Thirteen years of monthly counts of Oystercatchers from the Oosterschelde and Westerschelde, SW. Netherlands, contained approximately 13% missing values. Several log-linear Poisson models which assumed independence of the data were used to fill in (impute) these missing values. Yet these models either filled poorly (when only main effects year, month, and site were considered) or were too complicated (when first order interaction effects were also taken into account) to provide a reliable description. Furthermore, the errors appeared to be dependent. Consequently, regression models which consider simple deterministic functions of time and temporal dependence between the errors, were used for a more parsimonious description of the underlying process. The model which assumed that after some point in time all monthly means are increased or decreased by some factor yielded the best fit. This factor differed among months. Using this model, a sudden change was estimated to have occurred in June/July 1983, three years before the completion of hydrotechnical works that resulted in a loss of about 17% of the intertidal area in the Oosterschelde. Winter and spring numbers have decreased, while summer and autumn numbers have increased.
All data in the Integrated Marine Information System (IMIS) is subject to the VLIZ privacy policy
