Volume 28, Issue 12 p. 1413-1423
Research Article
Free Access

Inherited hypoxia: A new challenge for reoligotrophicated lakes under global warming

Jean-Philippe Jenny

Corresponding Author

Jean-Philippe Jenny

Centre Eau Terre Environnement, Institut National de la Recherche Scientifique, Quebec City, Quebec, Canada

Environnements Dynamiques Territoires de Montagne, Université de Savoie-Mont Blanc, CNRS, Le Bourget du Lac, France

INRA, UMR 042 Centre Alpin de Recherche sur les Réseaux Trophiques des Ecosystèmes Limniques, Université de Savoie-Mont Blanc, Thonon-les-bains, France

Correspondence to: J.-P. Jenny,

[email protected]

Search for more papers by this author
Fabien Arnaud

Fabien Arnaud

Environnements Dynamiques Territoires de Montagne, Université de Savoie-Mont Blanc, CNRS, Le Bourget du Lac, France

Search for more papers by this author
Benjamin Alric

Benjamin Alric

INRA, UMR 042 Centre Alpin de Recherche sur les Réseaux Trophiques des Ecosystèmes Limniques, Université de Savoie-Mont Blanc, Thonon-les-bains, France

Institut Méditerranéen de Biodiversité et d'Ecologie Marine et Continentale, CNRS/IRD, Université d'Avignon, Marseille, France

Search for more papers by this author
Jean-Marcel Dorioz

Jean-Marcel Dorioz

INRA, UMR 042 Centre Alpin de Recherche sur les Réseaux Trophiques des Ecosystèmes Limniques, Université de Savoie-Mont Blanc, Thonon-les-bains, France

Search for more papers by this author
Pierre Sabatier

Pierre Sabatier

Environnements Dynamiques Territoires de Montagne, Université de Savoie-Mont Blanc, CNRS, Le Bourget du Lac, France

Search for more papers by this author
Michel Meybeck

Michel Meybeck

Structure et fonctionnement des systèmes hydriques continentaux, Université de Paris 6, CNRS, Paris, France

Search for more papers by this author
Marie-Elodie Perga

Marie-Elodie Perga

INRA, UMR 042 Centre Alpin de Recherche sur les Réseaux Trophiques des Ecosystèmes Limniques, Université de Savoie-Mont Blanc, Thonon-les-bains, France

Search for more papers by this author
First published: 14 November 2014
Citations: 36

Abstract

The Anthropocene is characterized by a worldwide spread of hypoxia, among other manifestations, which threatens aquatic ecosystem functions, services, and biodiversity. The primary cause of hypoxia onset in recent decades is human-triggered eutrophication. Global warming has also been demonstrated to contribute to the increase of hypoxic conditions. However, the precise role of both environmental forcings on hypoxia dynamics over the long term remains mainly unknown due to a lack of historical monitoring. In this study, we used an innovative paleolimnological approach on three large European lakes to quantify past hypoxia dynamics and to hierarchies the contributions of climate and nutrients. Even for lake ecosystems that have been well oxygenated over a millennia-long period, and regardless of past climatic fluctuations, a shift to hypoxic conditions occurred in the 1950s in response to an unprecedented rise in total phosphorus concentrations above 10 ± 5 µg P L−1. Following this shift, hypoxia never disappeared despite the fact that environmental policies succeeded in drastically reducing lake phosphorus concentrations. During that period, decadal fluctuations in hypoxic volume were great, ranging between 0.5 and 8% of the total lake volumes. We demonstrate, through statistical modeling, that these fluctuations were essentially driven by climatic factors, such as river discharge and air temperature. In lakes Geneva and Bourget, which are fed by large river systems, fluctuations in hypoxic volume were negatively correlated with river discharge. In contrast, the expansion of hypoxia has been related only to warmer air temperatures at Annecy, which is fed by small river systems. Hence, we outline a theoretical framework assuming that restored lake ecosystems have inherited hypoxia from the eutrophication period and have shifted to a new stable state with new key controls of water and ecosystem quality. We suggest that controlling river discharge may be a complementary strategy for local management of lakes fed by large river systems.

