Publications (1)0 Total impact
[Show abstract] [Hide abstract]
ABSTRACT: This study developed a statistical approach to predicting the probability of the occurrence of phytoplankton species from the general properties of a water-body and the specific conditions at the time of sampling. Here, we report the results for the two Dinophytes Gymnodinium uberrimum and Peridinium willei which frequently cause problems in drinking-water treatment. Our statistical approach addresses the problem of uneven distribution of samples across parameters. For this purpose we used probability calculation to assess under which conditions the species' occurrence is a likely or an unlikely event. The ecological niches of both species turned out to be similar, as both occur in oligo- to mesotrophic, deep, stratified reservoirs at low phosphorus concentrations. However, their pH and temperature range differed: G. uberrimum occurred predominantly between pH 6 and pH 7 and at a temperature range up to 16 °C, whereas P. willei showed a wider pH range and a temperature range below 12 °C. With these results, the classification of P. willei within the functional group Lo proposed by Reynolds et al. (Reynolds et al., 2002) is now underpinned with a substantial data base. The results suggest G. uberrimum to fit into this association as well. Introduction Predicting which phytoplankton species are likely to occur under which environmental conditions remains a challenge. The growth rates of numerous phytoplankton species in relation to different environmental parameters have been analysed in culture experiments for specific conditions, and for a number of species, optimum growth conditions have thus been determined (see e.g. Reynolds (Reynolds, 1984) for an overview). However, in lakes or reservoirs these optimum conditions seldom occur, and net population growth in the ecosystem is further determined by interspecific competition and loss factors. Mechanistic modelling to forecast the occurrence of individual species at a given point in time and space therefore struggles with the complexity of interactions that determine the process and the outcome of species competition. Descriptive approaches to predicting occurrence are an alternative. Numerous field observations on species composition and the conditions under which selected species occur, dominate or proliferate are published, but most data are limited to a small number of lakes or reservoirs (e.g. Rott, 1988; Fott et al., 1999; Carrias et al., 2001; Queimalinos et al., 2002; Tolotti et al., 2003). Generic information about the conditions under which a given species is likely to occur or proliferate is accordingly sparse. Reynolds et al. (Reynolds, 1980; Reynolds et al., 2002) have systematised such descriptive approaches by defining associations of species that tend to co-occur in relation to the habitats in which they are typically found, similarly to the sociological groupings for terrestrial plants. These authors base their definition of 31 phytoplankton associations on the functional role of the species, determined by similar sensitivities and tolerances to environmental conditions which are often conferred by similarities in morphology. The authors emphasise the still provisional character of their classification and explicitly invite contributions to consolidating this functional groups concept. The aim of the study presented in the following is to generalise and predict the probability of species occurrence in relation to specific environmental conditions by using a statistical approach. For this purpose, we analysed a large data set from 24 reservoirs covering at least 5 years of monthly sampling. To develop this approach, for a start, we chose two Dinophyte species which frequently cause problems in drinking-water treatment plants in Germany and other European locations, thus rendering predictive models based on accurate descriptions of their occurrence useful for practitioners: Peridinium willei (Huitfeld-Kaas 1900) is known for its taste and odour producing metabolites (Jüttner, 2002) and Gymnodinium uberrimum ((Allman) Kofoid and Swezy 1921) is a nuisance for water supplies because it accelerates clogging of filter systems in drinking-water treatment, but may also break through these filters with the consequence of elevating the DOC concentrations of the purified water and thus enhancing microbial growth (Hoehn, 2000). A further aim is to use this large data set in order to follow the invitation expressed by Reynolds et al. (Reynolds et al., 2002) for testing the validity of the phytoplankton associations that these authors propose. Among their 31 trait-differentiated functional groups P. willei belongs to the association L 0 which also includes the Cyanobacteria genera Woronichinia and Merismopedia. Their habitat is described as “summer epilimnia in mesotrophic lakes” and their traits as “tolerance to segregated nutrients” as well as “sensitive to prolonged or deep mixing”. More specifically, these species do not benefit (in competition against others) from low light availability, low temperature, low concentrations of dissolved inorganic nitrogen or carbon dioxide, or high grazing pressure, and are tolerant of a shallow epilimnion (< 3 m), low concentrations of phosphorus and silicate. G. uberrimum is not included in this classification. A specific feature of our statistical approach is the maximum exploitation of the data in the data set. As in many studies, our phytoplankton data are not evenly distributed over time, reservoirs and the environmental parameters analysed. Such imbalance interferes with the application of many statistical methods because the more frequently sampled water-bodies would bias the results. To solve this problem, means of seasons or years have been calculated to obtain equally distributed values. However, this would lead to an enormous loss of information. To avoid this loss, we used probability calculation, i.e. we tested if the occurrence of the species under a specific condition is a probable or improbable event. This relative approach has the advantage of not requiring homogeneous distribution of the data and thus provides an alternative to data aggregation.