Search

Forage fish populations often undergo large and rapid fluctuations in abundance. However, most of their predators are buffered against such fluctuations owing to their slower pace of life, which allows them to maintain more stable populations, at least during short periods of food scarcity. In this study, we investigated top-down processes exerted by seabirds on forage fish stocks in five contrasted marine ecosystems, compiling numerous data sets on seabird counts, diets, energetic needs and prey energy content and abundance. Off Norway, South Africa, Peru, Sweden and Scotland, we found that predation pressure-estimated as the proportion of a fish stock consumed by seabirds-was generally low (median = 1%), but increased sharply at low levels of prey abundance. When prey biomass decreased below 15-18% of its maximum recorded value, predation by seabirds became a source of important additional pressure on prey stocks (similar to 20% of prey biomass is consumed by seabirds). An earlier empirical study advocated for keeping forage stocks from falling below a threshold of 33% of long-term maximum prey biomass in order to safeguard seabird breeding success, but here we further suggest that a threshold of 18% should be considered as a limit not to be exceeded for the sake of the forage fish themselves, and below which extra cautious management of fisheries may be required. Nevertheless, despite exceptionally high rates of predation on some occasions, predation pressure was not correlated with prey dynamics, suggesting an absence of prey entrapment due to seabirds alone in these five ecosystems.

Suppose you are given some dataset drawn from an underlying probability distribution ¤ and you want to estimate a “simple” subset ¥ of input space such that the probability that a test point drawn from ¤ lies outside of ¥ equals some a priori specified ¦ between § and ¨. We propose a method to approach this problem by trying to estimate a function © which is positive on ¥ and negative on the complement. The functional form of © is given by a kernel expansion in terms of a potentially small subset of the training data; it is regularized by controlling the length of the weight vector in an associated feature space. We provide a theoretical analysis of the statistical performance of our algorithm. The algorithm is a natural extension of the support vector algorithm to the case of unlabelled data.