Home > Research > Services > Report Abstracts > 139

Abstract from BTO Research Report No 139:

Gregory, R.D. & Baillie, S.R. (1994)

Evaluation of sampling strategies for 1-km squares for inclusion in the Breeding Bird Survey.

Executive Summary

1. This report summarises the findings of an evaluation project assessing a number of possible sampling strategies for the Breeding Bird Survey. The main aim of the survey is to provide population indices that are representative of the United Kingdom as a whole and of particular regions and habitats for a large number of widespread and abundant species (excluding seabirds). This necessitates some form of random or stratified random sampling.

2. Fieldwork trials, within the Pilot Census Project 1992-93, suggest that the proposed methods of fieldwork, based on randomly selected 1-km squares and time-efficient transect or point count approaches, were acceptable to volunteer fieldworkers and had great potential in meeting the aims of the survey.

3. Four basic strategies were considered for the selection of 1-km survey squares and each was assessed at three levels of observer input. The later comprised both volunteer and professional fieldworkers. The non-random distribution of volunteers across the UK necessitates the use of professional fieldworkers in remote areas. Their input was modelled at three levels, the lowest representing minimum, acceptable coverage (in terms of the survey aims).

4. The four strategies were selection of squares:

(1) Proportional to observer density;

(2) Proportional to observer density based on a regular grid;

(3) Proportional to observer density and stratified by ITE land class with sampling proportional to the total number of squares in each class;

(4) Proportional to observer density and stratified by ITE land class with a constant number of squares per class.

Note that the 32 ITE land classes were reduced to 8 within strategies 3 and 4 for practical reasons.

5. Strategies were assessed separately by the selection of around 1000 squares to be surveyed by volunteers and comparison with data from the New Atlas of Breeding Birds in Britain and Ireland, the Pilot Census Project and moorland bird surveys, to estimate the range of species which would be monitored at given level of precision at the scale of the UK, individual countries and EC regions. The calibration curves necessary to make such comparison are described.

6. The findings are discussed in relation to the aims of the Breeding Bird Survey. Overall, the four strategies were similar in the range of species monitored. The selection of squares based on a regular grid (2) performed worst of the four strategies. Stratification by land class (3 & 4), which is a measure of landscape type, provided monitoring which was comparable with the simplest strategy (1), that is, stratification by land class did not significantly improve the range of species monitored over random sampling contrary to what might have been expected. This is probably because land class did not provide a sufficiently sensitive measure of bird habitats within the models we assessed.

7. The results give a broad indication of likely species coverage within the Breeding Bird Survey under different sampling designs and at differing sampling intensities. The results should not be taken as a precise statement of likely coverage within the new scheme.

8. We conclude that the selection of Breeding Bird Survey squares should be based on random sampling with sample sizes proportion to observer densities.

Back to Research Reports 71-158