Tracing alpha, beta, and gamma diversity responses to environmental change in boreal lakes
- First Online:
- Cite this article as:
- Angeler, D.G. & Drakare, S. Oecologia (2013) 172: 1191. doi:10.1007/s00442-012-2554-y
- 539 Views
Boreal lakes undergo broad-scale environmental change over time, but biodiversity responses to these changes, particularly at macroecological scales, are not well known. We studied long-term trends (1992–2009) of environmental variables and assessed α, β, and γ diversity responses of phytoplankton and littoral invertebrates to these changes. Diversity was assessed based on taxon richness (“richness”) and the exponentiated Shannon entropy (“diversity”). Almost all environmental variables underwent significant monotonic change over time, indicating mainly decreasing acidification, water clarity and nutrient concentrations in the lakes. These variables explained about 54 and 38 % of variance in regression models of invertebrates and phytoplankton, respectively. Despite this, most diversity-related variables fluctuated around a long-term mean. Only α and γ richness and diversity of invertebrates increased monotonically through time, and these patterns correlated significantly with local and regional abundances. Results suggest that biodiversity in boreal lakes is currently stable, with no evidence of regional biotic homogenization or local diversity loss. Results also show that richness trends between phytoplankton and invertebrates were widely uncorrelated, and the same was found for diversity trends. Also, within each taxonomic group, temporal patterns of richness and diversity were largely uncorrelated with each other. From an applied perspective, this suggest that long-term trends of biodiversity in boreal lakes at a macroecological scale cannot be accurately assessed without multiple lines of evidence, i.e. through the use of multiple taxa and diversity-related variables in the analyses.
KeywordsCommunity ecologyConservationMacroecologyMultiplicative partitioningTaxon richnessShannon entropy
Current rates of environmental change are unprecedented (Millennium Ecosystem Assessment 2005). Ecologists and conservation biologists are, therefore, increasingly concerned about the negative consequences for local and regional biodiversity and the provision of deriving ecosystem goods and services. Biodiversity loss is occurring widely at local and regional scales, leading to changes in the composition of biological communities (Poiani et al. 2000). Depauperate biodiversity can result from local species loss, which reduces the number of species present within habitats (decreased α richness). At larger spatial scales, such processes can lead to biotic homogenization, whereby communities become less differentiated in terms of taxonomic richness and diversity, function or genetic structure (decreased β diversity) (Olden et al. 2004; Devictor et al. 2008; Spear and Chown 2008). However, biodiversity is multifaceted (Purvis and Hector 2000), and many diversity-related variables, particularly α and β diversity, are mathematically independent from one another (Jost 2007; Tuomisto 2010). It follows that local and regional diversity must not necessarily show the same patterns in response to changing environmental conditions. Indeed, biotic homogenization has been shown to occur in response to eutrophication, despite local diversity being unchanged (Keith et al. 2009). Thus, local and regional diversity responses to environmental change must be assessed simultaneously, given that extrapolations from one scale of observation to another is fraught with the risk of making the wrong inference.
