Zooplankton cohort analysis using systems identification techniques

The linear-transfer and lag-Manly models of zooplankton cohortdevelopment were examined using data generated from a thirdmore realistic model. The more realistic multi-transfer modelincluded variance in development rate among individuals. Thelinear-transfer model produced highly biased estimates of developmentrate under conditions of rapidly changing recruitment. Althoughits performance was improved by increasing the number of modeledstages and thus decreasing the rate of change in recruitmentcompared to stage duration, a positive bias remained. The lag-Manlymodel also produced positively biased estimates of stage durationgiven non-zero variance in development rates. A comparison ofthe models' performances under different simulated samplingregimes recommended the multi-transfer model. Use of the multi-transfermodel was illustrated by determining the development and mortalityrates of the brine shrimp, Artemia monica reared under threedifferent conditions of food and temperature corresponding tonatural regimes in Mono Lake, California. The experimental conditionsand sampling regime resulted in high relative standard errors(mean, 33%) in stage abundance estimates not atypical of zooplanktonsampling regimes in lakes. A Monte Carlo analysis was used todetermine the uncertainty in estimated parameters and determinethe level of stage aggregation which maximized the amount ofinformation derived from the experiments.