Euphausiacea
Production

Copepods display determinate growth, and somatic production ceases on sexual maturity. Consequently, adult production can be monitored simply as the rate of egg production. By contrast, euphausiids display indeterminate patterns of growth and continue to invest in somatic production, as well as reproduction on attaining sexual maturity. Larval production can be estimated by ship-board incubation studies using much the same methodology employed to determine copepodite production (e.g. Hutchings et al., 1995). Adult production can be estimated from both laboratory incubation experiments (e.g. Stuart, 1992) and studies of population dynamics (Barange and Stuart, 1991; Tarling, 1995).
Methods for analysing population dynamics generally assume that recruitment takes place in discrete pulses, resulting in cohorts that can be traced through time as modes in length frequency diagrams. Many subtropical and temperate euphausiid species have one year life cycles and generally show unimodal length frequency distributions. Subpolar and polar species may survive and breed for several years and cohorts often overlap in size as a result of variation in growth rates. Where the overlap is small , one can apply the "Petersen" method and assume that each of the modes represents a single age class. Greater overlap requires more detailed graphical techniques to separate cohorts. The first techniques were developed by Harding (1949) and Cassie (1954) and involve making projections on probability paper to determine the proportion of similarly sized individuals belonging to adjacent cohorts, assuming that cohorts contain normal (or log-normal) size distributions. These techniques are simple and still commonly used but an alternative method was provided by MacDonald and Pitcher (1979) which fits the most probable underlying cohort structure to a polymodal size distribution. The problem is that the number of age classes needs to be already known which, in the case of euphausiids, is sometimes difficult. In a subsequent version of their distribution fitting software, "MIX; Ichthus data systems" (MacDonald and Green, 1988), one can avoid this problem by assuming that growth follows the von Bertallanffy function, but this curve may not always be the best fit to natural euphausiid growth patterns (Labat and Cuzin-Roudy, 1996). Making assumptions about growth and reproduction is unavoidable in size frequency analysis and it is important that they are detailed and justified in each case (Grant et al., 1987).

The desired optimum frequency of sampling an identifiable population depends on the rates of change of the population. For most euphausiid species in the South Atlantic region, sampling about every 10 to 14 days should be sufficient to describe the changes taking place. Samples must be integrated over the whole water column in neritic areas and up to 500 m in oceanic regions to avoid any bias caused by possible vertical segregation of different size categories (e.g. Williams and Lindley, 1982; Pillar et al., 1989). Length measurements must be kept standard, the most generally recognised dimension being the base of the eye stalk and the posterior end of the uropods, excluding their terminal setae, although carapace length may be more reliable in preserved specimens. Between 200 and 500 animals should be measured per sample to meet the statistical requirements of subsequent analysis (France et al., 1991).

Calculating the integrated production (IP) of a cohort (Kimmerer, 1987) firstly involves converting lengths to dry weight through the use of an empirically derived length weight equation (see Bird and Prairie, 1985; Giguere et al., 1989) and estimating densities of individuals per sample (Pommeranz et al., 1980). A number of methods may then be used (see Edmondson and Winberg, 1971; Omori and Ikeda, 1984) such as Allen curves, successfully applied to various euphausiid species by Lindley (1978, 1980, 1982), and removal summation, used by Berkes (1977) on Thysanoessa raschii. The exponential equation method (Gillespie and Benke, 1979; LeBlond and Parsons, 1977) is useful when sampling intervals are large.

In areas where reproduction is continuous, it is not always possible to trace cohorts through time and IP values may otherwise be obtained through "size frequency" methods. Winberg (1971) developed an equation requiring information on the developmental times of specified weight increments. Stuart and Pillar (1988) inferred developmental times from a combination of larval and juvenile laboratory rearing experiments and a limited number of adult size frequency distributions which showed modal progression over time. IP may also be calculated through applying weight dependent physiological models to size frequency distributions as carried out by Tarling (1995) on southwest Atlantic euphausiid species. However, a greater number of physiological studies on euphausiids is necessary to determine the sensitivity of such models.

Population dynamic techniques generally underestimate production because periods of intense production are often missed. "Size frequency" methods generally incur larger biases than "cohort" methods because the conditions for size frequency methods are frequently violated (Morin et al., 1987). Assessing the potential error of any production estimate is difficult but it is possible to calculate reliable confidence intervals for Allen curves using the "bootstrap" method (Efron, 1979) as shown by Morin et al., (1987). Therefore, if sampling intervals are suitably short and cohorts can be identified, the Allen curve method with added confidence intervals currently presents the best approach to estimating the integrated production of euphausiid populations.