Space-Time Geostatistical Analysis of King Prawn Catch Rate
Faculty of Computing, Health and Science
School of Engineering
We integrate traditional time series modelling of annual king prawn catch rate trends with space-time geostatistical interpolation and extrapolation to obtain estimates for the king prawn catch rate data for the 2004 fishing seasons. Classical decomposition is performed on weekly aggregated data to obtain a multiplicative model for the combined 2001 to 2003 seasons using a polynomial trend model and (lunar) weekly seasonal indices. The spatiotemporal semivariogram of the combined detrended and deseasonalised data for 2001 to 2003 is computed and modelled using a product-sum covariance model. We assess the ability of the model for the combined 2001 to 2003 seasons to predict the king prawn catch rate for May and July 2004 and show that short term catch rate prediction is possible with the use of the product-sum covariance model and the subsequent spatiotemporal kriging estimation process.