Effects of atmospheric nitrogen deposition on nutrient limitation and phytoplankton biomass in unproductive Swedish lakes

We used chemical data (3,907 lakes) and phytoplankton biomass (chlorophyll a) data (225 lakes) from Swedish lake monitoring programs to assess the effects of atmospheric nitrogen (N) deposition on nutrient limitation and phytoplankton biomass in unproductive Swedish lakes. There was a clear north–south gradient of increasing lake concentrations of dissolved inorganic nitrogen, which was related to the pattern of atmospheric N input. On the basis of positive relationships between total phosphorus (P) concentrations and phytoplankton biomass we conclude that lakes in areas of enhanced N deposition are mainly P limited during summer. This relationship was not detected in lakes in pristine areas with low N deposition, which, together with experimental evidence from the literature, suggest possible N limitation. During summer, lakes in high N-deposition areas had clearly higher phytoplankton biomass relative to the total phosphorus concentrations compared to lakes in low N-deposition areas. Thus, in Swedish unproductive lakes, high atmospheric N input is reflected by increased lake concentrations of dissolved inorganic nitrogen and, possibly, by a shift from natural N limitation of phytoplankton to P limitation. Our results also reveal that increased N input has caused a eutrophication with higher phytoplankton biomass as the result. Increased atmospheric deposition of nitrogen (N), as a result of anthropogenic activities such as fossil fuel combustion and agricultural fertilizer application, is a common phenomenon in large areas of the world. The global rate at which reactive nitrogen is produced has doubled during the last century, which has led to an increased amount of excess nitrogen in nature (Galloway and Cowling 2002). Most aquatic research related to N deposition has concerned acidification of freshwater ecosystems. However, N is not only an acidifying component in freshwater ecosystems; it is also a nutrient, which together with phosphorus (P) regulates the growth of phytoplankton in lakes (Wetzel 2001). N limitation can be expected when the ratio of bioavailable N to bioavailable P on a mass basis is lower than 7:1, which is the mean ratio of N and P in phytoplankton nutrient demand (Redfield 1958). Thus, increased atmospheric sources of N, and especially the inorganic forms, can affect the N : P ratio of lakes and the balance between limiting nutrients. If lakes were N-limited in their natural state, excess input of N may cause a eutrophication. 1 Corresponding author ([email protected]). 2 Passed away during the course of the project. Acknowledgments We thank Gunnar Persson, at the Department of Environmental Assessment Analyses, Swedish University of Agricultural Sciences, Uppsala, Sweden, for help with data from the Swedish Lake Inventory Programs. This study was conducted with funds from the Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning. Increased N loading due to atmospheric deposition has not been regarded as a eutrophication process since P is generally thought to be the most limiting factor to phytoplankton growth in northern temperate lakes. The concept of lakes being almost exclusively P-limited is to a large extent based on the results from whole-lake nutrient enrichment experiments conducted at the Experimental Lakes Area in Ontario, Canada (Schindler and Fee 1974; Schindler 1977), and on the positive relationships found between P loading and biomass of phytoplankton in mainly eutrophic lakes (Vollenweider 1968; Schindler 1978). However, there are several reasons to question the generality of the P-limitation concept when discussing unproductive northern temperate lakes. There is increasing evidence that phytoplankton in natural unproductive lakes, situated in more pristine areas with low N deposition, can be N-limited (Jansson et al. 1996; Levine and Whalen 2001; Fenn et al. 2003 and references therein), and enrichment with N or N 1 P often causes larger responses in phytoplankton growth than P (cf., Elser et al. 1990). Moreover, P limitation in unproductive lakes may be a derived character, evolved from increased N loading from the atmosphere during the last decades, as has been illustrated for Lake Tahoe (California–Nevada) (Goldman 1988; Jassby et al. 1995). Against this background we used monitoring data to assess the effects of N deposition on nutrient limitation and phytoplankton biomass in Swedish unproductive lakes. In Sweden, there is a clear gradient, with the highest deposition of N in the southwestern part of the country (.1,500 kg km22 yr21), and the lowest in the north (,100 kg km22 yr21), 988 Bergström et al. Fig. 1. The different Swedish regions used in this study. a pattern related to precipitation and distance to continental and local sources of emissions. Atmospheric deposition increased rapidly between the 1950s and the 1980s, and has since then remained at a fairly constant level. A large part of the total N deposition is in inorganic form (www.internat.environ.se). We compiled data from 3,907 Swedish unproductive lakes and compared the nutrient