Two disturbed aquatic systems, Bowland Lake, a strongly acidic lake (pH=4.9) in south-central Ontario, and the eutrophic Bay of Quinte at the northeastern end of Lake Ontario were selected as case histories for demonstration of an analytical protocol for defining and evaluating phytoplankton community rehabilitation targets. Bowland Lake was experimentally neutralized in 1983 to evaluate the efficacy of an application of a powdered limestone slurry for rehabilitating biological communities in the lake. The Bay of Quinte was identified by the International Joint Commission in 1985 as one of 42 ‘Areas of Concern’ in American and Canadian waters of the Great Lakes for which a Remedial Action Plan is required to restore beneficial uses. Phosphorus loading controls were implemented in the Bay of Quinte drainage basin in late 1977 to early 1978. In both cases, pre- and post-treatment phytoplankton community structure data were used to assess the degree of phytoplankton community impairment and the response to remediation by comparing the phytoplankton communities to those from suitable reference locations. The rationale, development and application of model phytoplankton communities intended to serve as ‘targets’ for phytoplankton community rehabilitation are presented within a multivariate framework of community structure. It is recommended that these, or similar analytical protocols based on multivariate methods for defining community structure, be more widely applied when corrective actions are taken to rehabilitate aquatic systems.

Introduction

Phytoplankton assessments have long been typical components of limnological investigations of aquatic ecosystems experiencing eutrophication, acidification or other aquatic pollution problems. As corrective management is applied to solve these problems, the response of the phytoplankton community is an important component of the total ecosystem response and can help to assess the success of remedial actions (Willén, 2001). Assessments to this time have been largely qualitative, relying on the response of so-called ‘indicator species’ for evaluation of phytoplankton community impairment and recovery (Makarewicz, 1993; Willén, 2000). While quantitative standards and targets have existed for some time for chemical pollutants (Peeters and Gardeniers, 1998; Anderson et al., 2001), quantitative targets for biological communities, and plankton communities in particular, have been slow to develop (Hartig et al., 1997; but see Reynoldson et al., (1997) for benthos and Sayer and Roberts (2001) for diatoms).

In this paper we propose an analytical protocol for defining phytoplankton community restoration targets and assessing subsequent compliance for two very different ecosystems undergoing rehabilitation. One, the Bay of Quinte, is a hardwater eutrophied embayment of Lake Ontario for which phosphorus loading control was implemented to ameliorate algal-related water quality problems; the other is an acidified lake in central Ontario that was neutralized by aerial application of a powdered limestone slurry. In each case, phytoplankton data from a nearby lake with comparable hydrologic and morphometric features, as well as biological communities that were believed to represent much less perturbed states, served as the starting point for defining target phytoplankton community structures. Also, in both cases, the phytoplankton communities were monitored over a number of years before and after the management intervention. In the Bay of Quinte, the events were complicated by invasion of zebra mussels (Dreissena spp.), but because the P-loading controls and the Dreissena invasion were separated by several years, it was possible to isolate the separate effects (Nicholls et al., 2002).

Case history background

Bowland Lake

Bowland Lake (47° 05′ N, 80° 50′ W; Figure 1), on Ontario's Precambrian Shield (surface area = 109 ha; mean depth = 7.6 m; pH = 4.9; total P = 5 μg l−1), was an acidified, oligotrophic lake prior to its experimental neutralization in August 1983. The pH of Bowland Lake increased to 6.8 immediately after liming. After this treatment, pH dropped back in subsequent years into the range of 5.8–6.5 (Gray, 1989). Other changes associated with the neutralization of Bowland Lake were described previously: for water chemistry by Molot et al. (1990a); for phytoplankton, by Molot et al. (1990b); for zoobenthos, by Keller et al. (1990); and for fish, by Gunn et al. (1990). Regarding phytoplankton changes, there were minimal changes in total biomass (total cell biovolumes and chlorophyll a concentrations were not significantly different after neutralization); however, there were major changes in relative abundance of certain taxa. In particular, the cyanophyte Rhabdoderma relinquished its pre-liming domination of the community, and the haptophyte Chrysochromulina breviturrita (a producer of offensive odours) became abundant the second year after liming (Molot et al., 1990b).

Figure 1.

Map showing the locations of Bowland Lake, Miskokway Lake and the Bay of Quinte relative to three of the Laurentian Great Lakes. The locations of the two main sampling stations in the upper Bay of Quinte (B and N) and of West Lake, the nearby reference lake, are also shown.

Figure 1.

Map showing the locations of Bowland Lake, Miskokway Lake and the Bay of Quinte relative to three of the Laurentian Great Lakes. The locations of the two main sampling stations in the upper Bay of Quinte (B and N) and of West Lake, the nearby reference lake, are also shown.

Bay of Quinte

The Bay of Quinte, at the northeastern end of Lake Ontario (Figure 1), was identified by the International Joint Commission in 1985 as one of 42 ‘Areas of Concern’ (AOC) in American and Canadian waters of the Great Lakes. Under Annex 2 of the revised (1987) bi-national Great Lakes Water Quality Agreement, specific water quality issues and ecosystem impairments in each of these AOC's were to be addressed under Remedial Action Plans (RAPs) (Hartig and Vallentyne 1989; Hartig et al., 1997). As with several of the other Great Lakes AOCs, the Bay of Quinte AOC included ‘eutrophication or undesirable algae’ and ‘degradation of phytoplankton and zooplankton populations’ on its list of impaired beneficial uses.

One typical manifestation of eutrophication is the proliferation of phytoplankton and other algae along with changes in the composition of these communities towards species that 1) are less desirable for efficient food chain function, 2) impair drinking water quality by clogging filters and producing off-flavours (objectionable tastes and odours), and 3) impair aesthetic qualities (surface ‘blooms’). By the late 1970s, in response to local concerns about Bay of Quinte water quality, phosphorus loading to the upper bay from local sewage treatment plants was decreased by nearly 70% (Minns et al., 1986b; Nicholls, 1999). The Bay of Quinte RAP Stage 2 Report, however, advocated further control of phosphorus sources and outlined 14 specific remedial actions required to achieve reductions from a number of urban and agricultural sources (Bay of Quinte RAP, 1993).

In the mid-1990's, the Bay of Quinte was among the last of the lower Great Lakes areas to be invaded by the exotic zebra mussel (Dreissena spp.). Populations elsewhere in the Great Lakes had already had dramatic effects on water clarity (Leach, 1993) and on phytoplankton communities (Holland, 1993; Nicholls and Hopkins, 1993), so the potential existed for similar effects to be manifested in the Bay of Quinte. Nicholls et al. (2002) used univariate step-trend analysis and multivariate analysis of phytoplankton community structure to demonstrate major changes in the Bay of Quinte phytoplankton community coinciding with point-source phosphorus loading controls in the late 1970s and with the establishment of zebra mussels (Dreissena spp.) nearly two decades later in the Bay of Quinte.

Methods

Sampling and phytoplankton analyses

The analyses reported here for the Bay of Quinte are based on phytoplankton data collected over a 28 year period under the auspices of Project Quinte, a multi-agency collaborative study of long-term change in the Bay of Quinte (Minns et al., 1986a). The main upper bay Stations B and N (Figure 1) were sampled weekly (1972–1982) and biweekly (1983–1999), except for Station N which was not sampled during 1983 through 1989. Methodology for sampling, laboratory analyses and data organization is described in Nicholls et al. (2002). Briefly, samples were collected as composites of the euphotic zone, fixed with Lugol's iodine solution and concentrated in the laboratory by sedimentation. Algal genera were identified and enumerated by inverted microscopy and expressed as biovolume (Nicholls and Carney, 1979).

Phytoplankon samples from Bowland Lake were collected during the ice-free periods of 1982 to 1989 at successive 2 m depth intervals. From these, a single volume-weighted composite sample was prepared which represented the euphotic zone (assumed to extend from the lake surface to a depth corresponding to 2 times the Secchi disk visibility). Sample analyses (similar to that described above for the Bay of Quinte samples) and other methodological details are given in Molot et al. (1990b).