Boreal lakes are not exempt from regional environmental change that results from the combined effects of alterations in land use, catchment hydrology, acid deposition rates, and climatic change (Evans et al. 2005; Monteith et al. 2007). This causes regionally decreasing water clarity and ion concentrations (Angeler and Johnson 2012), and facilitates the range expansion of nuisance species (Angeler et al. 2012). Recent long-term studies have shown different biodiversity responses to these changes at local and regional scales. For instance, Stendera and Johnson (2008) reported an increase of local phytoplankton and littoral invertebrate diversity in eight lakes in southern Sweden with the rate of increase being stronger for invertebrates than phytoplankton. Johnson and Angeler (2010) have shown that the interannual variability in community composition of invertebrates, but not phytoplankton, correlated with the North Atlantic Oscillation winter index in the same lakes. More recently, Angeler et al. (2011), using multivariate time series modeling, found that species groups within phytoplankton and invertebrate communities responded distinctly to environmental factors; that is, species groups of invertebrates with monotonic change over time tracked mainly decreasing acid deposition and climatic variability, while phytoplankton groups with monotonic change through time responded more to decreasing water clarity and increasing trophic status. Such taxon-specific responses may not be surprising because invertebrate assemblages comprise many taxa that are sensitive to acid stress (McFarland et al. 2010), while phytoplankton, as primary producers, may be more sensitive to changes in nutrient conditions and water clarity. Goedkoop and Angeler (2011) observed trends of taxonomic homogenization of invertebrates through time in arctic and alpine lakes while such trends were not evident for phytoplankton, highlighting that these taxonomic groups also track environmental change distinctly at regional scales. These taxon-specific patterns are in agreement with those found in other studies in terrestrial (Santia et al. 2010; Axmacher et al. 2011) and aquatic (Tolonen et al. 2005; Lopes et al. 2011) environments. This clearly has implications for applied ecology because it underscores the need to infer biodiversity responses to environmental change using multiple lines of evidence (Johnson and Hering 2009; Heino 2010).
Recent advances in our understanding of diversity concepts and the ecological phenomena they address (Anderson et al. 2011; Tuomisto 2010, 2012) also highlights the need to extend inference making beyond the consideration of multiple taxa, to assess environmental change impacts from different diversity-related variables. Biodiversity has often been inferred from species distributions (presence–absence data) (Koleff et al. 2003; Magurran 2004), which ignore species dominance patterns that are an important facet of community structure. Hillebrand et al. (2008) stressed the role of species dominance in ecological processes, because ecological change is often manifested in community evenness rather than taxon richness. The most common notion about evenness in the ecological literature is that richness and evenness are independent components of diversity (Jost 2010; Tuomisto 2012), which implies that they are supposed to be uncorrelated. Inferences derived from richness versus evenness should therefore be expected to differ in many cases. Thus, metrics with different weights on species abundance in the analysis need to be studied to provide complementary but ecologically more realistic information about diversity patterns.
Here, we use long-term (1992–2009) time series data from the National Swedish lake monitoring programme to study diversity dynamics of phytoplankton and littoral invertebrates over an extended time period. Contrary to previous studies with a focus on a few lakes at small spatial scales, our main aim is to focus on temporal diversity dynamics at a broad spatial extent. Most analyses on broad-scale environmental change have focused at a rather small spatial scale (Kerr et al. 2007); however, macroecological approaches that are ideally suited to address large-scale patterns have less frequently been applied to global change problems (Kühn et al. 2008). We use multiplicative partitioning (Whittaker 1960, 1972) to reconstruct the temporal trends of taxon richness and the exponentiated Shannon–Wiener index (or Shannon entropy that considers evenness in addition to richness) for both metacommunities. Partitioning methods to study α, β, and γ components of diversity will generate ecologically realistic patterns of diversity and provide useful information for management and conservation (e.g., Stendera and Johnson 2005; Hendrickx et al. 2007; Flohre et al. 2011; Golodets et al. 2011). Given the need to consider different taxonomic groups and diversity-related variables to strengthen inference about biodiversity response to environmental change, we test the following (not mutually exclusive) hypotheses:
(1) Phytoplankton and invertebrate diversity changes vary with the scale of observation (local vs. regional) and may or may not show patterns of monotonic change; (2) both metacommunities show different qualitative and quantitative responses to environmental change. As a result, (3) biodiversity patterns of phytoplankton are uncorrelated from those of invertebrates, and (4) biodiversity patterns derived from species richness do not correlate with those of Shannon entropy within these communities.
