Does artificial intelligence modelling have anything to offer traditional management of freshwater food resources?

Abstract

Management of freshwater systems and the ecosystem services they provide has become a multi-stakeholder activity. This requires information on resources and how to manage them to be disseminated to a wide range of users. While artificial intelligence modelling can provide a powerful tool in managing and understanding resources and their drivers, they can be confusing to many users. In this study, we explored the potential use of two alternative modelling approaches (Boosted Regression Trees (BRT) and Bayesian Belief Networks (BBN)) for managing three species of freshwater mahinga kai species-kākahi or kaeo (freshwater mussel), koura (freshwater crayfish) and tuna (freshwater eel). While the BBN model I sbetter for stakeholder communication, the BRT produced more accurate models for all species. However, variables identified as being important for predicting abundance and biomass of these species were often environmental parameters that cannot be managed to improve yield. The artificial intelligence modelling does provide some accurate linkages between the target species and their environmental drivers. Nevertheless, translating these relationships into management plans remains challenging. The models are clearly not a panacea for better resource management, but provide one more tool that might assist multi-stakeholder understanding of how best to manage freshwater resources.