Defining drivers of nitrogen stable isotopes (δ15N) of surface sediments in temperate lakes

ORIGINAL PAPER

Defining drivers of nitrogen stable isotopes (d15N) of surfacesediments in temperate lakes

Morgan Botrel • Irene Gregory-Eaves •

Roxane Maranger

Received: 21 January 2013 / Accepted: 20 September 2014 / Published online: 7 October 2014

� Springer Science+Business Media Dordrecht 2014

Abstract The nitrogen stable isotopic signature

(d15N) of sediment is a powerful tool to understand

eutrophication history, but its interpretation remains a

challenge. In a large-scale comparative approach, we

identified the most important drivers influencing

surface sediments d15N of 65 lakes from two regions

of Canada using proxies that reflect watershed nitro-

gen (N) sources, internal lake microbial cycling

and productivity. Across regions, we found that

water column total nitrogen (TN), %N in the

sediments and lake morphometric variables were the

best predictors of sedimentary d15N, explaining 66 %

of its variation. Significant relationships were also

found between sediment d15N and human-derived N

load (R2adj: = 0.23, p \ 0.001), the latter being a strong

predictor of TN (R2adj: = 0.68, p \ 0.001). Despite a

relatively strong overall relationship, variation parti-

tioning revealed an interesting difference in the

dominant variable that influenced regional d15N.

Alberta lake sedimentary d15N signature was domi-

nated by human derived N load. In contrast, internal

processing appeared to be more important in Quebec

lakes, where sediment d15N was best explained by

%N in the sediments and lake volume. Overall, our

findings support the use of d15N in paleolimnological

investigations to reconstruct changing N sources to

lakes but also highlight that regions may have

distinctive drivers. Interpretations of sediment d15N

are likely to be strongest when multiple lines of

evidence are employed and when placed in a regional

context.

Keywords Nitrogen stable isotopes �Nitrogen load �Lake sediments � Organic matter degradation �Diagenesis � Eutrophication

Introduction

Eutrophication is one of the greatest threats to

freshwater ecosystems and ultimately results in the

dominance of noxious algal blooms, decreases in

oxygen concentrations and losses in biodiversity

(Smith and Schindler 2009). Despite the widespread

consensus that phosphorus (P) loads need to be

controlled to mitigate eutrophication in lakes (Schin-

dler 2006), there is growing evidence that enhanced

nitrogen (N) inputs can further degrade water quality

M. Botrel (&) � R. Maranger

Departement des sciences biologiques, Universite de

Montreal, C.P. 6128 suc. Centre-ville, Montreal,

QC H3C 3J7, Canada

e-mail: [email protected]

M. Botrel � I. Gregory-Eaves � R. Maranger

Groupe de recherche interuniversitaire en limnologie et en

environnement aquatique (GRIL), Montreal, QC, Canada

I. Gregory-Eaves

Department of Biology, McGill University, Montreal, QC,

Canada

123

J Paleolimnol (2014) 52:419–433

DOI 10.1007/s10933-014-9802-6

through the proliferation of toxic cyanobacteria (Giani

et al. 2005; Monchamp et al. 2014). Paleolimnology

provides a unique long-term understanding of the

impact of N pollution on lake ecosystems. Sedimen-

tary nitrogen stable isotopes (d15N) of organic matter

have served as an informative metric of such impacts.

However, the d15N sediment signature is a complex

indicator with many possible interpretations : it can

serve to identify the dominant source of N to lakes, the

influence of internal microbial cycling or changes in

lake productivity (Fig. 1).

Most often the d15N of bulk sediment organic

matter is used to reconstruct past changes in N source.

Such inferences are based on the notion that each N

source has a distinctive d15N signature. For instance,

human-derived N typically has a high d15N composi-

tion. As such, human and animal wastes d15N signal

ranges between ?10 and ?20 % because of the high

degree of fractionation related to ammonia volatiliza-

tion in wastes (Heaton 1986; Kendall et al. 2007).

Synthetic fertilizer however produced via industrial

atmospheric N2 fixation has a low d15N (between -4

and ?4 %). Regardless of the fertilizer source

(manure or industrial fertilizer) agricultural soil nitrate

d15N varies between ?15 and ?30 % due to elevated

inputs and high fractionation associated with intense

soil denitrification. Meanwhile, atmospheric N has

lower d15N, in the range of -15 to 15 % for dissolved

inorganic N (DIN) in wet deposition (Kendall et al.

2007). Using these distinct signatures, several previ-

ously published paleolimnological studies have

attempted to track different N sources coming from

wasterwater (Leavitt et al. 2006), agricultural organic

or inorganic fertilizers (Bunting et al. 2007) and

atmospheric deposition (Jones et al. 2004; Holtgrieve

et al. 2011). Conceptually, the expected sediment d15N

Fig. 1 Conceptual

diagrams showing different

processes and their

hypothetical effects on

sediment d15N. Panels

express the expected

relationship when a

dominant factor regulates

sediment d15N: a N source

(human derived or

atmospheric deposition),

b internal microbial cycling

(expressed as proportion of

organic matter and

morphometry indices) and

c productivity (TN shown as

example). Suggested

relationship patterns are

based on literature

observations

420 J Paleolimnol (2014) 52:419–433

123

signature will depend on the dominant N source, with

N derived from agriculture or human waste increasing

the d15N signature while atmospheric inputs results in

a lower one (Fig. 1a).

Since lakes are intense biological reactors and act

as sinks for reactive N through sediment burial and

loss via denitrification (Harrison et al. 2009), the d15N

signature of the externally derived N sources can be

altered by N retention in two ways. First, bacteria can

indirectly change sediment d15N through the remin-

eralization of organic matter resulting in the release of

ammonium to the water column which can further be

nitrified and denitrified. This remineralized DIN

available to primary producers could have a different

d15N compared to the original external source signa-

ture. Evidence of this indirect effect on sediment d15N

has been previously observed (Hodell and Schelske

1998; Hadas et al. 2009). Second, bacteria can directly

alter the sediment d15N through the degradation of

organic matter. The d15N changes can be caused by

bacterial consumption and the fractionation of the

organic N pool, via the preferential degradation of

certain N forms or through the increase of bacterial

biomass (Lehmann et al. 2002; Kohzu et al. 2011).

Yet, there is conflicting evidence on the effect of

degradation on sedimentary d15N where some studies

have found it resulted in an increase in the isotopic

signature (Ostrom et al. 1998; Freudenthal et al. 2001),

whereas others have found that it resulted in a decrease

(Galman et al. 2009; Kohzu et al. 2011). Another

potential proxy for the influence of internal N

processing by microbes on sediment d15N is lake

morphometry (Fig. 1b). Indeed, lake N retention is

strongly linked to primary morphometric features such

as water residence time and lake shape (Finlay et al.