Data reduction

For the Bay of Quinte case (including the reference lake, West Lake, see below), a 20-variable set (listed in Tables 2 and 3) was generated by combining related species and genera and deleting rare taxa. Although all phytoplankton classes were represented in the Quinte reduced data set, more emphasis was placed on the diatoms and cyanophytes because these two groups comprised between 83 and 95% (median = 90%) of total phytoplankton biovolume over the 1972 to 1999 period in the upper bay. Dominant genera were selected, but attention was also given to the indicator potential of certain other sub-dominant taxa which could reflect changing trophic states (Stoermer, 1978; Nicholls et al., 1986): N-fixing cyanophytes, the Chrysophyceae, Cyclotella, and so on. More details are provided in Nicholls et al. (2002).

For the Bowland Lake case (including Miskokway Lake, the reference lake; see below), the original listing of 107 taxa was reduced to 63 through combinations of related taxa and by deletions of rare taxa. For example, some records contained biovolumes for Peridinium inconspicuum, P. limbatum, and/or P. wisconsinense), while others were left at the genus level (Peridinium). In this and other similar cases involving other genera, the lowest common denominator was always sought (here, all species grouped under Peridinium). A few xanthophytes (e.g., Isthmochloron, Ophiocytium), choanoflagellates (Salpingoeca, Stelexomonas) and the haptophyte, Chrysochromulina, were combined with the Chrysophyceae, for convenience (as were Mallomonas and Synura of the Synurophyceae). All taxa with a frequency of occurrence of < 11% and/or with seasonal mean biovolume of < 0.01% of the total biovolume across all sampling units (SUs) were removed.

Data analyses

Each sample location-year was designated as a sampling unit (SU) where each SU represented a May–October average phytoplankton community structure as defined by the reduced data set described above. The term ‘treatment’ SUs refers to the aquatic system under study (in this case, either Bowland Lake or the Bay of Quinte), while ‘target’ SUs refer to a set of SUs developed from reference lake data to serve as a collective target or objective for determining rehabilitative progress of the treatment phytoplankton communities. Calculation of phytoplankton community similarities and follow-up analyses followed Clarke (1993, 1999), utilizing the Bray-Curtis similarity measure on 4th root-transformed phytoplankton biovolume data, cluster analysis by a group averaging algorithm and ordination by non-metric multidimensional scaling (NMDS). For the Bay of Quinte, the final application of the testing protocol included replicates for the treatment SUs based on both Stations B and N (Figure 1) on the understanding that these two stations were representative of the upper bay and that the tests for meeting the target objective community structure were to be relevant to the whole upper bay and not just the immediate areas represented by either Station B or Station N.

The statistical differences between groups of SUs (e.g., treatment vs target SUs) were determined by ‘within vs between’ cluster comparisons of community similarity utilizing the ‘Analysis of Similarities’ (ANOSIM) protocol of Clarke and Green (1988; see also Clarke, 1999). Briefly, a test statistic, R, was computed that reflected the differences in rank similarities between SUs contrasted with the differences among SUs. The statistic R ranges between −1 and +1; R = +1 only if all replicates within-treatment SUs are more similar to each other than to any other replicate from any other cluster of SUs; R = 0 if Ho is true (i.e., there is no difference in community structure between sites). The test for significant difference from 0 involved Mantel-type permutation/randomization tests, whereby R was calculated a large number of times as the SU labels were randomly re-assigned (all possible permutations, to a maximum of 10,000 times). A frequency distribution of R values was generated from this permutation of SU labels to which the original calculated R was compared for probability inference. This was run in the ANOSIM routine in ‘Plymouth Routines in Multivariate Ecological Research’ (PRIMER; Carr, 1997).

The taxa contributing most to significantly different groups of SUs were determined by the ‘Similarity Percentages’ (SIMPER) routine in PRIMER (Carr, 1997) as discussed in Clarke (1993, 1999). Briefly, the contribution of each taxon to the average dissimilarity between all pairs of inter-group SUs () was calculated. Many pairs of samples contributed to , and therefore, for a particular taxon (i), if i is large and its standard deviation (SD) is low (i.e., i /SD(δ) is large), then this taxon was an important (and consistent, because of the low SD(δ)) contributor to the dissimilarity between the groups.

Throughout the multivariate approach described above, different years within a treatment (e.g., pre-P-loading control, post-zebra mussel establishment, etc.) at each sampling site were all treated as within-site replicates. Similarly, different years at each of the reference locations (or each of the target SUs) were treated as replicates of the target ‘treatment.’

Reference lakes and the definition of target community structure

Sampling and statistical considerations

Ideally, data should be drawn from a large number of suitable reference lakes, thus creating the statistical power required to determine within defined confidence intervals the difference between treatment and target SU clusters. For phytoplankton this is not easily achieved owing to the large numbers of samples required to characterize the ‘average’ target condition in the presence of (often dramatic) seasonal changes in biomass and composition. A more feasible alternative to a set of ‘real’ reference sites, is a set of ‘constructed’ target phytoplankton communities based on measured phytoplankton community structure in just one or two suitable (minimally impacted) reference locations.

The multivariate ANOSIM test (Clarke and Green, 1988), which has some analogies to univariate analysis of variance (but is based on permutation testing; see Methods, above), is used here to determine if the phytoplankton composition of the treatment location is different from the target composition. A suitable number of replicate SUs (for reliable ANOSIM permutation tests of significance) can be created by randomly drawing values for each taxon from a range of values (e.g., ±50% of measured values from the reference location) deemed acceptable for the target community structure. Another way of achieving a similar product is to reorder randomly, or in systematic fashion, a suitable number of times the quantitative data for each taxon in the measured set, if the number of reference location SUs is greater than two. There may be some risk in creating taxon biovolumes by the random number generation process because rarely will there be evidence that any of the taxa comprising the reference lake SU, upon which the new constructed target communities are based, are independent of any other taxon (as is implied by the random assignment of values). By restricting the random number selection to values found for a few different years (or some other realistic range, e.g., ±50%, if multiple years for the reference lake are not available), disruption of any inter-taxon relationships that may have existed in the original data should be minimal. This problem is likely of minimal concern was indicated by high degrees of similarity of all nine of the Bowland Lake, and all 10 of the Bay of Quinte target communities, respectively (see Results). The use of these constructed ‘target replicates’ in a combined average fashion for objective-setting purposes is also legitimate statistically because a 1-way ANOSIM deals with cluster averages.

The ANOSIM multivariate test that determines whether or not significant differences exist between treatment and target phytoplankton compositions suffers a loss of power when the number of replicates is low. For example, in a preliminary assessment (unpublished) of the statistical difference between the upper Bay of Quinte phytoplankton community and a 2-replicate reference SU group (Balsam and Cameron Lakes from the upper Trent-Severn Waterway), the maximum number of permutations possible was only 21. In that case, only one (i.e., 4.76%) of the R values resulting from these 21 permutations was larger than the original ANOSIM R-statistic; we therefore concluded that R was statistically significant (i.e., P = 0.0476). Had the same percentage of extreme R values resulted from, for example, 10,000 permutations, (i.e., 476 permuted R values greater than the original R), then considerably more confidence could have been attached to the conclusion that the clusters were significantly different. A reference target community structure should, therefore, have enough replicates to generate a suitably high number of permutations, thus allowing a higher level of confidence in concluding when and if the treatment site community structures are no longer significantly different from those of the target.

By setting the target cluster replicates relatively high, the number of treatment cluster replicates can be relatively low (if, for example, measured data from the treatment site are scarce). A treatment site, however, also needs a certain minimal number of replicate SUs, if not for purposes of statistical power of the ANOSIM permutation tests as described above, then to generate a sufficient level of confidence in the stability of rehabilitated community structure after compliance has apparently been achieved. A remedial action intended to rehabilitate a phytoplankton community needs time to take effect and additional time to demonstrate stability of the rehabilitated community structure. Such time will undoubtedly vary with the nature of the original perturbation and the extent of the remedial measures, along with other more intrinsic factors such as interannual climate (hydrologic) variability, lake flushing rates, and others. The data requirements (number of annual SUs) must, therefore, be left somewhat open-ended, except for the statistical requirements of the tests.

