CCAMLR

The distributional features and physical characteristics of some 1500 krill (Euphausia superba Dana) aggregations detected and sized acoustically in the south-west Indian Ocean are described. The aggregations were on average considerably smaller than elsewhere in the Antarctic. Aggregation biomass was log-normally distributed with a mean of 0.18 t. Only 3% of the aggregations were heavier than 1 t. The aggregations were on average 4.2 km apart along-track, and on 10 of the 15 survey days were randomly distributed, judged by a Chi-square test based on expected Poisson statistics. The distribution of aggregations heavier than 1 t, and of aggregations 20 km on either side of such aggregations was similarly judged to be random. These observations suggest widespread randomness throughout the survey area, which is consistent with the absence of pronounced hydrographic or topographic concentrating mechanisms in the region. On the 5 non-random days, aggregations were smaller than average and were clustered on a scale of I km, suggesting that they could have been in the process of fragmenting or amalgamating. Estimates of the mean nearest-neighbour distance were derived for the random distributions. It is suggested that these could be useful in the development of fisheries-dependent indices of krill abundance.

Polar Biology (in press)

In this paper, I show how data that are routinely collected by survey vessels can be used to estimate the number of Concentrations of krill (in the sense of Butterworth (1988), Mangel (1988)) in a given region of the southern ocean. Sample computations are performed, using data collected by Soviet research/survey vessels in the early 1980s. These examples highlight the need for a navigational log as well as a fishing log in order to make accurate inferences.

Published in: Selected Scientific Papers, Volume 1, pp. 127-252

A model is set up for the operation (which includes both searching and fishing) of a Japanese krill trawler over a half-month period. It is based on an underlying krill distribution model whose parameters are determined primarily from the scientific FIBEX surveys. Output from the model of the operation is compared with (and partially tuned to) statistics for a sample of data from the commercial fishery. A major inconsistency is found: haul times are a factor of 4-5 times greater in reality than in the model. Two ad hoc model modifications are introduced to eliminate this inconsistency: artificially elongating krill swarms, and allowing hauls to continue through more than one swarm. Twenty four candidate abundance indices (generally of a CPUE form) for krill biomass in the 600 n.mile square oceanic sector modelled are considered, and their performance in response to a variety of ways in which the overall krill biomass might decline is investigated. Generally the indices respond by dropping relatively less than the proportional biomass decrease. Catch statistics collected at present (centred primarily on catch per fishing time) are of low utility in detecting biomass decline. Combination catch rate indices incorporating within-concentration search time give improved performances, but are able to monitor changes in within-concentration krill distribution parameters only. Indices that distinguish primary searching time from secondary searching time (searching while waiting to finish processing) within concentrations perform better, but collection of the requisite data may not be practical. Other approaches (e.g. research vessel surveys) need to be considered to monitor changes in the number, distribution and size of krill concentrations, both because there are doubts about the reliability of indices based on concentration searching time (which do respond to such changes), and because such indices are relatively imprecise. Priority needs to be given to improving the krill distribution model underlying the analysis; this probably requires that scientific surveys be planned to operate in small areas concurrently with fishing vessels.