Key Points

  • Factors driving hypoxia are assessed over the Holocene and the last 130 years
  • Hypoxia in the twentieth century has been triggered by anthropogenic P supplies
  • The current dynamics of anthropogenic hypoxia has been driven by climate

1 Introduction

In marine and lake environments, hypoxia ([O2] <2 mg L−1) [Roberts et al., 2009] is forecast to increase in extent, severity, and duration [Diaz and Rosenberg, 2008; Posch et al., 2012; Meire et al., 2013] in response to the combined effects of excessive external nutrient inputs [Nixon, 1995; Rabalais et al., 2001] and global warming [Posch et al., 2012; Meire et al., 2013]. Hypoxia is a mounting problem with severe consequences for aquatic life, including death and catastrophic changes [Vaquer-Sunyer and Duarte, 2008; Villnäs et al., 2012]. Over the past century, the consequences of anthropogenic nutrient emissions in lacustrine ecosystems include an increase in planktonic biomass and a decrease in bottom oxygen concentrations deriving from the respiration and consumption of sinking organic matter [Nixon, 1995; Müller et al., 2012]. In addition, increased air temperatures (T°) can cause water warming and therefore decrease oxygen solubility [Deutsch et al., 2011] or contribute to decreased water mixing efficiency and oxygen renewal [Straile et al., 2003; Coma et al., 2009]. In contrast, during seasonal flooding events, high river discharges have been documented to promote oxygenation either by introducing well-oxygenated water masses into lakes, weakening the residual stratification during the winter, or by increasing the trapping of internal phosphorus (P) load in sediments [Livingstone, 1997; Ambrosetti et al., 2003; Hupfer and Lewandowski, 2005]. Such multiple regulators of hypoxia have been observed separately using either individual site investigations or short time series of data (i.e., less than 60 years) [Friedrich et al., 2013]. However, when considered together, the interactions of these regulators present complex and unpredictable responses [Jenny et al., 2013; Friedrich et al., 2013]. For example, hypolimnetic hypoxia generally persists in most large, deep lakes of Western Europe long after phosphorus was reduced to low concentrations following the abatement measures of the 1970s and 1980s. In some cases, such as in lakes Bourget and Annecy, hypoxia persists even when pelagic symptoms of eutrophication (e.g., a decrease in the nitrogen (N):P ratio, a decrease in the water column transparency in the spring, and an increase in chlorophyll a concentrations) have disappeared [Jacquet et al., 2005].

The persistence of hypoxia raises the question of the long-term role of phosphorus in interaction with climatic forcings; for example, phosphorous might be a trigger, a contributor, or an amplifier of oxygen depletion. Here we used a paleolimnological approach (Figure S1 in the supporting information) to reconstruct the long-term trends (i.e., 130 and 10,000 years) of hypoxia in order to address the effects of P, T°, and river discharge before, during, and after historical P enrichments on bottom oxygen depletion. Wind was not included as a driver despite its impact on water oxygen levels [Michalski and Lemmin, 2012] because there is no paleoproxy for historical wind dynamics from sediments. We used a modeling approach for three large, deep lakes based on the use of additive models [Simpson and Anderson, 2009] to provide greater insight into complex multivariate interactions. These European lakes, located on the northwestern edge of the Alps (Figure 1), were selected because they are within the same ecoclimatic region and because they share a common history of air temperatures [Auer et al., 2007]. Furthermore, as they were carved by the same glacial processes, the three lakes arose from a similar morphology and are among the few lakes in the world that have been monitored for > 60 years.

Details are in the caption following the image
(a) Location map of lakes Geneva, Bourget, and Annecy at the western edge of the Alps (i.e., France and Switzerland). The classification of the three lakes is based on hydrological indicators: river discharges (Q) and river connections to the lakes (permanent or intermittent), the ratio of the drainage basin area (dotted line) against the total lake area (Ad:Ao), and the mean terrigenous fluxes of each lake (Figure 2). (b) The evolution over the Holocene of hypoxia dynamic in Lake Bourget, local phosphorus, local sediment supplies, and global air temperatures.

2 Materials and Methods

2.1 Study Area

Lakes Geneva (Léman), Bourget, and Annecy are young lakes located on the northwest edge of the French Alps (Figure 1a) that originated from the last deglaciation, 11,500 years ago. They belong to a similar ecoclimatic zone. They are large, deep temperate lakes (i.e., free from ice during winter) with a relatively fast water renewal rate (Table 1). They also share a common trophic history and undergo similar climatic variability (see results).

