v.adehabitat.clusthr computes the home range of one animal using a clustering algorithm (Kenward et al. 2001). This method has been "translated" for GRASS from the R function "clusthr" in the package adehabitat (Calenge, 2006). This method estimates home range using a modification of single-linkage cluster analysis developped by Kenward et al. (2001). The clustering process is described hereafter: the three locations with the minimum mean of nearest-neighbour joining distances (NNJD) form the first cluster. At each step, two distances are computed: (i) the minimum mean NNJD between three locations (which corresponds to the next potential cluster) and (ii) the minimum of the NNJD between a cluster "c" and the closest location. If (i) is smaller that (ii), another cluster is defined with these three locations. If (ii) is smaller than (i), the cluster "c" gains a new location. If this new location belong to another cluster, the two cluster fuses. The process stop when all relocations are assigned to the same cluster.
At each step of the clustering process, the proportion of all relocations which are assigned to a cluster is computed (so that the home range can be defined to enclose a given proportion of the relocations at hand, i.e. to an uncomplete process). At a given step, the home range is defined as the set of minimum convex polygon enclosing the relocations in the clusters.
The user may either: (i) compute the home range at a given percentage: this corresponds to the set of clusters enclosing a given percentage of the relocations (default), (ii) compute the home range corresponding to a given step of the algorithm (flag -s, a first run of the module indicates the maximum number of steps supported by the data), or (iii) have a visual display of the whole clustering algorithm (flag -a). Note that for the first case (i), we may have different home range configuration for a given value of 'percent'. For example, if percent = 100, it is possible that all relocations have been assigned to one out of two clusters, just before the last step of the algorithm. So that the last step of the algorithm just consists into the "fusing" of the two clusters. The two choices (two clusters or one cluster) correspond to configurations where 100% of the relocations are enclosed in the home range. When such cases happen, the module returns the configuration with the largest number of clusters, i.e., the minimum home range size (in the example, two clusters of relocations).
Kenward R.E., Clarke R.T., Hodder K.H. and Walls S.S. (2001) Density and linkage estimators of homre range: nearest neighbor clustering defines multinuclear cores. Ecology, 82, 1905-1920.
Calenge, C. (2006) The package adehabitat for the R software: a tool for the analysis of space and habitat use by animals. Ecological Modelling, 197, 516-519.
v.adehabitat.clusthr input=localisations output=homerange percent=95
v.adehabitat.clusthr input=localisations output=homerange step=25 -s
v.adehabitat.clusthr input=localisations output=homerange -a
Last changed: $Date$