cm22 h21

and volumetric PPr ranged between 0.00119 and 381 mg Cm23 h21 in the field data set. Both volumetric and areal PPrwere closely correlated in this data set, and were bestexplained by positive correlations with TN and temperatureand a negative correlation with DOMwhen temperature wasincluded in the models (Tables 1, 2). Differences betweenvolumetric and areal PPr were caused by factors thataffected Zeu, e.g., DOM, lake shape, and latitude explainedmore variation in areal PPr than volumetric PPr, and TNwas a better predictor of volumetric PPr than areal PPr. Thebest model for areal PPr also included positive correlationswith lake shape and TP, although these did not explain muchof the variance (12% and 5%, respectively). Withouttemperature included, PPr was negatively correlated withincreasing latitude and DOM, and the negative correlationbetween PPr and latitude strengthened (from 36% to 50%for areal and 24% to 40% for volumetric PPr), suggestinglatitude accounted for some of the effects of temperature(Tables 1, 2). Without temperature PPr remained positivelycorrelated with TN, and was higher in larger, shallowerlakes. Models for both areal and volumetric PPr withouttemperature had higher explanatory power than withtemperature. PPr ranged between 0–904 mg C m23 h21 inthe experimental study. Experimental additions of Nincreased PPr slightly more than P additions (Fig. 3). WhenN and P were added simultaneously, PPr more thandoubled, suggesting co-limitation of N and P in experimentalsystems. There was no effect of G addition on PPr, andDOM addition had a weak positive effect on PPr.

There were marked differences between predictors ofareal and volumetric BP. Areal BP ranged between 0.00091and 6764 mg C m22 h21, was positively correlated withDOM and TP, and was higher in shallow large lakes(Table 1). Areal BP was negatively correlated with TN anduncorrelated with areal PPr and temperature. The best

multivariate model for areal BP had a low predictive powerand did not coincide with the hierarchical partitioningresults, suggesting a low predictive power of the responsevariables (Table 1). Volumetric BP ranged between 0.00021and 446 mg C m23 h21, and was positively correlated withPPr, TP, and latitude, and negatively correlated withtemperature (Table 2). Without temperature included inthe model, volumetric BP remained positively correlatedwith PPr, DOM, and TP, but was higher in shallow largelakes and was negatively correlated with TN. In theexperimental data set BP ranged between 0.000001 and171 mg C m23 h21. All treatment combinations increasedBP, with the greatest increases apparent for N and P; G, N,and P; and G and P (GP) treatments (Fig. 3).

The areal BP : PPr ratio ranged between 0.000002 and519. Volumetric BP : PPr ranged between 0.0000009 and85.3. Both volumetric and areal BP : PPr ratios werepositively correlated with DOM and latitude and negativelycorrelated with TN and temperature in the field study(Tables 1, 2). However, higher areal BP : PPr was associat-ed with small deep lakes, whereas volumetric BP : PPr wasnot associated with lake morphometry. The negativecorrelations with areal BP : PPr and lake shape, TP, andtemperature were not significant for the hierarchicalpartitioning results (Table 1). When we excluded temper-ature, the best model for areal BP : PPr showed positivecorrelations with DOM, latitude, and TP, and was higherin small deep lakes, but still conflicted somewhat with thehierarchical partitioning results, as TN was not included inthe model, but explained 15% of the variation in arealBP : PPr (Table 1). The best model for volumetric BP : PPrwas also found when excluding temperature, and includedpositive correlations with DOM, TP, and latitude, and anegative relationship with TN (Table 2). The variablesaccounting for the most variation in volumetric BP : PPrwere latitude (44%), TN (21%), DOM (15%), and lakeshape (14%), even though the latter was not included in themodel. The positive effect of DOM and TP on BP : PPrsuggests that this is driven by changes in BP. In theexperimental study BP : PPr ratios had a higher upper range(0.05–1410) than in natural systems, which is unsurprisingconsidering that the aim of many experiments was tomanipulate this ratio. The BP : PPr ratios tended to increasewith G addition, but this trend was not significant, andotherwise BP : PPr did not differ with any form of nutrientaddition. The small error bars indicate that effect sizes forBP : PPr were generally close to zero (Fig. 3).

In the experimental studies none of the other covariates(N and P addition rates, temperature, duration of experi-ment, size of experimental unit, latitude, presence or absenceof grazers, presence or absence of mixing, ratio of N : Paddition, concentration of G, or DOM additions) showedany significant relationship with BP, PPr, or BP : PPr. Eventhough TN, TP, and DOM are not totally bioavailable, therelationships between TN, TP, BP, and PPr are consistentwith the results from experimental additions of dissolvedinorganic N and P (Fig. 3). In freshwater systems, TN andTP are commonly measured indicators of trophic state, so itis reassuring that these parameters correlate with biologicalmeasurements of production.