Materials and methods
In the late 1980s, Sweden initiated a long-term monitoring program of multiple habitats and trophic levels of lakes. We chose 26 lakes from the monitoring data for this study based on the criteria of longest available time series and broadest spatial coverage (Electronic Appendix 1). These lakes are little impacted by direct human activity (Johnson 1999), and they are environmentally heterogeneous, spanning gradients in water clarity (Secchi depth, water color), nutrient conditions (concentrations of phosphorus and nitrogen), acidity status (pH, alkalinity), and lake size (Angeler et al. 2011). Samples for water chemistry, phytoplankton, and littoral invertebrate assemblages used in this study have been collected during 18 years (1992–2009) according to standardized protocols at the same sites from these lakes (see below) and counted by the same persons to reduce sampling bias as much as possible. Sampling was also carried out during the same months over the years, and although between-year variability in climatic conditions can be substantial, standardized sampling periods allowed for capturing the variability of biological processes without being confounded from the temporal variability in sampling regimes. More information is available at http://www.slu.se/vatten-miljo.
For this study, we used water quality data that were obtained from surface water samples (taken at 0.5 m depth), collected usually 4–8 times each year during the ice-free period at a mid-lake station in each lake. Water was collected with a Plexiglas sampler and kept cool during transport to the laboratory. Samples were analyzed for acidity (e.g., alkalinity, pH, and SO42− concentration), nutrient (e.g. total P, total N) and water clarity (e.g. Secchi depth) variables. All physicochemical analyses were done at the Department of Aquatic Sciences and Assessment following international (ISO) or European (EN) standards when available (Wilander et al. 2003). Because the studied lakes were relatively small (average surface area: 0.92 km2), previous research has shown that the acidity, nutrient, and water clarity conditions of the mid-lake sample are representative for other lake areas (Göransson et al. 2004).
We used phytoplankton data obtained through epilimnion samples (2–8 m depth depending on lake), which were usually taken during the second half of August of each year. Sampling was carried out in the middle of each lake using a 2-m-long Plexiglas tube sampler. Five random epilimnetic water samples were pooled to form a composite sample. A subsample was taken and preserved with Lugol’s iodine solution (2 g potassium iodide and 1 g iodide in 100 mL distilled water) supplemented with acetic acid. Phytoplankton counts and taxon identification (usually at the species level) were made using an inverted light microscope and the modified Utermöhl technique (Olrik et al. 1989). A total of 389 taxa were used in the analyses; 72 % of the taxa were identified to species level or below (variety and form), while 28 % comprised morphospecies that could not be assigned to a species. Taxa that could only be identified to family level (or above) were not included to not unduly influence the analyses.
Benthic invertebrates were collected in each lake from one wind-exposed, vegetation-free littoral habitat in late autumn (end-October–early November; end-September in the most northern, arctic lakes Abiskojaure and Jutsajaure to cover similar environmental and climatic conditions relative to the other lakes) each year. Five samples were taken within each habitat using standardized kick sampling with a hand net (0.5 mm mesh size). Each sample was taken by disturbing the bottom substratum for 20 s along a 1-m-long stretch of the littoral region at a depth of c. 0.5 m; thus, a total area of 1.25 m2 was sampled in each lake. Samples were preserved in 70 % ethanol in the field and processed in the laboratory by sorting against a white background with ×10 magnification. Invertebrates were identified to the finest taxonomic unit possible and counted using dissecting and light microscopes. For the analyses, we used 376 taxa of which 60 % were identified to species level and 40 % comprised morphospecies that deviated from the other identified species. Taxa with higher taxonomic-level resolution (family and above) were excluded from the analyses.
In this study, we calculated different metrics (total abundance, evenness, taxon richness, and exponentiated Shannon entropy) for assessing diversity dynamics. The flow chart in Electronic Appendix 2 summarizes the relationships between these metrics and the ecological rationale for use in this study. We partitioned diversity of invertebrates and phytoplankton into α, β, and γ components following Whittaker’s (1960, 1972) multiplicative formula (γ = α × β), which achieves mathematical and statistical independence of α and β diversity (Baselga 2010). We calculated two forms of α, β, and γ that represent different ecological phenomena (Tuomisto 2012): the first is based on taxon richness (i.e., the actual number of species); the second is based the exponentiated Shannon index (expH’) that is a measure of “true diversity” (also known as diversity of order 1 because geometric means of proportional species abundances are used in the calculations). Here, diversities are derived through the quantification of effective number of species; in practice, the effective number of species quantifies how many equally abundant species would give the observed mean proportional species abundance (Tuomisto 2010, 2012). Following Tuomisto (2012), we will refer to diversity based on taxon richness and expH’ as “richness” and “diversity”, respectively. Yearly α and γ richness and diversity, and total abundance were calculated in Primer 6 (Primer-E, Plymouth, UK). Tuomisto (2012) showed that the partitioning of diversity into α and β components is related to the partitioning of diversity into richness and evenness components. We therefore calculated evenness as the quotient between diversity and richness (evenness = diversity/richness). This allowed us obtaining a fraction of evenness that is unrelated to richness (Jost 2010; Tuomisto 2012).