An adequate number of SUs for confidence in the permutation output can be achieved by including data from two to three representative sampling stations for the treatment location over three to five years (= 6–15 SUs). It is possible that for many applications, the numbers suggested above will also be adequate for confidence in the stability of the rehabilitated phytoplankton community. The basis for the above suggested values is as follows: The number of permutations possible in a test of significance between two groups of SUs is: (n1 + n2)!/[(n1!)(n2!)], where n1 = the number of SUs in one group and n2 is the number of SU's in the other group (Clarke and Warwick, 1994). Ideally, the number of permutations should be at least 1000 for confidence in the power of the test. Therefore, if the target cluster contains 10 replicate SUs, the treatment cluster should contain at least 5 or 6 replicates for adequate power (the total number of possible permutations of the SU labels in the ANOSIM test is 8008 when n1 = 10 and n2 = 6, and is 3003 when n1 = 10 and n2 = 5).

For both the Quinte-West Lake P control case and the Bowland-Miskokway neutralization case, our proposed procedure uses SU's based on May to October mean phytoplankton composition. For others strictly following this protocol, replicates should be constituted from an adequate number of samples representing the May to October season (we suggest a minimum of monthly sampling for lakes in the north temperate climate zone). However, there is no reason why SUs could not be defined on the basis of some other time period (e.g., mid-summer peak of thermal stratification in dimictic lakes).

The Miskokway Lake reference for Bowland Lake

Phytoplankton samples collected from Miskokway Lake (45° 40′ N; 80° 14′ W; Burton Township) over the ice-free periods of 1983 to 1985 were analyzed using the same methods as those used for Bowland Lake (Molot et al., 1990b). Miskokway was selected as a ‘control’ mainly for fisheries studies on-going in another acidic lake (Trout Lake; Gunn et al., 1990). Although also acidic, Miskokway Lake's higher pH (5.6) contrasted sharply with Bowland Lake's pre-neutralization pH of 4.9 and this suggested that its phytoplankton community might serve as a target for Bowland Lake after its neutralization. Both lakes are alike in water chemistry and are situated in drainage basins of comparable geology where soils and vegetation are similar. These factors, together with their limited human access (neither is accessible by road) lend credence to Miskokway Lake's selection as a reference location for a phytoplankton target community of a rehabilitated Bowland Lake.

Initially, a range of target phytoplankton community structures was developed from the three available Miskokway Lake SUs (ice-free seasonal means of 1983, 1984 and 1985) by selecting 10 random values (Knodt, 1999) in the range of ±50% for each taxon in the list. Relative to the interannual variability in Bowland Lake phytoplankton composition, however, this procedure did not produce enough variability within the 10 target replicates for a meaningful comparison with the pre- and post-liming Bowland Lake data (the 10 replicates thus produced were located virtually on top of each other in the NMDS ordination). Clearly, another approach was needed. Added to the three original SUs for Miskokway Lake (Mi83, Mi84 and Mi85) were three new target SUs constructed as follows: the values for variables (taxa) 1 and 2, 1 and 3 and 2 and 3 were interchanged for Mi83 and Mi84, Mi83 and Mi85 and Mi84 and Mi85, respectively. This pattern was repeated for variables 5 through 7, 9 through 11, and so on to the end of the variable list (taxon number 67). Variable numbers 4, 8, 12, and on were not interchanged (but were transferred directly to the three new target SUs in the same sequence from the original Mi83, Mi84 and Mi85 SUs). Three additional replicates were generated by subjecting the original three Miskokway SUs to the same routine, but starting at variable number 2 (no interchanges at variable numbers 1, 5, 9, 13, etc.). The new target cluster consisted of nine SUs: the original three, plus an additional six generated by the systematic reassignment of variable values from the original three.

The West Lake reference for the upper Bay of Quinte

Phytoplankton data (sampling and analysis methods identical to those used in the 28 years of monitoring the Bay of Quinte, see above) were available from West Lake (Hollowell Twp., Prince Edward County), located approximately 8 km S of Station B in the upper Bay of Quinte (Figure 1), for 1980, 1985, 1986 and 1987. For a number of reasons, West Lake was suitable as a starting point for the definition of a target phytoplankton community for the upper Bay of Quinte. Morphometric, hydrologic and basic chemical characteristics of West Lake (mean depth = 2.8 m, flushing rate = approx. 4 times per year, specific conductance = 290–300 μmhos cm−1; Ontario Ministry of the Environment, Kingston, unpubl.) were comparable to values for the upper Bay of Quinte. Both the upper bay and West Lake support diverse warm-water fish communities with many common characteristics. There are, however, some notable differences in fish communities of these two systems that have been attributed to the much more extensive human influence in the Bay of Quinte watershed and the resulting eutrophication of the bay: ‘West Lake is comparable in morphometry and limnology to the bay, and presumably it supports a fish community that resembles that of the bay at an earlier stage in the eutrophication sequence’ (Hurley and Christie, 1977). As with the fish community, the macrophyte communities of West Lake have some similarities to remnant macrophyte beds in parts of the upper Bay of Quinte, but again there are many differences that are best explained by the much more advanced state of eutrophy in the upper bay (Bristow, 1987).

There are some 120 permanent residences, 80 cottages (summer use) and 22 resorts on the shores of West Lake, all of which utilize septic tank-tile bed systems for wastewater treatment. There is undoubtedly some translocation of nutrients from these systems to the lake, thus creating a comparable but less severe parallel to the upper Bay of Quinte where, despite dramatic reductions in P-loading from wastewater treatment facilities (Nicholls, 1999), total exclusion of nutrients of human origin is an unrealistic expectation. There are, therefore, compelling reasons, based on the contrasts in the degree of human influence and on analogues in physical, chemical, and biotic components of these two ecosystems, to suggest that the phytoplankton community structure of West Lake might serve as the starting point for a rehabilitation goal and the construction of a ‘target’ phytoplankton community for the upper Bay of Quinte.

A target phytoplankton composition based on West Lake was defined on the basis of the four (1980, 1985–1987) May to October means for each of the same 20 taxa used to define Bay of Quinte community structure. Six additional replicates of this target composition were created by randomly selecting additional values from within the range of values represented by the existing four years of data (Table 2, values in parentheses).

The original objective for May–October mean total phytoplankton biomass for the upper bay was 4 to 5 mm3 l− 1 (Bay of Quinte RAP, 1993). This value was determined from the total P-phytoplankton relationship determined for the period 1972 to 1981 (Nicholls et al., 1986). It was based on achievement of a May to October total P concentration of 30 μg l− 1, which was derived from mass balance calculations assuming an achievable P load reduction from major point and non-point sources. There is no obvious reason why the 4 to 5 mm3 l− 1 total phytoplankton biovolume objective is no longer useful, but it is considerably higher than the four-year average for West Lake of 1.12 mm3 l− 1 (Table 2). So, although the community composition of West Lake may be appropriate as an objective, West Lake's total phytoplankton biomass was likely too low to serve realistically as an upper Bay of Quinte objective. Our proposed solution was a scaling up of the West Lake phytoplankton community structure (keeping relative proportions of all taxa the same) to achieve the Bay of Quinte RAP total phytoplankton biovolume objective of 4.5 mm3 l− 1. This was done for each of the 20 taxa used to describe the phytoplankton community structure by first assigning 10 random replicate values (Knodt, 1999) drawn from within the boundaries of the original 4-year West Lake range for each of these taxa. The total phytoplankton biovolume for each of these 10 communities was then calculated along with a scaling factor to adjust each of the taxon biovolumes so that the overall total phytoplankton biovolume of 4.5 mm3 l− 1 was achieved.

Results and discussion

Lake neutralization (Bowland Lake)

Cluster analysis revealed that in 1983 to 1985 the phytoplankton community structures of the less acidic and untreated reference lake (Miskokway Lake) were more similar among themselves (Figure 2) than with any of the years from Bowland Lake, either before or after Bowland Lake's experimental neutralization. The lower than expected similarity (72%) between the 1982 and 1983 Bowland Lake phytoplankton communities was perhaps caused by sudden and disruptive physical and chemical effects (Molot et al., 1990b) on the 1983 community due to the whole-lake neutralization process in mid-summer 1983. The reference lake phytoplankton communities were clearly more similar to the late post-neutralization Bowland Lake communities than they were to those of the pre- and early post-neutralization years in Bowland Lake. The 3-year Miskokway Lake community cluster joined with the post-neutralization Bowland Lake communities of 1986, 1987, 1988 and 1989 at a similarity level of 70%. In contrast, the pre- (1982) and partial post-neutralization (1983) communities joined the other Bowland Lake and Miskokway Lake SUs at a much lower 60% level of community similarity (Figure 2).