Reviews in L&O 1707

Discussion

The most important predictors of areal and volumetricPPr in the field data set were latitude, TN, and lake shape.Latitude alone explained the most variation in areal (50%)and volumetric (40%) PPr. Much of this variation wasaccounted for by temperature, as latitude became moreimportant when we excluded temperature from the models.Positive correlations with PPr and temperature have been

found in field studies, and temperature can also affectthermocline and therefore epilimnion depth (Vrede et al.1999; Lomas et al. 2002). However, latitude explained morevariation in PPr than temperature, and previous studies havefound that increasing latitude is negatively correlated withPPr in northern hemisphere lakes (Hakanson and Boulion2001; Flanagan et al. 2003). PPr can vary with many factorsthat are correlated with latitude, i.e., mean annual temper-ature, variation in solar angle or photosynthetically active

Table 1. Multiple regression and hierarchical partitioning results for areal production. Model selection with and withouttemperature was based on backwards selection using Akaike information criterion values. All values are log10 transformed, except forlatitude. Z-scores indicate statistical significance, based on an upper 0.95 confidence limit Z $ 1.65. I-values represent the percentagevariance accounted for by each explanatory variable. Hierarchical partitioning results significant at p , 0.05 are highlighted in bold type.

Mixed model Confidence intervals Hierarchical partitioning

Parameter R2 F p Slope p 2.5% 97.5% Z I

PPr 0.526 F5,41 5 11.2 ,0.001

Intercept 24.71 ,0.001 28.55 25.59DOM 22.80 ,0.001 22.56 20.932 2.57 15.3

TP 1.04 0.04 20.744 0.748 0.38 5.1TN 1.47 ,0.001 1.58 2.97 1.52 12.9

Lake shape 0.17 0.057 20.792 0.240 1.63 11.9Latitude 5.94 35.6

Temperature 2.47 ,0.001 0.952 3.230 3.84 19.2

logPPr 5 24.71 2 2.80(logDOM) + 1.04(logTP) + 1.47(logTN) + 0.17(log Lake shape) + 2.47(logTemperature)

PPr 0.769 F4,70 5 62.4 ,0.001

Intercept 1.041 0.17 20.465 2.547DOM 21.366 ,0.001 21.947 20.786 8.87 14.3

TP 0.29 2.3TN 1.085 ,0.001 0.571 1.600 5.08 14.5

Lake shape 0.313 ,0.001 0.157 0.468 9.25 19.3

Latitude 20.034 ,0.001 20.046 20.023 21.75 49.6

logPPr 5 1.04 2 1.37(logDOM) + 1.09(logTN) + 0.313(log Lake shape) 2 0.034(Latitude)

BP 0.357 F4,94 5 14.6 ,0.001

Intercept 0.154 0.600 20.419 0.727PPr 0.26 12.9DOM 0.629 ,0.001 0.154 1.105 0.43 12.7TP 1.24 ,0.001 0.724 1.756 0.1 7.30TN 20.582 ,0.001 20.893 20.271 0.57 12.6Lake shape 0.274 ,0.001 0.150 0.398 20.41 3.57Latitude 4.35 47.6

Temperature 20.45 3.31logBP 5 0.154 + 0.629(logDOM) + 1.24(logTP) 2 0.582(logTN) + 0.274(log Lake shape)

BP : PPr 0.796 F5,19 5 19.7 ,0.001

Intercept 25.21 0.056 210.58 0.156DOM 2.57 0.020 0.445 4.70 2.88 21.8

TP 22.48 0.104 25.53 0.562 0.49 10.7TN 1.49 14.7Lake shape 20.740 ,0.001 21.16 20.323 0.83 12.4Latitude 0.137 ,0.001 0.087 0.187 3.79 34.2

Temperature 23.03 0.120 26.88 0.820 0.05 6.3logBP : PPr 5 2 5.21 + 2.57(logDOM) 2 2.48(logTP) 2 0.274(log Lake shape) + 0.137(Latitude) + 3.03(logTemperature)

BP : PPr 0.651 F4,57 5 29.5 ,0.001

Intercept 24.60 ,0.001 25.712 23.487DOM 1.28 0.0032 0.445 2.111 5.21 16.0

TP 0.485 0.171 20.216 1.185 0.83 4.0TN 5.16 15.1

Lake shape 20.261 0.0452 20.516 20.006 3.28 13.5

Latitude 0.049 ,0.001 0.032 0.065 13.1 51.5

logBP : PPr 5 2 4.60 + 1.28(logDOM) 2 0.485(logTP) 2 0.261(log Lake shape) + 0.049(Latitude)

1708 Faithfull et al.

Table 2. Multiple regression and hierarchical partitioning results for volumetric production. Model selection with and withouttemperature was based on backwards selection using Akaike information criterion values. All values are log10 transformed, except forlatitude. Z-scores indicate statistical significance, based on an upper 0.95 confidence limit Z $ 1.65. I-values represent the percentagevariance accounted for by each explanatory variable. Hierarchical partitioning results significant at p 5 0.05 are highlighted in bold type.

