Home > Research > Services > Report Abstracts > 320

Abstract from BTO Research Report No. 320:

Freeman, S.N., Noble, D.G., Newson, S.E. & Baillie, S.R. (2003)

Importance of geographical location and local habitat features for species abundance: analyses using Breeding Bird Survey data.

Executive summary

1. Numbers of birds recorded in the BTO/JNCC/RSPB Breeding Birds Survey (BBS) in 2000 are related to three sources of habitat data, which utilise different methods of landscape categorisation and are recorded at different spatial scales.

2. BBS data routinely provide valuable information on temporal trends in abundance. We consider here the extent to which their conservation value may be increased by the development of spatial models for bird distributions and numbers, based on the habitat data.

3. One of the three data sets is gathered by the BBS observers themselves, and therefore most closely reflects the spatial distribution and resolution of the bird count data. The other two sources, ‘landclass’ and ‘landcover’, are taken from CEH databases and are available on a national scale. Habitat-based models based on the latter sources of information therefore permit prediction of bird distributions a national scale.

4. Predictably, most species showed pronounced differences in abundance between sites of different habitats and landscapes. Reflecting large-scale differences in range, a modelling approach showed relationships with latitude, longitude and altitude for most species. Habitat-based models were then considered in detail for seven species of widely differing ecology (Buzzard, Skylark, Meadow Pipit, Dunnock, Sedge Warbler, Nuthatch, and House Sparrow).

5. After adjusting for Northing, Easting and altitude, similar distributions were predicted whether habitat was represented by data from the BBS (at the scale of the 200m. transect section) or data from the CEH databases (at the scale of the 1 km square).

6. We split the data for species of interest into two random halves, and employed a model fitted to one half to predict the numbers of birds in squares of the other half. A comparison of residuals suggests that the BBS-based habitat data, due to their more direct equivalence in scale to the bird data, are marginally superior to the coarser CEH data. The latter, however, permit the mapping of predicted numbers on a national scale.

7. Predictions based upon the two CEH datasets were also similar. Of the two, landclass (a single category, to which a square is assigned) proved more likely to produce predicted values close to zero although a species may have been present. As the landcover data set assigns to each square the proportion of the land surface of a number of land cover types, these models were less prone to this severe underestimation.

8. Areas for future research into the spatial distributions of birds, and their relationships with habitat, including the utilisation of Neural Network models, are discussed.

Back to Research Reports 289-344