Figure 2.

Dendrogram of Bray-Curtis percent similarities of May–October average phytoplankton community structures in Miskokway Lake (Mi), 1983 to 1985, and in Bowland Lake (Bo), 1982–1989.

Figure 2.

Dendrogram of Bray-Curtis percent similarities of May–October average phytoplankton community structures in Miskokway Lake (Mi), 1983 to 1985, and in Bowland Lake (Bo), 1982–1989.

The interannual similarity patterns identified above were largely maintained when the nine target SUs (which included the three original Miskokway Lake SUs and six additional constructed replicates) were included in the ordination analysis (Figure 3). The target phytoplankton community (SUs 1-9 bounded by the dashed circle in Figure 3) was most dissimilar to that characterizing Bowland Lake's first full pre-treatment year (1982), and was most similar to that of Bowland Lake during the final year of the study (1989), six years after neutralization. ANOSIM analysis revealed that the phytoplankton communities comprising the 9-SU target group were significantly different from the communities found in post-neutralization (1985-1989) Bowland Lake (Global R = 0.851; P < 0.001).

Figure 3.

NMDS ordination of the Bowland Lake (Bo) May–October average phytoplankton community structures relative to the target structures (SUs 1–9, enclosed in dashed circle). The Bowland Lake years 1985 to 1989 (enclosed by the dashed ellipse) represent the most recent of post-neutralization period for which data are available for statistical comparison with the target community structures.

Figure 3.

NMDS ordination of the Bowland Lake (Bo) May–October average phytoplankton community structures relative to the target structures (SUs 1–9, enclosed in dashed circle). The Bowland Lake years 1985 to 1989 (enclosed by the dashed ellipse) represent the most recent of post-neutralization period for which data are available for statistical comparison with the target community structures.

The SIMPER analysis revealed that 17 of the 63 taxa used to define community structure (27%) contributed more than 50% of the dissimilarity in phytoplankton community structure between the target structure and post-neutralization Bowland Lake (Table 1). Of those 17, 11 taxa need to increase their biovolumes further (i.e., all taxa with lower mean biovolumes in Bowland Lake in Table 1) in order to increase the resemblance between the Bowland Lake and target community structures. Conversely, six taxa (those listed in Table 1 with higher mean biovolumes in Bowland Lake) need to decrease further their biovolumes in Bowland Lake in order that greater similarity with the target composition be achieved. Therefore, although considerable progress appears to have been made towards achievement of a target community structure by 1989 in Bowland Lake, it was not fully achieved by the sixth year after the calcite addition. Sampling ended in 1989 and so subsequent changes are not known.

Table 1.

Mean biovolumes of the most important phytoplankton taxa in the target group (SUs 1–9 in Figure 3) and in Bowland Lake after experimental neutralization, 1985 to 1989 (Bo85–Bo89 in Figure 3) arranged in order of their percentage (%) contributions (to an accumulated total of 50%) to the average dissimilarity between the two groups. Also listed are the values of i/SD(δi), where i is the average contribution of taxon i to the overall dissimilarity between the two groups. Where ‘%’ and the ratio i-to-SD(δi) are both relatively high, taxon i is consistently an important contributor to the inter-group dissimilarity (see Methods section)

 Mean biovolume (mm3 m− 3  
   Target and Bowland Lake 
Taxon Target SUs Bowland Lake (1985 to 1989) i/SD(δi
Rhizosolenia 47.29 0.90 3.58 7.11 
Asterionella 17.59 41.51 1.77 3.90 
Synura 119.57 25.79 1.65 3.78 
Chrysosphaerella 7.95 1.53 1.49 3.51 
Monoraphidium 1.29 4.08 3.05 
Cosmarium 0.12 4.98 2.06 3.02 
Uroglena 4.96 28.69 1.39 2.80 
1C. breviturrita 2.27 19.65 1.16 2.71 
Cryptomonas 16.27 67.24 3.37 2.63 
Merismopedia 2.55 0.42 1.64 2.42 
Rhabdoderma 3.64 2.16 1.53 2.32 
Coelastrum 0.29 12.06 2.27 
Coelosphaerium 2.41 0.47 1.58 2.26 
Spondylosium 0.21 2.09 1.25 2.25 
Aulacoseira 0.81 0.68 2.44 2.24 
Chroococcus 2.56 1.29 1.22 2.10 
Aphanothece 2.18 0.84 1.74 2.01 
   Total = 50.4% 

1One of two recorded species of Chrysochromulina. The other, Chrysochromulina parva, contributed only moderately (1.39%) to the intergroup dissimilarity.

 
 Mean biovolume (mm3 m− 3  
   Target and Bowland Lake 
Taxon Target SUs Bowland Lake (1985 to 1989) i/SD(δi
Rhizosolenia 47.29 0.90 3.58 7.11 
Asterionella 17.59 41.51 1.77 3.90 
Synura 119.57 25.79 1.65 3.78 
Chrysosphaerella 7.95 1.53 1.49 3.51 
Monoraphidium 1.29 4.08 3.05 
Cosmarium 0.12 4.98 2.06 3.02 
Uroglena 4.96 28.69 1.39 2.80 
1C. breviturrita 2.27 19.65 1.16 2.71 
Cryptomonas 16.27 67.24 3.37 2.63 
Merismopedia 2.55 0.42 1.64 2.42 
Rhabdoderma 3.64 2.16 1.53 2.32 
Coelastrum 0.29 12.06 2.27 
Coelosphaerium 2.41 0.47 1.58 2.26 
Spondylosium 0.21 2.09 1.25 2.25 
Aulacoseira 0.81 0.68 2.44 2.24 
Chroococcus 2.56 1.29 1.22 2.10 
Aphanothece 2.18 0.84 1.74 2.01 
   Total = 50.4% 

1One of two recorded species of Chrysochromulina. The other, Chrysochromulina parva, contributed only moderately (1.39%) to the intergroup dissimilarity.

 

Eutrophication control (Bay of Quinte)

For the Bay of Quinte, both cluster analysis and NMDS ordinations revealed clear separation of SUs into the pre- and post-P control, pre- and post-Dreissena treatments (see Nicholls et al., 2002 for details). The inclusion here of the four original West Lake reference phytoplankton SUs with the Bay of Quinte data set did not alter those basic patterns. There are some further insights to be gained from the NMDS ordination of phytoplankton community structures for Bay of Quinte Station B compared with those for the West Lake reference location (Figure 4). It clearly illustrates that Station B phytoplankton communities of the pre-Dreissena and post-P control era (1978–1994) were very different from those found in West Lake (Figure 4). These data also suggest that the phytoplankton community structures of the post-Dreissena period (1995–1999) were also quite different from those of West Lake. Had the post-Dreissena phytoplankton community been structured more like that found in West Lake, then the post-Dreissena clusters for Station B would have been positioned between the West Lake cluster and the group II cluster in Figure 4. The 1995 SU appears in this NMDS ordination as a transitional community reflecting the early effects of zebra mussel establishment.

Figure 4.

NMDS ordinations of Bray-Curtis percent similarities of phytoplankton communities for Bay of Quinte Station B and the West Lake reference location. Average phytoplankton community structure in SU groups bounded by the circles, ellipses and rounded polygons (including dashed lines) are statistically different from other groups (ANOSIM analyses, P < 0.05).

Figure 4.

NMDS ordinations of Bray-Curtis percent similarities of phytoplankton communities for Bay of Quinte Station B and the West Lake reference location. Average phytoplankton community structure in SU groups bounded by the circles, ellipses and rounded polygons (including dashed lines) are statistically different from other groups (ANOSIM analyses, P < 0.05).

