Detection of Dust Storms Using MODIS Reflective and Emissive Bands

Abstract

Dust storms are one of the natural phenomena, which have increased in frequency in recent years in North Africa, Australia and northern China. Satellite remote sensing is the common method for monitoring dust storms but its use for identifying dust storms over sandy ground is still limited as the two share similar characteristics. In this study, an artificial neural network (ANN) is used to detect dust storm using 46 sets of data acquired between 2001 and 2010 over North Africa by the Moderate Resolution Imaging Spectroradiometer (MODIS) instruments aboard the Terra and Aqua satellites. The ANN uses image data generated from Brightness Temperature Difference (BTD) between bands 23 and 31 and BTD between bands 31 and 32 with three bands 1, 3, and 4, to classify individual pixels on the basis of their multiple-band values. In comparison with the manually detection of dust storms, the ANN approach gave better result than the Thermal Infrared Integrated Dust Index approach for dust storms detection over the Sahara. The trained ANN using data from the Sahara desert gave an accuracy of 0.88 when tested on data from the Gobi desert and managed to detect 90 out of the 96 dust storm events captured worldwide by Terra and Aqua satellites in 2011 that were classified as dusty images on NASA Earth Observatory.