Cluster analysis of linear model coefficients under contiguity constraints for identifying spatial and temporal fishing effort patterns

Abstract

For fisheries management purposes, it is essential to take into account spatial and seasonal characteristics of fishing activities to allow a reliable assessment of fishing impact on resource. This paper presents a novel technique for describing spatial and temporal patterns in fishing effort. The spatial and seasonal fishing activity patterns of the French trawler fleet in the Celtic Sea during the period 1991–1998 were analysed by modelling fishing effort (fishing time) with generalised linear models. The linear model for fishing effort included fixed effects for both spatial (statistical rectangles) and temporal units (months). In addition, spatial correlations in any given month were modelled by an exponentially decreasing function. Temporal correlations were included using the previous month's fishing effort for a given spatial unit as predictor. A method based on cluster analysis of estimated model coefficients of spatial or temporal fixed effects is proposed for identifying groups of similar spatial and temporal units. A contiguity constraint is imposed in the clustering algorithm, ensuring that only neighbouring spatial units or consecutive temporal units are grouped. The cluster analysis identified 22 spatial and 9 temporal groups. Winter and spring months stood out as being more variable than the remaining months. Spatial groups were of varying size, and generally larger offshore. The proposed method is generic and could for example be used to analyse temporal and spatial patterns in catch or catch rate data.