Prior to all partitioning, diversity values were obtained through transformation of Shannon entropy values into their number equivalents through exponentiation, while no transformations were required for richness data (Jost 2007). This allowed calculating α and β components that were independent from each other and comparing richness and diversity at appropriate scales (Jost 2007; Baselga 2010). Proportional abundance values have been used for deriving diversity values to give each lake the same weight in the analyses. After calculations of α for each lake (grain: average size 0.92 km2) and regional γ (extent: 450,000 km2), we obtained β through division of γ by the averaged α from all lakes. Thus, in this study, α is the richness or diversity within each lake, and γ is the richness or diversity considering all lakes within an extent of 450,000 km2 (i.e. almost the whole of Sweden). β comprises the quotient of mean local to regional richness or diversity, i.e. it quantifies how many times as rich in effective species the entire dataset is than its constituent sampling units (individual lakes) are on average (Tuomisto 2010).
To test the conjecture that the temporal patterns of richness and diversity of phytoplankton and invertebrate differ over time (hypothesis 1), we used linear regressions were α, β, and γ richness and diversity calculated for each year comprised the dependent variables and the linear time vector (comprising the years between 1992 and 2009) the predictor variable. Assessing whether patterns of monotonic change are significant at one but not across scales of observation allows for estimating at which scale environmentally changing conditions are potentially strongest. A similar regression approach was used to test for monotonic change of environmental variables over time.
To test the conjecture that phytoplankton and invertebrate richness and diversity respond differently to environmental change, both qualitatively and quantitatively (hypothesis 2), trends of α, β, and γ richness and diversity were regressed against environmental variables. We used averaged environmental variables (water temperature, pH, yearly pH minima, electrical conductivity, alkalinity, sulphate, NH4–N, NO2 + NO3–N, total N, PO4, total P, water color, and total organic carbon) from all 26 lakes and for each year to obtain regional predictors of α, β, and γ for the studied period. α, β, and γ richness and diversity for phytoplankton and invertebrates comprised the dependent variable in the regressions and the environmental variables the predictors. We used a multimodel inference approach and tested several models to cover a broad variety of potential environmental effects on diversity dynamics. Model comparison was based on Akaike’s Information Criterion adjusted for sample size, AICc (Burnham and Anderson 2002). Models were ranked using ∆AICc, the difference in AICc between a candidate model and the model with the lowest (best) AICc.
Using the Durbin Watson statistic (Durbin and Watson 1950, 1951), we tested whether the linear models were biased through temporal autocorrelation. No serial correlation was detected for the regression models when testing for monotonic change of richness and diversity variables and their environmental correlates, and alternative modeling approaches were therefore not needed. However, serial correlation was significant when testing for monotonic change through time of environmental variables. Regressions were therefore based on generalized least square (GLS) models, which estimate model error using maximum likelihood procedures, thereby avoiding autocorrelation bias.
To test hypothesis 3 (richness and diversity trends of phytoplankton differ from those of invertebrates) and hypothesis 4 (richness trends differ from those of diversity within the taxonomic groups), non-parametric Kendall’s tau correlation analyses, which measures agreement between 2 rankings (Kendall 1938), were used. If diversity trends between components (richness vs. diversity) and taxonomic groups (phytoplankton vs. invertebrates) differ, these correlations will not be significant.
