On current compositing algorithms

Several compositing techniques have been employed to minimize the influences of external factors such as the clouds and atmosphere on the signal measured by the multitemporal Advanced Very High Resolution Radiometer (AVHRR). However, the composited products are not satisfactory because some of the noise remain and some valuable data are lost after the compositing process. This is due mainly to the fact that the current pixel selection criteria are not specific enough to retain all valuable pixels while eliminating noisy ones and that the index or classifier used in compositing is too vulnerable to external factors. To improve composite products, therefore, both the compositing algorithm itself and its classifier need to be modified or corrected for external effects. In this study, the current compositing algorithms are revised and an alternative set of pixel selection criteria is proposed. The proposed algorithms use the combinations of the red, near‐infrared, thermal channel, and biological characteristics of vegetation to select pixels. In addition, the traditional classifier, normalized difference vegetation index, was replaced with other vegetation indices that were proved to be less sensitive to external influences. The new approach was applied to a 1992 AVHRR data set collected from the Hapex study sites of varying biome systems in Niger. The results showed that the proposed algorithms improved the AVHRR time series quality and re‐claimed the data lost by current compositing algorithms, while noise was further reduced.