Mapping shallow, benthic communities using hyperspectral, remotely sensed data: A test of habitat classification using multiple sources of bathymetry

摘要

Optical remote sensing technologies are increasingly used to map shallow, marine benthic communities. This approach largely relies on the use of bathymetric data, which vary greatly in accuracy and spatial resolution according to the source. We explored the effect of varying the quality of bathymetry on the output of classification maps of the invasive green alga Codium fragile ssp. tomentosoides along the Atlantic coast of Nova Scotia, Canada, using high resolution (1-m pixel) airborne hyperspectral imaging of a shallow ( < 6 m depth) subtidal zone, scuba-based ground truthing, and bathymetric data at two spatial resolutions ( ～10 m and > 300 m). Our preliminary results indicate that the abundance of C. fragile derived using coarse resolution bathymetry can be up to 6 times greater than that based on a finer resolution. The omission of bathymetric data in an unsupervised classification (based on data streams retained by Principal Component Analysis) yielded only slightly higher estimates of abundance than classification using fine resolution bathymetry. These findings suggest a precautionary approach to the use of bathymetry in classification of shallow habitats in optically dense waters based on remotely-sensed data. This practice can result in substantial overestimation of the occurrence of specific habitat types or species, which may mislead those charged with the study and management of marine ecosystems and coastal resources.