Determining Area Affected by Dust Storms in Different Wind Speeds, Using Satellite Images

Abstract

The aim of this study is to determine the area affected by dust storms in different wind speeds using satellite<br />images. In the first step, windy conditions of the Sistan plain were analyzed using wind statistics data. Next, five<br />stormy days of Zabol city, indicating different wind speeds and horizontal visibilities during those storms, were<br />selected. Then, high temporal resolution MODIS data was used as appropriate satellite data in this study. After that, a<br />storm index was defined by means of analyses of storm radiance profile in bands with maximum and minimum storm<br />reflection. The index is the square of difference between visible and thermal infrared bands, which is able to segment<br />stormy confines with the range of reflection changes between 0 and 16. The reflection values were segmented in<br />center of a 1 km2 network using usual interpolation methods such as Local Polynomial, Radial Basis Function,<br />Inverse Distance Weighted, Ordinary Kriging and Universal Kriging. In order to assess the above mentioned<br />interpolation methods, validation techniques were applied using ArcGIS 9.2 software. The result of these assessments<br />such as standard deviation method indicates that the Ordinary Kriging had lower standard deviation. By analyzing the<br />variograms and spatial analysis of the data using GS+ software, the best mathematic model able to fit the points was<br />selected and classification was done by using this model. Finally, the stormy corridors with different dust densities<br />were determined and by calculating the area and determining the villages located in these corridors, the critical<br />regions were recognized. In this study the data from visible bands (4 and 9) and thermal band (21) of the MODIS<br />sensor shows better results compared with the other bands, to segment and classify relative density of dust storms.<br />Moreover, variographic analysis of the satellite data indicates that in most of the dust storms, power models with<br />spherical threshold is the best for interpolation.