Patterns of cyanobacterial abundance in a major drinking water reservoir: what 3 years of comprehensive monitoring data reveals?

The present study provides a detailed analysis of the factors influencing variation in cyanobacterial communities of a large shallow off-river drinking water reservoir on the east coast of Australia. Receiving multiple inflows from two unprotected mixed land-use catchments, the Grahamstown Reservoir is a model example of a reservoir which is highly vulnerable to adverse water quality issues, including phytoplankton blooms and the resulting filtration, toxin and taste and odour problems produced. The spatial and temporal distributions of cyanobacteria were assessed for a period of 3 years (January 2012-December 2014) based on samples collected from three monitoring stations within the reservoir. Relationships between cyanobacterial abundance and a range of environmental factors were evaluated by application of multivariate curve resolution-alternating least squares (MCR-ALS) analysis.Results of the analysis indicated that among the 22 physico-chemical variables and 14 cyanobacterial taxa measured, the vertical temperature gradient within the water column and nutrient availability were the most powerful explanatory factors for the observed temporal and spatial distribution patterns in the densities of cyanobacterial taxa. The abundance patterns of the dominant cyanobacterial taxa-Aphanocapsa, Aphanothece, Microcystis and Pseudanabaena-were strongly linked with rainfall and run-off patterns into the reservoir, while Coelosphaerium and Microcystis were the taxa most influenced by the apparent occurrence of thermal stratification. The findings demonstrate the capacity of rigorous multivariate data analysis to identify more subtle relationships between water quality variables, catchment factors and cyanobacterial growth in drinking water reservoirs.

A review of recent methods for the determination of ranges of feasible solutions resulting from soft modelling analyses of multivariate data.

Soft modelling or multivariate curve resolution (MCR) are well-known methodologies for the analysis of multivariate data in many different application fields. Results obtained by soft modelling methods are very likely impaired by rotational and scaling ambiguities, i.e. a full range of feasible solutions can describe the data equally well while fulfilling the constraints of the system. These issues are severely limiting the applicability of these methods and therefore, they can be considered as the most challenging ones. The purpose of the current review is to describe and critically compare the available methods that attempt at determining the range of ambiguity for the case of 3-component systems. Theoretical and practical aspects are discussed, based on a collection of simulated examples containing noise-free and noisy data sets as well as an experimental example.

Determination and visualization of rotational ambiguity in four-component systems.

One of the main problems that limit the use of model-free analysis methods for the resolution of multivariate data is that usually there is rotational ambiguity in the result. While methods for the complete definition of rotational ambiguity for two- and three-component systems have been published recently, the comprehensive and general resolution of rotational ambiguity for four-component systems has eluded chemists for several decades. We have developed an extension of self-modelling curve resolution for a mixture of four-components. The performance of the method was verified by applying it to resolve simulated and real data sets.

The reduction of rotational ambiguity in soft-modeling by introducing hard models.

Rotational ambiguity is a major problem in the application of soft-modeling analysis to a variety of multivariate mixture resolution problems and particularly important in the analysis of kinetic data. Soft-modeling analyses rely on constraints that restrict the concentration profiles and/or the spectral responses of all components. The main goal of this work is to demonstrate how a hard-modeling constraint on concentration profiles drastically decreases the extent of the rotational ambiguity. Therefore, in the present paper the discussion is focused on systems in which hard-modeling information is available. The results of simulated examples reveal that the utilized hard constraint decreases the rotational ambiguity in estimated concentration profile even components that do not take part in the explicit model. In addition, the rate constant of known reaction is determined in this method.

Rank annihilation factor analysis method for spectrophotometric study of second-order reaction kinetics.

In this study several methods are described to determine the rate constant of a second-order reaction in the form of A+B→C. These approaches allow circumventing a rank deficiency inherent of a second-order reaction when the spectroscopic data is influenced by additional source of variance. Classically, to determine the unknown rate constant in this kind of systems, one needs to have extra knowledge about the system, including the spectra of the reactants or product and the exact kinetics. In the case of the presence of an unknown phenomenon in the data set that cannot be explained by the model, such as baseline drift, the estimated rate constant might be erroneous. Present work is a modification of the rank annihilation factor analysis (RAFA) algorithm by inclusion of I) pure spectra of reactants, or IIA) mean centering step, or IIB) mean spectrum. The proposed methods can interestingly be applied on a single kinetic run. The performances of the new methods have been evaluated by applying them to analysis of simulated and experimental data.

Resolution of rotational ambiguity for three-component systems.

Soft modeling of multivariate data is a powerful method for the analysis of processes that cannot be described quantitatively by a chemical model. Soft modeling usually does not result in unique solutions. Thus, the determination of the range of feasible solutions is important. For two-component systems the determination of that range is well-understood; for three-component systems the task is remarkably more complex. We present a novel method that can be applied to any multivariate data set, irrespective of overlap or realistic noise level. The expansion to four components is indicated.