Table 1. Lake Geomorphological Characteristics
Parameters Lake Geneva (Léman) Lake Bourget Lake Annecy
Total lake area (Ao) (km2) 582 45 27
Trainage basin area (Ad) (km2) 7395 560 273
Ad:Ao 12.7 12.4 10.1
Volumes (km3) 89 3.6 1.12
Maximum depth (m) 309.7 147 65
Length (m) 72.3 18 14.6
Width (m) 13.8 3.2 3.2
Latitude N46°26 N45°51 N45°44
Altitude (m) 372 231 447
Renewal rate (yr) 12 8 4

However, the three lakes present contrasting hydrological conditions because water and sediment supplies from the river systems differ greatly among the three lakes. The supplies are the highest from the large “Alpine-Rhône,” with a mean discharge of 181 m3 s−1, which flows directly into Lake Geneva. The supplies are lower from the “Geneva-Rhône,” downstream of Lake Geneva, which flows at a rate of 10 to 100 m3 s−1 and discharges into Lake Bourget through the Savière Chanel only during major flooding events [Giguet-Covex et al., 2010a, 2010b; Jenny et al., 2014]. The supplies are lowest in Lake Annecy because it is supplied by small tributaries, with mean discharges of less than 3 m3 s−1. The ratio of the drainage basin area (Ad) to the total lake area (Ao) (Ad:Ao) [Meybeck, 1995] confirmed the differences in the local river inputs for the three lakes (12.7, 11.8, and 10.5 in lakes Geneva, Bourget, and Annecy, respectively).

2.2 Core Sampling and Monitoring

A total of 81 sediment cores were collected from depth gradients in the three lakes during eight field campaigns conducted between 2009 and 2012 (Figure S2). Cores were split lengthwise, photographed, and described. All three lakes have been monitored for decades; monitoring began in 1955 for Lake Geneva (managed by the International Commission for the Protection of Lake Geneva Waters and the INRA, French National Institute for Agronomical Research INRA), 1973 for Lake Bourget (Lake Bourget water agency and INRA), and 1966 for Lake Annecy (managed by the Intercommunal Association of Lake Annecy and INRA). These three lakes belong to the French Long Term Ecological Observatory of Lakes (SOERE OLA http://www6.inra.fr/soere-ola), which provided the monitoring data.

2.3 Sediment Dating

Age-depth relationships were previously modeled with 210Pb, 226Ra, 137Cs, and 241Am activities [Alric et al., 2013]. Activities were measured by gamma spectrometry on the deepest core of each lake. Sediments from other depths were precisely correlated in this study using annual lamination and flood deposits as lithological markers, and ages were then interpolated between dated lithological markers according to Jenny et al. [2013, 2014].

2.4 Hypoxia Volume Reconstruction

Lakes Geneva (Léman), Bourget, and Annecy are warm-monomictic lakes; their hypolimnia are partially reoxygenated during winter water mixing (Figure S3). Oxygen is then consumed progressively during summer stratification through microbial respiration (Figure S3). When annual oxygen concentrations fall below a critical point, determined by a combination of time and intensity, macrobenthic life disappears [Cicchetti et al., 2006], which prevents bioturbation and its related sediment mixing [Zolitschka, 2009]. This threshold is well recorded by a sedimentological proxy of hypoxia through the formation of varves (i.e., annually laminated sediments) [Francus et al., 2013; Giguet-Covex et al., 2010a, 2010b]. The annual volumes of hypoxic waters were therefore reconstructed for each lake according to an innovative approach [Jenny et al., 2013]. The approach involved (i) using the geographical mapping of varve extension and (ii) integrating the volume between the lake bottom and the depth of the shallowest varve-bearing core for each year via SURFER 9 software (Rockware earth science software, US). Based on 81 well-dated sediments, volumes of hypoxic waters in the three deep lakes have been reconstructed for the past 130 years at an annual resolution. A long core collected in 2001 [Arnaud et al., 2012] also provided insight into Lake Bourget's natural history with respect to oxygen conditions over the last 11,500 years (Figure 1b).

