Marine Habitat Mapping Using High Spatial Resolution Multispectral Satellite Data

摘要

The application of high spatial resolution and multi-sensorrnsatellite was assessed to map the sensitive marine habitat ofrnTarut Bay in the western Arabian Gulf. The coastal zone ofrnthe Gulf is becoming more populated and industrialized withrntime, which requires an integrated monitoring approach tornmanage the valuable marine resources. In this study, SpectralrnMixture Analysis (SMA) and a water depth removal modelrnwere applied to derive sea-bottom reflectance data. Highrnspatial resolution Ikonos multispectral and panchromaticrnimagery are being integrated with SeaWiFS ocean coveragerndata to produce biotope maps for the project area. The imagernprocessing sequence includes linear spectral mixture analysisrnusing the first four Ikonos bands, application of the waterrnremoval procedure, another stage of spectral mixture analysis,rnand then machine learning based feature extraction of marinernecosystem biotopes including coastal beaches, coastalrnvegetation, mud flats, rock flats, tidal channels, seagrass beds,rnsubtidal sands, subtidal mud, and algal mats. Spectral MixturernAnalysis models the recorded radiance from each pixel as arnlinear combination of endmember materials. The result is arnset of fraction images that relate digital numbers to materialrnabundance. Water attenuation coefficient k is being extractedrnfrom the SeaWiFS eight spectral bands for depth modelingrnover the study area. The goals of this study include: anrnassessment of the capability of high-resolution multispectralrndata and new remote sensing processing techniques for coastalrnmonitoring, evaluation of the depth modeling algorithm forrnmitigating effects of the water column to map target features,rndevelopment of a prototype geodatabase model for integratingrndiverse data about targets in a large enterprise database, andrnassessment of whether these techniques will lower the costs ofrnexpensive field data collection programs in thernmarine environment.