The comparison of remotely sensed and ground-based habitat area data using species-area models.
As long as high resolution landscape data are more expensive than low resolution data, there needs to be some form of cost-benefit analysis to inform the choice of data for a particular landscape ecological application. An approach is suggested whereby models describing a common ecological data set are compared using high and low resolution landscape data. First, the detection of habitat units is compared on the two scales, and second, an ecological model with a common set of species data is fitted to both sets of landscape data. For the control, the model was fitted using the higher resolution data but including only those habitat patches that could be detected on the low resolution map. The approach is demonstrated using data from a study relating the number of species of resident birds to habitat area in East Anglia, UK. Woodland areas were calculated using Ordnance Survey (OS) 1:25 000 maps supplemented with field survey data, and these were compared with areas estimated from the remotely-sensed ITE (institute of Terrestrial Ecology) Land Cover Map, which was derived from Landsat TM (thematic mapper) data with a pixel size of 25×25 m. The discrepancies between the two sets of area values were greatest at the smallest woodland size. Only 18% of woodlands <1 ha (as measured from the ground-based methods), but almost all woodlands >2 ha, could be detected on the Land Cover Map. Estimates of woodland area tended to be smaller according to the remotely sensed map, especially at the smaller sizes. Species-area relationships using ground-based data described the bird species number better than those from remotely sensed data, even when comparing only those woodlands found on both maps. At this scale, ground-based maps gave a better quality of information than the Land Cover Map, but at greater cost. However, the remotely sensed data are of sufficient resolution for coarse estimates of species-area relationships.