||Reconstruction of data from the aquatic environment
||Lophaven, Søren Nymand (Informatics and Mathematical Modelling, Technical University of Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark)
||Technical University of Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark
||This thesis describes, applies and compares statistical methods for reconstruction of concentrations of Dissolved Inorganic Nitrogen (DIN) and Dissolved Inorganic Phosphorus (DIP) in Kattegat. The measurements are taken in the period from 1993 to 1997 within the monitoring program, which was implemented by the adoption of the Action Plan on the Aquatic Environment in 1987.
The aim of the reconstruction methods is to estimate the concentration of the two species for each week in the five year period, and at any location in Kattegat. The methods are general and could be applied to other parameters.
The spatial distribution of DIN and DIP have been computed by three different variants of kriging, i.e. ordinary kriging, universal kriging and cokriging. In order to have a sufficient number of observations per week, methods for temporal reconstruction of data have to be applied prior to the computation of spatial predictions. Two methods have been used for this purpose, these are the General Linear Model and locally weighted regression. The methods are compared from a statistical and a physical point of view.
Furthermore, the thesis applies different 3 dimensional approaches. These are 3 dimensional kriging, locally weighted regression and an ARIMA model. The first two methods are applied to raw data, i.e. measurements, while the ARIMA model is applied to different stations, where the time series are temporally reconstructed by the General Linear Model. Kriging is applied to the parameters of the model, in order identify these at any location in Kattegat. Such 3 dimensional methods are much less developed and described in literature compared to methods for analysis of strictly temporal or spatial data.
||Dissolved Inorganic Nitrogen; Dissolved Inorganic Phosphorus; the General Linear Model; locally weighted regression; (cross) semivariogram; anisotropy; ordinary kriging; universal kriging; cokriging; ARIMA modelling
Creation date: 2006-06-22
Update date: 2007-09-11