Key points are not available for this paper at this time.
Abstract In the Dutch sector of the North Sea, sea bird densities are recorded bi‐monthly by using airborne strip‐transect monitoring. From these data we try to estimate: (i) high‐resolution spatial patterns of sea bird densities; (ii) low‐resolution spatial‐average bird densities for large areas; and (iii) temporal changes in (i) and (ii), using data on Fulmaris glacialis as an example. For spatial estimation, we combined Poisson regression for modelling the trend as a function of water depth and distance to coast with kriging interpolation of the residual variability, assuming spatial (co)variances to be proportional to the trend value. Spatial averages were estimated by block kriging. For estimating temporal differences we used residual cokriging for two consecutive years, and show how this can be extended to analyse trends over multiple years. Approximate standard errors are obtained for all estimates. A comparison with a residual simple kriging approach reveals that ignoring temporal cross‐correlations leads to a severe loss of statistical accuracy when assessing the significance of temporal changes.This article shows results for Fulmaris glacialis monitored during August/September in 1998 and 1999. Copyright © 2005 John Wiley & Sons, Ltd.
Pebesma et al. (Thu,) studied this question.