2013; Harrison et al. 2009), but no studies to our

knowledge have considered how morphometry influ-

ences sediment d15N.

The organic matter, and therefore, the sediment

d15N also depend on the assimilation of nutrients by

primary producers. From this perspective, there would

be an expected positive unimodal relationship

between d15N and trophic state explained by fraction-

ation during N assimilation and N fixation (Fig. 1c)

(Gu 2009). During assimilation, phytoplankton tends

to fractionate N and preferentially incorporate the 14N

isotope. As N becomes more limiting with increasing

trophic state, phytoplankton would have a lower

fractionation effect and thus incorporate more of the

heavier isotope 15N resulting in higher d15N signature

(Wada and Hattori 1978; Fogel and Cifuentes 1993).

In hypereutrophic lakes, low N:P ratios could lead to

atmospheric N fixation (-1 to -2 %) as the main N

source, resulting in a marked decrease in planktonic

d15N (Gu et al. 1996; Brenner et al. 1999). Several

paleolimnological studies have used sedimentary d15N

as an indicator of past productivity (Brenner et al.

1999; Wu et al. 2006) and N fixation (Riedinger-

Whitmore et al. 2005; Schindler et al. 2008).

To assess what are the most important drivers of

sediment d15N, we used a comparative approach to

quantify the relative influences of three broad factors:

external N sources, internal microbial cycling and

productivity. Specifically, we examined the relation-

ships between surface sediment d15N of 65 lakes and

their respective watershed N inputs, land use, mor-

phometric indices, sediment organic matter content

and productivity metrics as proxies for the three

factors of classification mentionned. We address the

question in temperate lakes from two geographically

distinct regions impacted by human activities.

Study sites

To capture a broad range in land use, lake shape and

trophic state, data were collected from two contrasting

regions of Canada in 40 lakes from southern Quebec

and 25 lakes across Alberta (Fig. 2). In our subset of

Southern Quebec, the majority of lakes lie in the

Appalachian Mountains with some located in the Saint

Lawrence Lowlands, all of them being in a region

dominated by mixed forest where agriculture is

present at different intensities. Our Quebec lake have

circumneutral pH, span large P and N gradients and

are predominantly shallow (Vermaire et al. 2011). Our

sampled Alberta lakes are located in the Boreal Plain

and the Aspen Parkland, a transition zone between the

prairies and boreal forest biomes. The exceptions to

this are Elkwater, Reesor and Spruce Coulee reservoir,

which are located on a plateau rising 200 m above the

Canadian prairies and fall within the Cypress Hill

Provincial Park. Alberta sites are generally low

elevation lakes, with large watersheds and are natu-

rally productive as their watershed soils are rich in P

(Prepas and Trew 1983). Agriculture is well developed

in the parkland region thus contributing to the cultural

eutrophication of lakes (Taranu and Gregory-Eaves

2008).

J Paleolimnol (2014) 52:419–433 421

123

Materials and methods

Collection of sediment samples and isotopic

analysis

To measure N stable isotopes, surface sediments (top

0–1 cm) were retrieved from the deep-water deposi-

tional zone of the lakes (Glew et al. 2001). Sediments

from Quebec were collected in summers 2007 and

2008 and sediments from Alberta in summer 2006.

Based on 210Pb age models, the upper 1 cm of Quebec

lakes represents * 2 years of sedimentation (Verm-

aire 2011) and for Alberta lakes, between * 2 years

to a maximum of 5 years (Koster et al. 2008). N

isotopic composition was measured on freeze-dried

homogenized bulk sediment using a continuous flow

Micromass IsoprimeTM IRMS coupled to a Carlo Erba

NC 1500 elemental analyzer located at the GEOTOP

Research Center (Universite du Quebec a Montreal).

For most samples, the analytical precision was ±

0.2 %, but for samples with low N (\ 1 %N) the

analytical precision was ± 0.4 %. Isotopic values are

reported as the d notation (d = ([Rsample-Rstandard]/

Rstandard) 9 1000, where R = 15N/14N). Values for

d15N are referenced against atmospheric N2 gas. d15N

duplicates were run on 10 % of samples; coefficient of

variation was less than 10 %.

Watershed and land use data acquisition

To determine watershed characteristics and quantify

the coverage of land use, we used geographic infor-

mation system (GIS) techniques. The watersheds were

first delineated using digital elevation models (DEM)

Fig. 2 Map of the study sites and the major river systems from Southern Quebec and Alberta provinces of Canada. Region location

presented in inset

422 J Paleolimnol (2014) 52:419–433

123

at a scale of 1:50,000 (Canadian Digital Elevation

Data Level 1 available at www.geobase.ca) loaded in

ESRI ArcGIS 9.3 and processed through the spatial

analyst hydrology tools. To obtain the final watershed

overlays, we substracted the lakes’ outlines taken from

digitized bathymetric maps when available or from the

Canadian National Topographic Database (NTDB;

www.geogratis.ca) to the watershed layers computed

with the DEM. We then calculated the lake and

watershed areas using the Canadian Albers equal-area

conic projection to minimize distortion. The water-

shed maximum and mean slope statistics were esti-

mated using the spatial analyst surface tool on DEM

data and the drainage ratio was calculated as the

watershed area divided by the lake area (WA:LA). The

proportions of land use type cover was finally com-

puted by clipping the watershed overlay on the land

use layers. For Quebec lakes, the vegetation cover and

number of buildings were taken from the spatial

database CanVec (CanVec; www.geogratis.ca) and

agricultural cover was acquired from the Base de

donnee sur les cultures generalisees (BDCG:: www.

fadq.qc.ca/). For Alberta lakes, land coverage of these

same land use types was made available by Alberta

Environment (Z. E. Taranu pers. comm.) and number

of buildings extracted from CanVec.

Estimation of watershed N load

GIS techniques were also used to calculate the

watershed N load coming from human and from

atmospheric deposition. Our approach to calculate

watershed human derived N load was to combine

census data on fertilizer, livestock and human popu-

lation that were applied to each watershed and convert

them in kg of N. County level data on fertilizer

expense and farm animal numbers were acquired from

the 2006 Canadian census of agriculture (Statistics

Canada 2006a,b) and the number of inhabitants from

the 2006 census of population (Statistics Canada:

http://www12.statcan.ca/). Fertilizer expenditure in

dollars were converted to kg of N using provincial

total amount of fertilizer and nutrient sold (Canadian

Fertilizer Institute: http://www.cfi.ca/elibrary/). Live-

stock N excretion reported in Culley and Barnett

(1984) were used to estimate the total N from manure

since they best represent agricultural practices in

Canada. N produced by the human population was

estimated to be 5.6 kg per individual per year (Castro

et al. 2003), representative of a typical North Ameri-

can diet and the use of septic tanks.

