ABSTRACT
Data from long term annual monitoring of (137)Cs concentrations in harvested moose (Alces alces) were empirically modelled by applying multivariate data analysis that is able to from relatively raw datasets show how the many various impact variables are related (Principal component analysis, PCA). In the later stage regression modelling (Partial least squares, PLS) was applied to analyse which environmental and physiological factors were significant (i.e. of predictive value) based on the measured (X) activity concentrations in moose meat. The data sets originate from two different forest dominated areas in Sweden. One area is located inland (Heby municipality) and the other borders to the Baltic Sea (Gävle municipality). In inland with 20% farmland, GIS-software was used to calculate the proportion of different habitat types and (137)Cs deposition around individual killing spots. This model reveals that the proportions of farmland and forest around the killing spot were significant parameters, second to deposition and years since fallout. Significance was also obtained for the proportions of mire and water bodies, the amount of rain in summer and the age of the moose. In the other model based on data from the coastal area with only about 4% farmland, the coordinates of the moose killing spots were not recorded in the data sets. In the resulting model the temperature in July was the most important parameter, second to years since fallout. Significance was also found for the following parameters: temperature and rainfall in several summer months, the approximate north- and eastward location of the killing spot and to which age category (adult/calf) the harvested moose belonged.
