Algal blooms in lake waters are frequently associated with elevated nutrient levels, in particular phosphorus. Here we show, with long-term data from Sweden’s largest lake, Lake Vänern, that total phytoplankton biomass has significantly increased since the 1980s at the same time as total phosphorus and inorganic nitrogen concentrations have significantly decreased (non-parametric Mann-Kendall test: P < 0.01). Lake Vänern was not the only lake showing such a pattern: evaluating long-term data from 13 small oligotrophic reference lakes revealed the same pattern. Other synchronous changes observed in Sweden’s oligotrophic lakes included strongly decreasing sulfate concentrations and increasing water temperatures. We found that alkalinity, water temperatures, silica and sulfate concentrations were the strongest predictors for total phytoplankton biomass at all three sites in Lake Vänern when we used partial least squares regression models with 11 input variables. From our results, we suggest that decreasing acidity in combination with increasing water temperatures has been more important for phytoplankton growth than total phosphorous and nitrogen concentrations. Thus, we conclude that phytoplankton biomass in oligotrophic small and large lakes is strongly associated with climate and atmospheric depositional changes, and consequently needs to be managed beyond catchment scales.
The paradigm of phosphorus as the limiting nutrient for freshwater algae can be rooted back as far as Neumann’s description from 1919 (Lewis and Wurtsbaugh, 2008), and was widely accepted after Schindler’s pioneering experimental work in Canada in the 1970s (Schindler, 1977). Nutrient reductions with focus on phosphorus have been performed in many lakes around the world and often with a positive result (Marsden, 1989; Jeppesen et al., 2005). Sweden’s largest lakes, Lake Vänern, Lake Vättern and Lake Mälaren, are no exceptions. A sharp phosphorus reduction in the 1970s resulted in a rapid decline of algal biomass in these large lakes (Willen, 2001), a re-oligotrophication process that is known from many large lakes around the world.
However, despite sharp phosphorus reductions in freshwater ecosystems, algal blooms still frequently occur. World-wide discussions are ongoing on management strategies how to reduce algal blooms. Several studies suggest that a decrease of phosphorus alone might not be the only management solution since the growth of algae is not only limited by phosphorus but also by nitrogen (Howarth and Paerl, 2008; Conley et al., 2009). Management strategies become even more complex when additional factors that are known to control algal growth are taken into consideration including temperature, light, mixing regimes, and other nutrients than phosphorus and nitrogen (Reynolds, 1997; Elliott et al., 2005). A variety of modeling approaches are available that simulate algal growth in aquatic ecosystems under climate change (Reynolds et al., 2001; Elliott et al., 2005) but temporal variations of algal biomass are generally rather lake-specific despite synchronous changes in physical and chemical lake variables among lake ecosystems (Bloch and Weyhenmeyer, 2012). Finding common management strategies is, therefore, not always possible.
Generally, it is important for the management of aquatic ecosystems to differentiate between eutrophic and oligotrophic systems. Most management strategies have been developed for eutrophic systems where algal blooms frequently occur. The results from these systems have then commonly been extrapolated to oligotrophic systems without further calibration and validation. A recent study now revealed that the paradigm of nutrient limitation for the productivity of a lake ecosystem does not hold for oligotrophic lakes (Karlsson et al., 2009). Instead of nutrient limitation there was clear light limitation in the nutrient-poor lakes. Since drivers for light and nutrients in a lake ecosystem can largely differ, it is highly important to know the limiting factors for the productivity of a lake ecosystem. Considering that eutrophic systems might have other limiting factors than oligotrophic systems, the question arises which factors influence the productivity of a system that has become oligotrophic after extensive nutrient reduction measures in the catchment.
In order to better understand drivers of lake productivity when lakes have reached oligotrophic conditions after nutrient reduction measures we analyzed the temporal development of phytoplankton biomass in Sweden’s largest Lake Vänern and related this to physical and chemical changes in the lake. We hypothesized that the phytoplankton biomass in Lake Vänern is no longer strongly related to nutrient concentrations but follows variables that are relevant for the development of phytoplankton biomass in nutrient-poor lakes. To test the hypothesis we used physical, chemical and phytoplankton long-term data from 13 nutrient-poor lakes and from Lake Vänern.
