The objective of this contribution is to further the understanding of long-term (37 years) changes in the composition and biomass of the phytoplankton of the Bay of Quinte (northeastern Lake Ontario), especially those changes associated with the simultaneous point-source phosphorus loading reduction/white perch winter kill of early 1978 and the establishment of dreissenid mussels in the mid-1990’s. The relatively shallow and polymictic upper bay has facilitated the ice-free period domination of the phytoplankton by meroplanktonic diatoms (especially Aulacoseira spp.); while a more balanced representation by several algal Divisions has characterized the thermally stratified, dimictic lower bay. At all three stations (upper bay, middle bay and lower bay), phytoplankton biomass declined and community similarity decreased after both the phosphorus loading and the Dreissena interventions, but the biomass changes associated with the P load reduction were greater than those associated with Dreissena establishment. Conversely, the loss of phytoplankton community similarity after Dreissena establishment was greater than that associated with P loading reduction at all three stations. The Remedial Action Plan phytoplankton objective of 4–5 mm3 l−1 (May-October mean) has been met inconsistently since the establishment of Dreissena. The post-Dreissena period, however, was also characterized by occasional very high biomass values for the potentially toxic cyanoprokaryote (blue-green alga) Microcystis, as well as by a dramatic decline in the bloom-forming blue-green Aphanizomenon, and the near extirpation of the diatoms, Tabellaria and Synedra spp. A partially synthetic phytoplankton community was constructed using data taken from three local aquatic systems (Trenton Bay, upper Hay Bay and West Lake). This might be used as a reference against which past and future changes in the upper Bay of Quinte phytoplankton can be compared and evaluated.
The phytoplankton of the Bay of Quinte has been investigated several times in the past 60 years (Tucker, 1948; McCombie, 1967; Nicholls and Carney, 1979, 1986; Nicholls and Heintsch, 1986; Nicholls et al., 1986). While much of this earlier work, including that reported under “Project Quinte” begun in 1972, was descriptive in scope, a preliminary attempt was made more recently using multivariate analyses (Nicholls et al., 2002), to determine the relative importance of changes in Bay of Quinte phytoplankton associated with the point-source phosphorus loading controls of early 1978 and the mid-1990's establishment of zebra and quagga mussels (Dreissena spp.).
Nicholls and Hurley (1989) provided evidence to suggest that the “top-down” food web effects of the white perch die-off may well have had a greater impact on phytoplankton than the P loading reductions. Further testing of this hypothesis is not pursued here. Although the P loading reduction of the winter of 1977–1978 is used as a point in time for describing phytoplankton community change, the reader should recognize that a potentially major food web disruption occurred simultaneously with the P loading reduction, and that references herein to conditions “before and after point-source P loading reductions” are made for convenience only and are not meant to suggest that effects on phytoplankton through the food web (so-called “top-down” effects) are necessarily less important than the P loading reductions.
This paper includes an update of the Nicholls et al. (2002) assessment for the upper bay, by including new data from nine more recent years and by extending the analysis to similar phytoplankton data sets collected from the middle and lower Bay of Quinte over the entire 37-year period of the project. Because Nicholls et al. (2002) had identified and assessed, with non-parametric statistical tests, the step-trends for several taxa associated with the P loading reductions in 1978, similar testing was not repeated here. Additionally, with the recent international emphasis on the inclusion of community structure and other biotic attributes in objective setting for aquatic ecosystem rehabilitation (Nielsen et al., 2003; Padisák et al., 2006), a proposal for defining the structure of an objective/reference/target phytoplankton community for the upper Bay of Quinte is advanced, furthering the initial work of Nicholls et al. (2004).
Sampling, laboratory and multivariate statistical methods
The phytoplankton sampling, analyses with inverted microscopes and multivariate statistical methods were as outlined in Nicholls et al. (2002, 2004). Some new information relevant to phytoplankton sample pooling or “recombining” is, however, important at this time.
