Identification of longitudinal and temporal patterns of phosphorus fractionation in river sediments by non-parametric statistics and pattern recognition techniques

A combination of non-parametric explanatory procedures and pattern recognition techniques is introduced in this study as an alternative tool to the assessment of sedimentary P fractionation. The complex raw data matrix was treated with a series of parametric and non-parametric techniques such as correlation analysis, cluster analysis, factor analysis/principle components analysis, discriminant analysis, multidimensional scaling, Kruskal–Wallis analysis of variance and non-parametric median test. Each of these procedures has been directed towards the identification of specific mechanisms that control P biogeochemistry in the river sediment. Non-parametric tests were proven as a very powerful tool in interpreting temporal variations of the data in a short-term basis as opposed to previous procedures predicting seasonal or annual changes. On the other hand, parametric techniques were not always able to delineate a reduced number of indicator variables responsible for the large variations in sediment quality, necessitating the application of various techniques in order to interpret the intricate mechanisms occurring in the sediment phase. The examined procedures provide the basis for identifying the patterns of sedimentary P associated with seasonal and spatial patterns, pollution sources as well as differentiate between natural and non-natural inputs, presenting an alternative approach for environmental quality assessments.

The combination of parametric and nonparametric statistical examinations was presented in this study as means for identifying the complex behavior of P in river sediments. Non-parametric ANOVA and median tests were proven as a valuable tool in studying short-term temporal variations of sediment quality as opposed to parametric FA/PCA procedures which could not satisfactorily assign any variations of this kind. Furthermore, FA/PCA could not satisfactorily describe spatial changes. The adaptation of some new modifications did not improve the situation which indicates that the application of this technique to sediment data merits contemplation. On the other hand, other methods like correlation analysis, DA, CA and MDS contributed to the interpretation of the observed behavior while afforded a considerable data reduction by isolating those variables that exert a significant effect. As far as Louros river is concerned, the concentrations in its sediments indicates relatively moderate accumulation, especially of inorganic forms, along the river although several sources that could maintain an elevated concentration profile are present. Two counteracting mechanisms seem to be responsible for this behavior. The significant calcite-P co-precipitation and the organic matterinorganic P interactions that provide an important link of P to the sediment and the high portion of sandy material which acts as a natural barrier for P accumulation due to P evasion (dissolution). The relative importance of each process is affected by seasonal and spatial variations which follows the magnitude of the overbank and river-side activities. Higher inputs are observed during spring and autumn months especially close to agricultural areas while winter and summer months exhibit the smallest variation as an outcome of the low P input.