Robust model-based indicators of regional differences in food-web structure in the Southern Ocean

Efforts to model marine food-webs are generally undertaken by small teams working separately on specific regions (<106 km2) and making independent decisions about how to deal with data gaps and uncertainties. Differences in these largely arbitrary decisions (which we call ‘model personality’) can potentially obscure true differences between regional food-webs or lead to spurious differences. Here we explore the influence of model personality on a comparison of four Southern Ocean regional food-web models. We construct alternative model versions which sequentially remove aspects of personality (alternative model ‘currencies’, schemes for aggregating organisms into functional groups, and energetic parameter values). These alternative versions preserve regional differences in biomass and feeding relationships. Variation in a set of model metrics that are insensitive to absolute biomass and production identifies multiple regional contrasts, a subset of which are robust to differences in model personality. These contrasts imply real differences in ecosystem structure which, in conjunction with differences in primary production and consumer biomass (spanning two and four orders of magnitude respectively), underpin differences in function. Existing regional models are therefore a useful resource for comparing ecosystem structure, function and response to change if comparative studies assess and report the influence of model personality.