ABSTRACT
Cloud and cloud shadow detection is an inevitable preprocessing step for analyzing Landsat time series. This chapter provides a comprehensive review of all the relevant algorithms. Based on the number of Landsat images used in the algorithm, we categorize the algorithms into two groups: single-date algorithms and multitemporal algorithms. Within the single-date algorithms, we further categorize them into two subgroups: physical-rules-based algorithms and machine-learning-based algorithms. Based directly on the physical characteristics of clouds and their shadows, the physical-rules-based approaches derive a set of static or dynamic rules from Landsat spectral bands and apply them for the detection of clouds and cloud shadows. Clouds and cloud shadows can also be identified using machine learning models, but these kinds of methods generally require large amounts of high-quality training data, which limits their usage. Finally, with the recent policy shift to no-cost, open-access Landsat data, cloud and cloud shadow detection algorithms may now also be based on multitemporal Landsat images. These methods can provide more accurate cloud and cloud shadow masks, especially for thin clouds. This review provides guidance on the selection of cloud and cloud shadow detection algorithms for various applications using Landsat time series.
