Identifying declines in waterbirds: The effects of missing data, population variability and count period on the interpretation of long-term survey data

Abstract To manage and conserve wildlife populations effectively it is necessary to use methods that identify the often non-linear trends in populations, have an inbuilt assessment of trend quality and can analyse count data from a range of spatial scales. We present a method of trend analysis using generalised additive models. These produce smoothed indices of abundance that can be used to assess population change from one or more sites or time periods, with any number of estimates of abundance per index period. We apply this method to count data collected under the Wetland Bird Survey, a national scheme that monitors waterbirds in the United Kingdom. To highlight declining populations, ‘alerts’ were raised if the population decline was equal to or greater than 50%. Significance was determined using bootstrapped confidence intervals for analyses that included many sites, or a novel Monte-Carlo method for single site analyses. The impact of missing data, species count variability and the number of months used to calculate the population change was greater at individual sites than for national datasets, which were relatively insensitive to changes in the above parameters. For single sites it is essential that three or more counts be made per index period if reliable estimates of population change are required. We propose that the method presented could be applied to a wide range of national or other monitoring schemes for a variety of taxa.