A Model for Clustering Fish Community Structure with Application to Songkhla Lake Bi-monthly Catches 2003-2006

Abstract

Monthly catch weights in Songkhla Lake were collected over the period from January 2003 to December 2006, for a total of 126 species. Catch weights were first aggregated by species and combination of bi-monthly season of year and catching gear (set bag net, trap, or gill net) and were log-transformed to remove skewness. A regression model containing three species-season/gear components was then used to predict these outcomes. The first component was represented by the most species of estuarine and marine vertebrates as well as some invertebrates and reflected the fact that set bag net was the gear that resulted in the highest catches. The second component mainly represented freshwater fish and some marine invertebrates, and reflected the fact that most of these species were caught by gill nets. The third component focused on the seasonal fluctuations in catch weight. Such models can provide further tools in understanding of fish community structure clusterings in fishery landings.