Abstract from BTO Research Report No 251:

Freeman, S.F., Baillie, S.R. & Gregory, R.D. (2001)

Statistical analyses of an indicator of population trends in farmland birds.

Executive summary

1. Methods for the statistical analysis of trends in a farmland bird indicator are developed in order to enable MAFF to make a more effective assessment of whether they are progressing towards their target of reversing the declines in farmland bird populations by 2020.

2. The indicator analysed here is based on 18 of the 20 farmland species included in an indicator published by DETR. Suitable data were not available for Rook and Barn Owl.

3. A smoothed indicator is produced by fitting generalized additive models to the individual species trends and then taking the geometric means of the resulting annual indices. Smoothing removes short-term fluctuations due to weather effects and sampling error.

4. The last point in such a time series is often a poor estimate, as can be demonstrated by comparing values of the indicator for the final year of a time series with those obtained for the same year once further years of data have been added into the analyses (Figure 3). We therefore recommend that the final estimate from any time series should not be used to assess the performance of farmland bird populations.

5. Confidence limits for the smoothed indicator and other statistics were calculated using a bootstrapping technique which avoids restrictive assumptions about the underlying distribution of the data. The indicator was measured with a high level of precision (Figure 5).

6. Estimates of rates of change of the smoothed indicator and their 95% confidence intervals provide a straightforward means of assessing trends in the indicator (Figure 4b).

7. An objective method of identifying turning points in the indicator is also presented. This will allow changes in the rate of change to be identified, and will thus provide a further aid to interpreting changes in the indicator.

8. The sensitivity of the indicator to individual species trends was examined. The indicator changed most when either Tree Sparrow or Whitethroat was omitted, dependent to a degree upon which year was chosen as a baseline for comparison. However, the indicator was not particularly sensitive to the omission of any of the 18 species used, despite the inclusion of species showing a wide range of population trends.

9. The Common Birds Census (CBC) will be replaced by the BTO/JNCC/RSPB Breeding Bird Survey (BBS) from 2001. CBC and BBS have both been operated between 1994 and 2000 providing 7 years of overlap between the two schemes. Methods of producing long-term population trends based on combining data from the two schemes are now being developed by the BBS partnership. Provided that this work is successful it will be straightforward to apply the techniques presented here to such a combined index.

Back to Research Reports 248 - 286