Land Use and Land Cover Trajectory Classification and Analysis Methods in the Amazon: Implications for Forest Regeneration Studies

Abstract

The analysis of forest regeneration processes provides information used to estimate atmospheric carbon assimilation, soil fertility recovery, hydrological cycles and biodiversity maintenance, among other environmental services. These studies require information that is usually found in land use and land cover trajectories mapped in long time intervals and with annual observations. These trajectories are usually obtained by processing remote sensing image time series. In this review article we first identify and describe the main methods used to classify and analyze land use and land cover trajectories based on orbital remote sensing data. We then discuss these methods based on their applicability in forest regeneration studies in the Amazon. Throughout this process we observe that analyzing land use and land cover trajectories in the Amazon is not a trivial task. Traditional change detection methods result in invalid trajectories or require many classification steps. Given the large volume of data, it is common to simplify the information contained within the trajectories to the point that analyses are reduced to one or two observed times. In these cases, important information about regeneration processes is lost, such as persistence of secondary vegetation and time of use before abandonment. Among the main observed limitations, we highlight the lack of available data, such as cloud free images and reference data.