ArtikelDe introductie van Autocluster in het BMP

The Breeding Bird Monitoring Program (BMP) assesses population trends of breeding birds in the Netherlands since 1984. It is based on intensive territory mapping in fixed study plots. Fieldwork and interpretation methods are highly standardized and are described in detail in a manual. All birds with territory-indicative behavior (e.g. song, pair bond, display, alarm, nests) are recorded on field maps. Species-specific interpretation critera are used to determine the number of territories per species by the end of the season. Interpretation criteria focus on the type of behavior observed, the number of observations required (taking into account the varying detction probability between species and within the breeding season), and the timing of observations (to exclude non-breeding migrants). This interpretation process is time-consuming and rather complicated. Therfore, we have developed the software tool Autocluster, which allows data-entry and automatic clustering of field observations into territories. In this paper the aims, techniques and effects on estimated population trends of the introduction of Autocluster are described. Autocluster primarily aims to facilitate (volunteer) observers by simplifying data-entry  and interpretation, and generating standard output such as territory maps. Since 50% and 58% of all observers used Autocluster in 2011 and 2012 respectively, we seem to have provided in a need. Furthermore, Autocluster further standardizes interpretation of the field data, therby increasing comparability of counts between sites and years and the quality of trend estimates. In addition, all individual observations become digitally available for additional analyses, such as studies on habitat use and detection probabilities. Finally, estimating population trends on the basis of individual observations instead of interpreted territories will be possible in the future, thereby accounting for possible time-trends in detection probabilities. Effects of introduction of Autocluster on population trends were assessed by comparing trends based on study plots using traditional 'manual' interpretation in both 2010 and 2011 with trends based on study plots that changed to Autocluster in 2011. We found significant differences for 12 out of 92 species, but these might bepartly caused by coincidence or by real differences in trends between both datasets. We provisionally conclude that the effects on national population trends are limited. However, effects on local trends may be larger, depending on the extent of deviations from standard interpretation guidelines that individual observers have been applying.