Study sites and data material
Our study sites were Sweden’s largest lake, Lake Vänern and 13 small reference lakes. Reference lakes are lakes that, apart from atmospheric deposition and climate changes, have experienced no or little anthropogenic changes in the catchment area during the past 20 years. The study lakes, including Sweden’s largest lake, can all be classified as oligotrophic lakes with mean total phosphorus concentrations <15 μ l−1. In contrast to Lake Vänern, the Swedish 13 reference lakes are relatively small (mean lake area 0.81 ± 0.77 km2), shallow (mean depth 5.4 ± 2.7) and humic (Table 1). The lakes are distributed all over Sweden (Figure 1).
All lakes are part of the Swedish lake inventory program and all data are freely available at http://www.slu.se/aquatic-sciences. We chose the 13 reference lakes that had complete monthly time series available during the open water season from May to October (in Abiskojaure and Jutsajaure from June to October) during 1988–2010. This approach gave us for each lake six (for Abiskojaure and Jutsajaure five) data points per year. In a very few cases a water sample was missing from a lake. Our statistical analyses were not sensitive to these few missing values which we tested by applying our statistical methods to both incomplete and complete data series. We achieved complete data series by simple linear interpolation techniques.
Sampling in Lake Vänern took place at three sites and in the reference lakes at one site in the middle of each lake. In this study we used the intensity of thermal stratification (Temp diff; calculated as the difference between surface and bottom temperatures), and the following variables measured in surface waters (0.5 m): water temperature (Temp), pH, sulfate (SO4), alkalinity (Alk), ammonium-nitrogen (NH4-N), nitrate-nitrogen (NO3-N), total phosphorous (TP), phosphate-phosphorus (PO4-P), absorbance at 420 nm of 0.45 μm filtered water in a 5 cm cuvette (AbsF420nm/5cm), and reactive silica (Si). Water chemical variables were analyzed by a certified laboratory according to international (ISO) or European (EN) standards (Wilander et al., 2003). We did not use total nitrogen concentrations since there was a method change during the time period. We also disregarded chlorophyll a measurements for the evaluation of changes over time due to incomplete time series during the 1990s. We however compared chlorophyll a concentrations between the time periods 1980–1989 and 2000–2009. For further information on methods see http://www.slu.se/aquatic-sciences.
In addition to water chemical samples, we had phytoplankton samples that were taken from the epilimnetic layer, determined by a temperature profile, or from 0–8 m in case the lakes were not stratified. The samples were preserved with Lugol’s iodine solution (2 g potassium iodide [KI] and 1 g iodine [I] in 100 ml distilled water) supplemented with acetic acid. Phytoplankton counts at the finest possible taxonomic resolution (usually at species level) were made using an inverted light microscope and the modified Utermöhl technique modified (Olrik et al., 1998). We divided our phytoplankton data into the six main phytoplankton groups that occurred in our lakes: chlorophytes (Chloro), diatoms (Diatoms), chrysophytes (Chryso), cryptophytes (Crypto), dinoflagellates (Dino) and cyanobacteria (Cyano). These phytoplankton groups were also used when we evaluated changes in phytoplankton biovolume, referred to as biomass. Sampling and analyses of phytoplankton were carried out by one and the same certified laboratory and the same few people for all years. A detailed description of analyses and lake characteristics is available at http://www.slu.se/aquatic-sciences.