Finally, Kendall’s tau correlations were also used to correlate the temporal trends of α and γ richness with raw total abundances. This was necessary because we observed an increase of α and γ richness of invertebrates over time and therefore had to test whether this increase was due to the more individuals effect (increased detection of species with increasing abundance over time; Scheiner et al. 2011; Electronic Appendix 2). Likewise, we tested if trends of α and γ diversity were correlated with evenness and richness to evaluate the relative contributions of these independent fractions to the temporal dynamics of diversity (Electronic Appendix 2). If diversity trends correlate predominantly with richness trends, an assessment of richness trends alone could serve as a surrogate of diversity trends; however, if evenness or a combination of evenness and richness correlates with the temporal patterns of diversity, richness alone will be unable to cover a broader range of potential diversity responses to environmental change. Regression and correlation analyses were carried out in R 2.10.1 (nlme package, v.3.1-96; AICcmodavg package, v.1.07; R Development Core Team 2009).
Temporal patterns of environmental variables
Temporal patterns of abundance, evenness, richness and diversity
Similar to richness, α and γ diversity of invertebrates increased significantly through time (Fig. 3a, e). Evenness increased monotonically only regionally but not locally (Fig. 2c). Consequently, correlation of diversity patterns with its independent fractions (i.e. richness and evenness) brought different results. Local diversity correlated with local richness (Kendall’s τ = 0.525, P = 0.002) but not evenness (P > 0.05), and regional diversity correlated marginally with regional richness (Kendall’s τ = 0.31, P = 0.075) but stronger with evenness (Kendall’s τ = 0.673, P < 0.001). This highlights that diversity is dominated by different fractions (richness, evenness) over time, depending on the scale of observation.
Despite α and γ richness and diversity of invertebrates changing monotonically through time, β richness and diversity did not show the same patterns. Both β richness and diversity fluctuated around a long-term mean (Fig. 3), indicating no trends of biotic homogenization over time.
Contrary to the invertebrates, none of the phytoplankton diversity variables change monotonically through time. Abundances and evenness (Fig. 2b, d) and α, β, and γ richness and diversity (Fig. 3b, d, f) fluctuated around a long-term mean. Abundances were uncorrelated from α and γ richness (P > 0.05), deviating from the results observed for invertebrates. However, a similar finding to invertebrates was that evenness and richness contributed with different strengths to α and γ diversity: α diversity correlated similarly with richness and evenness (both components: Kendall’s τ = 0.399, P = 0.021), while γ diversity of phytoplankton correlated strongly with evenness (Kendall’s τ = 0.98, P < 0.001) but not richness (P > 0.05).
Environmental correlates of temporal richness and diversity patterns
Summary of results (adjusted R2, F ratios, P levels, corrected AIC values, and Durbin-Watson statistics) from best linear models showing associations between α, β, and γ richness and diversity and environmental variables
Intercept −0.97 ns
log(pH) 6.25**; log(SO4) −0.41(*);
log(TN) −0.61**; log(Color) 0.42*
Adj. R2 0.671; F4,13 9.663; P = 0.0007
AICc −43.09; Durbin-Watson 2.07, P = 0.33
Intercept −3.1 ns
