| Popis: |
The occurrence of toxic algal blooms and the environmental and commercial damage associated with these blooms appears to be on the increase globally. Toxic algal blooms are often known as red tides, as high enough concentrations of these phytoplankton may cause surface discolorations of the water body they are blooming in. Dinoflagellates and diatoms are commonly associated with red tide blooms, although other phytoplankton species are known to cause them. A number of reasons have been suggested for the apparent increase in these red tide blooms: increased utilization of coastal waters for aquaculture; stimulation of blooms by eutrophication and changes in climatological conditions; transport of toxic species worldwide in ballast water; and increased awareness of toxic species. Even though there is an increasing awareness and understanding of these species, current methods for detection and identification of toxic phytoplankton are inadequate.The classical method for identifying toxic algae is through the use of optical microscopy. However, size limitations and difficulties in observing and identifying morphological features means that even an expert can find it difficult to identify some of the smaller or more fragile species. Scanning Electron Microscopy is generally needed for unambiguous identification of toxic phytoplankton. However, difficulties with fixation of fragile algae, expense and limited access to SEM means that this method of identification is rarely used and is subject to some fundamental methodological problems. To establish an efficient monitoring system for waterways, it is imperative that a cheap, automated method of identification is developed.The most commonly used automated method of analysis for algae, flow cytometiy, is useful only for enumeration, sorting and identification of samples down to phylum. However, this thesis has demonstrated that the use of FT-IR microscopy combined with chemometrics (in particular PCA and linear discriminant analysis) shows promise as a useful tool for the rapid identification of toxic phytoplankton. Eight species of blooming phytoplankton were successfully differentiated using the method: Nitzschia closterium, Nitzschia frustulum, Skeletonema costatum, Skeletonema psuedocostatum, Nitzschia longissima, Peridinium balticum, Amphidinium caterae and Heterosigma carterae. Two groups of species that are almost identical morphologically (i.e. Nitzschia closterium / Nitzschia longissima and Skeletonema costatum/Skeletonema psuedocostatum were differentiated using this technique. The method is rapid and relatively inexpensive, and does not require the knowledge of an expert to undertake analysis as it is computer aided.Seven species of phytoplankton were grown in culture and one species (Nitzschia longissima) was collected from a bloom episode that occurred approximately 26 km downstream from the mouth of the Brisbane River. Spectra from early stationary stage cells were obtained for each species using an FT-IR microscope. Spectra were sorted into two groups depending upon the presence or absence of a strong peak at 1068 cm1. This peak is indicative of the large silica component contained in the diatom frustule, and hence was an initial method of sorting between the two groups for ease of data handling. PCA was then applied to each group and a number of classification methods were attempted to differentiate between the two groups. Mahalanobis distances, centroid and furtherest neighbour cluster analysis and k-means cluster analysis were of little use in differentiating between the species. However, discriminant analysis was able to differentiate between the species. 100 % of spectra of Nitzschia longissima, Skeletomena costatum, Skeletonema psuedocostatum, Peridinium balticum, Amphidinium caterae and Heterosigma carterae were classified correctly using this method. 69% of Nitszchia closterium and 61% of Nitzschia frustulum were correctly classified. First derivative and second derivative specthttps://espace.library.uq.edu.au/js/fckeditor/editor/images/spacer.gifra were also analysed, to little benefit. |