All our statistical tests were carried out in JMP, version 9.0. Due to the non-normal distribution of many of our variables, tested by a Shapiro-Wilk test for normality, we restricted our statistical analyses to those that are robust against non-normal distributions, i.e. non-parametric tests. To evaluate changes over time we applied the non-parametric Mann-Kendall test (Helsel and Hirsch, 1992). This test gives a measure whether long-term changes of a variable are significant (P < 0.05) or not (P ≥ 0.05). In addition to Mann-Kendall we used partial least squares regressions (PLS) to find important drivers for the total phytoplankton biomass in Lake Vänern. PLS-analyses were chosen because of the method’s insensitivity to X-variable’s interdependency and the insensitivity to deviations from normality (Wold et al., 2001). PLS is commonly used to find fundamental relations between two matrices (X and Y) where the variance in X is taken to explain the variance in Y. In PLS, X-variables are ranked according to their relevance in explaining Y, commonly expressed as VIP-values (Wold et al., 2001). The higher the VIP values are the higher is the contribution of an X-variable to the model performance. VIP-values exceeding 1 are considered as important X-variables. To make our X-variables more comparable we used a logarithmic transformation for our highly skewed nutrient and phytoplankton data before we applied our PLS models. The performance of a PLS model is expressed in terms of R2Y and Q2. R2Y is comparable to R2 in linear regressions while Q2 is a measure of the predictive power of the PLS model. The closer Q2 is to R2Y, the better and more robust is the model. We performed a permutation test for model validation. The Y-variable was permutated 50 times, new models were fitted and for each of the 50 new models R2Y and Q2 were assessed. The intercept of a linear regression of the original R2Ys and Q2s against the 50 permutated R2Ys and Q2s provided an estimate of the background correlation due to random chance. Since we always received small model background correlations we considered our models as robust models (Wold et al., 2001).
Physical, chemical and phytoplankton changes in Lake Vänern since the 1980s
Using a non-parametric seasonal Mann-Kendall test, we found all three sites in Lake Vänern showed significant changes during 1980–2011 for surface water Temp (increase, P < 0.05), pH (increase, P < 0.01), Alk (increase, P < 0.0001), SO4 (decrease, P < 0.0001), NH4-N (decrease, P < 0.01), NO3-N (decrease, P < 0.0001), PO4-P (decrease, P < 0.05), TP (decrease, P < 0.001), total phytoplankton biomass (increase, P < 0.01), Cyano (increase, P < 0.0001), Dino (increase, P < 0.0001), Chryso (increase, P < 0.001) and Chloro (increase, P < 0.001). Strongest were the changes for Alk, SO4, NO3-N, TP, Cyano, Dino, Chryso and Chloro (Figure 2). The intensity of thermal stratification, AbsF420 nm/5 cm, Si, Crypto and Diatoms were the only variables that did not show significant changes from 1980 to 2011 (P > 0.05). Comparing chlorophyll a concentrations between the time periods 1980–1989 and 2000–2009 we found significant increases at all three sites from on average 2.3 ± 1.0 to 2.9 ± 1.1 μg l−1, from 1.8 ± 0.8 to 2.4 ± 0.8 μg l−1 and from 1.9 ± 0.9 to 2.2 ± 0.8 μg l−1, respectively (non-parametric Wilcoxon test: P < 0.05). Variations in chlorophyll a concentrations are generally low in Lake Vänern and in 90% of all samples taken since 1980 they stayed below 3.3 μg l−1. During the peak eutrophication in the 1960s and 1970s chlorophyll a concentrations had reached values of up to 10 μg l−1.
Prediction of the phytoplankton biomass in Lake Vänern
The total phytoplankton biomass had a similar median value between the three different sites in Lake Vänern from 1980 to 2011 with 0.24, 0.20 and 0.20 mg l−1, respectively. The coefficient of variance, however, varied from 73% at Tärnan to 105% at Dagskärsgrund. Using partial least squares regressions with Temp, Temp diff, pH, SO4, Alk, NH4-N, NO3-N, TP, PO4-P, AbsF420 nm/5 cm and Si as X-variables the phytoplankton biomass could best be predicted for Dagskärsgrund (R2Y = 0.61). The model performance was slightly lower for Megrundet data from all three sites we were able to explain 45% of phytoplankton biomass variations in Lake Vänern. The X-variables that turned out being most important for the phytoplankton biomass differed slightly between the three sites but Alk, Temp, Si and SO4 were important at all three sites (Figure 3). TP and inorganic nitrogen were moderately important for the prediction of phytoplankton biomass in Lake Vänern at some sites but not at all sites (Figure 3).
