Mapping sand dune fields in abu dhabi emirate over the period of 1992–2013 using landsat data

Abstract

Up to 90% of the United Arab Emirates’ (UAE) surface is covered by sand dunes and intervening inter-dune belts. The country is severely affected by problems related to sand dunes movement and Aeolian deposits, recognized as a major contributor to desertification. This study discusses the use of publicly available Landsat TM and ETM+ data to detect sand dunes fields and enable monitoring of their movements in the Emirate of Abu Dhabi, UAE. The study focuses on developing a classification approach and applying it to historical Landsat data to produce consistent Land cover maps useable in subsequent change detection studies. Landsat scenes acquired over the period 1992-2013 are used to evaluate different multispectral classification approaches and determine the accuracy of resulting maps. The methodology uses several configurations of supervised classification techniques that include different band combinations to determine those that produce the highest accuracy in mapping the predominant land cover classes in the area. Results indicate that the tested configurations exhibit an unacceptable level of confusion in detecting the built-up class and that the use of surface reflectance as input to supervised classification yields adequate results for sand detection. All configurations also exhibit a certain level of confusion between sparse vegetation and other classes. The use of a vegetation index as a discriminator helps improve the classification accuracy.