Quantitatively, the dissimilarity between West Lake phytoplankton communities and those of Station B in 1978 to 1994 was caused mainly by much lower biovolume of virtually all West Lake taxa used to define the community structure. The only exceptions were slightly higher biovolume of Cyclotella, Dinophyceae and Chrysophyceae in West Lake (compare ‘mean biovolume’ columns for West Lake and Stn B, Group II in Table 2). The most important contributors to the West Lake Station B group II dissimilarity were Aphanizomenon + Raphidiopsis (13.74%), Anabaena (9.85%) and Oscillatoria (6.94%) (Table 2). The dissimilarity between West Lake and Station B group III (post-Dreissena) phytoplankton communities (compare ‘mean biovolume’ column for West Lake and Stn B, group III in Table 2) was caused by higher biovolume of several taxa in West Lake, including Dinophyceae, Euglenophyceae, Coelosphaerium, Lyngbya, remaining cyanophytes, Cyclotella, Synedra and remaining diatoms. There were also some taxa that were much less abundant in West Lake, including Anabaena, Aphanizomenon + Raphidiopsis, Microcystis and Aulacoseira, which together contributed a total of 46% of the dissimilarity between the West Lake and the post-Dreissena Station B phytoplankton communities (Table 2). It is clear therefore, that the post Dreissena phytoplankton composition in the Bay of Quinte is not necessarily ‘desirable’ in that it does not show a clear transitional tendency towards the West Lake structure, which is presumably more desirable; see below.

Table 2.

Mean biovolume (and range in parentheses) of West Lake phytoplankton taxa (May–October of 1980, 1985–1987) and for Bay of Quinte Station B post-P-control, pre-Dreissena (Group II) and post-Dreissena (Group III), as identified in Figure 4. Also shown is the percentage (%) contribution of each taxon to the average dissimilarity between the group pairs (i.e. West lake compared with Bay of Quinte Stations B before and after Dreissena establishment). The comparisons also include the values of i/SD(δi), where i is the average contribution of taxon i to the overall dissimilarity between the two groups. Where ‘%’ and the ratio i-to-SD(δi) are both relatively high, taxon i is an important and consistent contributor to the inter-group dissimilarity (see Methods section)

 Mean Biovolume1 (mm3 m− 3    
    West L. vs Stn B Group II West L. vs Stn B Group III 
Taxon West Lake Stn B Group II Stn B Group III i/SD(δii/SD(δi
Dinophyceae 138.70 (73.8–201.3) 135.7 46.7 1.15 1.52 1.74 4.04 
Cryptophyceae 66.30 (22.8–149.1) 243.9 204.2 2.00 3.80 1.81 4.55 
Euglenophyceae 2.35 (0–4.3) 3.4 1.3 1.44 2.83 1.46 3.81 
Chrysophyceae 77.81 (33.0–142.6) 73.1 86.3 1.35 1.38 1.48 1.70 
Chlorophyceae 157.59 (56.0–240.1) 294.8 183.5 1.21 2.20 1.42 2.41 
Anabaena 39.65 (33.0–43.8) 1026.3 439.0 5.24 9.85 3.58 8.48 
Aphanizomenon Raphidiopsis 1.00 (0–4) 548.7 68.0 4.02 13.70 2.68 10.26 
Coelosphaerium 5.58 (1.0–17.4) 62.1 2.8 1.94 4.15 1.43 2.17 
Gomphosphaeria 4.97 (0–10.0) 1.5 0.3 1.49 3.46 1.49 4.65 
Microcystis 29.87 (13.0–61.3) 82.6 698.9 1.53 2.48 1.29 8.97 
Lyngbya 17.54 (0–31.1) 113.4 0.3 1.28 4.93 1.72 5.97 
Oscillatoria 0.80 (0–2.8) 76.2 4.5 2.74 6.94 1.45 3.25 
Gloeotrichia 1.2 1.05 2.77 
Remaining cyanophytes 60.15 (19.0–103.7) 149.6 25.1 1.45 3.12 1.30 3.29 
Aulacoseira 214.26 (50.0–291.7) 3673.0 4117.5 2.94 12.61 4.10 18.36 
Cyclotella 25.73 (1.0–72.7) 24.6 8.8 1.33 2.22 1.41 3.11 
Stephanodiscus 23.10 (6.5–41.0) 572.1 104.7 3.76 8.45 1.61 4.25 
Synedra 29.29 (1.0–55.4) 69.7 20.8 1.24 2.96 1.34 3.3 
Tabellaria 69.5 4.4 7.08 8.98 0.79 3.06 
Remaining diatoms 223.46 (173.0–263.0) 321.9 192.0 1.03 4.37 1.09 1.62 

1For West Lake, the range over the four available years, 1980 and 1985 to 1987, are also given in parentheses. The mean total phytoplankton biovolume (and range) for West Lake = 1.12 (0.7–1.4) mm3 l−1.

 
 Mean Biovolume1 (mm3 m− 3    
    West L. vs Stn B Group II West L. vs Stn B Group III 
Taxon West Lake Stn B Group II Stn B Group III i/SD(δii/SD(δi
Dinophyceae 138.70 (73.8–201.3) 135.7 46.7 1.15 1.52 1.74 4.04 
Cryptophyceae 66.30 (22.8–149.1) 243.9 204.2 2.00 3.80 1.81 4.55 
Euglenophyceae 2.35 (0–4.3) 3.4 1.3 1.44 2.83 1.46 3.81 
Chrysophyceae 77.81 (33.0–142.6) 73.1 86.3 1.35 1.38 1.48 1.70 
Chlorophyceae 157.59 (56.0–240.1) 294.8 183.5 1.21 2.20 1.42 2.41 
Anabaena 39.65 (33.0–43.8) 1026.3 439.0 5.24 9.85 3.58 8.48 
Aphanizomenon Raphidiopsis 1.00 (0–4) 548.7 68.0 4.02 13.70 2.68 10.26 
Coelosphaerium 5.58 (1.0–17.4) 62.1 2.8 1.94 4.15 1.43 2.17 
Gomphosphaeria 4.97 (0–10.0) 1.5 0.3 1.49 3.46 1.49 4.65 
Microcystis 29.87 (13.0–61.3) 82.6 698.9 1.53 2.48 1.29 8.97 
Lyngbya 17.54 (0–31.1) 113.4 0.3 1.28 4.93 1.72 5.97 
Oscillatoria 0.80 (0–2.8) 76.2 4.5 2.74 6.94 1.45 3.25 
Gloeotrichia 1.2 1.05 2.77 
Remaining cyanophytes 60.15 (19.0–103.7) 149.6 25.1 1.45 3.12 1.30 3.29 
Aulacoseira 214.26 (50.0–291.7) 3673.0 4117.5 2.94 12.61 4.10 18.36 
Cyclotella 25.73 (1.0–72.7) 24.6 8.8 1.33 2.22 1.41 3.11 
Stephanodiscus 23.10 (6.5–41.0) 572.1 104.7 3.76 8.45 1.61 4.25 
Synedra 29.29 (1.0–55.4) 69.7 20.8 1.24 2.96 1.34 3.3 
Tabellaria 69.5 4.4 7.08 8.98 0.79 3.06 
Remaining diatoms 223.46 (173.0–263.0) 321.9 192.0 1.03 4.37 1.09 1.62 

1For West Lake, the range over the four available years, 1980 and 1985 to 1987, are also given in parentheses. The mean total phytoplankton biovolume (and range) for West Lake = 1.12 (0.7–1.4) mm3 l−1.

 

The main response to P-loading control was a rapid reduction in total phytoplankton biovolume achieved by a general decline in most important taxa, except for a tendency for some oligotrophic indicators (Cyclotella, chrysophytes) to increase or at least not decline to as great an extent. Other details of the shift in composition between group I and II (pre- and post-P control) and the later shifts in phytoplankton community structure associated with the establishment of zebra mussels are discussed in detail in Nicholls et al. (2002).