Sample pooling (for purposes of analyzing a single sample representative of the entire May–October sampling period), for which 13 samples were consistently obtained for all of the “new” years reported here, 2000–2008, was done only for Station HB. Individual samples collected at biweekly intervals during 2000–2008 at Stations B and C were all analyzed as individual samples. An evaluation was done of this pooling or “recombination” procedure, which is intended to save on laboratory analytical costs, and included a total of 41 station-years representing samples from the four main Bay of Quinte stations: B and N in the upper bay, HB at the mouth of Hay Bay in the middle Bay of Quinte and C, near Conway in the lower bay. For each of these station-years, phytoplankton analyses were completed on both a single sample constituted from aliquots of all May–October samples for the given year, as well as on each of those samples, individually, and for which May–October means were calculated arithmetically from the individual sample results. The comparison of May to October means from the two methods revealed good agreement at the Division level with, for example, correlation coefficients for total phytoplankton, diatoms and blue-green algae of 0.966, 0.922 and 0.965, respectively. Agreement between the two methods was lower at the genus level (r = 0.21–0.75) and, in fact was not statistically significant for some rarely occurring genera.
About one-half of all Bay of Quinte phytoplankton samples, 1972–2008, were analyzed as pooled seasonal composites, with the exception of Station B, where 32 of the 37 years’ samples were analyzed individually. For the Bay of Quinte pooled samples, enumeration errors were reduced by increasing the number of “pieces” or algal units enumerated, to totals in the 500–1000 range that were expected to yield adequate precision for the totals (Lund et al., 1958).
Phytoplankton taxonomy in Project Quinte has undergone minor revisions over the years with the splitting of several genera and species. We have, however, maintained the same major groupings in order to protect the continuity and integrity of the Project Quinte phytoplankton database, thus enabling year-to-year evaluation of changes; more details were provided in Nicholls et al. (2002). Generally, the laboratory methods have remained unchanged over the years. E.C. performed most of the analyses since the beginning of the project in 1972 and had trained others responsible for a minority of sample analyses.
For multivariate analyses, the 20 “bins” (individual genera, groups of genera or Classes/Divisions) used by Nicholls et al. (2004) were used here as follows: (i) total cryptophytes, (ii) total chlorophytes (incl. tribophytes, prasinophytes, and raphidophytes), (iii) total dinophytes, (iv) total euglenophytes, (v) total chrysophytes (incl. haptophytes), (vi) Anabaena, (vii) Aphanizomenon/ Cuspidothrix/Raphidiopsis, (viii) Gloeotrichia, (ix) Microcystis, (x) Lyngbya/ Planktolyngbya/ Leptolyngbya, (xi) Oscillatoria/ Planktothrix/ Limnothrix/Phormidium/Pseudanabaena, (xii) Gomphosphaeria/Coelomoron/ Snowella, (xiii) Coelosphaerium/Woronichinia, (xiv) all remaining Cyanoprokaryota, (xv) Stephanodiscus/ Cyclostephanos, (xvi) Synedra/ Ctenophora/Oxyneis, (xvii) Tabellaria, (xviii) Aulacoseira/Melosira, (xix) Cyclotella/Stephanocyclus, and finally, (xx) all remaining diatoms.
Every phytoplankton taxon encountered during the phytoplankton enumerations completed over the 1972–2008 period was included in the multivariate analyses either in its own genus bin or a bin that grouped it with other related taxa. No taxon was double-counted (e.g. a diatom assigned to a genus bin was not included in e.g. the “remaining diatoms” bin).
The terms “biovolume” and “biomass” are used here interchangeably, assuming unit density for planktonic algal cells; i.e. 1 mm3 has a mass of 1 mg.