To obtain spatial data describing N from fertilizer

and animal wastes at the watershed scale, county level

data were scaled by weighting each county by the

agricultural land area included within the watershed

boundaries to the total area of agricultural land of each

county. For this calculation we assumed that the

fertilizer sold in one county is applied within that same

county. Similarly, population data were scaled to

watershed by weighing with number of buildings.

Atmospheric N deposition was estimated by multi-

plying lake and watershed area by an average N

deposition rate of 8 kg N ha-1 year-1 for Quebec and

4 kg N ha-1 year-1 for Alberta (Schindler et al. 2006;

Krzyzanowski 2010). The human-derived and atmo-

spheric estimated N loads are all reported as

kg N year-1.

Sediment geochemical analyses

To test wether degradation of the organic matter can

alter sediment d15N, we measured elemental propor-

tion of sedimentary N and carbon (C) on the same

samples as for isotope analysis with a Fisons EA1108

Elemental Analyzer according to standard techniques

(Meyers and Ishiwatari 1993). All %C measurements

reported represent organic carbon content since car-

bonates were removed prior to analyses by fumigation

with 1 M HCl (Helie 2009). This method was suitable

for our samples as none of the sites were located in

regions dominated with dolomite which resist the

fumigation treatment (J.-F Helie pers. comm.). Ana-

lytical precision for organic C and for N was ± 0.02

and ± 0.05 % respectively and replicate were run on

10 % of samples with a coefficient of variation less

than 10 %.

Morphometry data acquisition

Lake morphometry was characterized using GIS

methods to further test possible effect of N retention

on sediment d15N. Using raster overlay of the lake

bathymetry, lake volume was first calculated as well as

shoreline length. For Quebec lakes, bathymetric maps

(provided by Service des bibliotheques et archives,

Universite de Sherbrooke) were digitized in point

J Paleolimnol (2014) 52:419–433 423

123

vector layer that were rasterized to 1-m cell size using

the nearest neighbour algorithm. For most Alberta

lakes, rasterized bathymetric map were obtained from

the Alberta Geological Survey (AGS; www.ags.gov.

ab.ca) except for four lakes where bathymetric maps

were obtained from Alberta Lake Management Soci-

ety (ALMS; http://www.alms.ca) and digitized. Lake

volume was calculated following the formula

V(x1,x2) = ((Ax1 ? Ax2)/2) 9 (X2 - X1) where A is

area and X is depth interval (Kalff 2002); volume was

summed for all depth increments. For six lakes,

bathymetric maps were not available and regression

models between lake area and volume was applied to

account for missing data (R2adj: [ 0.77, Botrel 2011).

All other morphometric indices were derived from

these measurements. Mean depth was calculated as the

ratio of volume to lake area (V/LA), the index of basin

permanence (IBP; Kerekes 1977) was computed as the

Table 1 Summary table of the N sources, internal cycling and productivity variables used in this study

#

* indicates significant regional differences at p \ 0.05, 4 are the variables kept for the MLR, (poly) indicates the variable kept for

MLR was the second order polynome. % Agr represents proportion of agriculture area on watershed, % Forest proportion of forest

area on watershed, Building # number of buildings on watershed, % Water proportion of water area on watershed, WA watershed

area, WA:LA watershed area to lake area ratio, Human N watershed human derived N load, Atm. dep. atmospheric deposition on the

watershed, %N sed proportion of sediment N, %C sed proportion of sediment C C:N sed elemental sediment C:N ratio, LA lake

area, SL shoreline length, Zmax maximum lake depth, Zmean mean lake depth, IBP index of basin permanence, V:WA lake volume to

watershed area ratio, TN total nitrogen concentration, TP total phosphorus concentration, DP dissolved phosphorus concentration,

DOC dissolved organic carbon concentration, chla chlorophyll a, N:P ratio of lake total N to total P

424 J Paleolimnol (2014) 52:419–433

123

ratio of volume to shore line length (V/SL) and V:WA

refers to the ratio of volume to watershed area.

Retrieval of water chemistry data

To test our hypotheses of the influence of productivity

on sediment d15N, we acquired lake water chemistry

data (Table 1). For all Quebec lakes, physicochemical

variables were primarily measured at a single time

point in August when the sediment samples were

originally collected (Vermaire et al. 2011). For some

of these lakes, growing season means were also

available through a monitoring program (three to four

visits from May to November, Prairie and Parkes

2007). For Alberta, all lake chemistry data represent-

ing average growing season values (one to five

visit from May to September) was acquired from

publicly available monitoring programs (Lake Water

Quality Data: http://environment.alberta.ca/; Provin-

cial Parks Lake Monitoring Program: http://www.

environnement.gov.ab.ca/info/; Alberta Lake Man-

agement Society: http://www.alms.ca). In order to

remain consistent among regions, we ran simple least

square regression models (R2adj: [ 0.60, Botrel 2011)

between Quebec single point measures from August

and the growing season averages to correct for the

potential differences. In a few instances, a regression

model approach was also applied, (R2adj: [ 0.77, Bo-

trel 2011) to account for missing data in the Quebec

and Alberta datasets.

Statistical analyses

To first assess the overall difference between the two

regions, a multivariate ANOVA with a redundancy

analysis (RDA) was conducted on the isotopes and

environmental data (the explanatory variables) and a

binary factor coding for the regions (the response

variable) using the ‘‘vegan’’ package in R (Legendre

and Anderson 1999). To determine if individual

variables were significantly different between regions,

multiple T-tests with a Holm correction for multiple

comparisons were done. To further described the

environmental data structure we used principle com-

ponent analysis (PCA).

To test whether N sources, organic matter, mor-

phometry or productivity are significant predictors of

sediment d15N and explain maximum variation, a least

square multiple linear regression (MLR) approach was

used. Because of high colinearity in the dataset,

variables were first screened to remove those that were

highly correlated with each other. Simple linear

regressions were computed between d15N and each

explanatory variable. Since one of our working

hypotheses included a non-linear response model,

polynomial regressions were also computed between

d15N and each variable. Only significant polynomial

terms were included as new variables into our models

(volume2 and IBP2). Variables with the highest

adjusted R2 (R2adj:) were first selected and the variable

related to the selection with an r [ 0.6 were consid-

ered collinear and eliminated. We then ran forward

selection on the pre-selected variables (‘‘packfor’’

package in R) and the MLR (‘‘ape’’ package in R).

On the explanatory variables selected by MLR, we

ran analysis of covariance (ANCOVA) to compare

relationships (slopes and intercepts) between the two

regions. Finally, to determine what is the most impor-

tant predictor of d15N when removing the effect of other

predictors, variation partitioning was computed on

these same explanatory variable of d15N (Borcard et al.

1992). Prior to all analyses, variables were transformed

(square-root or natural logarithm) to satisfy the

assumption of normality and a multivariate Levene’s

test was conducted to satisfy homoscedasticity. Anal-

yses were performed in R statistical software.