Mixed model Confidence intervals Hierarchical partitioning

Parameter R2 F p Slope p 2.5% 97.5% Z I

PPr 0.566 F3,67 5 31.4 ,0.001

Intercept 26.79 ,0.001 28.08 25.51DOM 21.75 ,0.001 22.35 21.15 1.82 11.6

TP 20.08 2.15TN 2.17 ,0.001 1.57 2.78 7.59 45.7

Lake shape 20.2 1.91Latitude 6.08 23.7

Temperature 2.04 ,0.001 1.00 3.08 2.77 15.1

logPPr 5 26.79 2 1.75(logDOM) + 2.17(logTN) + 2.04(logTemperature)

PPr 0.736 F4,87 5 64.5 ,0.001

Intercept 21.79 0.0202 23.30 20.29DOM 21.26 ,0.001 21.86 20.67 4.82 9.84

TP 2.54 5.57

TN 1.69 ,0.001 1.17 2.21 17.3 29.0

Lake shape 0.31 ,0.001 0.16 0.47 9.5 15.1

Latitude 20.03 ,0.001 20.04 20.01 25.8 40.5

logPPr 5 21.79 2 1.26(logDOM) + 1.69(logTN) + 0.31(logLake shape) 2 0.03(Latitude)

BP 0.528 F4,39 5 13.0 ,0.001

Intercept 24.27 ,0.001 26.07 22.472PPr 0.24 0.0579 20.01 0.491 1.08 12.6DOM 1.71 19.3

TP 1.15 ,0.001 0.57 1.737 4.26 39.4

TN 0.5 6.44Lake shape 20.63 1.05Latitude 0.04 ,0.001 21.84 1.655 1.13 13.51Temperature 20.09 0.9146 0.02 0.063 0.69 7.73logBP 5 24.27 + 0.24(logPPr) + 1.15(logTP) + 0.04(Latitude) 2 0.09(logTemperature)

BP

Intercept 0.606 F6,66 5 19.5 ,0.001 21.06 0.1062 22.36 0.233PPr 0.29 0.0033 0.10 0.483 5.55 18.0

DOM 0.68 0.0125 0.15 1.218 2.15 7.46

TP 1.29 ,0.001 0.84 1.745 12.58 42.3

TN 20.70 0.0260 21.32 20.086 2.36 10.9

Lake shape 0.22 0.0095 0.06 0.386 3.59 15.9

Latitude 0.01 0.0185 0.00 0.022 1.36 5.32logBP 5 21.06 + 0.29(logPPr) + 0.68(logDOM) + 1.29(logTP) 2 0.70(logTN) + 0.22(log Lake shape) + 0.01(Latitude)

BP : PPr 0.660 F5,34 5 16.1 ,0.001

Intercept 1.45 0.354 21.69 4.58DOM 1.33 0.0101 0.337 2.32 1.62 10.8TP 0.94 0.0297 0.098 1.78 0.44 6.83TN 21.34 0.0101 22.33 20.34 4.05 24.7

Lake shape 2.16 16.6

Latitude 0.05 0.0088 0.012 0.08 5.25 30.0

Temperature 23.02 0.0145 25.41 20.64 1.12 11.0logBP : PPr 5 1.45 + 1.33(logDOM) + 0.94(logTP) 2 1.34(logTN) + 0.05(Latitude) 2 3.02(logTemperature)

BP : PPr 0.707 F4,72 5 46.8 ,0.001

Intercept 0.16 0.840 21.40 1.71DOM 1.20 ,0.001 0.631 1.76 7.74 14.6

TP 1.07 ,0.001 22.33 21.06 2.28 6.4

TN 21.69 ,0.001 0.521 1.63 7.47 20.8

Lake shape 5.02 14.0

Latitude 0.03 ,0.001 0.020 0.04 23.6 44.2

logBP : PPr 5 0.16 + 1.20(logDOM) + 1.07(logTP) 2 1.69(logTN) + 0.03(Latitude)

Reviews in L&O 1709

radiation (PAR), terrestrial productivity, or atmospheric Ndeposition rates (Campbell and Aarup 1989; Hakanson andBoulion 2001; Bergstrom and Jansson 2006). However, solarangle does not directly affect the amount of PAR available athigh latitudes in summer (Campbell and Aarup 1989). Thecorrelation of a large suite of climatic factors with latitudecould explain why latitude accounted for such a highproportion of the variance in PPr, and thus provides us withan easily measurable parameter to predict PPr.

Latitude is also correlated with TN if we remove the 16Brazilian lakes and 2 New Zealand lakes, which are in areasof low N deposition (slope 5 20.349, p , 0.001). AlthoughN deposition may have accounted for some of the variationin PPr explained by latitude, TN was also an importantpredictor of areal (15%) and volumetric PPr (29%). Researchhas shown that in areas with low atmospheric N deposition(such as at high latitudes) N strongly limits PPr, especially inunproductive lakes (Vitousek and Howarth 1991; Bergstromand Jansson 2006; Elser et al. 2009). In productive lakes Nfixation by cyanobacteria can potentially compensate for Ndeficiency, as N-fixing cyanobacteria often become dominantat lowN :P ratios, thus alleviating N limitation of PPr (Vredeet al. 2009). However, in nutrient-rich lakes, N limitation isalso common (Downing and McCauley 1992; Elser et al.2007), and in some cases N fixation may not totallycompensate for N limitation of PPr (Scott and McCarthy2010). In the experimental studies PPr was also most stronglyN limited. Positive synergistic effects of combined N and Paddition on PPr were evident in the nutrient enrichmentexperiments, because single enrichment of either N or P likelyinduces limitation by the other element (Elser et al. 2007).