composition in the lakes with the atmospheric N deposition. We also evaluated biological data from 225 unproductive lakes and compared the phytoplankton biomass (chlorophyll a [Chl a]) with N deposition. The following hypotheses were tested: (1) Atmospheric N deposition has affected the N : P ratios in Swedish unproductive lakes; (2) the phytoplankton production in natural unproductive Swedish lakes, outside areas with enhanced atmospheric N deposition, is N-limited; (3) increased atmospheric N deposition has changed the growth conditions for phytoplankton so that phytoplankton production has become P-limited; (4) enhanced atmospheric N deposition has caused eutrophication of originally N-limited lakes. Materials and methods Lake databases—Two lake databases were used, with kind permission by the Department of Environmental Assessment Analyses, Swedish University of Agriculture Sciences, Uppsala, Sweden. The first database, the Swedish National Lake Survey (SNLS), is a national inventory of the chemical and physical conditions in Swedish lakes (Bernes 1991). The Swedish Environmental Protection Agency (SEPA) has conducted the SNLS every fifth year since 1975. We used data from the two latest inventories (1995 and 2000), each comprising approximately 4,000 lakes. Lakes included in the SNLS were randomly selected from four different size classes. The lakes in the largest size class (.10 km2; 380 lakes) were all included in the SNLS. One thousand lakes in each of the three smaller-size classes were randomly selected and evenly distributed throughout the country. In the SNLS lakes, surface samples were taken (depths between 1 and 2 m) on one occasion during late autumn/winter (October to February). The second database is the Swedish Reference Lake Monitoring Program (SRLMP), also run by SEPA. Approximately 350 lakes are included in SRLMP. They are evenly distributed throughout the country, with somewhat higher numbers in southern Sweden. The SRLMP lakes vary in sizes between 0.03 and 30 km2, and the mean lake surface area per catchment area for the SRLMP lakes is 9% (Wilander 1997). The SRLMP lakes were sampled at higher intensity, 4–8 times per year, and samples were taken at different water depths. In addition to chemical and physical data, the SRLMP also includes analyses of biological variables, such as Chl a. In this study we used data from SRLMP from the time period 1996–2001. To compensate for the somewhat higher distribution of SRLMP lakes in southern Sweden, we complemented our data set with published results from different research projects (Jansson et al. 1996 and 2000; Blomqvist et al. 2001; Karlsson 2001; Sobek et al. 2003) where sampling occasions and procedures, as well as analytical procedures, are comparable to the procedures in the SRLMP lakes. The contribution from different research projects was 37 lakes distributed as: 6 lakes in region 3, 5 lakes in region 9, 10 lakes in region 12, and 16 lakes in region 13. Lake monitoring was administered by Swedish county administration, organizing lake databases following county borders (Fig. 1). As the size of the Swedish counties varies, small counties in southern Sweden situated close to each other were pooled into larger regions, to avoid large differences between total numbers of lakes within regions used in this study. In addition, two counties in northern Sweden were pooled into one region, region 11, to make this region comparable with regions 12 and 13, which both are geographically similar to region 11. We checked for possible influences of differences in lake morphometry and lake water renewal times on water chemical characteristics between different regions. Data on lake area and catchment area were available for 10–30% of the lakes in each region in the Swedish Lake Register of the Swedish Meteorological and Hydrological Institute. We used these data and calculated the mean drainage ratio (DR 5 catchment area : lake area) in each region. We found no sta989 Effects of atmospheric nitrogen on lakes Table 1. Drainage ratio (DR), specific runoff (SR), and proxy of water renewal time (SR 3 log DR) for lakes in each Swedish region (mean values with standard deviations for DR presented within parentheses). Region n log DR SR (mm yr21) Water renewal time proxy (SR 3 log DR) 1 2 3 4 5 32 39 67 32 26 1.4(0.7) 1.5(0.6) 1.5(0.6) 2 (0.6) 1.4(0.6) 400 400 300 150 100 560 600 450 300 140 6 7 8 9 10 53 47 52 103 64 1.5(0.6) 1.7(0.7) 1.5(0.5) 1.5(0.6) 1.9(0.8) 100 250 300 400 300 150 425 450 600 570 11 12 13 80 43 52 1.8(0.7) 1.7(0.5) 1.6(0.6) 500 500 500 900 850 800 tistically significant differences in DR between regions (Table 1). To assess possible impacts of water renewal times on lake water chemistry we calculated the product between mean specific runoff (mm yr21) (www.sna.se) and mean DR for each studied region as a proxy for lake water renewal time (Table 1). We found no correlation between the proxy of lake water renewal time and mean values of Chl a, total nitrogen (Tot-N), dissolved inorganic nitrogen (DIN), and total phosphorus (Tot-P) estimated for each region (see below: Chemical status in unproductive Swedish lakes); i.e., the parameters that are used in our analyses of the effects of N deposition on nutrient limitation and phytoplankton biomass in Swedish unproductive lakes (cf., below). The correlation coefficients (r2) were 0.13, 0.20, 0.10, and 0.30, respectively, and p . 