Multi-Image and Multi-Sensor Change Detection for Long-Term Monitoring of Arid Environments With Landsat Series

Abstract

An automated procedure has been proposed to monitor by multispectral satellite imagery the cultivation expansion between 1987 and 2013 in the arid environment of the Fayyum Oasis (Egypt), which is subject to land reclamation. A change detection procedure was applied to the four years investigated (1987, 1998, 2003 and 2013). This long-term analysis is based on images from the Landsat series, adopting a classification strategy relying on vegetation index computations. In particular: (a) the consequences of the radiometric differences of three Landsat sensors on the vegetation index values were analyzed using data simulated by a hyperspectral Hyperion image; (b) the problems resulting from harvesting cycles were minimized using five images per year, after a preliminary analysis on the effects deriving from the number of processed images; (c) an accuracy assessment was carried out on the 2003 and 2013 maps using high resolution images for a portion of the investigated area, with an estimated overall accuracy of 91% for the change detection. The method is implemented in a batch procedure and can be applied to other similar environmental contexts, supporting analyses for sustainable development and exploitation of soil and water resources.

Keywords

• land reclamation
• sensor comparison
• change detection
• multiple image classification
• Fayyum Oasis (Egypt)