Summary: | The monitoring of wildlife populations is considered from the point of view of spatial statistics. The motivation for the approach is the inherent spatial dependence in practically any wildlife population. This dependence affects also the observations, and hence it should not be omitted in the monitoring schemes. The first part of this monograph focuses on line transect sampling, which is used to assess the sizes or intensities of wildlife populations. The line transect method is modelled as sampling from a marked point process, where the spatial distribution of the population is regarded as a point process and the detectability of the individuals is described by the marks of the points. As mains results a general covariance function for the sampling and a new intensity estimator are derived. They are used in comparing the variances of intensity estimators for five types of transects. Furthermore, stereological interpretations of the line transect method are given. The second part concentrates on the estimation of biogeographical distributions. The estimation is based on atlas data, that is, observations recorded in a grid over the study area. The estimation of the true grid map is regarded as a restoration of a binary pixel image, yet the present problem has many features which make it deviate from the standard applications of image analysis. The apparent contextuality in the distribution maps is modelled by using Markov random fields. Several restoration techniques are discussed, and the iterated conditional modes algorithm and the maximum marginal posterior estimation are applied. A fully Bayesian approach is suggested in the restoration of heavily degraded images to overcome difficulties in the estimation of spatial interaction. Two examples are given; one uses data of an atlas survey on breeding birds, and the other uses data of a herpetological atlas.
|