We discuss methodological and implementation issues of spatial, temporal and combined spatio-temporal backtracking and illustrate larval backtracking for North Sea lesser sandeel Ammodytes marinus larvae, using a combined hydrodynamical and individual-based model. It was found that dispersal effects are important for larval backtracking predictions. Our results show large differences in average transport distance, as well as in shape and extent of predicted hatch areas, when backtracking advected larval cohorts in different regions of the North Sea, thus emphasizing the importance of using realistic, spatially and temporally resolved diffusivity fields in simulations of larval transport. In all cases, biologically likely hatching areas have been predicted. We discuss issues of methodological consistency and present a new scheme for including life-history stochasticity effects on growth in backtracking in a consistent way, as well as procedures for assessing the effects of larval mortality. Finally, fundamental limitations of larval backtracking are clarified, most importantly the time horizon and spatial resolution limit for backward prediction.KEY WORDS: Individual-based modelling · Backtracking · Larval transport · Inverse stochastic methods · Hatch area identification 347: 221-232, 2007 hatching sites. We go one step further and consider growth backtracking as well in a fully realistic hydrodynamic setup. Compared to passive particles, larvae have ontogenetic development as well as active behaviour, which complicates backtracking, due to the stochastic and nonlinear nature of these processes.
Resale or republication not permitted without written consent of the publisher
OPEN PEN ACCESS CCESSMar Ecol Prog SerAt first sight, the issue of tracing backwards the state of a larval sample in time appears simple: just advect larvae in a direction opposite the currents and shrink their size by the same amount as their growth, if time is running forward (Pedersen et al. 2000, Allain 2004); however, dispersion processes are present in the ocean and must be considered, as we will show later, because otherwise we have no information on the error margin in results and further simple backtracking opposite current lines may contain an error bias, since dispersion processes in the ocean are not spatially uniform. Batchelder (2006) similarly found that diffusion cannot be ignored in backtracking. Alternatively, larval origin might be estimated by a forward approach by releasing a vast number of particles in a forward simulation at potential spawning grounds and focussing on the small number that arrives in the area of interest (Allain et al. 2003). While it is straightforward and circumvents some of the methodological issues of backtracking, forward tracking remains inefficient for spawningground identification because the majority of particle trajectories are useless (they end up at a position different from the area of interest). This is especially true if arrival in the area of interest is a rare event (e.g. i...