@article { author = {Ataei, Shima and Abkar, Ali Akbar and Mohammadzadeh, Ali}, title = {Dust detection using improved TIIDI and applying MODIS sensor data}, journal = {Journal of Environmental Studies}, volume = {41}, number = {3}, pages = {563-572}, year = {2015}, publisher = {دانشگاه تهران}, issn = {1025-8620}, eissn = {2345-6922}, doi = {10.22059/jes.2015.55897}, abstract = {Introduction Thermal Infrared Integrated Dust Index (TIIDI) is one of indices that is presented in this direction. The index operates based on four brightness temperature difference including Wavelengths 3.7, 8.6, 11 and 12 micrometer. According to the studies, BTD (BT12-BT11) is used for cloud detection. BTD (BT8.6-BT11) is the index for separation for dust from sandy surface and BTD (BT3.7-BT11) is employed for differentiating dust from vegetation in addition it shows dust intensity. In general, according to Weather Meteorological Organization (WMO) protocol, dust phenomena is divided four categories based on reducing of horizontal visibility: Dust-in-Suspension: Its development is suspended , visibility of less than 10 km Blowing Storm: Reduced visibility from 1 to 10 km Dust Storm: Reduced visibility from 200 to 1000 m Severe Dust Storm: strong gust of wind with large dust particles and reduced visibility from less than 200 m.  Yang by combining these tree parameters created the TIIDI index.   (1) Dust Sky without dust and cloud Cloud In formula (1), if the get the positive value,‘a’ would be 10 and in otherwise it would be 5. Based on the  has a small value for sky without dust and cloud, the index TIIDI get less value for sky without dust and cloud than dust.   Therefore, using this index appears reasonable according to topographic complexity that exists in west and southwest of Iran. Because in these areas there is a combination of mountainous terrain with vegetation or bare land and vegetation. Consequently, the suggested index by Yang is developed and customized. The modified can estimate dust storm intensity with improved accuracy.   Data In this research we used MODIS products including MOD021, MCD12Q1, MOD35, topographic map 1:25000 of Ilam ( )and Khuzestan ( ) to obtain land cover types and also meteorological data of the two provincesduring (2005-2012) years are used to implement the algorithmsand evaluate it. The ground data is for the 30 stations of the two mentioned provinces, thus all analysis related to computing and image processing is done on the pixel location of the 30 stations in the image.   Method As mentioned before, the aim of this study is detection of dust pixels using corrected TIIDI. Considering the fact that in the study area there are three types of classes namely cloud, dust and sky without clouds and dust. As a results, the radiative behavior of these three classes in thermal infrared bands of MODIS should be checked. Hence, in the first step appropriate training data must be collected for all three classes and then diagram of three classes’ radiation in thermal bands is obtained and analyzed. According to the results of previous research for sky without cloud and dust, BTD (BT31-BT32) is negative. Which means the slope line of band 31 to band 32 must be negative but it is positive like cloud due to the complexity of study area. Since a pixel  in study area not only contains one class but also it contains more than one class such as building, vegetation, desert lands and wasteland. Because the spectral changes in surface emissivity will also cause change in the behavior of the index. If BTD (BT31-BT32) is positive for sky without cloud and dust, then cannot be concluded that TIIDI is positive. In this case, the dust and the sky without cloud and dust will have not a similar behavior. Dust based on the horizontal visibility is separated into four groups(See the introduction section). As previously mentioned dust intensity is defined based on four categories. Because one of the challenges that exist in dust detection is lower intensity dust event and horizontal visibility of more than a kilometer and also inability to detect them on true color MODIS image.According to what was said,TIIDI index must be improved for study area. Figure 1 shows that slope line of band 29 to band 32 for both dust and sky without cloud and dust is same and for both cases is positive. As a result, in the formula (2) BTD (BT32-BT29) can be replaced by BTD (BT31-BT32).   (2)   Dust   Sky without dust and cloud Cloud           Figure 1: reflectance behavior diagram of cloud, dust and sky without dust and cloud pixels in thermal infraredbands of MODIS         Our studies on the iTIIDI for cloudy days, days of dust and the days when the sky is free of clouds and dust reveals that this index have smaller value (10-25) for sky without cloud and dust but in during dust event the values will be increased (more than 25). Also, iTIIDI index have negative value for cloud pixel. Therefore, first, by choosing zero threshold on iTIIDI cloud pixels can be eliminated then those have values more than 25 is classified as the dust pixels. From pixels that are known as dust pixels, if corrected index value is between 25 to 50 it shows weaker dust otherwise, if iTIIDI is more than 50 it specifies the dust with more intensity.   Discussion of Results To assess the accuracy of the proposed indices and its success rate in dust detection, the indices on 6 July, 2009 (15 Tir, 1388) is calculated, where a severe dust storm occurred over the region. In this image that was taken at 10:35 AM local time, meteorological data shows the minimum and maximum horizontal visibility of 200 m and 6000 m, respectively in the study area. Based on beginning categories (See the introduction section), in this day severe dust storm is occurred. In the figure 4, results of enforcement iTIIDI index is shown on the image and is compared with the true color MODIS image for same day. As can be seen there is good agreement between image (a) and image (b). In addition, improved index can be able to detect dust as well as the severity of dust. In other hand, the accuracy of the developed method is evaluated using the ground observation data. 30 synoptic stations of meteorological data located in the provinces of the study area are listed in Table 1. These information includes horizontal visibility and meteorological codes.  According to meteorological organization standards, during dust event and reducing the horizontal visibility, this parameter reaches less than 10 km and meteorological codes determine with value 05 or 06 or 07. The overall accuracy of corrected index corrected iTIIDI is about 65% for detection of dust pixel and the accuracy of 64 percent is achieved by using the TIIDI index. Figure (2) shows result of index implementation (iTIIDI)on image that is compared with RGB image on the same day (6 July, 2009).       A C B     Figure 2: Comparison of (A) MODIS RGB image that was taken on 6 July, 2009 (B) TIIDI index by Yang and (C) Improved index (iTIIDI) for dust detection for same day.         Conclusions In conclusion because of the land cover presented in the west and southwest of Iran that there are combined different classes such as Vegetation, Bare land or Sandy land, and Mountainous areas, a review of the indices proposed at a global level is inevitable. Since in this review, on the one hand, the former index is improved and on the other hand its threshold is customized. ThereforeThermal Infrared Integrated Dust Index (TIIDI)has been developed for dust detection and the improved index is presented. Although the accuracy achieved in this study is not more than 65 percent, the results obtained demonstrate the simplicity and accessibility of the method and by having an extensive coverage of MODIS data and its products would all increase the speed of the algorithm in dust detection. Also by comparing two indices the result show that the improved method not only is able to detect sever dust storm but also is able to detect less intense dust storm.}, keywords = {dust detection,improved TIIDI index and synoptic station data}, title_fa = {شناسایی گرد و غبار با استفاده از شاخص TIIDI بهبودیافته و به کارگیری داده‌های سنجندۀ مادیس}, abstract_fa = {در تحقیق پیش رو با به کارگیری شاخص یکپارچۀ مادون قرمز حرارتی مادیس به شناسایی اولیۀ ذرات گرد و غبار و بررسی میزان شدت آن‌ها در استان‌های ایلام و خوزستان طی سال‌های 1384 تا 1391 پرداخته شده است. به این منظور ابتدا با توجه به آنالیزهای آماری روی داده‌های آموزشی، شاخص به‌کاررفته بهبود یافته است و TIIDI بهبودیافته معرفی می‌شود. سپس، این شاخص برای منطقۀ موردنظر بومی‌سازی شده است. بنابراین، در مرحلۀ اول داده‌های آموزشی مناسب برای سه کلاس ابر، گرد و غبار و آسمان عاری از گرد و غبار و ابر جمع‌آوری شده است. سپس، رفتار تابشی سه کلاس مذکور در باندهای مادون قرمز حرارتی مادیس بررسی و شاخص TIIDI برای شناسایی گرد و غبار روی داده در منطقۀ مطالعاتی محاسبه می‌شود. در مرحلۀ بعد، دقت کلی شاخص موردنظر با استفاده از داده‌های ایستگاه‌های هواشناسی به دست آمده است که به دلیل پایین‌بودن این کمیت، شاخص مذکور با بررسی تعدادی از نمودارهای آماری در منطقۀ مطالعاتی بهبود یافته است و دقت کلی شاخص پیشنهادی با شاخص پیشین مقایسه می‌شود. نتایج حاکی از آن است که در صورت استفاده از شاخص بهبودیافته علاوه بر افزایش دقت (از 64 به 65 درصد) می‌توان شدت گرد و غبار را نیز تخمین زد.}, keywords_fa = {شناسایی گرد و غبار,شاخص TIIDI بهبودیافته,دادۀ ایستگاه هواشناسی}, url = {https://jes.ut.ac.ir/article_55897.html}, eprint = {https://jes.ut.ac.ir/article_55897_a7a9164c71f2577166bd554689f80988.pdf} }