0.05. The proxy of lake water renewal time was not correlated to mean conductivity (r2 5 0.14; p . 0.05), which shows that differences in lake water renewal times between regions do not to any large extent influence lake water chemistry. Chemical status in unproductive Swedish lakes—Chemical and physical data were compiled from both lake databases. Data were gathered from late autumn and winter sampling from water depths of 1–2 m. Thus, the data represent the chemical status when phytoplankton production is low, and can be regarded to represent the potentially available pools of nutrients for phytoplankton. If a lake was included in both inventory programs, data from the latest sampling occasion was chosen; i.e., all compiled lake data represent one sampling occasion in each lake in late autumn/early winter during the time period 1995–2001. Only unproductive lakes, with Tot-P and total organic carbon (TOC) concentrations of #25 mg L21 and #25 mg L21, respectively, were selected from the two databases. Tot-P in such lakes generally covaries with TOC (humic content) (Meili 1992). For each region we therefore plotted the Tot-P concentrations against the TOC concentrations to assure that increased P concentrations were related to an increasing lake humic content (cf., Meili 1992; Nürnberg and Shaw 1998). We excluded data from lakes that did not follow the Tot-P–TOC relationship to avoid lakes with unnaturally high P concentrations or TOC concentrations possibly affected by agricultural activities or sewage water. On the basis of the mentioned criteria, 200 lakes were excluded from the original database of approximately 4,100 lakes. Our data set then comprised 3,907 lakes, i.e., 4% of the total number of lakes in Sweden. We did not exclude limed lakes from our selection of lakes, as the variations in nutrient concentrations between limed and unlimed lakes were insignificant (data not shown). Thus, except for liming, the only anthropogenic influence on the selected lakes was from forestry (which is performed at similar intensity throughout the country) and from atmospheric deposition. Mean values with standard deviation of the chemical and physical data were then calculated for lakes in each of the different Swedish regions (Fig. 1). The mean chemical and physical status of lakes in each region was then compared with the atmospheric N deposition in Sweden. Atmospheric deposition—Wet deposition of DIN has been monitored monthly since 1983 at 25 stations distributed over all of Sweden within the national environmental surveillance program administered by SEPA (www.internat.environ.se). We used data (www.ivl.se) for the time period 1995–2001 and calculated the mean annual wet deposition of DIN for the different regions (Fig. 1) used in this study. Tot-N deposition is not included in regular monitoring of atmospheric deposition in Sweden. However, we estimated Tot-N deposition from the different regions by using data on Tot-N deposition in Sweden presented by the SEPA (which is performed using the so-called MATCH model) (www.internat.environ.se). Comparison of Tot-N and DIN deposition (Table 2) showed that regional variation of organic N (Tot-N minus DIN) was very small compared to DIN variation. Similar results have been reported in a large national survey in Finland where N deposition has the same sources and similar north–south gradients as in Sweden (Anttila et al. 1995). Therefore, we consider that DIN represents the major anthropogenic impact on atmospheric N deposition. P deposition is not included in Swedish monitoring programs. However, P deposition was reported to be low in Sweden (range between 10 and 100 kg km22 yr21) (Knulst 2001), and the N : P ratios (weight) of the deposition is considerably higher (range between 20 and 80) than, e.g., the Redfield ratio of phytoplankton. Similarly the DIN : TotP ratio in bulk wet deposition in Finland varied between 20 and 100 (weight) (Anttila et al. 1995) and regional variation in P deposition was small. Therefore, the regional N : P balance should have been changed mainly by increasing DIN deposition and not by deposition of organic N and P. Nutrient limitation—To evaluate spatial differences in nutrient limitation, linear regression analyses were performed between Chl a and Tot-P concentrations with the data from lakes for each of the different Swedish regions. Thus, with P-limited lakes, the Chl a concentrations should be positively related to Tot-P concentrations (cf., Vollenweider 1968; 990 Bergström et al. Ta bl e 2. C he m ic al ch ar ac te ri st ic s (3 ,9 07 la ke s) an d C hl a da ta (2 25 la ke s) of la ke s an d th e at m os ph er ic N de po si ti on fr om di ff er en t S w ed is h re gi on s du ri ng th e ti m e pe ri od 19 95 –2 00 1 (m ea n va lu es w it h st an da rd de vi at io ns w it hi n pa re nt he se s ar e gi ve n fo r al l pa ra m et er s ex ce pt fo r to ta l N de po si ti on w he re th e ra ng e is gi ve n) . R eg io n pH C on d.