Environmental monitoring datasets often contain a large amount of missing values, and are characterized as being sampled over time on a distinct number of locations in the area of interest. This paper proposes a stochastic approach for modelling such data in space and time, by taking the spatial and temporal correlations in data into account. It has been applied to observations of dissolved inorganic nitrogen in the Kattegat during the period 1993-1997. Modelling results are shown as maps of the spatial distribution of dissolved inorganic nitrogen (DIN) in 4 weeks, representing the four seasons, and as time series of DIN at three different locations. However, the model approach could be applied to any space-time point given by a location in the Kattegat area and a week in the 5-year period 1993-1997. The results can be interpreted from a biological and physical point of view. Thus for the specific application the approach seems to perform very well. The results obtained could be used to improve status reporting of the environment, or as forcing functions for time series models and deterministic, hydrodynamic ecosystem models.
|Publication status||Published - 2006|