log(Secchi) −1.23*; log(pH) 8.91**; log(Conductivity) 1.15*; log(pHmin) −4.62**
Adj. R2 0.347; F4,13 3.26; P = 0.047
AICc −39.62; Durbin-Watson 2.58, P = 0.79
Adj. R2 0.541; F1,16 21.05; P = 0.0003
AICc −13.25; Durbin-Watson 1.93, P = 0.33
Intercept 8.14 ns
log(Conductivity) 4.21**; log(TOC) 3.31*; log(pHmin) −10.66(*)
Adj. R2 0.344; F3,14 3.97; P = 0.03
AICc −23.57; Durbin-Watson 2.29, P = 0.59
Intercept −0.94 ns
log(Conductivity) 1.08**; log(SO4) −0.58*; log(TOC) −0.36(*); log(TP) 0.16*
Adj. R2 0.562; F4,13 6.452; P = 0.004
AICc −52.65; Durbin-Watson 2.56, P = 0.73
log(Secchi) −1.02*; log(pH) 9.68**;
log(Conductivity) 0.92*; log(Alkalinity) −0.37(*); log(pHmin) −3.77*
Adj. R2 0.384; F5,12 3.119; P = 0.05
AICc −42.77; Durbin-Watson 1.82, P = 0.09
log(Secchi) 2.96**; log(Conductivity) 2.88**;
log(Alkalinity)−1.33**; log(SO4) −3.16***; log(pHmin) −6.25**
Adj. R2 0.677; F5,12 8.118; P = 0.001
AICc −28.65; Durbin-Watson 2.25, P = 0.37
Mean Alpha richness
Adj. R2 0.664; F1,16 34.64; P < 0.0001
AICc −43.46; Durbin-Watson 1.92, P = 0.33
log(Secchi) −8.53**; log(TOC) −6.38*
Adj. R2 0.283; F2,15 4.35; P = 0.032
AICc 43.23; Durbin-Watson 2.95, P = 0.97
Mean Alpha diversity
Intercept −3.16 ns
log(Temperature) −0.98*; log(pH) 6.79***; log(TN) −0.74***; log(PO4) −0.19*
Adj. R2 0.796; F4,13 17.53; P < 0.0001
AICc −37.19; Durbin-Watson 2.3, P = 0.65
Congruency between phytoplankton and invertebrates
Because of the different temporal patterns between phytoplankton and invertebrate diversity measures and their distinct correlation with environmental variables, richness and diversity patterns between phytoplankton and invertebrates were mostly uncorrelated (P > 0.05). Only γ richness trends between phytoplankton and invertebrates were correlated significantly over the study period (Kendall’s τ = 0.367, P = 0.036). This highlights that the patterns of taxon congruence in response to environmental change is rather weak.
Congruency between richness and diversity
Weak patterns of congruence between richness and diversity patterns were found within the invertebrates and phytoplankton. Significant correlations were detected between α richness and α diversity of invertebrates (Kendall’s τ = 0.525, P = 0.002) and phytoplankton (Kendall’s τ = 0.399, P = 0.021) but not between β and γ richness and diversity.
This study supports the conjecture that boreal lakes undergo broad-scale environmental change. Many abiotic variables, especially those related to water clarity and acidity, showed patterns of monotonic change during the 18-year study period. The observed patterns are consistent with the abiotic changes expected for these lakes because decreasing acid deposition has been related to the recovery of organic carbon concentrations, and thus levels of water clarity, in surface waters to pre-industrial levels (Monteith et al. 2007). Also, global warming and land-use change have been suggested to influence these patterns (Evans et al. 2005; Sucker and Krause 2010).
We highlight the broad latitudinal gradient covered in this study along which the patterns and rates of environmental change can be spatially contingent, thereby causing spatial patterns in biodiversity dynamics. However, Angeler and Johnson (2012) have shown that community dynamics along this latitudinal gradient are very similar. Building on this finding, our main interest in this study was to track α, β, and γ richness and diversity changes at the macroecological scale rather than elucidating spatial patterns within it.