Physical, chemical and phytoplankton changes in 13 reference lakes
Analyzing trends over time from 1988 to 2010 of 11 water physical and chemical and seven phytoplankton variables we found significant changes for all variables. However, only some of the changes changed synchronously over Sweden, which we define here as changes occurring in 7 or more lakes during 1988–2010. Among these variables were SO4 (decrease in all 13 lakes, P < 0.0001), TP (decrease in 12 lakes, P < 0.05), NO3-N (decrease in 11 lakes, P < 0.05), Crypto (increase in 11 lakes, P < 0.05), Chloro (increase in 11 lakes, P < 0.05), Chryso (increase in 10 lakes, P < 0.01), Cyano (increase in 10 lakes, P < 0.05), total phytoplankton biomass (increase in 10 lakes, P < 0.05), pH (increase in 10 lakes, P < 0.05), AbsF420 nm/5 cm (increase in 10 lakes, P < 0.05), Temp (increase in 7 lakes, P < 0.05), Dino (increase in 7 lakes, P < 0.05) and Diatoms (increase in 7 lakes, P < 0.05). The remaining lake variables, i.e. Alk, NH4-N, Temp diff and Si, showed significant trends over time in less than 7 lakes during 1988–2010.
Similarities in trends over time between Lake Vänern and reference lakes
Considering the time period 1988–2010 we observed significant changes over time (Mann-Kendall test, P < 0.05) at all three Vänern sites and in the majority of the 13 reference lakes (see above) for total phytoplankton biomass (increase), Crypto (increase), Chryso (increase), SO4 (decrease), TP (decrease), NO3-N (decrease) and pH (increase). Thus, almost all lakes showed an increase in the total phytoplankton biomass, in particular in the biomass of cryptophytes and chrysophytes despite decreasing total phosphorus and nitrate-nitrogen concentrations.
According to our hypothesis, we did not find a relationship between changes over time in the phytoplankton biomass and in total phosphorus and inorganic nitrogen concentrations in Lake Vänern. In contrast to the nutrient limitation paradigm we found that the total phytoplankton biomass, and in particular the biomass of cryptophytes and chrysophytes, has significantly increased since the late 1980s despite significantly decreasing nutrient concentrations. This pattern was coherent with the pattern that we observed in 13 nutrient-poor reference lakes. Thus, the development of phytoplankton in Lake Vänern that underwent re-oligotrophication (Wilander and Persson, 2001) now follows the development that is typical for oligotrophic lakes.
It has been suggested that the productivity in oligotrophic lakes is limited by light (Karlsson et al., 2009). The underwater light availability in nutrient-poor small Swedish lakes is presently decreasing, here indicated by increasing AbsF420 nm/5 cm values in most of the lakes. Despite this decrease we observed an increase in the phytoplankton biomass. Consequently, we suggest that light is presently not limiting the development of phytoplankton in Swedish nutrient-poor lakes. This statement applies especially for cryptophytes and chrysophytes that showed a significant increase all over Sweden, including Lake Vänern. Cryptophytes and chrysophytes usually have flagellates (Reynolds, 1984). These flagellates might be an explanation for an increase in biomass despite decreasing light availability. Flagellates allow phytoplankton to escape light constrains by maintaining an elevated position in the water column (Reynolds, 1984). Another reason for the lack of phytoplankton response to light constrains might be that many of the species that are present in our lakes are mixotrophic (Bloch and Weyhenmeyer, 2012). Such species can move in the water column and graze on bacteria as a carbon source, independently on light availability for photosynthesis (Isaksson et al., 1999). Mixotrophic phytoplankton are also usually less prone to be grazed by zooplankton due to their mobility, size and protective plates (Findlay et al., 2001).