Results

Interregional difference in response

and explanatory variables

Several environmental variables in the three broad

classification factors of interest and the sedimentary

d15N signature were significantly different between

regions (Table 1; MANOVA p \ 0.05). Alberta lakes

exibited significantly higher d15N sedimentary signa-

ture than Quebec lakes (mean d15N = 4.1 and 2.5 %respectively, p \ 0.001). Most of our N source

variables were also higher for Alberta. For example,

Alberta lakes had greater human derived N load,

watershed area and atmospheric deposition but Que-

bec lakes were characterised by steeper slopes. These

differences in N load were not reflected in land-use

coverage which was similar between regions. Strong

J Paleolimnol (2014) 52:419–433 425

123

differences were also observed in some sediment

variables where Alberta had higher %N, although

%C and C:N was similar to Quebec lakes. Several

morphometric variables also differed significantly

where Quebec lakes generally had smaller surface

areas and volumes and were shallower. Despite the

differences in basin and watershed morphometries, the

drainage ratio and V:WA were similar between the

two regions. The majority of the productivity variables

(TN, TP, pH, DP, chla) that reflect lake trophic status

were higher for Alberta, but secchi depth and N:P

ratios were similar across regions.

Using a PCA analysis we observed strong colin-

earity among multiple variables reflective of different

broad factors of classification: sources, internal

cycling and productivity (Fig. 3). The first PCA axis

accounted for 37 % of the variance and illustrated that

lakes were located along a trophic gradient with

Alberta lakes being more nutrient rich compared to

Quebec lakes. The small angles between the produc-

tivity variables on the PCA (TN, DP, TP, DOC, chla,

pH, secchi depth) was in agreement with correlation

analysis revealing that these variables were highly and

significantly correlated with each other and with the N

source variables (p [ 0.6, p \ 0.01). The second PCA

axis accounted for 15 % of the variance and showed

that lakes were on a size gradient with Alberta lakes

being in general bigger and deeper. Again, we found

-2 -1 0 1

-2

-1

0

1

2

PCA axis 1 (37%)

PCA

axis

2(1

5%)

WAMax slope

Volume

ZmaxZmean

SLLA

WA:LA

IBP

TN

TPDP

DOC

Secchi

%C sedHuman N

Atm.dep.

chla

Build.%N sed

% Agr% Water

pHC:Nsed Mean slope

%Forest

N:P

IBP2

V:WA

QuebecAlberta

Productivity variablesInternal cycling variablesN source variables

Fig. 3 Principal component analysis (PCA) of the environ-

mental data, scaling type 2. First axis explains 37 % of the

variation and second axis 15 %. Environmental variables:

angles between arrows are proportional to the correlations

between variables. Arrow styles represent our three broad

variable classifications : gray represents N source, dashed

internal cycling and black productivity. Shorter arrows are

variables that contribute less to the formation of the reduce

space; angles in that case are less representative. Points

represent sites from both regions and where they are located

along the gradients

426 J Paleolimnol (2014) 52:419–433

123

strong correlation between the morphometric vari-

ables (r [ 0.6, p \ 0.01) representative of internal

cycling. Given observed high colinearity, only ten

environmental variables were kept for MLR (Table 1).

Predictors of sedimentary d15N variation

Simple regression analyses across the two regions

showed that nutrient concentrations were the strongest

explanatory variables of d15N. Total nitrogen (TN)

and dissolved phosphorus (DP) concentrations were

the strongest predictors (respectively R2adj: = 0.32 and

0.29, p \ 0.01, n = 65;). Morphometric variables also

explained a significant fraction of the variation

whereby volume and IBP were related in a unimodal

fashion to d15N (respectively R2adj: = 0.28 and 0.25,

p \ 0.01). Finally, sediment N content (%N) and

watershed human derived N load were also found to be

significant predictors of d15N (respectively R2adj: =

0.15 and 0.23, p \ 0.01). Overall our MLR model

showed that 66 % of the variation of the sedimentary

d15N signature was explained by TN, %N in the

sediment and morphometry (Table 2). Most of the

variation was explained by TN and %N in the sediment.

IBP2 was only marginally significant and thus elimi-

nated from further analyses.

Numerous regional differences in the relationship

between the selected environment variables and the

sediment d15N were detected using ANCOVA. For

example, both the slope and intercept for the models

describing the relationship between sedimentary d15N

and TN were significantly different (interaction of

region and d15N: F1,61 = 8.6, p = 0.005; Fig. 4a).

Part of this observed difference could be explained by

the smaller range of TN concentrations in Quebec

lakes. Nonetheless, TN was significantly correlated

with human derived N load for both regions (Fig. 4b).

The relationship between human derived N load and

TN had similar slopes (F1,61 = 0.6, p = 0.43) but

significantly different intercepts (F1,62 = 19.5,

p \ 0.01), where Alberta lakes had 2.5 mg L-1 higher

TN concentrations on average for a similar human

derived N load.

A regional difference was also observed when

considering internal lake cycling variables. In the

sedimentary d15N– %N relationship (R2adj: = 0.47,

p \ 0.001, Fig. 4c) we found similar slopes (F1,61 =

2.6, p = 0.11) but significantly different intercepts

between regions, with the d15N signature being

* 2.6 % higher on average in Alberta lakes for a

similar %N sediment content (F1,62 = 40.5, p \0.01). When considering lake volume, both slopes

and intercepts were regionally different (interaction of

region and d15N: F1,61 = 3.3, p = 0.07) and the

relationship is best modelled by seperate regressions

(Fig. 4d). In this case, lake volume was a significant

predictor of the d15N signature for Quebec lakes only.

Variation partitioning

In order to characterize the relative importance of the

broad factors of classification to the overall d15N of the

sediment, variation partitioning was computed inde-

pendently for the two regions using lake volume, TN

and %N in the sediments as predictor variables

(Fig. 5). In Quebec lakes, we found that %N in the

sediments explained most variation (34 %, p = 0.05),

with an additional 20 % shared with lake volume

(p = 0.05, Fig. 5a). A modest fraction of the variation

in d15N was explained by TN, but this variation

overlapped completely with lake volume (2 %) and

%N in the sediments (1 %). A different pattern was

observed for Alberta lakes where TN was the most

important predictor explaining uniquely 48 %

(p = 0.01) of the d15N variation (Fig. 5b). %N in

the sediments explained an additional 19 % (p =0.05)

of the variation in d15N from Alberta lakes, whereas

lake volume only explained 3 % (Fig. 5b).