Large shallow lakes had higher PPr, both for volumetric,but especially for areal PPr. Although we standardized PPrfor Zeu (cf. Methods), this did not affect the relationship ofPPr with lake shape (unstandardized volumetric logPPr 520.223 + 0.976(logTN) + 0.371(logLake shape) 20.035(Latitude), F3,93 5 86.6, R2 5 0.728, p , 0.001).Large shallow lakes tend to have more sediment resuspen-sion, and therefore higher turbidity and nutrient concen-trations (Kalff 2003). However, in our data set lake shapewas not correlated with TN, TP, or any other explanatoryvariable. In addition, large shallow lakes often have a largerZeu relative to mean depth, which is the case in this data set(R2 5 0.254, p 5 0.0027, n 5 135). This could increase lightand nutrient availability and reduce sinking rates ofphytoplankton cells (Thebault et al. 1999; Jager et al. 2008).

DOM concentrations negatively influenced areal PPr in thefield studies. However, DOM and G showed little or no effecton volumetric PPr measured in the field and experimentalstudies. One explanation for this is that DOM reduced lightlevels for photosynthesis for in situ areal PPr (Jones 1992) butnot volumetric PPr. In the experimental studies light wasusuallymaintained at a saturating level across treatments evenwhenDOMwas added (except for Berglund et al. 2007 andK.Dahlgren unpubl., who manipulated light levels). In the fieldstudies Zeu standardized for mean depth was negativelycorrelated with increasing DOM (Pearson’s productionmoment correlation: 20.266, p 5 0.006, n 5 104). Previousstudies using large data sets (del Giorgio and Peters 1994;Carpenter et al. 1998; Nurnberg and Shaw 1998) have alsoshown that PPr was negatively correlated with decreasingtransparency and increasing water color, which were bothpositively correlated with DOM. All in all, areal andvolumetric measurements of PPr showed consistent relation-ships with the explanatory variables (latitude, TN, lake shape,temperature, and DOM) in this study. The small effect sizesfor DOM and G addition on PPr in the experimental studiessuggests that bacteria do not outcompete phytoplankton fornutrients across a wide trophic gradient when there is an

Fig. 3. Averageweighted effect sizes (Hedges g) and correspond-ing 95% confidence intervals for PPr, BP (confidence interval forDOMP is 2.37 6 6.67), the ratio of bacterial to phytoplanktonproduction (BP :PPr) (confidence intervals forGNandGPare20.2146 27.2 and 5.06 6 70.9, respectively, but are not shown), across alltreatment combinations of N, P, DOM, and G addition. Numbersdenote the number of studies attributing to the effect size. Abbrevi-ations are explained in the text.

1710 Faithfull et al.

external C source (Jansson 1998). Alternatively, it appearsthat other properties of the system, such as trophic state (Jones1992), nutrient stoichiometry (Elser et al. 2009), lightlimitation (Jones 1992), and the proportions of autotrophicand mixotrophic phytoplankton biomass (Bergstrom et al.2003), may regulate the interactions between PPr and BP.

The main determinate of volumetric BP and areal BP inthe field data set and experimental studies was TP. Thisreflects earlier studies of bacterial stoichiometry that confirmbacterial cells have a high P :C ratio (Vadstein 2000), thusrequiring high P concentrations for growth. DOM waspositively correlated with BP in the experimental studies,and with areal and volumetric BP, as has been welldocumented in the literature for temperate lakes (Tranvik1988; Jones 1992). Similar to PPr, large shallow lakes hadhigher BP, which could be due to increased PPr in thesesystems, or increased turbidity providing a supply ofnutrients (i.e., P) and DOM from the sediments (Kalff2003). The positive correlation between volumetric BP andPPr was not as steep and did not explain as much variationas previously reported correlations for BP and PPr (Cole etal. 1988; Fouilland and Mostajir 2010). Our data set had alarger range in BP and PPr values than both aforementionedstudies and possibly also had a higher variation in nutrientand environmental factors. Consequently, at the lowest andhighest values of BP and PPr or at the extremes of thetrophic gradient the log-linear relationship between BP andPPr may disintegrate. Volumetric BP was better explainedby our explanatory variables and had much strongercorrelations with TP and PPr than areal BP. This seemslogical if we consider that nutrient concentrations andvolumetric production rates were both calculated for theepilimnion, where bacteria and phytoplankton are probablymost tightly coupled because of proximity. Areal BP and PPrwere not correlated in our study, most likely because arealmeasurements of BP and PPr are not coupled in space, asthey are determined by different factors over depth (Ochs etal. 1995). Areal BP is expected to be heavily influenced byprocesses such as decomposition and aging of DOM(Berggren et al. 2009), and lower temperature and anoxiaoccurring in the meta- and hypolimnion (Ochs et al. 1995).

