Abstract
The Landsat Thematic Mapper (TM) has provided fine spatial resolution data spanning two decades. These data are useful for long‐term studies of environmental change. However, temporal factors such as sensor degradation, variation in Sun–target–satellite geometry, and variable atmospheric conditions can create inconsistencies in multi‐temporal images and complicate data analysis. This study investigated temporal influences on satellite data. The methodology was developed based on a theoretically derived relationship between pixel values of pseudo‐invariant features (PIFs) across time. The relationship was validated using multi‐temporal Landsat‐5 TM images, which showed that temporal factors contribute to PIF pixel values in both multiplicative and additive ways. For Landsat‐5 TM level‐0 data, temporal influences were simulated in terms of multiplicative and additive components. The results showed that atmospheric variation is the most influential factor, followed by variation in the Sun–target–satellite geometry and TM sensor degradation.
Acknowledgements
The Remote Sensing Technology Center of Japan provided the Landsat TM data. Here we express our thanks to Prof. S. Nishiyama for his kind assistance in data acquisition. This work was sponsored by a research fellowship at Arid Land Research Center, Tottori University, Japan. It is also supported by a grant from the Ministry of Education, Culture, Sports, Science and Technology of Japan, “Dynamics of the Sun‐Earth‐Life Interactive System, No.G‐4, the 21st Century COE Program”.