Automatic Extraction of Mega Dune of Yealan Erg, Lut Desert by Self Organizing Maps

Abstract

85 % of Total moving sands in the world are in ergs which cover about 32000 Km2. Traditional qualitative methods for geomorphological surveys are based on the field observations which due to harsh condition in deserts are time and money consuming. Lut desert in eastern part of Iran includes the mega dunes which are very interesting landforms but unreachable. This paper presents a new approach using a Self Organizing Map (SOM) as an unsupervised algorithm of artificial neural networks for analysis and identification of Lut mega dunes. The latest version 1 arc second SRTM/X band DEM were re-projected 30 m UTM grid and used to generate 22 morphometric parameters of first order (e.g. slope), second order (e.g cross-sectional curvature, maximum curvatures and minimum curvature) and third order( extreme curvature)by using a bivariate quadratic surface. The five optimum parameters are selected based on Optimum Index factor (OIF) and used in a SOM to identify morphometric features (or landform). The number of landforms was determined by Davies-Bouldin Index. The ETM+ bands of landsat satellite data also included to input data for increasing the potential of SOM to extract features. The result showed that all mega dunes could clearly be recognized and classified by this method when their width is larger than the DEM resolution. The results also demonstrate that a SOM is a very efficient tool for analyzing geo-morphometric features as Aeolian landforms under a hyper-arid environmental condition.

Keywords

• desert landforms
• lut
• mega dunes
• satellite data
• srtm