FACTORS INFLUENCING DETECTION OF DENSITY-DEPENDENCE IN BRITISH BIRDS .1. POPULATION TRENDS

Citation
M. Holyoak et Sr. Baillie, FACTORS INFLUENCING DETECTION OF DENSITY-DEPENDENCE IN BRITISH BIRDS .1. POPULATION TRENDS, Oecologia, 108(1), 1996, pp. 47-53
Citations number
25
Categorie Soggetti
Ecology
Journal title
ISSN journal
00298549
Volume
108
Issue
1
Year of publication
1996
Pages
47 - 53
Database
ISI
SICI code
0029-8549(1996)108:1<47:FIDODI>2.0.ZU;2-R
Abstract
We question why density dependence has remained elusive in series of a nnual abundances of British birds. In particular an earlier study repo rted that significant temporal trends in abundances occur in up to 74% of time series from the Common Birds Census. Several studies showed t hat such trends can hinder detection of density dependence. Temporal t rends do not preclude the presence of density dependence and two publi shed tests for density dependence include temporal trends in the null hypothesis model. We explore the extent to which detection of density dependence was hindered by temporal trends in bird abundance data. We used a conservative method to test for trends. which found significant (P < 0.05) linear population trends in only 7 of 60 time series of ab undances (of 17-31 years) compiled from the Common Birds Census data. However both of the tests for density dependence that allow for trends and a third method gave P-values that were strongly influenced by the strength of trends, including trends that were not significant (P > 0 .05). This shows that density dependence may be falsely rejected or de tected when trends are present, even when these trends are weak and no t statistically significant. To circumvent this problem we detrended t he time-series prior to testing for the presence of density dependence . To minimize subjectivity we used simulated time series to check that this procedure did not increase the level of type I error (false reje ction of density independence). Additionally, we confirmed that the me thod gave acceptable levels of type II error where the test fails to r eject density independence in series generated using a density depende nt model. This showed that the detrending method was acceptable and re presents a major improvement in our ability to detect density dependen ce in time series that contain temporal trends. Detrending the bird ti me series increased the number of series in which significant (P < 0.0 5) density dependence was found from 10 (17%), when trends are ignored , to 27 (45%) when series are detrended. However, this rate of 45% is still surprisingly low by comparison to other taxa, and we believe tha t other factors may contribute to this, which we explore in the second of this pair of papers.