기상위성 3차산출물 생성을 위한 시공간 통계기법 연구

Abstract

Recently, As Korea began to operate COMS (Communication, Ocean and Meteorological Satellite), it can continually collect large amounts of real-time information over land, ocean and atmosphere at East-Asia and Oceania. COMS collect various weather data which is used for producing 16 species products by using CMDPS (COMS Meteorological Data Processing System) through one visible channel and four infrared channels. 16 species products can be used in weather forecasting, numerical weather prediction and climate change predictions. Although these important position of 16 species products, there is no countermeasure to check and correct error could be occurred by aging sensor or false alarm. Thus research about the methods to increase data quality and to reproduce level 3 products refined from level 2 products is needed. A lot of survey about outlier detection and removal to intensify these vulnerable points performed in various field. This paper presents the algorithm to detect outliers that occur in the COMS data using spatio-temporal Statistic Methods. Detected outliers are removed for the purposes of the Level 3 product because it is not a definite value. Instead, as the basis of the removal, we were calculated estimation using the spatio-temporal interpolation methods. Estimation was added as a reference information to the metadata after testing their effectiveness. In order to solve the problem of over or underestimates by the missing data, the synthesis about time was used ARIMA and timeshare average. Resampling of space was used Nearest neighbor and Bicubic. This dissertation investigate the availability of outlier detection algorithm and estimation applying the selected 7 species possessed a numeric type value, of 16 species level 2 products. Each algorithm is implemented by using the JAVA and R language to process numerous satellite data efficiently. As this study suggests standard of creating COMS level 3 products, it could contribute to enhancing the applicability of COMS data.