Guaranteeing safety of navigation within the Netherlands Continental Shelf (NCS), while efficiently using its ocean mapping resources, is a key task of Netherlands Hydrographic Service (NLHS) and Rijkswaterstaat (RWS). Resurvey frequencies depend on seafloor dynamics and the aim of this research is to model the seafloor dynamics to predict changes in seafloor depth that would require resurveying. Characterisation of the seafloor dynamics is based on available time series of bathymetry data obtained from the acoustic remote sensing method of both single-beam echosounding (SBES) and multibeam echosounding (MBES). This time series is used to define a library of mathematical models describing the seafloor dynamics in relation to spatial and temporal changes in depth. An adaptive, functional model selection procedure is developed using a nodal analysis (0D) approach, based on statistical hypothesis testing using a combination of the Overall Model Test (OMT) statistic and Generalised Likelihood Ratio Test (GLRT). This approach ensures that each model has an equal chance of being selected, when more than one hypothesis is plausible for areas that exhibit varying seafloor dynamics. This ensures a more flexible and rigorous decision on the choice of the nominal model assumption. The addition of piecewise linear models to the library offers another characterisation of the trends in the nodal time series. This has led to an optimised model selection procedure and parameterisation of each nodal time series, which is used for the spatial and temporal predictions of the changes in the depths and associated uncertainties. The model selection results show that the models can detect the changes in the seafloor depths with spatial consistency and similarity, particularly in the shoaling areas where tidal sandwaves are present. The predicted changes in depths and uncertainties are translated into a probability risk-alert map by evaluating the probabilities of an indicator variable exceeding a certain decision threshold. This research can further support the decision-making process when optimising resurvey frequencies.

The spatial sampling often used to process and represent bathymetric data are of fixed grid resolution where the least depth value is stored in each grid cell. This results in Digital Elevation Models (DEMs) that are used to depict the underlying features of the seafloor. With the discretion of the user, the resulting DEMs used may either be of coarse resolution or a very fine resolution surface which provides as many details as possible. However, depending on the resolution of the data collected and the variability of the seafloor, the arbitrary user defined grid resolution is not the best option. Hence we address the problem of finding an optimal grid resolution for representing and processing the bathymetric data for the application of bathymetric risk assessment whilst maintaining computational efficiency. Here we adopt the quadtree decomposition approach. In addition, the research suggests the optimal criteria and standard deviation threshold, σ t h {\sigma -{th}} values for this particular application. These suggestions are still flexible and can be optimized for this application depending on the end user requirements. Previous studies have focused only on the splitting criteria or the constrained criteria to ensure that there is homogeneous accuracy over the entire dataset. However, an investigation into the threshold selection for the standard deviation, σ t h {\sigma -{th}} which describes the variability in the dataset is one of the most important splitting criterion, that is still lacking. Also, a new approach to store the depths in the grid in a time ordered approach for each epoch is shown. By optimizing the criteria for the quadtree decomposition and time series algorithm, the approaches shown in this paper provide the adaptive, accurate DEM which makes optimal use of the available bathymetric data for the Netherlands Continental Shelf (NCS) as the study area. This data preparation step forms the basis for developing a probabilistic approach to assigning hydrographic resurvey frequencies in the NCS.