2.5 Trophic State, Air Temperature, and High River Discharge

The seasonal total P concentrations ([TP]) in lake water were analyzed over the monitored period using the acid molybdate method (Association Française de Normalisation). To extend the data collection to the previous unmonitored period, starting in 1880, changes in the annual water [TP] of the three lakes were reconstructed using a diatom transfer function [Berthon et al., 2013]. The data set of annual and seasonal air temperatures was extracted from the historical instrumental climatological surface time series of the Alpine region (Histalp) database (www.zamg.ac.at/histalp/) using lake geographical coordinates [Auer et al., 2007]. Finally, X-ray fluorescence measurements using an Avaatech-type core scanner at the Environnement, Dynamiques et Territoires de la Montagne Laboratory were used to track the yearly evolution of detrital inputs through time, evidenced by the flux of terrigenous supplies recorded by lake sediments, considered herein as a tracer of the Rhône River's hydrological activity [Arnaud et al., 2005, 2012; Giguet-Covex et al., 2010a, 2010b].

2.6 Statistical Analysis

To investigate the relationships between oxygen depletion and external forcings, i.e., lake nutrient concentration, seasonal temperatures, and flood regime, as well as to separate and quantify their influences, general additive models (GAMs) were used [Simpson and Anderson, 2009; Alric et al., 2013]. In GAM, covariates are assumed to affect the response variable through an additive sum of smooth functions. GAM parameterization and adaptation to paleoecological data were performed following technical recommendations from Simpson and Anderson [2009]. Model selection with GAMs involves choosing explanatory variables. A starting model that included all predictor variables as smooth terms was then simplified as required using a backward/forward stepwise procedure to drop terms that did not significantly contribute to model fit based on P values, while Akaike information criterion (AIC) and the adjusted coefficient of determination (R2 adjust) were used to select the best model. Linear models were used to explore the relationships between the volume of hypoxic waters and two relevant environmental forcings (i.e., river discharge and winter temperature). All statistical analyses were performed in R 2.15.1 (R Development Core Team, 2009) using the packages mgcv [Wood, 2008, 2011] and nlme [Pinheiro et al., 2009].

3 Results

3.1 Past Hypoxia Dynamics

The reconstructed hypoxia volumes (HVs) successfully matched 52 years of monitoring data from Lake Geneva (Figure 2), confirming the reliability of the methods for the studied lakes. In Lake Geneva, overall monitored oxygen concentrations were the lowest for the 1970–1980 and 1990–2000 time periods. Consistently, these two time periods are also those for which volumes of reconstructed hypoxia were the highest.

Details are in the caption following the image
The evolution of hypoxia volumes (HVs) in Lake Geneva. The annual HV (106 m3) reconstructed from sediment archives is indicated by the red line. Error bars correspond to 95% confidence intervals on age. Dashed lines correspond to the error on HV that takes into account the mean sampling depth between cores. Monthly monitored oxygen concentrations are indicated by the color scale (top right corner). The monthly oxygen profiles from monitoring match the reconstruction and validate our method.

The reconstructed hypoxia dynamics demonstrated that the studied lakes were well oxygenated before the second half of the twentieth century (Figure 3a). Reconstruction from the long core indicated that over 11,500 years, Lake Bourget was well oxygenated and phosphorus concentrations were low. Hypoxia and phosphorus increases were only recorded in recent decades. According to the sediment records, hypoxia first appeared in 1952 ± 2 years in Lake Geneva, in 1933 ± 1 year in Lake Bourget, and in 1950 ± 2 years in Lake Annecy and persisted in the three lakes with no interruption until today (i.e., sediments from the deep zones always record varves since the onset of hypoxia). The hypoxia dynamics has fluctuated over the last 60 years, exhibiting a series of decadal to multidecadal contractions and expansions (Figures 3a and 4). Since hypoxia has settled in all three lakes, it has reached up to 20% of the total lake area in Lake Geneva and up to 60% in lakes Bourget and Annecy (Figure 4). A first expansion was recorded for the three lakes in the 1950–1960s, which we expect to be related to anthropogenic activities. However, the later contractions and expansions in HV occurred at time intervals that varied among lakes. Periods of contraction occurred in the 1980s in lakes Geneva and Bourget and in the 1970s in Lake Annecy. Hypoxia has increased to higher volumes again for lakes Bourget and Geneva since the 1990s. Thus, investigating potential forcers of the last 130 years should provide information regarding the factors that contribute to the onset of and latest fluctuations in hypoxia for each of the three lakes.

