Kim, H.; Lee, J., and Kim, Y., 2021. Tidal creek mapping from airborne LiDAR data using multi-resolution cloth simulation filtering. In: Lee, J.L.; Suh, K.-S.; Lee, B.; Shin, S., and Lee, J. (eds.), Crisis and Integrated Management for Coastal and Marine Safety. Journal of Coastal Research, Special Issue No. 114, pp. 86–90. Coconut Creek (Florida), ISSN 0749-0208.
Tidal creeks are transitional waterways promoting the evolution and expansion of tidal flats. The precise and objective delineation of tidal creeks is critical for monitoring the characteristics, formation, and evolution of tidal flats. Airborne light detection and ranging (LiDAR) data are the most widely used source of tidal topography information, since they can provide precise terrain information over wide areas. However, existing tidal creek extraction methods using airborne LiDAR data have limitations, such as the necessity of excessive user intervention and a lack of adaptability to the various shapes and widths of tidal creeks. The morphological irregularities, complexity, and diverse widths (from a few centimeters to several kilometers) of tidal creeks, complicates their automatic or manual extraction from LiDAR data. In this study, we propose an effective and practical method for the mapping of tidal creeks with a wide range of widths. Here, cloth simulation filtering (CSF), a verified ground filtering technique used to filter off-ground objects from point cloud in land LiDAR surveys, was adopted and modified. By sequentially filtering the creek points through a hypothetical cloth with an increasingly larger grid resolution, the proposed multi-resolution CSF can extract huge tidal creeks without compromising the details of narrow creeks. This sequential filtering is based on local thresholds calculated using the creek points extracted during the previous filtering and does not require empirical parameter adjustments. The results of an experimental evaluation based on airborne LiDAR data collected over the west coast of South Korea indicate that the accuracy of the proposed method is high (Kappa > 0.8) and superior to that of results obtained through an user parameter.