Synthesis of a reference community structure
We suggest that a useful way to evaluate long-term changes in phytoplankton community structure of the upper bay (Stations B and N) is to measure its similarity with a reference community that is considered more desirable than that which existed pre-P control (before 1978). Furthermore, this reference community likely should not be based on phytoplankton communities in existence after the arrival of zebra mussels owing to an associated apparent loss of certain desirable algal species and a domination by potentially toxic Microcystis, as seen in the Bay of Quinte and elsewhere in the Great Lakes (Vanderploeg et al., 2001; Nicholls et al., 2002). Other criteria used in constructing this reference phytoplankton community included (1) selection of nearby locations, thus ensuring local geology, basic mineral chemistry and land uses similar to those influencing the upper Bay of Quinte (Ecologistics, 1988), and (2) selection of shallow, well-mixed and highly flushed systems with hydrologic and physical characteristics comparable to those present in the upper bay between Belleville and Napanee, but with lower point-source loads of total P.
The rationale for developing a reference phytoplankton community derived from the phytoplankton communities of West Lake, the inner Bay of Quinte near Trenton (old Station T of Project Quinte), and upper Hay Bay (old station U-HB of Project Quinte; Nicholls, 1999) was based on the belief that no single location can best serve this purpose. The averaging of phytoplankton community structures at these locations during 1978–1987 has generated a synthetic phytoplankton community with some blended attributes of those existing at all three locations that are believed to typify a less degraded ecosystem than that characterizing the upper bay (Stations B and N) during the pre-1978 period. The rationale for including West Lake as a component of this reference was detailed in Nicholls et al. (2004) and will not be repeated here to save space. The inner bay near Trenton and the upper Hay Bay location meet all of the criteria outlined above for a reference site. In particular, point source loading of total P to the Trenton area in the inner bay (averaging about 10 kg d−1) was considerably lower than the point-source loading influencing phytoplankton at Stations B and N, where additional inputs from Belleville, Deseronto and Napanee totalled about 55 kg d−1 during the period 1978–1986 (Owen, 1989). Similarly, the upper Hay Bay location experienced insignificant point-source loads, being located in a predominantly forested and rural agricultural watershed.
May–October means for all taxa were calculated for each of the three locations listed above, which were themselves averaged to produce a “master reference row” of the 20 column data matrix, thus constituting the synthesized reference phytoplankton community structure. (Note—the mean May-October phytoplankton biomasses at the inner bay (Station T), upper Hay Bay and West Lake were 4.03, 4.93 and 1.11, respectively. Had the mean phytoplankton biomass for any of these three locations been above 5 mm3 l−1 (the upper limit of the RAP upper bay phytoplankton objective), that might have rendered it inadmissible as a candidate reference location).
The SIMPER routine [SIMilarity PERcentages (PRIMER software, as described by Clarke and Warwick (1994)] was used to determine which taxa contributed most to the between-treatment dissimilarites (where “treatments,” were the pre- and post-intervention community structures, with the division points being 1978 for P control and 1995 for the Dreissena intervention. Significant difference between treatments was determined by the ANOSIM routine (ANalysis Of SIMilarities) - run in PAST (Hammer et al., 2001) and PRIMER (Carr, 1997; Clarke and Warwick, 1994).
In order to facilitate statistical analysis of difference between the reference community and any other multi-year group of Bay of Quinte phytoplankton communities, the “master reference” row of the data matrix needed to be replicated in a fashion that provided a level of within-group variation similar to that inherent in the treatment group to which it was intended to be compared (e.g. Station B, 1972–1978 or 1978–1994). This is because the measure of statistical difference between two groups of communities (i.e. treatment and reference groups) is determined by the ANOSIM procedure, which is based on the differences between the within-group community similarities and the between-group similarities. A reference group of community similarities that has either much greater or much lower within-group similarity values than those of the treatment group will bias the test of significant difference between groups towards either statistical non-significance (if variance is high) or significance (if variance is low). The adjustment of the within-group variance in community similarity values of the reference group was achieved with 10 sets of 20 random numbers [generated in Modstat (Knodt, 1999)] over the range 0.25 to 4.0 were that were, in turn, applied to the “master reference row”, thus generating 10 new rows of reference communities in the data matrix. Because the Remedial Action Plan for the Bay of Quinte had set an upper bay May–October average phytoplankton biomass rehabilitation target of 5 mm3 l−1, each of the 20 variables included in the now 10 rows of the reference data matrix was multiplied by a single, empirically derived factor (0.61) that ensured that the highest total phytoplankton biomass value of the 10 reference communities did not exceed 5 mm3 l−1. As a consequence, the 10 generated reference means ranged between 2.80 and 4.98 with an average of 4.15 mm3 l−1.
