A literature review on satellite image time series forecasting: Methods and applications for remote sensing

Satellite image time-series are time series produced from remote sensing images; they generally correspond to features or indicators extracted from those images. With the increasing availability of remote sensing images and new methodologies to process such data, image time-series methods have been used extensively for assessing temporal pattern detection, monitoring, classification, object detection, and feature estimation. Since the study of time series is broad, this article focuses on analyzing articles related to forecasting the value of one or more attributes of the image time-series. The image time series forecasting (ITSF) problem appears in different disciplines; most focus on improving the quality of life by harnessing natural resources for sustainable development and minimizing the lethality of dangerous natural phenomena. Scientists tackle these problems using different tools or methods depending on the application. This review analyzes the field's leading, most recent contributions, grouping them by application area and solution methods. Our findings indicate that artificial neural networks, regression trees, support vector regression, and cellular automata are the most common methods for ITSF. Application areas address this problem as renewable energy, agriculture, and land-use change. This study retrieved and analyzed relevant information about the recent activity of image time series forecasting, generating a reproducible list of the most pertinent articles in the field published from 2009 to 2021. To the author's best knowledge, this is the first review presenting and analyzing a reproducible list of the most relevant state-of-the-art articles focusing on the applications, techniques, and research trends for ITSF. This article is categorized under: Algorithmic Development > Spatial and Temporal Data Mining Technologies > Machine Learning Technologies > Prediction