The study on cloud masking of GOCI optical images by using adaptive threshold and BRDF-based NDVI profiles for the land application

Abstract:

Geostationary Ocean Color Imager (GOCI) which was launched on 27 June 2010, developed to detect, monitor, and predict the ocean phenomena around Korea. Although GOCI was developed to observe the ocean environment, GOCI has also enormous scientific data for land surface. However, it is extremely important to extract the cloud pixels over the land surface to utilize its data for the land application. Over the land surface, the reflectance variation is higher and the characteristic of surface is more various than those over the ocean. Furthermore, the infra-red (IR) channel is not included in 8 GOCI bands, which is useful to detect the thin cloud and the water vapor with cloud top temperature. Nevertheless, GOCI has potential to detect the cloud using the temporal variation due to the characteristics of geostationary satellite observation. The purpose of this study is to estimate the cloud masking maps over the Korean Peninsula. For cloud masking with GOCI, following methods are used; simple threshold with reflectance and ratio of bands, adaptive threshold with multi-temporal images, and stable multi-temporal vegetation image. In the case of adaptive threshold, high variable cloudy when comparing with surface reflectance is used by comparing the surface reflectance by temporal based analysis. In this study, the multi-temporal NDVI data processed by the bi-directional reflectance distribution function (BRDF) modeling also used to reflect the relative solar-target-sensor geometry during the daytime. This result will have a substantial role for the land application using GOCI data.