Trends in major components
May to October average total phytoplankton biomass has fluctuated greatly over the 37-year record of Project Quinte. Dramatic declines immediately following point-source phosphorus loading reductions in the winter of 1977–1978 at the upper and middle bay sampling stations were followed by increases again during the 1980's with total biomass values occasionally comparable to the pre-P control years (Figure 1). More recently, similar sharp declines in total phytoplankton were experienced immediately following establishment of dreissenid mussel species in the Bay of Quinte, but, again, there was a rapid, but brief, increase in the late 1990's to levels similar to those seen pre-Dreissena. Thereafter, levels were lower and more stable with May–October biomass at Stations B and N averaging at about the level of the Remedial Action Plan Objective of 4–5 mm3 l−1, and Station HB in the middle bay somewhat lower, in the 2–3 mm3 l−1 range (Figure 1). May-October mean phytoplankton biomass at Station C in the lower bay has declined nearly 80%, from a 1972–1977 (pre-phosphorus control) mean of 2.5 mm3 l−l to a 2003–2008 value of 0.56 mm3 l−1.
Of the 6 or 7 major algal Divisions routinely encountered in the Bay of Quinte, by far the most important have been the diatoms (Bacillariophyta) and the blue-green algae (Cyanoprokaryota). The combined contribution from these two groups has routinely been between about 75 and 95% of the total phytoplankton biomass in the upper bay, and a slightly lower contribution in the middle bay, but a considerably lower contribution in the lower bay, where cryptomonads, dinoflagellates chlorophytes and chrysophytes and haptophytes assumed greater relative importance than in the upper and middle Bay of Quinte (as reported previously by Nicholls et al. (2002), but not illustrated here for the more recent data because of space limitations).
Of the blue-green algae, N-fixing species (predominantly of the genera Anabaena, Aphanizomenon and Gloeotrichia), have been an important part of the total community. Only in 1998 and 2008 was biomass of N-fixers less than 50%. During both of these years, and more generally, after the establishment of dreissenid mussels, the non-N-fixing genus Microcystis was atypically abundant, especially at the middle and lower bay locations (Figure 2f). Most other algal genera, especially in the middle and lower bay, registered well-defined reductions in biomass associated with the establishment of Dreissena, including bloom-forming Anabaena, Aphanizomenon, Lyngbya, Oscillatoria and Coelosphaerium (Figures 2a– e, respectively).
The dominant diatom genera Aulacoseira and Stephanodiscus generally declined over the long-term (Figures 2g and h), but the clearest example of “steps” associated with the interventions of 1978 and 1995 were in the upper bay where 1972–1977 May to October average biomass of Stephanodiscus of 1.26 mm3 l−1 fell to 0.46 mm3 l−1 for the 1978–1994 period, and 0.13 mm3 l−1 during 1995–2008. Biomass values for Synedra spp., and especially Tabellaria spp., have declined drastically since the establishment of Dreissena (Figures 2i and j).
Multivariate community structure
Based on 20 taxonomic input variables consisting of May to October averages of individual genera and groups of genera recorded in the Station B samples (see listing in Methods section), all years from 1972–2008 were correctly classified in a dendrogram into the three time periods surrounding the two interventions - point-source P control and Dreissena establishment (cophenetic correlation coefficient (ccc) = 0.759). Also, not shown here to save space, were dendrograms for Station HB with 100% classification of the years into the three time blocks (ccc = 0.883), and for Station C, wherein just one year (1990) was “incorrectly” classified with the 1972–1977 group (ccc = 0.854).