Although positive relationships between BP and tempera-ture have been found in the laboratory (Berggren et al. 2010)and across natural systems (White et al. 1991), BP andBP :PPr were negatively correlated with temperature in ourfield data set. Other studies have found that BP can still behigh at very low temperatures, and temperature only tends tolimit BP when other factors such as nutrient and Cconcentrations are not limiting or when temperature is ,

10–4uC (Laybourn-Parry et al. 2004; Kirchman et al. 2009;Berggren et al. 2010). However, bacterial communities tend tobe adapted to in situ conditions, and this adaptation mayoccur within days (Adams et al. 2010; Berggren et al. 2010). Inthe Baltic Sea, however, increasing BP andBP :PPr ratios havebeen found with increasing temperature (Muren et al. 2005;Hoppe et al. 2008). But this may reflect the local conditions,and not how BP :PPr is regulated on a wider latitudinal scale.Given the positive relationship with PPr and temperature andthe negative relationship of BP :PPr with temperature found inthis study, and rapid adaptation of the bacterial community

(Adams et al. 2010), we would rather predict that increasingtemperature could decrease the importance of energy mobi-lized through the microbial food web on a global scale.

The ratio of BP : PPr was positively correlated with DOMin the field studies, which confirms that lakes with highDOMconcentrations have relatively higher BP (Karlsson et al.2002), and reflects the individual correlations of PPr and BPwith DOM. This relationship was highest for areal BP : PPr,and small deep lakes also had higher areal BP : PPr. Areal BPwas higher in large shallow lakes, so the change in BP : PPrmust be driven by the larger effect of lake morphometry onPPr than BP, and not increased BP below the euphotic zone(Ochs et al. 1995). In this study volumetric PPr tends to bedriven by TN concentrations and volumetric BP by TPconcentrations; therefore, bacteria and phytoplankton donot seem to be competing for the same nutrients. The bestpredictor of areal and volumetric BP : PPr was increasinglatitude, which must have been driven by changes mainly inPPr, as latitude was not an important predictor of BP.Therefore, BP : PPr appeared to be regulated by changes inPPr rather than changes in BP, so at high latitudes and lowTN :TP ratios BP makes up a larger part of the availablebasal production. This has also been shown for oligotrophiclakes in low atmospheric N deposition areas in northernSweden (Karlsson et al. 2002). Consequently, the microbialfood web may be more important for energy transfer tohigher trophic levels at low TN :TP ratios and at highlatitudes (Karlsson et al. 2002). Alternatively, high latitudelakes tend to be clearer, and areal PPr may be driven bybenthic instead of pelagic PPr (Karlsson et al. 2009).

The experimental results conflict with the field data set, asthey suggest that BP : PPr is not influenced by anycombination of treatment factors across studies. Theindividual studies with the largest effect sizes for BP : PPralso decreased light intensity, thus reducing PPr (Berglund etal. 2007). Across experimental studies BP and PPr respondedproportionally to nutrient additions, and thus appear to belimited by the same factors. Alternatively, BP may appear tobe stable in both experiments and field studies, becausemeasurement intervals are better suited to capture changes inPPr rather than the rapid peaks or delayed responses that canoccur in BP rates (Wetzel 2001). Experimental studiesmeasured changes in BP and PPr with intervals of 10 h–730 d, whereas the field data were based on summer meanvalues, so these studies were probably unable to detect short-lived BP peaks. This may explain why the models producedfrom the field data set have amuch higher explanatory powerfor PPr relative to BP. As seen from the field data in ourstudy, switches in the BP : PPr ratio are generally driven bychanges in PPr (del Giorgio and Peters 1994); therefore,experiments need to focus on manipulating PPr to manip-ulate BP : PPr. However, as illustrated by the experimentaldata (Fig. 3) there are very few experiments measuring BPand PPr that have manipulated DOM, and even fewer thatexamined the effects of DOM addition at different N : Pratios. From the field data set we would expect higherBP : PPr in experiments with additions of colored DOM andlow N : P ratios. Thus, there is clearly a gap in ourunderstanding of how DOM interacts with nutrients toaffect BP : PPr in lakes of different trophic states.

Reviews in L&O 1711

Both field studies and experimental results showed tightagreement with regard to N and P as predictors of volumetricPPr andBP, respectively. This was despite the large variation instudy duration, size, and nutrient addition rates. Hence,phytoplankton and bacteria do not seem to be competing forthe samenutrients. In contrast to volumetricmeasurements,wefound that areal rates of PPr and BP were less influenced bynutrient concentrations, and more dependent on physicalvariables such as latitude and lakemorphometry. Temperaturewas positively correlated with PPr, but not with BP. Hence,increased temperatures and N availability due to climatechange could lead to higher PPr and lower BP :PPr, potentiallydecreasing the importance of energy mobilized through themicrobial food web. Although there was a large range inBP :PPr ratios in the field study, none of the experimentaladditions of nutrients, G, or DOM successfully managed tochange the BP :PPr ratio. Changes in the BP :PPr ratio in thefield data set were generally driven by changes in PPr;therefore, experiments need to focus on manipulating PPr(i.e., light, N additions) to see changes in BP :PPr.