Discussion

Defining the controls of sedimentary d15N variation is

of primary importance for its application to paleolim-

nological records and identifying causes of eutrophi-

cation. Our analysis of surface sediments from 65

temperate lakes demonstrated that over a large spatial

Table 2 Least squares MLR results with forward selection

using permutation test to model sediment d15N

Significant

explanatory variables

Standardized

coefficient

p value

(two-tail)R2

adj:

TN (ln) 0.52 0.001 0.66

%N sediment (H) -0.59 0.001

Volume (ln) 0.27 0.008

IBP2 (ln) -0.20 0.011

J Paleolimnol (2014) 52:419–433 427

123

scale, lake TN concentration is a strong predictor

variable of sedimentary d15N. Given that TN was

highly correlated to human derived N load, this

suggests the sediment isotopic signature of our

systems was most likely regulated by N sources.

These findings are in agreement with spatial (Selbie

et al. 2009; Bunting et al. 2010) and temporal studies

(Elliott and Brush 2006, Bunting et al. 2007) of

sedimentary d15N control. Interestingly, the %N in

sediment explained a similar fraction of the variance

in sedimentary d15N, whereby lakes with lower

sediment N content tend to have higher d15N signa-

tures. This result indicates internal microbial

transformations are important in shaping sediment

d15N and is consistent with studies that observed a 15N

enrichment of the residual organic material as a

function of degradation in oxygenated waters (Altabet

1988; Freudenthal et al. 2001). Volume also entered

the model as a significant variable highlighting the

importance of N retention (Harrison et al. 2009).

Despite a wide range in productivity in our study

lakes, we failed to support the idea that d15N serves as

an indicator of increasing productivity and fixation by

cyanobacteria (Gu et al. 1996; Gu 2009; Fig. 1c). Our

analysis also showed strong regional differences in the

dominant variables influencing sedimentary d15N,

-2

0

2

4

6

8

ln(TN)

δ15N

(o/ oo

)

QC: δ15N = 1.1 ln(TN) + 3.5 R2adj. = 0.02, p = 0.17

AB: δ15N = 1.9 ln(TN) + 3.7R2adj. = 0.40, p < 0.001

-1

0

1

2

ln(human derived N)

ln(T

N)

QC: ln(TN) = 0.1 ln(hum. N) - 2.1 AB: ln(TN) = 0.1 ln(hum. N) -1.2

R2adj. = 0.68, p < 0.0001

(B) TN and human derived N

Quebec

Alberta

SitesLinear regression

Linear regression

-2

0

2

4

6

8

sqrt(%N sediments)

δ15N

(o / oo)

QC: δ15N = -3.4 sqrt(%N sed) + 6.2AB: δ15N = -3.4 sqrt(%Nsed) + 8.8

R2adj. = 0.47, p < 0.0001

-1 0 1 2 6 8 10 12 14

0. 5 1.0 1.5 12 14 16 18 20

-2

0

2

4

6

8

ln(lake volume)

δ15N

(o / oo)

QC: δ15N = 0.6 ln(volume) - 5.8R2adj. = 0.24, p < 0.001

AB: δ15N = 0.3 ln(volume) - 0.2R2adj. = 0.00, p = 0.63

(A) δ15N and total nitrogen

(C) δ15N and %N sediments (D) δ15N and lake volume

Sites

Fig. 4 Analysis of covariance on the variables selected by the

forward selection to compare the linear relationships between

regions: a relationship between d15N and TN, with significant

regional differences in slopes and intercepts; b relationship

between TN and watershed human derived N load, showing

similar slopes but different intercepts between regions; c rela-

tionship between d15N and %N in the sediments, having similar

slopes but different intercepts between regions; d relationship

between d15N and lake volume, differing regionally in both

slopes and intercepts

428 J Paleolimnol (2014) 52:419–433

123

where proxies of internal cycling dominated in

Quebec and external sources in Alberta.

The role of N sources on sedimentary d15N

Studies on d15N distribution in ecosystems have long

recognized the use of d15N natural abundance as a

tracer of different anthropogenic N sources (Heaton

1986). Our analysis provides broad support for the

presence of a positive relationship between TN and

sediment d15N. TN was itself highly correlated to

human derived N load, suggesting TN is reflective of

the watershed N sources and sediment d15N is

dependent upon the magnitude of N inputs (Fig. 4a,b).

Paleolimnological reconstructions often report

increasing trends in d15N coincident with human

population growth or more intense agricultural

practice since post-industrialisation (Bunting et al.

2007, Engstrom et al. 2006). Indeed this pattern is

what we observed across regions, but was driven

primarily by the strong significant relationship in

Alberta lakes.

Regionally, the average TN concentrations and

watershed human derived N load were higher in the

Alberta lakes relative to Quebec, which may explain

part of the observed higher average d15N sediment

signatures in Alberta. However, these regions also

differed in the dominant anthropogenic N source type,

which can alter the average sediment d15N signature at

the landscape scale (Kendall et al. 2007). Alberta lakes

experience more substantial agricultural N loads, 6.8

times more on average, whereas atmospheric N depo-

sition is relatively more important in Quebec (Botrel

2011). The human watershed N signal may be lost in

Quebec lakes due to the effect of mixing N from

agricultural and human wastes (high d15N) with equally

important inputs of N from atmospheric deposition with

low d15N (Table 1). The dominance of certain crop

types such as N fixing legumes which are more

prevalent in Quebec would further contribute to lower

the d15N signature (BDCG:: www.fadq.qc.ca/).

Potential effect of degradation on sedimentary d15N

Our study revealed a significant relationship between

N content in sediments and the d15N signature across

both regions where lake sediments with low %N

tended to have higher d15N. Bunting et al. (2010)

found the same relationship between sediment N and

d15N in surface sediments of 75 lakes from the Rocky

mountains in Alberta (R2 = 0.45). The consistency of

this relationship in various lake systems across Canada

suggests that it is a common feature of temperate

lakes, although very few studies report it. In Quebec

lakes, %N in the sediments was the strongest

predictor variable of the sediment d15N isotopic

signature.

The spatially consistent relationship between sed-

iment %N and d15N could be explained by nitrogen

consumption via bacterial degradation during sedi-

mentation. High amounts of primary organic matter

consumed by bacteria in the water column will result

in a low proportion of total reaching the sediment

surface (Meyers 1997). Increases in the d15N of the

residual organic matter in sediment has been observed

in marine studies when particles settle through an

0.34

0.03

0.02

0.200.01

Residuals = 0.40

0.48

0.19

0.010.02

Residuals = 0.41

TN

%N sediments

Volume

Volume

TN

%N sediments

(B) Total sedimentary δ15N variation, Alberta

(A) Total sedimentary δ15N variation, Quebec

Fig. 5 Variation partitioning of the explanatory variables of

sedimentary d15N signature: a for Quebec, b for Alberta. Sum of

b does not total 1 because the best model of d15N includes both

TN and %N in the sediment

J Paleolimnol (2014) 52:419–433 429

123

oxygentated water column (Altabet 1988; Altabet and

Francois 1994; Freudenthal et al. 2001). Laboratory

experiments exposing particulate organic matter

(POM) to oxygenated conditions confirm that rapid

degradation results in an increase of d15N (Lehmann

et al. 2002). The causal mechanism would be the

preferential release of 14N dissolved N from bacterial

metabolism.

