Population connectivity of many marine organisms is dependent on the extent of along- and across-shelf larval transport. We use a high resolution (100 m to 4 km) unstructured grid coastal circulation model based on FVCOM coupled with the offline particle tracking code FISCM to simulate movement and development of larval blue mussels (Mytilus edulis) along the coast of eastern Maine, which is an area of interest by itself due to its high tide, complex coastline, many bays and small rivers, strong coastal current, and numerous mussel populations spread across the bays. The circulation model resolves individual bays along the coast, and the combination of temporal and spatial validation improved confidence in the model's ability to recreate realistic flow patterns that determine particle movements. The model also demonstrates how temporally and spatially variable along- and across-shelf transports interact with biological traits to create dispersal patterns of blue mussel larvae. In agreement with the recruitment data, the model predicted settlement shows primarily two clusters: the Frenchman Bay cluster that appears to be self-seeding and the Pleasant Bay cluster that contributes both locally and downstream. The two clusters appear to be preconditioned by the reproductive output, while the longer distance connectivity relies on the coastal current, which requires the cross-shelf transport to bring larvae out to the coastal current and later on from the coastal current back to the neashore. The connectivity is less sensitive to timing within the bed specific two-week spawning windows. Lastly, the model suggests an increase in settlement in a warming scenario.