Based on our findings we make the following recommen-dations for future research. The global distribution of studiesmeasuring BP and PPr were strongly skewed towards Europeand North America. Thus, we have very little understandingof how basal production varies with nutrient concentrations,DOM, latitude, temperature, and lake morphometry inAfrica, Asia, and South America. Especially interesting arethe tropics, as the few studies conducted in these areas havealready revealed that paradigms from northern temperatelakes may not apply (Farjalla et al. 2009). Further studies arerequired examining the effects of natural DOM additioncombined withN and P additions and realistic manipulationsof light climate, which would help us understand how thesefactors individually contribute to the relative rates of BP andPPr in pelagic systems at different levels of productivity. Thesurprising relationship we found here between BP and PPrand temperature suggests this needs further investigationalong trophic and latitudinal gradients.

AcknowledgmentsWe thank Goran Englund for statistical discussions and advice,

especially regarding the analysis of the experimental data. JennyAsk, Sebastian Diehl, Jonathan Cole, and two anonymousreviewers gave valuable comments on the draft manuscript. Thisstudy was conducted as part of the Lake Ecosystem Response toEnvironmental Change (LEREC) project and supported by agrant from the Swedish Research Council for Environment,Agricultural Sciences, and Spatial Planning (Formas).

References

ADAMS, H. E., B. C. CRUMP, AND G. W. KLING. 2010. Temperaturecontrols on aquatic bacterial production and communitydynamics in arctic lakes and streams. Environ. Microbiol. 12:1319–1333, doi:10.1111/j.1462-2920.2010.02176.x

BERGGREN, M., H. LAUDON, AND M. JANSSON. 2009. Aging ofallochthonous organic carbon regulates bacterial productionin unproductive boreal lakes. Limnol. Oceanogr. 54: 1333–1342, doi:10.4319/lo.2009.54.4.1333

———, ———, A. JONSSON, AND M. JANSSON. 2010. Nutrientconstraints on metabolism affect the temperature regulationof aquatic bacterial growth efficiency. Microb. Ecol. 60:

894–902, doi:10.1007/s00248-010-9751-1

BERGLUND, J., U. MUREN, U. BAMSTEDT, AND A. ANDERSSON. 2007.Efficiency of a phytoplankton-based and a bacteria-basedfood web in a pelagic marine system. Limnol. Oceanogr. 52:121–131, doi:10.4319/lo.2007.52.1.0121

BERGSTROM, A. K., AND M. JANSSON. 2006. Atmospheric nitrogendeposition has caused nitrogen enrichment and eutrophica-tion of lakes in the northern hemisphere. Global Change Biol.12: 635–643, doi:10.1111/j.1365-2486.2006.01129.x

———, ———, S. DRAKARE, AND P. BLOMQVIST. 2003. Occurrenceof mixotrophic flagellates in relation to bacterioplanktonproduction, light regime and availability of inorganicnutrients in unproductive lakes with differing humic contents.Freshw. Biol. 48: 868–877, doi:10.1046/j.1365-2427.2003.01061.x

CAMPBELL, J. W., AND T. AARUP. 1989. Photosyntheticallyavailable radiation at high latitudes. Limnol. Oceanogr. 34:1490–1499, doi:10.4319/lo.1989.34.8.1490

CARPENTER, S. R., J. J. COLE, J. F. KITCHELL, AND M. L. PACE.1998. Impact of dissolved organic carbon, phosphorus, andgrazing on phytoplankton biomass and production inexperimental lakes. Limnol. Oceanogr. 43: 73–80,doi:10.4319/lo.1998.43.1.0073

COLE, J. J., S. FINDLAY, ANDM. L. PACE. 1988. Bacterial productionin fresh and saltwater ecosystems—a cross system overview.Mar. Ecol. Prog. Ser. 43: 1–10, doi:10.3354/meps043001

CRAWLEY, M. J. 2002. Statistical computing: An introduction todata analysis using S-Plus. Wiley.

DEL GIORGIO, P. A., AND R. H. PETERS. 1994. Patterns in planktonic P–Rratios in lakes—influence of lake trophy and dissolved organic car-bon. Limnol. Oceanogr. 39: 772–787, doi:10.4319/lo.1994.39.4.0772

DOWNING, J. A., AND E. MCCAULEY. 1992. The nitrogen-phosphorus relationship in lakes. Limnol. Oceanogr. 37:

936–945, doi:10.4319/lo.1992.37.5.0936ELSER, J. J., AND oTHERS. 2007. Global analysis of nitrogen and

phosphorus limitation of primary producers in freshwater,marine and terrestrial ecosystems. Ecol. Lett. 10: 1135–1142,doi:10.1111/j.1461-0248.2007.01113.x

———, AND OTHERS. 2009. Shifts in lake N : P stoichiometry andnutrient limitation driven by atmospheric nitrogen deposition.Science 326: 835–837, doi:10.1126/science.1176199

FALKOWSKI, P., AND oTHERS. 2000. The global carbon cycle: A testof our knowledge of earth as a system. Science 290: 291–296,doi:10.1126/science.290.5490.291