The observed negative relationship between lake

volume and %N of sediments in Quebec lakes

suggests that residence time is also an important

factor explaining d15N shift during water column

particle degradation. Shallow, smaller lakes have

shorter water residence times and particles would be

subject to less processing (Kalff 2002). The oscillation

between oxic and anoxic conditions in the water

columns of large and deep lakes can further increase

the average d15N signature of the POM as observed in

Lake Kinneret (Hadas et al. 2009). We found some

support for this effect in our systems when considered

regionally, where the effect of increasing volume in

Quebec lakes resulted in overall higher signatures

(Fig. 4d).

Another possible factor explaining the relationship

between sediment d15N and %N in the sediment is the

composition of the organic matter itself. If the

relationship was driven by the organic matter compo-

sition, one would expect a negative relationship

between sediment C:N and d15N (Selbie et al. 2009;

Woodward et al. 2012). However, this hypothesis is

not suported by our data. The absence of a relationship

may be due to the low atomic sediment C:N ratio of

our sites (max = 18), being more representative of

aquatic organic matter relative to the higher C:N

values reported for vascular plants (Meyers and

Ishiwatari 1993). Indeed, previous studies have shown

an increase in the variability of the relationship

between sediment C:N and d15N, when C:N is less

than 20 (Woodward et al. 2012).

Limitations of the use of d15N as an indicator

of productivity

Our analysis failed to support the hypothesis that

sediment d15N was driven by changes in phytoplank-

ton N assimilation and fixation along a trophic

gradient as outlined in Fig. 1c. In addition to the lack

of consistency between this response model and our

isotopic data, analyses of water chemistry support this

conclusion. For example, we detected significant

negative relationships between N:P and TP in both

regions, indicative of a declining contribution of N to

the total nutrient pool when fertilizer, sewage and

animal manure, as well as internal loading are the

predominant nutrient sources (Downing and McCau-

ley 1992). These data suggest that N limitation is

possible (Pick and Lean 1987), and yet we failed to

detect a significant relationship between sediment

d15N and the lake water N:P ratio, which would be

expected if N-fixation was a dominant process oper-

ating in lakes with low N:P. This result was surprising

for our Alberta lakes where phytoplankton communi-

ties were dominated by nitrogen-fixing taxa such as

Anabaena and Aphanizomenon, (Taranu et al. 2012).

A change in sediment isotopic signature as a function

of fixation has been observed in highly hyper-eutro-

phic subtropical lakes (Gu et al. 1996). Unfortunately,

we do not have any estimates of fixation for these

Alberta lakes but Lewis and Wurtsbaugh (2008) make

a convincing case that the N arising from fixation,

even in some eutrophic lakes, is relatively small. As

such, we suggest that the elevated signature of the

anthropogenic N load diluted the effect of low d15N

coming from fixation and/or fixation was a minor N

input in these Alberta lakes. However, it would be

interesting in the future to mechanistically model the

inputs required by N fixation that could influence

sediment d15N signature for lakes with a given N

watershed load.

The importance of N sources in our study also

suggests that fractionation by phytoplankton assimi-

lation is not a dominant process controlling sediment

d15N in our study lakes. The fractionation effect of

phytoplankton assimilation is typically observed by a

negative relationship between sediment d15N and

water column NO3- concentration. Previous studies

have observed this relationship in oceans (Altabet and

Francois 1994) and oligotrophic lakes with high

residence times (Owen et al. 1999; Teranes and

Bernasconi 2000). However, in a detailed study of a

subset of our Quebec lakes spanning a broad trophic

gradient, no such effect was observed as nitrification

was apparent in the dual isotopic NO3- signal

throughout the summer, despite low overall concen-

tration in the upper water column (Botrel 2011).

Interestingly, the strongest assimilation fractionation

effect observed in lake systems to date was reported in

Lake Lugano, where summer and fall sediment strata

430 J Paleolimnol (2014) 52:419–433

123

could be analyzed and interpreted seperately (Teranes

and Bernasconi 2000). However, the seasonally var-

iable assimilation effect (Hodell and Schelske 1998)

would be diluted in lakes where sediments are not

laminated and thus summer and winter sediment

deposition are integrated upon sediment core extru-

sion. Although DIN depletion over the course of

summer is likely an overall feature of our Quebec

lakes, we cannot say conclusively that some level of

fractionation did not occur as a function of incomplete

uptake in Alberta lakes.

Recommendations for paleolimnological

reconstruction

Paleolimnological reconstructions are powerful tools

to understand eutrophication histories in lake systems.

Several studies have clearly demonstrated recent shifts

in sediment d15N signatures due to changes in N

sources (Bunting et al. 2007) or changes in fraction-

ation during phytoplankton assimilation including a

shift in biomass of N fixers (Gu et al. 1996). Such

studies support a dominant driver influencing the

observed change is signature. The heterogeneity of

lakes captured in this comparative study present the

full spectrum of possible drivers of d15N sediment and

how these may vary across regions. Our study

highlights that since sediments represent the integra-

tion of the functional, spatial and temporal variation of

lake organic matter, paleolimnological reconstruction

looking at simply one driver should also consider

alternate hypotheses. Furthermore, sedimentary d15N

should be considered within a regional context, where

knowledge of the isotopic signature of the major N

sources is critical.

Our spatially explicit analysis also suggested that

degradation is a possible determinant of d15N from

surface sediment, but we cannot claim that this trend

would be transposed through time. Studies comparing

multiple cores taken at different years showed that

diagenesis tended to lower sediment d15N with aging

(Galman et al. 2009; Kohzu et al. 2011). However,

Kohzu et al. (2011) also showed the importance of the

oxidation–reduction boundary in determining the

direction of the d15N diagenesis patterns. Our study

points out that there should be careful interpretation of

upper sediment signatures that are subjected to

degradation in oxygenated waters. We recommend

the examination of the %N profile to determine the

depth at which %N is more constant, indicative of

reduced bacterial consumption. Given the multiple

factors that may influence it, sedimentary d15N should

thus be used as part of a multiproxy approach, in

conjunction with proxies of algal community compo-

sition such as fossil pigments and of the organic matter

sources such as sediment C:N ratio to assess changes

in N sources and productivity over time.

Acknowledgments We would like to thank members of the

Maranger and the Gregory-Eaves labs for technical help during

this study, in particular K. Simpson, Z. Taranu and J. Vermaire.

Thank you to J. Cardille for help with GIS, G. McDonald for

guidance on human derived N loading and to J.-F. Helie and A.

