Abstract
With use of the two-way classification model in variance analysis, spatial and temporal variabilities of water quality in a lake are separated additively. For that purpose, the so-called extended logarithmic transformation is applied to measured data. The values of parameters in the transformation equation and the model are determined by the maximum likelihood procedure. They can be obtained in no analytical way, but through an iterative computational procedure. By the aid of statistical hypothesis test, grouping of measurement points for water quality is carried out, through the parameters expressing the spatial variability. The method is developed first for a single item and it is then extended to a method for multi-items. The technique is applied to analyzing a set of time-series data measured over several years in Lake Biwa.
