Assessment of 3D Spatial Interpolation Methods for Study of the Marine Pelagic Environment

Given the volumetric nature of the ocean, 3D spatial modeling and interpolation could be a key to a better understanding of continuous abiotic and biotic phenomena that compose the marine ecosystem, although such techniques are rarely used and their actual performance is poorly studied. Here, we evaluate the performance of 3D spatial interpolation for five pelagic variables derived from a typical oceanographic campaign conducted in the southeastern Beaufort Sea (Canadian Arctic) in 2009. Our main objective is to evaluate and compare the performance of a deterministic interpolation method (inverse distance, IDW) and a geostatistical method (ordinary kriging, OK) with a variation of method input parameters (search neighborhood, weights) for variables with increasing complexity in terms of data anisotropy and sampling configuration. Performance of different 3D interpolation strategies is evaluated by cross-validation and a qualitative comparison of 3D spatial models. Our results show that OK was the optimal method. However, when the complexity of pelagic variables increased in terms of spatial autocorrelation and data variation, the error difference between OK and IDW was reduced. We recommend that recent advances in spatial 3D modeling tools developed primarily for geological modeling should be exploited to extend the usual interpretation of marine pelagic phenomena from a 2D to a 3D environment.