Adamowicz at the Centre de Recherche en Geochimie et en

Geodynamique (GEOTOP) for isotopic analysis and method

advice. Comments that improved the manuscript were provided

by M.A. Altabet, E. Galbraith and anonymous reviewers. This

project was supported by the FQRNT-funded strategic group

Groupe de recherche interuniversitaire en limnologie et en

environnement aquatique (GRIL) and Natural Sciences and

Engineering Research Council of Canada (NSERC) grants to

RM and IGE.

References

Altabet MA (1988) Variations in nitrogen isotopic composition

between sinking and suspended particles—implications for

nitrogen cycling and particle transformation in the open

ocean. Deep-Sea Res 35:535–554

Altabet MA, Francois R (1994) Sedimentary nitrogen isotopic

ratio as a recorder for surface ocean nitrate utilization.

Global Biogeochem Cy 8:103–116

Borcard D, Legendre P, Drapeau P (1992) Partialling out the

spatial component of ecological variation. Ecology 73:

1045–1055

Botrel M (2011) Caracterisation du cycle et des sources d’azote

dans les lacs temperes par l’utilisation d’isotopes stables.

M.Sc. thesis, Univ. de Montreal

Brenner M, Whitmore TJ, Curtis JH, Hodell DA, Schelske CL

(1999) Stable isotope (delta C-13 and delta N-15) signa-

tures of sedimented organic matter as indicators of historic

lake trophic state. J Paleolimnol 22:205–221

Bunting L, Leavitt PR, Gibson CE, McGee EJ, Hall VA (2007)

Degradation of water quality in Lough Neagh, Northern

Ireland, by diffuse nitrogen flux from a phosphorus-rich

catchment. Limnol Oceanogr 52:354–369

Bunting L, Leavitt PR, Weidman RP, Vinebrooke RD (2010)

Regulation of the nitrogen biogeochemistry of mountain

lakes by subsidies of terrestrial dissolved organic matter

and the implications for climate studies. Limnol Oceanogr

55:333–345

Castro MS, Driscoll CT, Jordan TE, Reay WG, Boynton WR

(2003) Sources of nitrogen to estuaries in the United States.

Estuaries 26:803–814

Culley JLB, Barnett GM (1984) Land disposal of manure in the

Province of Quebec. Can J Soil Sci 64:75–86

J Paleolimnol (2014) 52:419–433 431

123

Downing JA, McCauley E (1992) The nitrogen—phosphorus

relationship in lakes. Limnol Oceanogr 37:936–945

Elliott EM, Brush GS (2006) Sedimented organic nitrogen iso-

topes in freshwater wetlands record long-term changes in

watershed nitrogen source and land use. Environ Sci

Technol 40:2910–2916

Engstrom DR, Schottler SP, Leavitt PR, Havens KE (2006) A

reevaluation of the cultural eutrophication of lake Oke-

echobee using multiproxy sediment records. Ecol Appl

16:1194–1206

Finlay JC, Small GE, Sterner RW (2013) Human influences on

nitrogen removal in lakes. Sciences 342:247–250

Fogel ML, Cifuentes LA (1993) Isotope fractionation during

primary production. In: Engel MH, Macko SA (eds) Organic

Geochemistry. Plenum Press, New York, pp 73–98

Freudenthal T, Wagner T, Wenzhofer F, Zabel M, Wefer G

(2001) Early diagenesis of organic matter from sediments

of the eastern subtropical Atlantic: evidence from stable

nitrogen and carbon isotopes. Geochim Cosmochim Ac

65:1795–1808

Galman V, Rydberg J, Bigler C (2009) Decadal diagenetic

effects on delta C-13 and delta N-15 studied in varved lake

sediment. Limnol Oceanogr 54:917–924

Giani A, Bird DF, Prairie YT, Lawrence JF (2005) Empirical

study of cyanobacterial toxicity along a trophic gradient of

lakes. Can J Fish Aquat Sci 62:2100–2109

Glew JR, Smol JP, Last WM (2001) Sediment cores collection

and extrusion. In: Last WP, Smol JP (eds) Tracking envi-

ronmental changes using lake sediments, vol 1. Kluwer

Academic, Dordrecht, pp 73–106

Gu B (2009) Variations and controls of nitrogen stable isotopes

in particulate organic matter of lakes. Oecologia 160:

421–431

Gu BH, Schelske CL, Brenner M (1996) Relationship between

sediment and plankton isotope ratios (delta C-13 and delta

N-15) and primary productivity in Florida lakes. Can J Fish

Aquat Sci 53:875–883

Hadas O, Altabet MA, Agnihotri R (2009) Seasonally varying

nitrogen isotope biogeochemistry of particulate organic

matter in Lake Kinneret, Israel. Limnol Oceanogr 54:75–85

Harrison JA, Maranger RJ, Alexander RB, Giblin AE, Jacinthe

PA, Mayorga E, Seitzinger SP, Sobota DJ, Wollheim WM

(2009) The regional and global significance of nitrogen

removal in lakes and reservoirs. Biogeochemistry 93:

143–157

Heaton THE (1986) Isotopic studies of nitrogen pollution in the

hydrosphere and atmosphere - a review. Chem Geol 59:

87–102

Hodell DA, Schelske CL (1998) Production, sedimentation, and

isotopic composition of organic matter in lake Ontario.

Limnol Oceanogr 43:200–214

Holtgrieve GW, Schindler DE, Hobbs WO, Leavitt PR, Ward

EJ, Bunting L, Chen G, Finney BP, Gregory-Eaves I,

Holmgren S, Lisac MJ, Lisi PJ, Nydick K, Rogers LA,

Saros JE, Selbie DT, Shapley MD, Walsh PB, Wolfe AP

(2011) A coherent signature of anthropogenic nitrogen

deposition to remote watersheds of the Northern Hemi-

sphere. Science 334:1545–1548

Helie J (2009) Elemental and stable isotopic approaches for

studying the organic and inorganic carbon components in

natural samples. From Deep-Sea to Coastal Zones:

methods—techniques for studying paleoenvironmentsiop

conference series: earth and environmental science,

012005, 10.1088/1755-1307/5/1/012005

Jones RI, King L, Dent MM, Maberly SC, Gibson CE (2004)

Nitrogen stable isotope ratios in surface sediments, epili-

thon and macrophytes from upland lakes with differing

nutrient status. Freshwater Biol 49:382–391

Kalff J (2002) Limnology: inland water ecosystems. Prentice-

Hall. 592 p

Kendall C, Elliot EM, Wankel SD (2007) Tracing anthropo-

genic inputs of nitrogen to ecosystems. In: Michener R,

Lajtha K (eds) Stable isotope in ecology and environmental

science. Blackwell, Malden, pp 375–449

Kerekes J (1977) The index of lake Basin Permanence. Int Rev

Hydrobiol 62:291–293

Kohzu A, Imai A, Miyajima T, Fukushima T, Matsushige K,

Komatsu K, Kawasaki N, Miura S, Sato T (2011) Direct

evidence for nitrogen isotope discrimination during sedi-

mentation and early diagenesis in Lake Kasumigaura,

Japan. Org Geochem 42:173–183

Koster D, Taranu ZE, Hall RI, Gregory-Eaves I (2008) baseline

water quality information for the Moose lake watershed

management plan. Consultant report http://www.lica.ca/

attachments/065_Moose_Lake_Paleo_Study_Final_Oct_

08_McGil_%20University.pdf

Krzyzanowski J (2010) Review and Identification of Research Needs

to Address Key Issues Related to Reactive Nitrogen (RN)

