Spatiotemporal Continuous Shallow Water Bathymetry from a Kriged Kalman Filter

In GIScience, problems of missing data in space or time are nontrivial. We implemented a Kriged Kalman filter (KKF)–based data interpolation and assimilation technique and tested it for mapping bathymetry at unsampled locations and times. This technique integrates the Kriging and Kalman filter computation frameworks to perform spatiotemporal data assimilation, which can produce spatially and temporally continuous bathymetric fields from samples that are scarce in space and time. The spatiotemporal bathymetric field over the estuary of the Yangtze River was mapped based on the four boat-based depth echo-sounding surveys conducted in 1982, 1997, 2002, and 2010. Our validation and verification analyses showed that the KKF assimilation model can predict bathymetry accurately and reliably at unsampled locations and times. This paper demonstrates that KKF is superior to traditional spatial interpolation methods because it informs the interpolator with the temporal component that also extends the prediction to the time do-main. The experiments indicate that greater time intervals in conducting bathymetric surveys result in a more pronounced influence on the performance of KKF than the spatial sparsity of depth samples. The ability of space-time prediction of bathymetry allows underwater depth measure-ments to be accurately aligned with satellite images, which is essential for improving multispectral image inversion in bathymetry studies.