Date of Award


Document Type



Edith Cowan University

Degree Name

Doctor of Philosophy


School of Engineering


Faculty of Computing, Health and Science

First Supervisor

Associate Professor Ute Mueller


Geostatistical methodology has been employed in the modelling of spatiotemporal data from various scientific fields by viewing the data as realisations of space-time random functions. Traditional geostatistics aims to model the spatial variability of a process so, in order to incorporate a time dimension into a geostatistical model, the fundamental differences between the space and time dimensions must be acknowledged and addressed. The main conceptual viewpoint of geostatistical spatiotemporal modelling identified within the literature views the process as a single random function model utilising a joint space-time covariance function to model the spatiotemporal continuity. Geostatistical space-time modelling has been primarily data driven, resulting in models that are suited to the data under investigation, usually survey data involving fixed locations. Space-time geostatistical modelling of fish stocks within the fishing season is limited as the collection of fishery-independent survey data for the spatiotemporal sampling design is often costly or impractical. However, fishery-dependent commercial catch and effort data, throughout each season, are available for many fisheries as part of the ongoing monitoring program to support their stock assessment and fishery management. An example of such data is prawn catch and effort data from the Shark Bay managed prawn fishery in Western Australia. The data are densely informed in both the spatial and temporal dimensions and cover a range of locations at each time instant. Both catch and effort variables display an obvious spatiotemporal continuity across the fishing region and throughout the season. There is detailed spatial and temporal resolution as skippers record their daily fishing shots with associated latitudinal and longitudinal positions. In order to facilitate the ongoing management of the fishery, an understanding of the spatiotemporal dynamics of various prawn species within season is necessary. A suitable spatiotemporal model is required in order to effectively capture the joint space-time dependence of the prawn data. An exhaustive literature search suggests that this is the first application of geostatistical space-time modelling to commercial fishery data, with the development and evaluation of an integrated space-time geostatistical model that caters for the commercial logbook prawn catch and effort data for the Shark Bay fishery. The model developed in this study utilises the global temporal trend observed in the data to standardise the catch rates. Geostatistical spatiotemporal variogram modelling was shown to accurately represent the spatiotemporal continuity of the catch data, and was used to predict and simulate catch rates at unsampled locations and future time instants in a season. In addition, fishery-independent survey data were used to help improve the performance of catch rate estimates.


Paper Location