Dust detection algorithm based on the entropy of visible band brightness and brightness temperature difference

Dust, as the main component of aerosols, has numerous effects on a climate system. Simultaneously, dust is harmful to human health as an environmental pollutant. The dust weather generally erupts in spring, thereby significantly affecting the production and life of most regions in Northern China. In the past, many studies have been conducted to identify dust by remote sensing. However, the traditional method has a poor effect and can hardly recognize dust in several complex situations, such as cloud-dust mixing. The dust process is typically accompanied by clouds, which are the main interfering factors in identifying dust. The judgment of pure dust is improved in the thermal infrared band, but the effect is poor for the dust mixed with clouds. In terms of microwave, the sensor is carried mostly by a polar orbit satellite, which time and space resolutions are low. This sensor cannot display real-time dust monitoring and warning. In this study, a new method was proposed using the data from the Himawari-8 satellite. In fact, the dust mixed with clouds demonstrated better successive distribution characteristics in space than in medium clouds and fractocumulus. Thus, the dust mixed with clouds could be identified. A difference in the reflectivity of 0.46and 0.51 μm in a certain range could properly exhibit the continuity characteristics of dust and effectively distinguish clouds and most surfaces with dust by analyzing several visible channels of the Himawari-8. A threshold less than 10-15 could cover most dust mixed with clouds in accordance with the experimental statistics. However, an RDI value of broken cumulus was mainly distributed between 5 and 15, which was similar to the RDI value of dust. Therefore, we introduced the entropy of brightness. In this study, pure dust was identified using a BTD value that is less than 0, and the dust mixed with clouds was identified through the new method. In the spring and summer of 2017, several types of dust accumulated in Inner Mongolia in China and its surrounding areas. We used the satellite data for April 16, May 4, and August 2 combined with visual interpretation and ground observation results to analyze and verify the proposed method. In the two dust processes of April 16 and August 2, we selected three typical regions with mixed cloud and dust. The dust storm on May 4 was large. Therefore, this dust storm was used to analyze and validate the algorithm on a large scale. The verification of dust on May 4, 2017, showed that the observations of the 22 stations were consistent with the results of 27 stations located in the cloud-sand mixing region. The algorithm proposed in this study achieved a significant result in dust recognition under various cloud-sand mixing conditions. A new method based on the brightness entropy of RDI was proposed in this study. The method could effectively identify dust mixed with clouds in comparison with the results of traditional methods or using visual interpretation and ground observation. This method compensates for the limitations of existing algorithms and data to a large extent. However, significant complications in recognition of certain floating dust and large-thickness cloud still exist. The accuracy of recognition will also be affected by the complex condition of the surface. This study reveals that the method exerted a certain effect on pure dust, which can be further discussed in the other research. In addition, this method is still limited when identifying dust at night. © 2018, Science Press. All right reserved.