Despite regionally changing environmental conditions being clearly evident, the temporal patterns of phytoplankton and invertebrate richness and diversity in response to these changes showed mixed results. A significant monotonic increase over time was observed only for α and γ richness of invertebrates. The remaining diversity-related variables of phytoplankton and invertebrates fluctuated around a long-term mean. This finding supports hypothesis 1 for invertebrates but not for phytoplankton. Based on preliminary findings of previous studies showing that temporal patterns of local and regional diversity differ between phytoplankton and invertebrates (Stendera and Johnson 2008; Johnson and Angeler 2010; Goedkoop and Angeler 2011), we expected different patterns at least between α and β richness and diversity for both metacommunities. However, because α and β are mathematically independent of each other, temporal diversity trends must not necessarily be different. Despite the mixed support of hypothesis 1, which further underscores taxon-specific responses to environmental change (see below), we highlight an encouraging finding of this study. Both metacommunities showed little evidence of local species loss and a relatively stable β richness and diversity. The latter indicates that both communities do not suffer taxonomic homogenization over time at the macroecological scale of study. This finding contrasts with the results from other studies, which have identified environmental change as the main cause of biotic homogenization in, for instance, woodland plant (Keith et al. 2009), bird (Devictor et al. 2008), or ungulate communities (Spear and Chown 2008). Except for β diversity of phytoplankton, it is interesting to note that β richness of phytoplankton and β richness and diversity of invertebrates clearly bore the imprints of environmental variables that changed significantly over time. The apparent long-term stability of β richness and diversity of both metacommunities highlights that these diversity-related variables currently resist the direct (changes in water quality) and indirect effects (species invasions; Angeler et al. 2010) resulting from environmental change. However, it cannot be ruled out that taxonomic homogenization of both communities can happen in the future, when lag effects of environmental change potentially lead to disruptions of the ecological conditions that currently confer stability to biodiversity.
Given the long-term stability of many diversity-related variables studied here, we cannot clearly establish a mechanistic interpretation about the observed monotonic change of α and γ richness and diversity for invertebrates. That is, we cannot clearly determine whether monotonic changes through time were due exclusively to the association with the temporal patterns of abundances or environmental variables or a combination of both. Notwithstanding, the correlations between diversity and abundances can either be spurious due to variation in sampling efficiency or result from a real ecological phenomenon. We discard the former because identical sampling protocols and evaluation standards have been applied during the lifetime of our monitoring program. Our analyses and inference will therefore not be unduly influenced by variability in methodology. More generally, our findings support early ideas about an increase of species richness with increased sampling (Grinnell 1922), which has been referred to as a “more individuals” effect (i.e., an increased detection of species with increasing abundance over time; Scheiner et al. 2011). Stendera and Johnson (2008) observed that the rate of increase of invertebrate diversity in lakes recovering from acidification is higher than for phytoplankton. While this suggests that invertebrates recover faster than phytoplankton assemblages, the more individuals effect likely contributes to overestimate recovery success for invertebrates, because the likelihood of detecting more species increases with increased abundances. Thus, accounting for the temporal abundance structure of communities, as we did in this study, is necessary for assessing ecologically realistic biodiversity trends (Gotelli and Colwell 2001).
Regarding environmental correlates of richness and diversity patterns of both metacommunities, the models had different predictive capacity, with model fits being consistently higher for invertebrates than phytoplankton. This suggests that changes in measured abiotic niche conditions over time were more important for change of invertebrate than for phytoplankton diversity. In addition to the variance explained by environmental variables in the models, their residual variance or lack of significant correlations suggests that other unmeasured variables may also be potentially important. These variables may include biological interactions, spatial or stochastic processes, but these cannot be evaluated with the data at hand. Also, the sets of environmental variables correlating with diversity-related variables differed between metacommunities. Invertebrates were associated with acidity variables or combinations of acidity and water clarity variables. Phytoplankton patterns were correlated with variables related to ionic strength (conductivity/alkalinity) and water clarity. These results support our second hypothesis that associations of invertebrates and phytoplankton richness and diversity with environmental variables differ both quantitatively and qualitatively. Although it can be a priori expected that invertebrates should show strong association with variables related to acidity, to which they are sensitive (McFarland et al. 2010), it has been shown that changes in acidity co-vary strongly with water clarity variables (Erlandsson et al. 2008). That other variables than acidity were significant for invertebrates may therefore not be surprising. Interesting, however, was the finding that phytoplankton was mainly unrelated to nutrients, despite a recent study showing that regional but not local nutrient (phosphorus) conditions influence phytoplankton diversity in Scandinavian lakes (Ptacnik et al. 2010). This difference may be due to three (not mutually exclusive) factors. First, contrary to our study, Ptacnik et al. (2010) studied a broader longitudinal gradient consisting of ultraoligotrophic lakes in Norway and more nutrient-enriched sites in eastern Scandinavia, and they studied mainly relationships between phytoplankton genus richness and phosphorus but not other water quality variables. Second, the monitoring program on which our study builds was designed to assess recovery of lakes from acidification. Thus, lakes have been selected that are relatively pristine (i.e. little influenced by environmental stressors other than acidification). Although the lakes used in our study are heterogeneous regarding nutrient status, the resulting gradient is therefore much shorter than the one covered by Ptacnik et al. (2010). Third, we studied time series while Ptacnik et al. (2010) studied spatial aspects. From the temporal perspective, our results suggest that phytoplankton diversity dynamics may not necessarily be driven by nutrients alone, but rather by the coupled dynamics of acidity and water clarity variables in the long term. Several other studies have reported similar findings, showing also that phytoplankton communities are driven by a mix of variables related to acidity, water clarity, nutrients, and temperature (Lepistö et al. 2004; Marchetto et al. 2009; Salmaso et al. 2006; Angeler et al. 2011).
