Delineating the number of animal territories using digital mapping and spatial hierarchical clustering in GIS technology

Estimation of the population size and distribution of individuals is crucial in ecological studies. In many territorial animals, e.g., birds, territory mapping is the method believed to produce absolute and accurate estimators of population size (number of breeding territories). However, the results are highly variable among observers because interpreting maps is subjective and might depend on previous experience, map scale and bulk of data in the map. This technique is also highly laborious. In this work, we present a new approach to the analysis of bird species maps using spatial hierarchical clustering. Observers mapped birds and noted details of their behaviour in a 429 ha forest-grassland mosaic plot in the Beskid Mały Mountains, Poland. Spatial hierarchical clustering uses only a distance criterion to establish a cluster that is considered a proxy of bird territory. The mathematically established (CrimeStat software) number of territories is compared with the number of territories delineated in a traditional manner by three map analysts experienced in field surveys and the method. The map analysts differed in the number of delineated territories. The average coefficient of variation was 64% (range: 0–173%) and intraclass correlation coefficient was lower than for the computer algorithm and mean estimates of the map analysts. Spatial hierarchical clustering estimated the numbers of territories inside the 95% confidence interval for means calculated using the estimates for each species from map analysts. The relative rank abundance (order of the number of territories estimated by spatial hierarchical clustering) was similar to the results estimated by map analysts as indicated by the Wilcoxon paired test and intraclass correlation coefficient which exceeded a value of 0.9. Higher-order clusters (clusters of “territories”) were also found. Although the presented method uses only the distance criterion, it produces results comparable to the traditional interpretation from maps. The advantage of the presented approach is that it generates repeatable results that are free of subjectivity and independent of observer experience and might supply additional information on the distribution of the territories and perhaps their size. The method might also identify areas of high suitability and can be successfully applied to any territorial animal. Further development should be directed towards inclusion of animal behaviour and simultaneous records of different individuals in the algorithm that calculates clusters.