Effects of density dependence on diel vertical migration of populations of northern krill: a genetic algorithm model

Net and acoustic studies of diel vertical migration (DVM) in krill often show a degree of dispersion around the mean population depth, which becomes greater during night-time. Trade-off models can predict optimum depths over diet cycles but rarely explain why there is vertical scatter and why aggregations disperse at certain times. We examined density-dependent factors as a potential explanation for these phenomena. A Genetic Algorithm model was developed that predicted DVM,in a krill population based on internal state (i.e. levels of energy reserves), risk of predation and location of conspecifics. The modelling approach was designed to be dynamic in that optimal policies could respond to changing circumstances through time. Parameterisation of the model was achieved through measurements made in the Clyde Sea Area on northern krill Meganyctiphanes norvegica and its environment. Light intensity at depth was used to assess the level of risk of visual predation. Food provision was a mixture of vertically stratified phytoplankton and vertically migrating copepods. A negative exponential function was used to simulate density dependence in the food returns at each depth. Sensitivity analyses involved alterations to the level of density dependence and the metabolic rate. DVM was predicted in all sensitivity analyses and each correlated positively with net catch and acoustic observations. Increased density dependence in feeding success did not affect the mean depths chosen at night but did increase the spread of the population. The closest fit to observations was achieved when the metabolic rate was lowered and risk of mortality rate was assessed over a yearly rather than daily period. The model predicted that the population should spread more under low food conditions. We recommend that density-dependent factors be included in future state-dependent models predicting krill behaviour and life-cycle patterns