Matskási István (szerk.): A Magyar Természettudományi Múzeum évkönyve 86. (Budapest 1994)

Kemenes, I. ; Demeter, A.: Uni- and multivariate analyses of the effects of environmental factors on the occurrence of otters (Lutra lutra) in Hungary

probability of finding otters at a particular site, as the logistic regression coefficients associated with them were not significantly different from 0 (Wald statistic, p0.05). An indication of predictive power (fit of the model) is how correctly it classifies the dependent variable: 292 (79.13%) of the sites were correctly classified using the constructed model. DISCUSSION A qualitative assessment of the distribution of sites where otters were found indicated that low water levels and scarce bank vegetation had a strong detrimental effect on the occurrence of otters. The univariate correlation analyses indicated that the depth of the water and the density of the bank vegetation were positively correlated with the occurrence of otters at particular sites. This was also the conclusion from previous studies focusing on either the effect of the depth of water (GLIMERWEEN & OUWERKERK 1984) or the effect of the density of the bank vegetation (MACDONALD & MASON 1985) and our multivariant analysis confirmed these observations. A great advantage of the multiple logistic regression technique is that it can take into account the effect of non-scalable parameters as well as scalable ones. All the factors that, to our a priori judgement, had a disturbing effect on the otter populations, were non-scalable (i.e. they were either present or absent). However, a surprising new finding of the present work is that cultivation of the land around the aquatic habitat (a disturbing factor in our view) also appears to have a positive effect on the occurrence of otters. In our opinion this is due to the fact that in such areas the irrigation canals and natural watercourses are kept in good condition in order to supply sufficient water for irrigation. Also, apart from short-term pollution waves caused by spraying pesticides on the crop etc., in these rural areas there are no chemical and other industrial plants which would be the source of a more acute pollution. Another interesting result of the analysis is that the presence of various disturbing factors, such as the vicinity of towns, or even household garbage floating in the water have apparently no or very little effect on the occurrence of otters. More importantly, if we compare the degree by which a unit change in the water depth or vegetation density and in the disturbing factors change the odds of finding otters, we see that the former two have a much more profound effect. This means that otter populations are surprisingly tolerant to disturbances caused by human activity but extremely sensitive to the deterioration of the vegetation or reduction of water depth. From a practical point of view this means that it is not necessary to create isolated safe havens for otters trying to exclude all human disturbance but it is absolutely essential to try to preserve the bank vegetation and ensure that the water depth in aquatic habitats is sufficient to support otters. JAMES & MCCULLOUGH (1990) warn against inference of causation from regression analysis when there is no experimental design. We restrict ourselves to drawing preliminary conclusions from the exploratory analyses, although with the otters we cannot imagine at present an experimental design that would satisfy the very stringent requirements set forth by the above authors. In this large-scale, nationwide study it was not possible to determine whether a given site was inhabited permanently by otters or whether only transient individuals had passed through the given site previous to the visit. Ecologists distinguish populations living in "sink" areas that are maintained by immigrants from "source" areas (PULLIAM 1988). The multiple regression model constructed fitted the data fairly well. One of the causes tor departure from the model might have been classifying unsuitable sites as inhabited by otters when in fact they were low-density sink areas only with transient individuals. Repeated visits to the sites would clarify this issue. In this survey we employed simple descriptors of habitat dimensions, yet a fairly strong predictive power was achieved. DUBUC et al. (1990) attained greater proportion of correct