Spatiotemporal Forecasting of Mediterranean Blue Mussel Biofouling Settlement to Inform Aquaculture Management

Mediterranean blue mussel ( Mytilus galloprovincialis ) settlement significantly influences aquaculture productivity and biofouling management in New Zealand's coastal waters due to its rapid colonization of farming infrastructure and its competitive interactions with cultivated species. In this study, we develop a regression‐based forecasting framework to predict monthly Mediterranean blue mussel spat settlement using temperature‐driven temporal features. Key environmental predictors include mean air temperature and sea surface temperature (SST), along with their lagged and rolling derivatives, which capture both seasonal patterns and short‐term thermal dynamics. To account for spatial variability, geographically proximate monitoring sites are grouped into 11 representative clusters using a spatial clustering algorithm. Forecasting is performed with the XGBoost regression model, which is well suited to nonlinear ecological processes and feature interactions. Model performance is evaluated using a Leave‐One‐Group‐Out (LOGO) cross‐validation strategy to assess generalization to previously unseen locations. The results demonstrate strong predictive skill in multiple regions and highlight the potential of this approach to support proactive aquaculture management and mitigate settlement‐driven biofouling impacts.