With remote sensing information products becoming increasingly varied and arguably improved, scientific applications of such products rely on their quality assessment. In an operational context such as the NASA (Natio...With remote sensing information products becoming increasingly varied and arguably improved, scientific applications of such products rely on their quality assessment. In an operational context such as the NASA (National Aeronautics and Space Administration) information production based on the MODIS (Moderate Resolution Imaging Spectroradiometer) instrument on board Earth Observing System (EOS) Terra and Aqua satellites, efficient ways of detecting product anomaly, i.e., to discriminate between product artifacts and real changes in Earth processes being monitored, are extremely important to assist and inform the user communities about potential unreliability in the products. A technique for anomaly detection, known as MAD (the median of absolute deviate from the median), in MODIS land products via time series analysis is described, which can handle intra- and in-ter-annual variation in the data by using MAD statistics of the original data and their first-order difference. This method is shown to be robust and work across major land products, including NDVI, active fire, snow cover, and surface reflectance, and its applicabil-ity to multi-disciplinary products is anticipated.展开更多
基金Funded by the National 973 Program of China(No.2006CB701302).
文摘With remote sensing information products becoming increasingly varied and arguably improved, scientific applications of such products rely on their quality assessment. In an operational context such as the NASA (National Aeronautics and Space Administration) information production based on the MODIS (Moderate Resolution Imaging Spectroradiometer) instrument on board Earth Observing System (EOS) Terra and Aqua satellites, efficient ways of detecting product anomaly, i.e., to discriminate between product artifacts and real changes in Earth processes being monitored, are extremely important to assist and inform the user communities about potential unreliability in the products. A technique for anomaly detection, known as MAD (the median of absolute deviate from the median), in MODIS land products via time series analysis is described, which can handle intra- and in-ter-annual variation in the data by using MAD statistics of the original data and their first-order difference. This method is shown to be robust and work across major land products, including NDVI, active fire, snow cover, and surface reflectance, and its applicabil-ity to multi-disciplinary products is anticipated.
