Using Landsat TM and field data to produce maps of predicted bird densities in Latvian farmland

Models of farmland bird population densities established from field surveys were applied to classified satellite data for mapping of predicted bird numbers. The field survey system was based upon point counts of birds and descriptions of their habitat within a 200 m radius. The relationship between birds and habitats was analysed by means of multiple regression analysis. The resulting regression models were coded into classified Landsat Enhanced Thematic Mapper (ETM+) data which had similar land cover/use classes as the field observation data. With the use of a circular 200 m radius moving window approach, simulated maps of predicted bird population densities were produced. The results indicate that the method is best suited to species with small- and medium-sized home ranges and non-complex habitat relations. This approach could possibly be used for species other than birds, and could have implications for monitoring agro-environments by means of selected indicators.