FARJALLA, V. F., A. M. AMADO, A. L. SUHETT, AND F. MEIRELLES-PEREIRA. 2009. DOC removal paradigms in highly humicaquatic ecosystems. Environ. Sci. Pollut. Res. 16: 531–538,doi:10.1007/s11356-009-0165-x

FLANAGAN, K. M., E. MCCAULEY, F. WRONA, AND T. PROWSE.2003. Climate change: The potential for latitudinal effects onalgal biomass in aquatic ecosystems. Can. J. Fish. Aquat. Sci.60: 635–639, doi:10.1139/f03-062

FOUILLAND, E., AND MOSTAJIR, B. 2010. Revisited phytoplanktoniccarbon dependency of heterotrophic bacterial in freshwaters,transitional, coastal and oceanic waters. FEMS Microbiol.Ecol. 73: 419–429, doi:10.1111/j.1574-6941.2010.00896.x

HAKANSON, L., AND V. V. BOULION. 2001. Regularities in primaryproduction, Secchi depth and fish yield and a new system todefine trophic and humic state indices for lake ecosystems. Int.Rev. Hydrobiol. 86: 23–62, doi:10.1002/1522-2632(200101)86:1,23::AID-IROH23.3.0.CO;2-4

HOPPE, H. G., P. BREITHAUPT, K. WALTHER, R. KOPPE, S. BLECK,U. SOMMER, AND K. JURGENS. 2008. Climate warming inwinter affects the coupling between phytoplankton andbacteria during the spring bloom: A mesocosm study. Aquat.Microb. Ecol. 51: 105–115, doi:10.3354/ame01198

1712 Faithfull et al.

IPCC, 2007. Climate Change 2007: Synthesis Report. Contribu-tion of Working Groups I, II and III to the FourthAssessment Report of the Intergovernmental Panel onClimate Change [Core Writing Team, R. K. Pachauri andA. Reisinger (eds.)]. IPCC, Geneva, Switzerland.

JAGER, C. G., S. DIEHL, AND G. M. SCHMIDT. 2008. Influence ofwater-column depth and mixing on phytoplankton biomass,community composition, and nutrients. Limnol. Oceanogr.53: 2361–2373, doi:10.4319/lo.2008.53.6.2361

JANSSON, M. 1998. Nutrient limitation and bacteria–phytoplank-ton interactions in humic lakes, p. 177–194. In D. O. Hessenand L. Tranvik [eds.], Aquatic humic substances. Ecology andbiochemistry. Springer.

———, L. PERSSON, A. M. DE ROOS, R. I. JONES, AND L. J.TRANVIK. 2007. Terrestrial carbon and intraspecific size-variation shape lake ecosystems. Trends Ecol. Evol. 22:

316–322, doi:10.1016/j.tree.2007.02.015JONES, R. I. 1992. The influence of humic substances on lacustrine

planktonic food-chains. Hydrobiologia 229: 73–91, doi:10.1007/BF00006992

KALFF, J. 2003. Limnology: Inland water ecosystems. Prentice-Hall.KARLSSON, J., P. BYSTROM, J. ASK, P. ASK, L. PERSSON, AND M.

JANSSON. 2009. Light limitation of nutrient-poor lake ecosys-tems. Nature 460: 506–509, doi:10.1038/nature08179

———, M. JANSSON, AND A. JONSSON. 2002. Similar relationshipsbetween pelagic primary and bacterial production in clear-water and humic lakes. Ecology 83: 2902–2910, doi:10.1890/0012-9658(2002)083[2902:SRBPPA]2.0.CO;2

KIRCHMAN, D. L. 1994. The uptake of inorganic nutrients byheterotrophic bacteria.Microb. Ecol. 28: 255–271, doi:10.1007/BF00166816

———, X. A. G. MORAN, AND H. DUCKLOW. 2009. Microbialgrowth in the polar oceans—role of temperature and potentialimpact of climate change. Nat. Rev. Microbiol. 7: 451–459.

LAPERRIERE, J. D., AND J. A. EDMUNDSON. 2000. Limnology oftwo lake systems of Katmai national park and preserve,Alaska: Part II. Light penetration and Secchi depth. Hydro-biologia 418: 209–216, doi:10.1023/A:1003990600537

LAYBOURN-PARRY, J., T. HENSHAW, D. J. JONES, AND W. QUAYLE.2004. Bacterioplankton production in freshwater Antarcticlakes. Freshw. Biol. 49: 735–744, doi:10.1111/j.1365-2427.2004.01221.x

LEBAUER, D. S., AND K. K. TRESEDER. 2008. Nitrogen limitation ofnet primary productivity in terrestrial ecosystems is globallydistributed. Ecology 89: 371–379, doi:10.1890/06-2057.1

LIPSEY, M. W., AND D. B. WILSON. 2001. Practical meta-analysis.Sage.