Long-term change in the Bay of Quinte phytoplankton community structure was demonstrated well by a non-metric multidimensional scaling ordination of the stations representing the upper (Stn B), middle (Stn HB) and lower (Stn C) bay after summarizing as group means representing each of the time periods referred to above. Clearly, the largest changes in community structure occurred after establishment of dreissenid mussels (Figure 3).
The relative declines in community similarity were greater after the Dreissena intervention than after the P loading intervention; the opposite was true for biomass. For example, at Station B, the percentage change in phytoplankton biomass and in community similarity associated with the P loading reduction were −47% and −1.5%, respectively, but biomass declined 21% and community similarity declined 3.7% after Dreissena establishment. Similar patterns were found in the Station HB and Station C data (Table 1).
To facilitate a more direct comparison of these “before and after” (treatment) differences, community similarity coefficients were calculated for time periods representing the two interventions, but after restricting the number of years representing each treatment to six (this number set by the available data before phosphorus load reduction, 1972–1977). Within treatments, there was a high degree of community similarity (89.3–92.8% for the P loading intervention) at all three (upper, middle, and lower) bay locations. Within treatment similarities were lower (86.1–91.8%) before and after Dreissena establishment (Table 1). In all cases (both interventions, all three stations), the phytoplankton community similarities were lower after the intervention, than before the intervention, as were the phytoplankton biomass values.
There are three aspects of the phytoplankton structural changes over time that were well visualized in the NMDS ordination of the upper bay phytoplankton (Figure 4): (1) community structure changed suddenly in association with both the P-loading and the dreissenid mussel interventions, (2) within each of the three treatment periods, (i) pre-P control, (ii) post-P control/pre-Dreissena and (iii) post-Dreissena, phytoplankton structure remained within boundaries distinctly separated from other treatment boundaries, and (3) the inter-annual variability increased progressively through the treatment periods ((i), (ii) and (iii), listed above). ANOSIM results confirmed the statistical distinctiveness of each of the treatments (Table 2), and in particular that the proposed reference community structure was statistically different from the upper bay phytoplankton of recent years (2003–2008):
It is important to note that the phytoplankton taxa most responsible for the dissimilarity between pre- and post-P control treatments were largely different from those contributing most to the dissimilarity in phytoplankton communities before and after Dreissena establishment. Only two taxa (Coelosphaerium and Oscillatoria) were on the short-lists for both treatments (Table 3). In order to achieve an “improved” phytoplankton community structure (i.e. closer to that of the reference community), the contemporary upper bay structure as defined by the 2003–2008 community, will need to include an increased representation by Cyclotella, Lyngbya, Tabellaria and Aphanizomenon and a decreased representation by Aulacoseira and Microcystis (Table 3).
Several studies have documented declines in phytoplankton following the establishment of dreissenid mussels in the Great Lakes (Nicholls and Hopkins, 1993; Holland, 1993; Barbiero et al., 2006). In a preliminary assessment of the relative changes in Bay of Quinte phytoplankton associated with P load reduction and Dreissena establishment, Nicholls et al. (2002; their Table 4) organized the phytoplankton response in five categories. Their conclusions were based on statistical tests of differences pre- and post-intervention; those tests were not repeated here because only the data for the post-Dreissena period have changed (increased to 14 years from the five years available to Nicholls et al., 2002). The general conclusion from the additional nine years of phytoplankton data from Station B in the upper bay (and largely reinforced by the phytoplankton data summaries for Station HB in the middle bay) is that the five previously identified “response categories” appear still to be valid. In particular, the post-Dreissena increases in Microcystis (the sole occupant of Category 5—taxa that “did not change after P control, but increased after Dreissena establishment”), were dramatic and especially evident in the middle and lower bay. The taxa declining most after the arrival of dreissenid mussels included Aphanizomenon, Tabellaria, Synedra and Oscillatoria. Because some of these changes have been observed at other Great Lakes locations (e.g. Vanderploeg et al. (2001) re Microcystis increase), it suggests that at least some of the ecosystem changes associated with Dreissena establishment are perhaps reproducible and predictable. Declines in many phytoplankton taxa can be explained by the direct effects of the filter-feeding habit of Dreissena spp., but increases in other taxa offer opportunities for research on the possibly secondary ecosystem influences of Dreissena as modifiers of water column light climate, and of nutrient availability. Also of significance in the updated findings from the Bay of Quinte is the observation that post-Dreissena inter-annual phytoplankton stability (as measured by year-to-year community percentage similarities) declined to levels not seen in the pre-Dreissena era. This may support the idea that an increase inter-annual instability (as inferred from increased variance in community structure) might be a useful indicator of stress (e.g. Brock and Carpenter, 2006).
