Error estimation for gridded bathymetry

Estimating the uncertainty or predicted accuracy of gridded products that are generated from historical bathymetric survey data is of high interest to the maritime navigation community. Surface interpolation methods used for gridding survey data in practice are well established. This paper investigates error estimation methods for gridded bathymetry in terms of their practical utility. Of particular interest are: 1) assessing the quality of a prior uncertainty of random error in survey data; 2) the significance of autocorrelated random errors; 3) the relationship between survey point density and propagated or product uncertainty; 4) the computational feasibility of Monte Carlo (MC) methods over large regions; and 5) the value of cross-validation to estimate error in the absence of controlled truth. K-fold cross-validation is used as the basis for performance evaluation of our approach to propagate a priori random errors via MC perturbation with spline-in-tension surface interpolation. Experiments are conducted with test areas in the Norwegian archipelago of Svalbard.