The new target phytoplankton composition based on the West Lake data was evaluated against the phytoplankton community composition in the upper Bay of Quinte (Stations B and N combined), before and after the establishment of zebra mussels (Figure 5). Despite the outlying position of SU N96, three clearly defined groups were discernable. Group I contained all 10 of the synthesized target phytoplankton communities; Group II contained all of the pre-Dreissena phytoplankton communities for both Stations B and N, and Group III contained all of the post-Dreissena phytoplankton communities for Stations B and N. The one-way ANOSIM procedure generated statistically significant R statistics for all combinations of group comparisons (groups I and II, R = 0.996; group I and III, R = 0.859; group II and III, R = 0.807, with N96 included in group III). Not only are the phytoplankton communities highly dissimilar in these groups, but it is important to note that the alterations associated with the establishment of zebra mussels did not result in any progress towards a community structure more like the target group. The changes required to achieve the target composition now (post-Dreissena) are at least as great as they were before the arrival of zebra mussels. Specifically, the average dissimilarity between groups I and II was 25.0%, groups I and III, 30.7%, and groups II and III, 27.8%.

Figure 5.

Clustering (a) and NMDS ordination (b) of Bray-Curtis percentage similarities of Project Quinte Stations B and N 1989–1999, including 10 replicates (a–j, Group I) of reference ‘target’ phytoplankton community based on West Lake data. Station N for 1996 was an obvious outlier for several reasons including 1) the lowest ever chlorophyte, Aphanizomenon + Raphidiopsis and Stephanodiscus biovolumes, 2) 2nd highest ever Microcystis biovolume and 3) the only year when Gloeotrichia contributed to the community composition.

Figure 5.

Clustering (a) and NMDS ordination (b) of Bray-Curtis percentage similarities of Project Quinte Stations B and N 1989–1999, including 10 replicates (a–j, Group I) of reference ‘target’ phytoplankton community based on West Lake data. Station N for 1996 was an obvious outlier for several reasons including 1) the lowest ever chlorophyte, Aphanizomenon + Raphidiopsis and Stephanodiscus biovolumes, 2) 2nd highest ever Microcystis biovolume and 3) the only year when Gloeotrichia contributed to the community composition.

Aphanizomenon + Raphidiopsis, Anabaena and Tabellaria were responsible for 28% of the average dissimilarity between groups I and II (Table 3). Taxa contributing most to the average dissimilarity between groups I and III were the Dinophyceae, Aulacoseira, and Lyngbya (26%, Table 3). The changes needed in phytoplankton composition of the upper bay to bring it more in line with the target community structure include large decreases in Aulacoseira and Microcystis, smaller decreases in Anabaena and Aphanizomenon + Raphidiopsis and Gloeotrichia and modest increases in most other taxa, especially chlorophytes and chrysophytes, certain cyanophytes (Coelosphaerium, Gomphosphaeria, Lyngbya) and diatoms (Cyclotella and remaining diatoms).

Table 3.

Mean biovolume of the 20 phytoplankton taxa used to define the synthesized ‘target’ phytoplankton community (Group I) and those found at upper Bay of Quinte Stations B and N during the pre-zebra mussels years 1989–1994 (Group II) and post-zebra mussel years 1995–1999 (Group III—see also Figure 5) and the percentage (%) contribution by of each taxon to the average dissimilarity between the target composition and the average composition of other two groups. Also listed are the values of i/SD(δi), where i is the average contribution of taxon i to the overall dissimilarity between the two groups. Where ‘%’ and the ratio i-to-SD(δi) are both relatively high, taxon i is an important and consistent contributor to the inter-group dissimilarity (see Methods section)

 Mean Biovolume1 (mm3 m− 3Groups I and II Groups I and III 
Taxon Group I Group II Group III i/SD(δii/SD(δi
Dinophyceae 489.4 142.5 31.8 2.58 4.18 2.20 8.87 
Cryptophyceae 363.1 228.9 171.9 1.66 1.88 1.56 2.50 
Euglenophyceae 5.4 1.0 1.78 4.17 1.67 3.68 
Chrysophyceae 408.0 81.5 66.1 2.91 5.11 3.02 5.37 
Chlorophyceae 542.3 270.3 157.7 1.68 2.69 1.82 4.25 
Anabaena 160.9 1618.7 293.8 2.51 8.46 1.68 2.26 
Aphanizomenon + Raphidiopsis 9.3 732.7 47.3 4.43 11.07 1.59 2.99 
Coelosphaerium 45.4 60.8 1.7 1.40 1.83 2.29 5.73 
Gomphosphaeria 31.3 1.6 3.00 6.75 4.87 6.96 
Microcystis 145.0 66.3 204.6 1.61 2.53 0.99 5.19 
Lyngbya 69.4 127.9 0.2 1.58 2.48 4.83 8.34 
Oscillatoria 4.8 56.5 2.2 1.78 4.17 1.47 2.91 
Gloeotrichia 26.3 0.57 2.30 
Remaining cyanophytes 280.8 16.7 33.7 3.15 7.4 2.05 6.54 
Aulacoseira 694.3 2823.4 3868.9 1.88 7.03 3.08 8.82 
Cyclotella 180.3 8.8 6.7 2.77 6.83 2.45 7.35 
Stephanodiscus 90.6 801.1 74 1.79 6.43 0.99 2.81 
Synedra 80.1 39.3 10.4 1.41 5.67 1.70 6.18 
Tabellaria 70.3 2.2 2.98 8.59 0.50 1.06 
Remaining diatoms 899.7 461.7 196.7 1.13 3.34 2.56 5.88 
 Mean Biovolume1 (mm3 m− 3Groups I and II Groups I and III 
Taxon Group I Group II Group III i/SD(δii/SD(δi
Dinophyceae 489.4 142.5 31.8 2.58 4.18 2.20 8.87 
Cryptophyceae 363.1 228.9 171.9 1.66 1.88 1.56 2.50 
Euglenophyceae 5.4 1.0 1.78 4.17 1.67 3.68 
Chrysophyceae 408.0 81.5 66.1 2.91 5.11 3.02 5.37 
Chlorophyceae 542.3 270.3 157.7 1.68 2.69 1.82 4.25 
Anabaena 160.9 1618.7 293.8 2.51 8.46 1.68 2.26 
Aphanizomenon + Raphidiopsis 9.3 732.7 47.3 4.43 11.07 1.59 2.99 
Coelosphaerium 45.4 60.8 1.7 1.40 1.83 2.29 5.73 
Gomphosphaeria 31.3 1.6 3.00 6.75 4.87 6.96 
Microcystis 145.0 66.3 204.6 1.61 2.53 0.99 5.19 
Lyngbya 69.4 127.9 0.2 1.58 2.48 4.83 8.34 
Oscillatoria 4.8 56.5 2.2 1.78 4.17 1.47 2.91 
Gloeotrichia 26.3 0.57 2.30 
Remaining cyanophytes 280.8 16.7 33.7 3.15 7.4 2.05 6.54 
Aulacoseira 694.3 2823.4 3868.9 1.88 7.03 3.08 8.82 
Cyclotella 180.3 8.8 6.7 2.77 6.83 2.45 7.35 
Stephanodiscus 90.6 801.1 74 1.79 6.43 0.99 2.81 
Synedra 80.1 39.3 10.4 1.41 5.67 1.70 6.18 
Tabellaria 70.3 2.2 2.98 8.59 0.50 1.06 
Remaining diatoms 899.7 461.7 196.7 1.13 3.34 2.56 5.88 

Phosphorus loading controls apparently resulted in an upper Bay of Quinte phytoplankton community that was more like that of the target community, but as of 1994 still required considerable changes in the relative abundances of several key taxa before the upper bay phytoplankton could be considered ‘on target.’ The subsequent arrival of zebra mussels was associated with a sudden and dramatic alteration of this community. The more recent (1996–1999) community structure was very unlike the pre- and post-P control communities and was widely divergent from the target community structure. It is not known if the proposed phytoplankton community objective based on West Laker will ever be achievable as long as zebra mussels are a major component of the Bay of Quinte ecosystem.

An additional concern relates to the possibility of unforeseen perturbations acting on the treatment communities after the target community structure has been defined and which may impact the rate of recovery of the treatment community. In our case for the Bay of Quinte, the target phytoplankton community structure was defined on the basis of communities found in West Lake under 1980's climatic conditions. The 1990s were characterized by generally higher mean air and water temperatures in the upper Bay of Quinte (Belleville Utilities Commission, unpublished data for the Belleville municipal water intake). If we are now in a new (warmer) climate regime, a phytoplankton community target based on communities defined under conditions surrounding the lower water temperatures of the 1980s may not be achievable. Clearly, more needs to be learned about how temperature influences phytoplankton community structure before these climate change effects can be quantified separately from the rehabilitative affects of phosphorus loading controls.

