Classification and Change Detection of Yellow Sea Intertidal Sediment Using a Two-step PCA of Optical Reflectance Data

Abstract:

Grain size distributions of intertidal sediment in the Yellow Sea vary widely ranging from mud dominant to sand dominant with extensive seasonal changes. These are affected by tidal energy of the Yellow Sea, river flows, topography, shoreline gradient and human activities such as land reclamation, etc. Grain size of tidal flats are linked closely to fisheries, aquaculture and pollutant process. Therefore, it is necessary to monitor these areas continuously. It is, however, difficult to retrieve grain size information from remotely sensed data. Because the optical reflectance of intertidal sediment is not a function of single parameter but varies according to water content, grain size, topography, surface water, benthic algae and halophytes, etc. Among these parameters, grain size and water content play a key role in bare intertidal surface. Since water content of intertidal sediment are affected by tide, it is necessary to establish a water-independent grain size retrieval model. It is known that mud and sand sediment are well distinguished under dry condition on PCA (Principal Component Analysis) space but hardly distinguished under saturated condition. Here we introduce a new grain size retrieval model by removing the water content dependency from optical reflectance via a two-step PCA transform. To define the relationship between grain size, water content and optical reflectance, two different standard samples were made as per grain size by wet sieving. By exploiting simplified reflectance features of the standard samples, a two-step PCA transform model was established. This grain size retrieval model was applied to GOCI (Geostationary Ocean Color Imager) images for sediment classification within the Yellow Sea. The results demonstrate it might be possible to discriminate between sand-dominant and mud-dominant areas based upon the model. Seasonal changes of sediment distribution within the tidal flats are well observed from the results.