Optimisation of eDNA sampling strategies: A novel Lagrangian approach for identifying optimal sampling locations in marine environments

We present a novel modelling approach aimed at optimising sampling strategies for environmental DNA (eDNA) based ecosystem monitoring. We demonstrate this in a case study for the monitoring of an invasive mussel species in Auckland Harbour Bay. The chosen objective is to identify a set of ship-based sampling locations that provide the highest spatial coverage for a given number of samples collected by towing a filter through the water.
This objective is achieved by leveraging the high performance of the particle tracking model 'oceantracker', which simulates over a trillion (1e12) individual particles for this demonstration. We apply this particle tracking model - also known as a Lagrangian model - to an inverse problem by simulating a large number of potential mussel locations throughout the bay. The simulated mussels continuously shed cells containing their DNA into the surrounding waters, which drift with the currents and slowly decay, represented by a Poission process. By computing grid-based statistics on-the-fly, i.e. during runtime, we are able to reduce the unweildy amount of data generated by the trajectories of trillions of simulated particles to obtain a compact dataset containing eDNA counts and concentrations for each potential source location. This data set is then used in the optimisation problem to identify the sets of ideal sampling locations that exceed the detection threshold for the largest number of potential source locations, i.e. have the highest coverage.
As ship-based monitoring is an expensive task, we suggest that the use of a modelling framework such as the one demonstrated here could help to reduce this cost and provide an empirical solution to the selection of sampling locations.