If light presently is not the limiting factor for the development of phytoplankton in nutrient-poor Swedish lakes, we have to search for other factors. One factor which we were not able to further examine and discuss due to a lack of data is a change in the top-down control of phytoplankton. Disregarding such changes we restrict our discussion to bottom-up controls for the development of phytoplankton in our oligotrophic lakes. The most predominant change which we observed in the Swedish nutrient-poor lakes as well as in the large Lake Vänern during the last decades was a recovery from acidification, indicated by strongly decreasing SO4 and increasing pH and alkalinity. Although phytoplankton biomass does not necessarily increase with a recovery from acidification (Findlay, 2003), there are lakes where an increase has been observed (Vrba et al., 2003). Since SO4 appeared as an important variable in explaining the total phytoplankton biomass in Lake Vänern we suggest that the increase in phytoplankton biomass in nutrient-poor Swedish lakes can at least partly be attributed to a recovery from acidification.
Another variable that was highly influential on temporal variations in the total phytoplankton biomass was water temperature and the strength of stratification (Figure 3). Cyanobacteria are among the phytoplankton that are positively affected by temperature increases (Reynolds, 1997; Weyhenmeyer, 2001; Butterwick et al., 2005). Although the increase in cyanobacteria was apparent in the reference lakes and in Lake Vänern it was not strongly pronounced, probably due to the fact that cyanobacteria require relatively high phosphorus concentrations (Huszar and Caraco, 1998), prerequisites that are not given in our nutrient-poor lakes. We suggest that the strong water temperature effect in our model is mainly a result of high seasonal variations in water temperature and phytoplankton biomass. When we considered trends over time, the water temperature effect on the phytoplankton biomass was not equally apparent. We found strong and coherent increases in the phytoplankton biomass but relatively weak and not equally coherent increases in water temperatures in the other Swedish lakes. Thus, water temperatures influence the seasonal development of phytoplankton, and a variety of studies have reported on phenology changes in relation to climate change (Weyhenmeyer et al., 1999; Weyhenmeyer, 2004; Winder and Schindler, 2004; Blenckner et al., 2007; Adrian et al., 2009). When it comes to trends over time, however, other factors such as recovery from acidification need to be considered since their trends show more similarities to the trends of the phytoplankton biomass than to the trends of water temperature.
Apart from alkalinity, SO4 and water temperature, Si appeared as an important variable for the prediction of phytoplankton biomass (Figure 3). Like water temperatures, Si has a strong seasonality (Weyhenmeyer, 2009). Since also phytoplankton has a strong seasonality all variables that have a pronounced seasonality co-vary with phytoplankton. The seasonality effect disappears when we analyze trends over time. Since Si did not show a coherent trend in the Swedish lakes, this variable is not suited to explain increases in the phytoplankton biomass in Swedish nutrient-poor lakes.
From our study, we conclude that increases in the phytoplankton biomass in the large Lake Vänern follow increases in small nutrient-poor reference lakes. In these lakes, nutrient concentrations are presently not limiting for the development of phytoplankton since both phosphorus and nitrogen concentrations strongly decrease. The variables that showed most coherent changes over time to the phytoplankton biomass were variables expressing acidity. Even changes over time in water temperatures had some similarities to the changes over time in phytoplankton. Since both acidity and water temperatures are driven by large-scale processes, management of phytoplankton needs to go beyond catchment scales.
We would like to thank everyone at the Department of Aquatic Sciences and Assessment, Swedish University of Agricultural Sciences, who has been involved in sampling and analyzing chemical and phytoplankton water samples from the lakes.
Many thanks go to the Swedish Environmental Protection Agency for financing the lake inventory programme. Financial support for Gesa A. Weyhenmeyer was received from the Swedish Research Council (VR), the Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning (FORMAS), the Norwegian Research Council NORKLIMA-ECCO project and the NordForsk project DOMQUA.