LOMAS, M. W., P. M. GLIBERT, F. K. SHIAH, AND E. M. SMITH.2002. Microbial processes and temperature in ChesapeakeBay: Current relationships and potential impacts of regionalwarming. Global Change Biol. 8: 51–70, doi:10.1046/j.1365-2486.2002.00454.x

MUREN, U., J. BERGLUND, K. SAMUELSSON, AND A. ANDERSSON.2005. Potential effects of elevated sea-water temperature onpelagic food webs. Hydrobiologia 545: 153–166, doi:10.1007/s10750-005-2742-4

MURRAY, K., AND M. M. CONNER. 2009. Methods to quantifyvariable importance: Implications for the analysis of noisyecological data. Ecology 90: 348–355, doi:10.1890/07-1929.1

NURNBERG, G. K., AND M. SHAW. 1998. Productivity of clear andhumic lakes: Nutrients, phytoplankton, bacteria. Hydrobio-logia 382: 97–112, doi:10.1023/A:1003445406964

OCHS, C. A., J. J. COLE, AND G. E. LIKENS. 1995. Populationdynamics of bacterioplankton in an oligotrophic lake. J.Plankton Res. 17: 365–391, doi:10.1093/plankt/17.2.365

PETERSEN, J. E., J. C. CORNWELL, AND W. M. KEMP. 1999. Implicitscaling in the design of experimental aquatic ecosystems.Oikos 85: 3–18, doi:10.2307/3546786

R DEVELOPMENT CORE TEAM. 2009. R: A language and environ-ment for statistical computing. R Foundation for StatisticalComputing, Available from www.R-project.org

SCHINDLER, D. W. 1977. Evolution of phosphorus limitation inlakes. Science 195: 260–262, doi:10.1126/science.195.4275.260

———. 1998. Replication versus realism: The need for ecosystem-scale experiments. Ecosystems 1: 323–334, doi:10.1007/s100219900026

SCHLESINGER, W. H., AND J. A. ANDREWS. 2000. Soil respirationand the global carbon cycle. Biogeochemistry 48: 7–20,doi:10.1023/A:1006247623877

SCOTT, J. T., AND M. J. MCCARTHY. 2010. Nitrogen fixation maynot balance the nitrogen pool in lakes over timescales relevantto eutrophication management. Limnol. Oceanogr. 55: 1265–1270, doi:10.4319/lo.2010.55.3.1265

THEBAULT, J. M., J. CAPBLANCQ, AND M. PETIT. 1999. The influenceof morphometry and vertical mixing on the net phytoplank-ton primary production. C. R. Acad. Sci. Ser. III Sci. Vie/LifeSci. 322: 63–70.

TRANVIK, L. J. 1988. Availability of dissolved organic carbon forplanktonic bacteria in oligotrophic lakes of differing humiccontent. Microb. Ecol. 16: 311–322, doi:10.1007/BF02011702

VADEBONCOEUR, Y., E. JEPPESEN, M. J. VANDER ZANDEN, H. H.SCHIERUP, K. CHRISTOFFERSEN, AND D. M. LODGE. 2003. FromGreenland to green lakes: Cultural eutrophication and theloss of benthic pathways in lakes. Limnol. Oceanogr. 48:

1408–1418, doi:10.4319/lo.2003.48.4.1408VADSTEIN, O. 2000. Heterotrophic, planktonic bacteria and cycling

of phosphorus—phosphorus requirements, competitive abil-ity, and food web interactions, p. 115–167. In B. Schink [ed.],Advances in microbial ecology, vol. 16. Kluwer Academic/Plenum Publ.

VITOUSEK, P. M., AND R. W. HOWARTH. 1991. Nitrogen limitationon land and in the sea—how can it occur. Biogeochemistry 13:87–115, doi:10.1007/BF00002772

VREDE, K., T. VREDE, A. ISAKSSON, AND A. KARLSSON. 1999. Effectsof nutrients (phosphorous, nitrogen, and carbon) and zoo-plankton on bacterioplankton and phytoplankton—a seasonalstudy. Limnol. Oceanogr. 44: 1616–1624, doi:10.4319/lo.1999.44.7.1616

VREDE, T., A. BALLANTYNE, C. MILLE-LINDBLOM, G. ALGESTEN, C.GUDASZ, S. LINDAHL, AND A. K. BRUNBERG. 2009. Effects ofN : P loading ratios on phytoplankton community composition,primary production and N fixation in a eutrophic lake. Freshw.Biol. 54: 331–344, doi:10.1111/j.1365-2427.2008.02118.x

WALSH, C., AND R. MACNALLY. 2008. hier.part: Hierarchicalpartitioning. R package version 1.0-3. Available from http://cran.r-project.org/web/packages/hier.part/

WETZEL, R. G. 2001. Limnology, third edition: Lake and riverecosystems. Academic Press.

WHITE, P. A., J. KALFF, J. B. RASMUSSEN, AND J. M. GASOL. 1991.The effect of temperature and algal biomass on bacterialproduction and specific growth-rate in freshwater and marinehabitats. Microb. Ecol. 21: 99–118, doi:10.1007/BF02539147

Associate editor: Jonathan J. Cole

Received: 05 November 2010Accepted: 04 May 2011

Amended: 28 March 2011

Reviews in L&O 1713