A multiple regression model for prediction the food consumption of Marine Fish populations

Abstract

The construction of trophic (food web) models of ecosystems, as needed for both theoretical and practical purposes such as fisheries management, requires estimates of food consumption (Q) by each of the various species (groups) included in the model. These estimates are usually required on a per-biomass (B) basis, i.e. as estimates of the ratio of the food consumed to the weight of the consumers (Q/B) during a stated period. For estimates of Q/B to be most useful, they must take account of: (i) seasonal fluctuations of food intake; (ii) the age/size structure of the population; and (iii) the type of food consumed.

In this study, 33 estimates of Q/B are reviewed, and an empirical multiple regression model for prediction of Q/B is presented which incorporates points (i) to (iii) above. The predictor variables are: (a) the asymptotic weight of the fish of the study population, (b) the aspect ratio of their caudal fin (as a measure of the average activity and/or metabolic levels of the fish), (c) the mean habitat temperature and (d) the food type (a dummy variable, 0 ih,carnivores and 1 in herbivores).

The model explains nearly 75% of the variance in the data set used, which includes myctophids and tunas, flatfishes, rabbitfishes, and other groups from both tropical and temperate waters. The implications of this model for bioenergetics are discussed, along with its future extension, to be based on a much larger data set.