Details are in the caption following the image
Evolution of HV in the three lakes over the last 130 years. (a) Temporal changes in HV (106 m3) as reconstructed from a 4-D sediment investigation. Dashed lines correspond to the error on HV that takes into account the mean sampling depth between cores. (b) Temporal changes in forcings of HV: mean annual air temperatures anomalies (°C), mean diatom-inferred winter total phosphorus concentrations (µg L−1), and terrigenous supplies (TS), which are a proxy for high river discharges. Third-degree polynomials were incorporated in the temporal forcings of HV.
Details are in the caption following the image
Spatiotemporal extension of lake hypoxia volumes in lakes (a) Geneva, (b) Bourget, and (c) Annecy. According to literature and samplings, hypoxia (marked by the preservation of varves) was never recorded in Lake Geneva sediments except in the most recent samples. According to sediment-based dating, the onset of hypoxia occurred in 1950, 1933, and 1952 in lakes Geneva, Bourget, and Annecy, respectively. Lake Geneva underwent two phases of HV extension: in 1971–1980 and in 1990–1999; these periods were separated by two decadal contractions of HV in 1981–1989 and 2002–1910. Today, observations show signs of HV reexpansion, whereas no signs are currently visible in the sediments. Hypoxia has never been recorded in Lake Bourget Holocene sediments (~12,000 last years). Lake Bourget experienced two major phases of expansion, one from 1940 to 1975 and the other during the period since 1997. One pronounced contraction of HV was recorded between 1976 and 1997. Lake Annecy recorded hypoxia before 1950, but it was confined to only the most profound zones. According to a spatial investigation of sediments, our results demonstrate that HV increased for the first time in 1952 in Lake Annecy. Two major expansions were recorded in 1958–1968 and 1990–2010. A contraction occurred in the 1970s–1980s. Grey color corresponds to the maximum estimated HV (cf. legend in Figure 2).

3.2 Temporal Changes in External Forcings

Temporal changes in external forcings of hypoxia were reconstructed for the period 1880–2010 (Figure 3b). Regional air temperatures have increased by 2.0°C during the twentieth century, with a first warming period between 1940 and 1960 and a second in the early 1980s [Auer et al., 2007; Alric et al., 2013]. Over the last 11,500 years, the climate has also undergone warmer periods, although well-oxygenated conditions prevailed, with air temperatures as high as those observed in the twentieth century [Marcott et al., 2013] (Figure 1b).

Previous paleoreconstructions demonstrated that phosphorus inputs to Lake Bourget were low over the past millennia and, by extension, that these lakes are naturally oligotrophic (Figure 1b). However, these lakes have been impacted by high anthropogenic pressures from both urbanized and agricultural areas that led to dramatic changes in total phosphorus inputs and total phosphorus concentrations (TP) in the early twentieth century (Figure 1) [Jacquet et al., 2005; Giguet-Covex et al., 2010a, 2010b]. The three lakes share the same P enrichment history (Figure 3b) [Berthon et al., 2013]. Although they were oligotrophic at the end of the nineteenth century, all three lakes underwent phosphorous enrichment as early as the 1920s; this process intensified during the 1940s. The three lakes reached varying levels of maximum eutrophication. Lake Annecy did not exceed the oligomesotrophic status (i.e., mean annual concentration over the first 20 m depths: 18 µg P L−1 in 1969). Lakes Geneva and Bourget reached a eutrophic status by the late 1970s (i.e., mean annual concentration over the first 20 m depths: 55–90 µg P L−1). Following restoration programs during the past three decades, observational data demonstrate that mean dissolved P concentrations measured during winter have been successfully reduced to 17 µg P L−1 in Bourget, 19 µg P L−1 in Geneva, and 6 µg P L−1 in Annecy [Berthon et al., 2013] (Figure 3b).

The temporal change in terrigenous supplies (TSs), i.e., the detrital inputs from the watershed, was used as a proxy of high river discharge [Arnaud et al., 2012; Wilhelm et al., 2012] from the Alpine-Rhône (i.e., Lake Geneva tributary) and Geneva-Rhône (i.e., Lake Bourget tributary). Since 1880, terrigenous fluxes have decreased by 50%, 40%, and 45% for lakes Geneva, Bourget, and Annecy, respectively, indicating a decrease in the river discharge regime over the entire studied period (Figure 3b). However, recent increases were recorded in the early 1990s and the early 2000s in Lake Geneva and in the 1980s in Lake Bourget. The reconstructed terrigenous fluxes also confirm the contrasting hydrological conditions among the three lakes (Figure 3b). The river inputs are highest in Lake Geneva (TS = 0.055 mg cm−2 yr−1), medium in Lake Bourget (TS = 0.022 mg cm−2 yr−1), and lowest in Lake Annecy (TS = 0.013 mg cm−2 yr−1).