Comparisons with other approaches

The approach we used for generating a group of nine target SUs from an original set of three Miskokway Lake SUs (rearrangement and distribution of originally measured variable (taxon) values among a set of six replicates of the synthetic target SUs) is perhaps somewhat arbitrary, but it is highly regimented and repeatable. A different method for generating the target SUs was used for the Bay of Quinte (based on random selection of variable values from within prescribed ranges found in four years of West Lake data) because the random-within-range approach used for the Bay of Quinte did not produce a workable target group from the three original target Miskokway SUs owing to very low inter-SU variability (they were essentially piled on top of each other in the NMDS ordination). Two conclusions arise from this exercise. The first is that no single method for generating synthetic target community structures may be applicable in all circumstances. There undoubtedly are other approaches (in addition to the two we used) that are equally suitable for generating target SUs. Second, whatever approach is adopted for generation of target community structure, it is important to apply an identical level of taxonomic aggregation (whether class, family, genus or species) to both the treatment and target community structures. It is important to achieve a level of separation of the target SU replicates in the NMDS ordination (i.e., inter-SU dissimilarity) that is comparable to that characterizing the set of treatment SUs. The scatter among the target SUs within the 2-dimensional NMDS ordination determines the size of the envelope defining their group boundary, and hence determines to some extent how ‘easily’ or quickly the rehabilitated SUs enter the target zone. We have discussed some quantitative criteria relating to minimum numbers of SUs within target and treatment groups for assessing statistical significance of the change in similarity between treatment and target community structures over time. Additionally, there is a certain amount of intuition required to assess whether or not the product of the target SU generation exercise is realistic. In our case, the resulting target SUs for both the Bowland Lake and the Bay of Quinte cases were grouped with a moderate degree of inter-SU separation that was consistent with the degree of inter-annual differences in the treatment SUs. All indications were that these synthetic SUs represented reasonable collective ‘targets’ for their respective cases.

The goal of the Bay of Quinte RAP is: ‘To restore and maintain the trophic status of the Bay of Quinte so that phytoplankton densities are reduced, taste and odour problems in the drinking water supplies are improved and the trophic status of the bay is similar to 1930's conditions’ (Bay of Quinte RAP, 1993). This overall goal of the Bay of Quinte RAP is heavily oriented to phosphorus control and the expected response of phytoplankton. At the time the goal statement was formalized, it was understood that phosphorus concentrations set the upper limits on phytoplankton production, and thus were important determinants of the quantity and species composition of the phytoplankton of the Bay of Quinte. In turn, the species and densities of algae present determined the fate of many other biotic components (rooted macrophytes, secondary producers, fish community characteristics and human impacts relating to aesthetics, potable water supply and recreation).

One clear objective of phosphorus control and other remedial measures implemented by the Bay of Quinte RAP was to reduce the May to October mean phytoplankton densities to an interim target level of 4 to 5 mm3 l− 1 (Bay of Quinte RAP, 1993). Other less clear objectives were related to the restoration of a more ‘normal’ phytoplankton community structure. In this paper we believe we have strengthened and imparted a desirable level of objectivity into this proposed new phytoplankton objective. As presently constituted, the Bay of Quinte RAP objective for phytoplankton community restoration has been presented only in terms broadly relating to declines in total phytoplankton biomass: ‘Reduce the average algal density from an existing level of between 7 and 8 to a new range between 4–5 mm3ll’ (Bay of Quinte RAP, 1993). Although not explicitly described, this objective was set in the belief that other associated benefits relating to less severe algae blooms, lower incidence and intensity of algae-induced offensive tastes and odours in water supplies, more efficient food web function, improved water clarity and restoration of aquatic macrophytes, would also accrue as a result of the lower total phytoplankton biomass.

Multivariate approaches, utilizing benthic invertebrate communities to assess environmental damage (Clarke, 1993) or to assess compliance with sediment quality objectives (Reynoldson et al., 1995), are precedents for the approach taken here, wherein a community structure is defined as an objective for rehabilitation purposes. For benthos, a multivariate method for testing for compliance was advocated by Reynoldson et al. (1997), but their ‘reference condition’ was defined on the basis of data from a suite of ideal habitats (their BEAST model— BE nthic A ssessment of Sedimen T). Their evaluation is similar to the Clarke (1993) protocol in that NMDS ordination is an integral part of the display of treatment and reference SU similarities; however, the BEAST model predicts group-membership for a treatment site using a discriminant function analysis and places probability ellipses around replicates in NMDS ordination space. This differs from the ANOSIM approach of Clarke (1993, 1999) as it is used here for the Bowland Lake and Bay of Quinte cases. Unlike the BEAST protocol, the ANOSIM procedure does not use the NMDS ordination results directly (these may be a source of distortion if ordination stress is relatively high). Our purpose for the NMDS ordinations was simply to display the group similarities, recognizing that because stress was relatively low in all cases, these ordinations were good representations of the inter-SU similarities. ANOSIM is an a priori exercise wherein testing is performed on the rank similarities (original calculated data, not ordinated data) of SUs assigned to groups on the basis of hypotheses to be tested, not on the basis of groupings in an ordination plane or as results from a cluster analysis. Regardless of the differences in approach, it is clear that multivariate methods have value in setting biotic community objectives.

Some may suggest that multivariate methods on the whole are too complex to be expected to gain widespread use in ecosystem assessments and that managers and the lay public may have difficulty in understanding the products of such analytical protocols (Gerritsen, 1995). It is our contention, however, that although the multivariate methods used here are indeed complex, the end products (e.g., the NMDS ordinations) used to display the restructuring of the phytoplankton community over time and its increasing resemblance to defined target communities, are very clear, simple and easily conveyed to water managers and most members of the informed public. The multivariate methods described here were specifically employed to measure the significance of shifts in community composition and for the clarity with which these shifts could be illustrated (Clarke, 1993, 1999). The untrained observer need only understand that the distance between SUs in the NMDS ordination is related to community dissimilarity and that SUs that move closer together over time in the ordination are becoming more similar in their community composition. Attainment of, or movement towards, a more desirable (target) community structure is thus easily visualized. A phytoplankton community that is responding to water quality and other ecosystem improvements will demonstrate increasing resemblance to those community structures characterizing the target condition. This is simply and effectively portrayed in an NMDS ordination wherein the distances between SUs in the ordination plane are directly related to community dissimilarities.

As demonstrated here, it is now possible to define a realistic phytoplankton target based on modern multivariate methods of community composition. In contrast to an objective that may be based on just a single variable (e.g., total phytoplankton biomass), the suggested new approach is based on an overall measure of phytoplankton community structure and is therefore more holistic and ecosystem-orientated. In addition, the approach is objective and quantifiable and permits statistical confirmation of compliance of the phytoplankton community structure once the target has been met. We suggest that biotic community objectives of the kind demonstrated here for the phytoplankton of two very different aquatic systems are ecologically defensible and recommend that these and related methods be applied and tested more extensively in other disturbed ecosystems and for other trophic levels.

Acknowledgements

KN acknowledges the financial support provided to Project Quinte since 1972 by the Ministry of the Environment and more recently by the Department of Fisheries and Oceans. Financial support provided by Environment Canada and the Department of Fisheries and Oceans in 2001 allowed the data analyses and preparation of part of this paper. Of the many Project Quinte collaborators, Scott Millard, Dept. of Fisheries and Oceans deserves special mention for his leadership and inspiration, especially during the latter years of the Project.

