Nonparametric Assessment of Multimodality for Size Frequency Distributions

Abstract

The multimodal nature of size frequency distributions is a common occurrence in fisheries analysis with modes representing potential Gaussian components in a mixed distribution. This characteristic is exploited to estimate and assess every component using any of the several parametric models that have been proposed. In general, the success of these procedures is dependent on the smoothness of the frequency distribution and the intelligent (data base and complemented with additional biological information) input of initial values. In this study we utilized nonparametric density estimators in combination with several rules for smoothing parameter (bandwidth) selection and a smoothed bootstrapping of critical bandwidths procedure to investigate multimodality of the length distrubtion of Japanese sea bass "suzuki" (Lateolabrax japonicus). These methods led to useful data-based estimations with statistically signifcant number of modes that produced growth estimations consistent with those obtained from fish aged by means of hard-parts (scales and otoliths) reading, specially at smaller ages. The resulting von Bertalanffy growth expressions had parameters inside the range of those reported in the literature. These nonparametrics methos prove to be an alternative valuable tool for the analysis of mixed size distrubtions of fish.