3.3 Contribution of External Forcings

The predictions of the generalized additive models (GAMs) and the subsequent temporal contributions of the covariates (T°, TP, and TS) to the fitted hypoxia volumes are presented in Figure 5. Coefficients in the model >0 indicate that the associated covariates contribute to increases in the fitted HV, whereas coefficients <0 indicate contributions to the contraction of the fitted HV (Figure 5). The final models accounted for 78%, 43%, and 59% of the total variance for lakes Geneva, Bourget, and Annecy, respectively (Table 2). Hypoxia first appeared in all three lakes when diatom-inferred total phosphorus ([Ptot]DI) (annual P concentrations over the first 20 m of depth) crossed thresholds of 10 ± 2 µg P L−1 in Lake Bourget (in 1933 ± 1), 8 ± 2 µg P L−1 in Lake Annecy (in 1952 ± 2), and 8 ± 2 µg P L−1 in Lake Geneva (in 1950 ± 2). Monitoring in Lake Geneva confirmed the relatively low P concentrations at the time of the establishment of hypoxia (annual values from the first 0–20 m; Ptot = 12.4 µg L−1 in 1955) (Figure 3b).

Details are in the caption following the image
Evolution of forcings of HV over the past 130 years. Temporal contribution of considered forcings (average ± approximately 95% pointwise confidence interval) to the observed changes in HV extension. Contribution can be positive (white area) or negative (brown area). For example, in Lake Annecy, the expansion of HV in the 1960s is explained by the increase in P; the later decrease of HV in the 1980s is explained by the restoration of P, and the reexpansion of HV starting in the 1990s is explained by temperature increases.
Table 2. Factors and Coefficients of the Generalized Additive Modela
Lake Model Time Window Deviance Explained Factors Estimate Standard Error F P Significance
Geneva Model 1 1880–2009 78% s(P) 6.943 8.004 5.367 2.66e-05 ***
s(TS) 6.473 7.602 2.660 0.0143 *
Bourget Model 2 1880–2009 43.1% s(TS) 2.646 3.338 11.834 7.15e-07 ***
s(T°s) 1.171 1.325 4.556 0.0259 *
s(P) 1.001 1.002 4.750 0.0322 *
Annecy Model 3 1880–2009 59.4% s(T°s) 3.65 4.411 8.769 5.92e-05 ***
s(P) 1.00 1.000 6.73 0.0146 *
  • a Factors considered in the models: P, diatom-inferred phosphorus concentrations; TS, terrigenous supplies (proxy of high river discharge); and T°(S), mean summer temperatures. F, F statistic; P, associated probabilities. In GAM, covariates are assumed to affect the response variable through an additive sum of smooth functions. Model selection with GAMs involves choosing explanatory variables. A starting model that included all predictor variables as smooth terms was then simplified as required using a backward/forward stepwise procedure to dropping terms making a nonsignificant contribution (i.e., based on P values), while AIC and adjusted coefficients of determination (R2 adjust) were used to select the best model.
  • * Significant at 0.01.
  • ** Significant at 0.001.
  • *** Significant at 0.

In lakes Bourget and Annecy, HV has been expanding over the past few decades despite a decrease in total P content (i.e., by a factor of 5); the reexpansion of HV in both lakes since 1980 could be solely related to an increase in air temperatures (model results >0; Figure 4). Besides, periods of higher river discharge related to the HV contraction in the 1970s and in the 2000s in Lake Geneva and in the 1980s in Lake Bourget (GAM results <0; Figure 5 and Table 2). In contrast, river discharge was not significantly related to HV in Lake Annecy (Figure 5 and Table 2). Model 1 (Table 2) shows that fluctuations in HV are related to only one factor (i.e., the terrigenous supplies) in Lake Geneva; model 2 shows that fluctuations in HV are related to two factors (i.e., terrigenous supplies and air temperature) in Lake Bourget; and model 3 shows that fluctuations are related to only one factor (i.e., the air temperature) in Lake Annecy. It suggests that relationship between floods and hypoxia differs between lakes.