References

Anderson, R. M., Hobbs, B. F., Koonce, J. F. and Locci, A. B.
2001
.
Using decision analysis to choose phosphorus targets for Lake Erie
.
Environm. Manage.
,
27
:
235
252
.
Bay of Quinte RAP
.
1993
. “
Time to Act The Bay of Quinte Remedial Action Plan Stage 2 Report
”. In
Bay of Quinte RAP Coordinating Committee
,
Kingston, Ontario
:
Ontario Ministry of the Environment
.
September, 1993
Bristow, J. M.
1987
.
The aquatic macrophytes of West Lake, 1985.
,
Kingston
:
Ontario Ministry of the Environment
.
(unpublished MS)
Carr, M. R.
1997
.
PRIMER User Manual (Plymouth Routines in Multivariate Ecological Research)
,
Plymouth, UK
:
Plymouth Marine Laboratory
.
Clarke, K. R.
1993
.
Non-parametric multivariate analyses of changes in community structure
.
Aust. J. Ecol.
,
18
:
117
143
.
Clarke, K. R.
1999
.
Nonmetric multivariate analysis in community-level ecotoxicology
.
Environ. Toxicol. Chem.
,
18
:
118
127
.
Clarke, K. R. and Green, R. H.
1988
.
Statistical design and analysis for a “biological effects” study
.
Mar. Ecol. Prog. Ser.
,
46
:
213
226
.
Clarke, K. R. and Warwick, R. M.
1994
.
Similarity-based testing for community pattern: the 2-way layout with no replication
.
Mar. Biol.
,
118
:
167
176
.
Gerritsen, J.
1995
.
Additive biological indices for resource management
.
J. N. Am. Benthol. Soc.
,
14
:
451
457
.
Gray, I. M.
1989
.
Experimental Lake Neutralization Studies in Ontario; Limnology Report, 1988
,
Guelph, Ontario
:
B.A.R. Environmental
.
Gunn, J. M., Hamilton, J. G., Booth, G. M., Beggs, G. L., Wren, C. D., Rietveld, H. J. and Munro, J. R.
1990
.
Survival, growth and reproduction of lake trout (Salvelinus namaycush) and yellow perch (Perca flavescens) after neutralization of an acidic lake near Sudbury, Ontario
.
Can. J. Fish. Aquat. Sci.
,
47
:
446
453
.
Hartig, J. H. and Vallentyne, J. R.
1989
.
Use of an ecosystem approach to restore degraded areas of the Great Lakes
.
Ambio.
,
18
:
423
428
.
Hartig, J. H., Zarull, M. A., Reynoldson, T. B., Mikol, G., Harris, V. A., Randall, R. G. and Cairns, V. W.
1997
.
Quantifying targets for rehabilitating degraded areas of the Great Lakes
.
Environm. Manage.
,
21
:
713
723
.
Holland, R. E.
1993
.
Changes in planktonic diatoms and water transparency in Hatchery Bay, Bass Island area, western Lake Erie since the establishment of the zebra mussel
.
J. Great Lakes Res.
,
19
:
617
624
.
Hurley, D. A. and Christie, W. J.
1977
.
Depreciation of the warmwater fish community of the Bay of Quinte, Lake Ontario
.
J. Fish. Res. Board Can.
,
34
:
1849
1860
.
Keller, W., Molot, L. A., Griffiths, R. and Yan, N. D.
1990
.
Changes in the zoobenthos community of acidified Bowland Lake after whole-lake neutralization and lake trout (Salvelinus namaycush) reintroduction
.
Can. J. Fish. Aquat. Sci.
,
47
:
440
445
.
Knodt, R. C.
1999
.
MODSTAT—version 12. Copyright 1999.
,
Bellingham, WA
:
Statware
.
Leach, J. H.
1993
. “
Impacts of the zebra mussel (Dreissena polymorpha) on water quality and fish spawning reefs in western Lake Erie
”. In
Zebra Mussels: Biology, Impacts and Control
, Edited by: Nalepa, T. and Schloesser, D. pp.
381
397
.
Boca Raton, FL
:
Lewis Publishers/CRC Press
.
Makarewicz, J. C.
1993
.
Phytoplankton as indicators of environmental health
.
Verh. Internat. Verein. Limnol.
,
25
:
363
365
.
Minns, C. K., Hurley, D. A. and Nicholls, K. H., eds.
1986a
.
Project Quinte: point source phosphorus control and ecosystem response in the Bay of Quinte, Lake Ontario
.
Can. Spec. Publ. Fish. Aquat. Sci.
86
,
270
p.
Minns, C. K., Owen, G. E. and Johnnson, M. G.
1986b
. “
Nutrient loads and budgets in the Bay of Quinte, Lake Ontario, 1965–81
”. In
Project Quinte: Point Source Phosphorus Control and Ecosystem Response in the Bay of Quinte, Lake Ontario
, Edited by: Minns, C., Hurley, D. and Nicholls, K. pp.
86
76
. Can. Spec. Publ. Fish. Aquat. Sci.
86
.
Molot, L. A., Dillon, P. J. and Booth, G. M.
1990a
.
Whole-lake and near-shore melt water chemistry in Bowland Lake, before and after treatment with CaCO3
.
Can. J. Fish. Aquat. Sci.
,
47
:
412
421
.
Molot, L. A., Heintsch, L. and Nicholls, K. H.
1990b
.
Response of phytoplankton in acidic lakes in Ontario to whole-lake neutralization
.
Can. J. Fish. Aquat. Sci.
,
47
:
422
431
.
Nicholls, K. H.
1999
.
Effects of temperature and other factors on summer phosphorus in the Bay of Quinte, Lake Ontario: implications for climate warming
.
J. Great Lakes Res.
,
25
:
250
262
.
Nicholls, K. H. and Carney, E. C.
1979
.
The taxonomy of the Bay of Quinte phytoplankton and the relative importance of common and rare taxa
.
Can. J. Bot.
,
57
:
1591
1608
.
Nicholls, K. H. and Hopkins, G. J.
1993
.
Recent changes in Lake Erie (north shore) phytoplankton: cumulative impacts of phosphorus loading reductions and the zebra mussel introduction
.
J. Great Lakes Res.
,
19
:
637
647
.
Nicholls, K. H., Heintsch, L., Carney, E. C., Beaver, J. and Middleton, D.
1986
. “
Some effects of phosphorus loading reductions on phytoplankton in the Bay of Quinte, Lake Ontario
”. In
Project Quinte: Point Source Phosphorus Control and Ecosystem Response in the Bay of Quinte, Lake Ontario
, Edited by: Minns, C., Hurley, D. and Nicholls, K.
86
158
.
Can. Spec. Publ. Fish. Aquat. Sci
.
Nicholls, K. H., Heintsch, L. and Carney, E. C.
2002
.
Univariate step-trend and multivariate assessments of the apparent effects of P-loading reductions and zebra mussels on the phytoplankton of the Bay of Quinte, Lake Ontario
.
J. Great Lakes Res.
,
28
:
15
31
.
Peeters, E. T. H. M. and Gardeniers, J. J. P.
1998
.
Ecologically based standards for nutrients in streams and ditches in The Netherlands
.
Wat. Sci. Tech.
,
37
:
227
234
.
Reynoldson, T. B., Bailey, R. C., Day, K. E. and Norris, R. H.
1995
.
Biological guidelines for freshwater sediment based on BEnthic Assessment of SedimenT (the BEAST) using a multivariate approach for predicting biological state
.
Aust. J. Ecol.
,
20
:
198
219
.
Reynoldson, T. B., Norris, R. H., Resh, V. H., Day, K. E. and Rosenberg, D. M.
1997
.
The reference condition: a comparison of multimetric and multivariate approaches to assess water-quality impairment using benthic macroinvertebrates
.
J. N. Am. Benthol. Soc.
,
16
:
833
852
.
Sayer, C. D. and Roberts, N.
2001
.
Establishing realistic restoration targets for nutrient-enriched shallow lakes: Linking diatom ecology and paleoecology at the Attenborough Ponds, U.K
.
Hydrobiologia
,
448
:
117
142
.
Stoermer, E. F.
1978
.
Phytoplankton assemblages as indicators of water quality in the Laurentian Great Lakes
.
Trans. Amer. Micros. Soc.
,
97
:
2
16
.
Willén, E.
2000
. “
Phytoplankton in water quality assessment - an indicator concept
”. In
Hydrological and Limnological Aspects of Lake Monitoring
, Edited by: Heinonen, P., Ziglio, G. and van der Beken, A. pp.
58
80
.
NY
:
John Wiley and Sons Ltd.
.
Willén, E.
2001
.
Phytoplankton and water quality characterization: experiences from the Swedish large lakes Mälaren, Hjälmaren, Vättern and Vänern
.
Ambio
,
30
:
529
537
.