| Popis: |
Many monitoring programs provide annual indices of relative change over time in some quantitative measure of ecological status, such as population abundance or species richness. These indices are usually scaled relative to a reference year so that they represent change in ecological status compared to this particular year. An issue with this approach is that uncertainty about ecological status in the reference year can propagate into large uncertainty in all other index values. Taking instead the mean of the ecological status over several years as the reference—a reference period—may reduce uncertainty in indices. At present, this approach is not commonly used in practice. I quantitatively evaluate how the choice of reference period affects the uncertainty of two variants of population indices, either estimated independently each year or smoothed over several years, for 100 bird species using monitoring data. Short reference periods containing years early in the series lead to reduced uncertainty in independently estimated index values, but not in smoothed indices, compared to when using a single reference year. When a long reference period was used, uncertainty was substantially reduced for independently estimated annual indices in particular, but also for smoothed indices. An exception to the reduction in uncertainty with the length of the reference period was found when indices are constrained to be log-linear. Given an appropriate model and indices that are not strictly log-linear, using smoothing and/or reference the periods can be useful ways of reducing irrelevant uncertainty in the presentation of indices. |