As part of the Remedial Action Plan (RAP) for the rehabilitation of the 42 Areas of Concern (AOC’s) identified in the North American (Laurentian) Great Lakes, specific water quality and aquatic ecosystem objectives (so-call “delisting criteria”) for each AOC were identified. For the Bay of Quinte, the only quantitative phytoplankton objective was a May to October mean phytoplankton biovolume of 4–5 mm3 l−1 for the upper bay (Bay of Quinte RAP, 1993). No quantitative objectives relating to phytoplankton community composition were identified, beyond a general statement relating to the undesirability of algal species inducing potential toxicity and undesirable tastes and odours to municipal water supplies intended for human consumption. Using the Bay of Quinte as one of two test cases, Nicholls et al. (2004) explored a multivariate statistical approach to define a target or objective phytoplankton community based on phytoplankton data from nearby West Lake. The unexpected arrival in the Great Lakes (including the Bay of Quinte) of Dreissena spp. (zebra and quagga mussels) introduced another level of uncertainty in attempts to interpret change in the Bay of Quinte phytoplankton community (Nicholls et al., 2002).
The large changes in the Bay of Quinte phytoplankton community structure over the past many years beg the question: Is there direction to this change, and, if so, is the direction desirable (from trophic efficiency and ecosystem health perspectives)? Certainly, some of the changes associated with Dreissena establishment have been desirable (e.g. the decline in Aphanizomenon), but others such as the large increase in Microcystis are disturbing, as it relates to potential toxin production and the associated loss of more desirable food-web functional species. Key elements of aquatic ecosystem rehabilitation must include objectives for biotic communities. Such objectives must be ecologically sensible and practical. Management objectives are well documented for Great Lakes fish communities (DesJardines et al., 1995), and framework proposals have been discussed for benthic communities (Reynoldson et al., 1997). Specific objectives for Great Lakes zooplankton and phytoplankton communities, however, are in the early stages of evolution. Initiatives under the European Union's Water Framework Directive are at a similar early stage of development for these lower trophic level communities (Padisák et al., 2006).
We are reluctant to use the term “target” or “objective” for the proposed upper Bay of Quinte reference phytoplankton community because it suggests an end-point that is actively sought through “ecosystem engineering” and aggressive environmental management initiatives. While such approaches may be necessary at some future date, they have not been advocated under Project Quinte's science plan or within the Bay of Quinte RAP. Instead, Project Quinte has been more oriented towards long-term ecosystem monitoring and the resulting enhancement of understanding of aquatic ecosystem function. It is in this context that the proposed phytoplankton community reference has some value; it allows an easily visualized portrayal of community change over time that can be readily interpreted by the lay public in terms such as “improving” or “deteriorating.”
We are grateful for the helpful suggestions provided by two anonymous referees. This work was funded by the Ontario Ministry of the Environment, 1972—1999, and by Fisheries and Oceans Canada (Great Lakes Action Plan), 2000–2009.