4 Discussion

4.1 Hypoxia Onset Triggered by Anthropogenic P Supplies

The sedimentary record indicates that lakes Geneva, Bourget, and Annecy were well oxygenated before the middle of the twentieth century (Figures 1 and 3). The past well-oxygenated conditions and the midcentury decrease in oxygen were confirmed based on chironomid assemblages in lakes Bourget and Annecy [Millet et al., 2010; Frossard et al., 2013]. Following this period, once the three lakes began to be overenriched in nutrients, the absence of bioturbation in sediments of the three lakes suggests that hypoxia was strong enough to prevent macrobenthic life in the deep zone. Since then, sediments indicate that HV presented a series of expansions and contractions, but there was no restoration at any time of the past well-oxygenated conditions. The three lakes exhibit similar temporal patterns: (i) well-oxygenated conditions, (ii) a shift to hypoxic conditions, and (iii) maintenance and fluctuations of HV. In all three lakes, the shift occurred when total phosphorus concentrations first exceeded 10 µg P L−1. Nutrient overenrichment appears to be the only environmental forcing that could explain the onset of hypoxia at that time in the three lakes. In Lake Bourget, however, the onset of hypoxia occurred earlier (i.e., in the 1930s) than in the other two lakes (i.e., in the 1950s). This earlier onset of hypoxia is attributed to higher nutrient contents in Lake Bourget over the entire studied period. The higher P concentrations could have accentuated the vulnerability of Lake Bourget to the first increase in temperatures during the 1930s and 1940s (Figures 2 and 4; GAM results). However, data from 130 yearlong records on Lake Bourget do not fully support the conclusion that P is a triggering mechanism. GAM results indicate that air temperature was also involved in hypoxia settlement between 1930 and 1940. It is consistent with the increases in temperature at that time. However, small increases in P concentrations were probably not detected by the transfer function, and we suggest that without a small increase in P concentration, hypoxia would not have been triggered. Indeed, the reconstruction using the TP transfer function does not detect small fluctuations of TP concentrations at low TP levels (<10 µg L−1) [Berthon et al., 2013]. This was evidenced in Berthon et al. [2013] by the comparison with Daphnia biostratigraphies, which highlighted, in the case of Lake Bourget, evidence of increases in P since 1930. Furthermore, long-term records demonstrate that despite several climatic fluctuations, well-oxygenated conditions occurred over the Holocene in Lake Bourget (i.e., over the last ~12,000 years) and that hypoxic conditions settled a few decades ago (Figure 1b), as observed for other monomictic lakes in Europe [e.g., Hollander et al., 1992; Lotter et al., 1997; Tylmann et al., 2013]. Hence, this recent change in hypolimnetic oxygenation highlights an unprecedented shift in lake ecological status in the midtwentieth century (in the sense of Scheffer et al. [2001]). As climate was not the cause of hypoxia settlement over the entire Holocene, it seems reasonable to suggest that recent anthropogenic P enrichment (even at levels as low as they were in Lake Bourget) is to a great extent the cause of hypoxia onset through an increase in benthic respiration. In Lake Bourget, we argue that P was necessary to trigger hypoxia but that the early onset of hypoxia resulted from the combination of an increase in P and warming in this lake; this combination appears to be the ultimate cause of the appearance of hypoxia at that time.

4.2 Hypoxia Dynamics Driven by Climatic Factors

We suggest that based on paleolimnological reconstructions and GAM results, anthropogenic change has a direct impact on the onset of hypoxia through P enrichments. P concentrations appear to contribute to a trigger mechanism for hypoxia in all three lakes. However, once the P threshold has been exceeded, fluctuations in P content and HV appear to be decoupled. GAM results also indicate that decadal to multidecadal extensions of HV in these large temperate lakes are driven by increases of summer or winter air temperatures (results >0) and that decadal to multidecadal HV contractions are driven by enhanced flood regimes (results <0; Figure 4 and Table 2).

Reconstructions in lakes Geneva and Bourget show that contractions of hypoxic volumes are related to periods of high river discharge over the studied period (Figure 5). This highlights the link between lake hydrology and hypoxia extent over the past 130 years. The role of hydrology in hypoxia dynamics is explained as higher river discharge can reduce hypoxia by supplying oxygenated water into the lake and by increasing water mixing during winter. In Lake Geneva, 52 years of monitoring confirmed that high discharges have increased the depth of mixing during the winter (Figure 6; R2 = 0.53, p = 6.97 × 10−11), which may explain the reoxygenation recorded during periods of flooding. Note that cold winters also increased the depth of mixing (Figure 6; R2 = 0.6, p = 6.62 × 10−11) and probably acted in conjunction with high river discharges to produce better oxygen conditions. The effects of long-term interactions between temperatures and river discharge on hypoxia dynamic also require further investigation.