Deposition and Eutrophication in a Canadian Context (Final

Report). http://www.ccme.ca/assets/pdf/rn_eutrophication_

1450.pdf

Leavitt PR, Brock CS, Ebel C, Patoine A (2006) Landscape-

scale effects of urban nitrogen on a chain of freshwater

lakes in central North America. Limnol Oceanogr 51:

2262–2277

Legendre P, Anderson MJ (1999) Distance-based redundancy

analysis: testing multispecies responses in multifactorial

ecological experiments. Ecol Monogr 69:1–24

Lehmann MF, Bernasconi SM, Barbieri A, McKenzie JA (2002)

Preservation of organic matter and alteration of its carbon

and nitrogen isotope composition during simulated and

in situ early sedimentary diagenesis. Geochim Cosmochim

Ac 66:3573–3584

Lewis WM Jr, Wurtsbaugh WA (2008) Control of lacustrine

phytoplankton by nutrients: erosion of the phosphorus

paradigm. Internat Rev Hydrobiol 93:446–465

Meyers PA (1997) Organic geochemical proxies of paleocea-

nographic, paleolimnologic, and paleoclimatic processes.

Org Geochem 27:213–250

Meyers PA, Ishiwatari R (1993) Lacustrine organic geochem-

istry—An overview of organic matter sources and dia-

genesis in lake-sediments. Org Geochem 20:867–900

Monchamp M, Pick FR, Beisner BE, Maranger R (2014)

Nitrogen forms influence microcystin concentration and

composition via changes in cyanobacterial community

structure. PlosOne 9:e85573

Ostrom NE, Long DT, Bell EM, Beals T (1998) The origin and

cycling of particulate and sedimentary organic matter and

nitrate in lake Superior. Chem. Geol 152:13–28

Owen JS, Mitchell MJ, Michener RH (1999) Stable nitrogen and

carbon isotopic composition of seston and sediment in two

Adirondack lakes. Can J Fish Aquat Sci 56:2186–2192

432 J Paleolimnol (2014) 52:419–433

123

Pick FR, Lean DSR (1987) The role of macronutrients (C, N, P)

in controlling cyanobacterial dominance in temperate

lakes. New Zeal J Mar Fresh 21:425–434

Prairie YT, Parkes A (2007) Developpement d’outils de gestion

simples permettant d’evaluer la capacite de support des

lacs en regard de l’eutrophisation—Volet Estrie. (Unpub-

lished report), Projet de recherche CRSNG-PARDE:50

Prepas EE, Trew DO (1983) Evaluation of the phosphorus

chlorophyll relationship for lakes off the precambrian

shield in Western Canada. Can J Fish Aquat Sci 40:27–35

Riedinger-Whitmore MA, Whitmore TJ, Smoak JM, Brenner

M, Moore A, Curtis J, Schelske CL (2005) Cyanobacterial

proliferation is a recent response to eutrophication in many

Florida lakes: a paleolimnological assessment. Lake Res-

ervoir Manage 21:423–435

Schindler DW (2006) Recent advances in the understanding

and management of eutrophication. Limnol Oceanogr 51:

356–363

Schindler DW, Dillon PJ, Schreier H (2006) A review of

anthropogenic sources of nitrogen and their effects on

Canadian aquatic ecosystems. Biogeochemistry 79:25–44

Schindler DW, Wolfe AP, Vinebrooke R, Crowe A, Blais JM,

Miskimmin B, Freed R, Perren B (2008) The cultural

eutrophication of Lac la Biche, Alberta, Canada: a paleo-

ecological study. Can J Fish Aquat Sci 65:2211–2223

Selbie DT, Finney BP, Barto D, Bunting L, Chen GJ, Leavitt PR,

MacIsaac EA, Schindler DE, Shapley MD, Gregory-Eaves

I (2009) Ecological, landscape, and climatic regulation of

sediment geochemistry in North American sockeye salmon

nursery lakes: insights for paleoecological salmon inves-

tigations. Limnol Oceanogr 54:1733–1745

Smith VH, Schindler DW (2009) Eutrophication science:where

do we go from here? Trends Ecol and Evol 24:201–207

Statistics Canada (2006a) Farm business operating expenses and

paid agricultural labour. Census of agriculture. Ottawa:

Statistics Canada, agriculture division

Statistics Canada (2006b) Livestock, poultry and colonies of

bees. Census of agriculture. Ottawa: Statistics Canada,

agriculture division

Taranu ZE, Gregory-Eaves I (2008) Quantifying relationships

among phosphorus, agriculture, and lake depth at an inter-

regional scale. Ecosystems 11:715–725

Taranu ZE, Zurawell R, Pick FR, Gregory-Eaves I (2012) Pre-

dicting cyanobacterial dynamics in the face of global

change: the importance of scale and environmental con-

text. Glob Change Biol 18:3477–3490

Teranes JL, Bernasconi SM (2000) The record of nitrate utili-

zation and productivity limitation provided by delta N-15

values in lake organic matter—A study of sediment trap

and core sediments from Baldeggersee, Switzerland.

Limnol Oceanogr 45:801–813

Vermaire J (2011) Reconstructing macrophyte biomass

dynamics in temperate lakes of northeastern North Amer-

ica using paleolimnology. Ph.D. thesis. McGill Univ

Vermaire J, Prairie Y, Gregory-Eaves I (2011) The influence of

submerged macrophytes on sedimentary diatoms assem-

blages. J Phycol 47:1230–1240

Wada E, Hattori A (1978) Nitrogen isotope effect in the

assimilation of inorganic nitrogenous compounds by mar-

ine diatoms. Geomicrobiol J 1:85–101

Woodward CA, Potito AP, Beilman DW (2012) Carbon and

nitrogen stable isotope ratios in surface sediments from lakes

of western Ireland: implications for inferring past lake pro-

ductivity and nitrogen loading. J Paleolimnol 47:167–184

Wu JL, Lin L, Gagan MK, Schleser GH, Wang SM (2006)

Organic matter stable isotope (delta C-13, delta N-15)

response to historical eutrophication of lake Taihu, China.

Hydrobiologia 563:19–29

J Paleolimnol (2014) 52:419–433 433

123