The observed trends of richness and diversity and their environmental correlates are not only useful for understanding ecological dynamics of boreal lakes in the longer term but they also provide relevant management information. Conservation ecologists have a long interest in cost-effectively monitoring biodiversity using a surrogate approach; that is, comparing diversity patterns among taxonomic groups and selecting a few communities that serve to indicate overall diversity change. Research in aquatic ecosystems has so far not been very supportive of this approach (Johnson and Hering 2009; Heino 2010), although its usefulness depends on spatial and temporal scales under study and the taxa and measures of diversity that are compared (Xu et al. 2008; Lindenmeyer and Likens 2011). In this study, we found only a low number of significant correlations between diversity variables between both metacommunities, which highlights weak taxon congruence in long-term community dynamics in the lakes studied here. This weak congruence also supports our third hypothesis, and shows that inferring long-term biodiversity responses to environmental change at macroecological scales will benefit from the consideration of multiple lines of evidence to derive appropriate management decisions.
We conclude by also highlighting the relevance of choice of diversity measures for optimizing the information gain regarding long-term biodiversity dynamics at broader spatial scales. Richness and diversity patterns were largely uncorrelated within the metacommunities, which may not be surprising given that both measures focus on different ecological phenomena (Tuomisto 2010, 2012). Richness considers the number of species while diversity covers both richness and evenness patterns. Hillebrand et al. (2008) have highlighted the importance of community evenness for ecological processes because the species dominance patterns in communities are sensitive to competition (Rajaniemi 2011), predation (Addicott 1974), dispersal (Hill et al. 2001), environmental variability (Ma 2005), and disturbance (Limberger and Wickham 2012), or a combination of these factors (Matthiessen et al. 2010a, b). While their importance has been ascertained mainly by using experimental approaches, the individual or collective role of these factors is impossible to elucidate in correlative, macroecological studies. Notwithstanding, our study shows that evenness but not richness correlated strongly with γ diversity of invertebrates and phytoplankton while α diversity of phytoplankton correlated with evenness and richness. Thus, evenness patterns are relevant for diversity dynamics at broad spatial scales but their importance varies with the scale of observation. Not only do these different imprints of evenness on diversity dynamics explain the weak metric congruence observed in this study, the results support an increasing body of evidence that deriving biodiversity patterns using richness alone can be misleading. This can potentially compromise effective management and conservation planning.
The authors thank the Swedish Environmental Protection Agency and the many people involved in the monitoring program for making the analyses of these datasets possible. This work was supported by the DYNAMO project funded from the “Oscar and Lili Lamms Minne” Foundation. Additional support from the REFRESH (Adaptive Strategies to Mitigate the Impacts of Climate Change on European Freshwater Ecosystems, contract No 244121,) project funded by the European Union under the 7th Framework Programme, Theme 6 (Environment including Climate Change) is acknowledged. We thank the reviewers for providing constructive criticism that helped to improve the paper.