Details are in the caption following the image
Relation among the depth of mixing in Lake Geneva as defined in Figure S3, winter temperatures, and Alpine-Rhône discharge. The correlations are presented for time series averaged over 5 years. R2 values and the p values (p) of the linear models highlight the significant relationship between these two forcings and the depth of mixing. Red text corresponds to the period from 1989 to 2009, black corresponds to 1974–1988, and blue corresponds to 1961–1973.

Local river discharges have also interfered with forcings of hypoxia over the studied period. We demonstrated that climate forcings (i.e., T° and flood regime) do not have the same contribution to hypoxia dynamics in the three lakes over the last 60 years. We attribute the different impacts of climate on hypoxia to local hydrology. Depending on the local intensity of supplies from the tributaries (i.e., high, medium, or low), the contribution of climate to the extent of hypoxia varies among lakes. River discharges have the greatest impact on hypoxia contractions in Lake Geneva, where the local supplies from the tributaries are the highest. In contrast, temperatures have the greatest impact on the extension of hypoxia in Lake Annecy, where local supplies from tributaries are the lowest. Lake Bourget [Jenny et al., 2014] is fed only during severe flood events that exceed 600 m3 s−1. The drivers of Lake Bourget lie between the other two lakes as oxygenation is controlled both by temperature and flood regime. Hence, based on this investigation of three lakes, local hydrology seems to control the vulnerability of lake hypolimnion to global warming. This suggests that the lower the local supplies from tributaries are, the more vulnerable the lake is to hypoxia induced by global warming. Our study sets the stage for new hypotheses to be tested on a broader range of lakes.

4.3 Vulnerability to Global Warming Inferred by P

Our results suggest that historical eutrophication has led to persistent hypoxic conditions in lakes, and there has been no return of past well-oxygenated conditions despite local P management. Conventional restoration programs must therefore consider that reducing local phosphorus inputs alone will not systematically counteract hypoxia unless concentrations are lowered to levels that will likely depauperate pelagic habitats, an undesirable outcome. With respect to climate forcings, fluctuations of HV were apparently not affected by hydrological and temperature variations before anthropogenic eutrophication, whereas HV was affected by these forcings after the onset of local eutrophication. In the context of global warming, local management appears to be the most pertinent strategy to reduce the impacts of increases in T° over the coming decades. GAM results and monitoring data (Figures 3 and 4) suggest that high river discharge can supply oxygen to lakes during the winter, and together, they suggest that local management of river activity could potentially overcome some of the vulnerability of lakes to global warming. Hence, human control of flood regimes seems to be a promising support for local P management of lakes fed by large rivers. This hypothesis also needs to be evaluated on a broader range lakes.

5 Conclusion

Our findings demonstrate that early lake hypoxia onset occurred at the time of the first low increase of phosphorus concentrations (i.e., monitored or reconstructed). Because a threshold of low P concentration was exceeded, the variance in the hypoxia was no longer explained by phosphorus fluctuations but, rather, by climate forcing; warmer temperatures explain the expansion of hypoxia in the lakes, whereas high river discharge explains the contractions of hypoxia. Low excess P supplies from watersheds might then pose a challenge for aquatic management in the context of global warming. We suggest that restoration of natural river flows could foster positive consequences on downstream lake management. Oxygen management of lake hypolimnions also needs to take into account the lake system as a metaecosystem, including both the lake and its watersheds.

Acknowledgments

We are grateful to the Observatoire des Lacs alpins (OLA) and the Alpine centre for research on lake ecosystems and food webs (CARRTEL INRA, http://www6.dijon.inra.fr/thonon) for providing long-term monitoring data on oxygen and phosphorus concentrations from the three lakes. Air temperatures were collected from historical instrumental climatological surface time series of the greater alpine region (Histalp http://www.zamg.ac.at/histalp/). Diatom-inferred phosphorus concentrations from Berthon et al. [2013] and new data (i.e., from this study) on hypoxia volumes and on high river discharges for the three lakes are presented in the supporting information (Tables S1–S3). This study was supported by the Association des Pays de Savoie (APS) and the French National Research Agency (ANR-VMCS 008) through the project Perturbations Impacts on Lake Food Webs: A Paleo-Ecological Approach (IPERRETRO).