engleski

New approach in data analysis from areas with limited ornithological data: automated modelling of small passerine densities from approximate transects

Ornithological data, in comparison to data for other biological groups, is often regarded as one that is most easily obtainable and birds are among best known organisms. Additionally, birds are indicators of environmental changes and good surrogate for ecological value of an area. For that reason, in Europe, data on birds and habitats are used as main pillars to advocate the conservation of nature in whole, through the Bird’s and Habitat’s Directives. Number of scientific and non-governmental organisations developed variety of monitoring programs in order to resolve questions on bird distributions, population statuses and diverse population parameters. Some Western European countries and North America, this period in time regard as time “beyond data gathering” where the usage of novel statistical techniques developed, enables us to answer more complex ecological questions that was not possible in the recent past. In the same time, vast amount of available remotely sensed data opened possibility to monitor bird habitats on a large scale, and more useful planning of corridors and stop-over sites as well as facilitate multispecies management. However, the situation in Croatia and neighbouring countries is far from described one. Extreme lack of bird data collected adequately through the time and space disables scientists to answer diverse ecological questions and detect changes in the environment. This is especially pronounced for territory keeping bird species for which only data on presence/absence can be found in EBCC Atlas of European Breeding Bird at 50 km resolution. Lack of experienced ornithologist resulted in lack of local breeding/wintering bird atlas on at least 10 x 10 km resolution. We propose a way of analysing data on passerine bird communities obtained from transect studies where only coordinates on start and end point of transects are available with total number of territory keeping species, which is often the case with ornithologists/birdwatchers in the region. Automation procedure via R script is prepared to do the following procedures: 1) check the date from which bird data are available and adjoining corresponding environmental predictors ; 2) link start and end points according to transect code substitute real transects with straight line start-end points transect ; 3) split data according to the transect code and create KML and Shapefile for check and validation ; 4) prepared a set of environmental variables at 30 m resolution (e.g. elevation, slope, aspect, distance to the coastline, climate variable and Corine Land Cover or alternatively habitat classes if appropriate habitat mapping has been made for the area of interest) ; 5) extract information on environmental conditions along transect line using line profiles ; 6) testing for all possible combinations of models and selecting the one that suits the best to the data according to Akaike Information Criterion and 7) make spatial prediction on counts according to the best subset model using regression or regression-kriging technique. In the future we will test performance of this type of approximate transects with real ones on the areas where data will allow, and detection of bird guilds where technique performs the best.

automated modelling; geostatistics; bird densities

nije evidentirano