نوع مقاله : مقاله پژوهشی
نویسندگان
1
Hormozgan University
2
دانشگاه تربیت مدرس
3
Researcher Soil Conservation and Watershed Management Research Department, Isfahan Agricultural and Natural Resources
چکیده
As complex climatic events, dust storms could be managed by considering their nature and attributes. Therefore, this study sought to investigate interactions between the aerosol optical depth index and climatic land surface characteristics using data mining and zoning techniques in dust-prone regions of the Yazd province, Iran. To this end, the required data was collected from several climatic products of the University of Idaho and Modis Sensor for the 2000–2017 period using Google Earth Engine.
Moreover, the image of the maximum dust was processed using AOD Modis and ENVI 5.1 software. Then, the underlying correlation between the variables was identified through various data mining techniques. In addition, the ROC curve was used for cross-validation, and different metrics were applied to assess the model, including Square Root of Error, Absolute Normalized Error, Classification Error, Absolute Error, and Crucial Class Fraction Ratio. Finally, the best data mining approach was used to determine the location and zoning of dust-prone regions.
The findings of the study indicated that the decision tree outperformed the Bayesian theory with 89.53% accuracy and that it performed better than the nearest neighbor with an accuracy of 61.3% and 81.31%, respectively. As for the validation of the models, the decision tree methods, nearest neighbor search, and Bayesian network theory showed 74.21%, 64.39%, and 55.42% values, respectively.
Moreover, in regions with crucial harvest and dust ranges, wind speed and soil surface moisture were found to have the most significant role. On the other hand, the zoning of dust-prone regions revealed that 888,067067 Km2 of areas were covered with the highest concentrations of dust, most of which were located in the central and eastern parts of Yazd province, with the AOD values being 0.465, 0.309, 0.162, and 0.065 for the ranges of 0-0.036, 0.036-0.072, 0.072-0.107, and 0.107-0.3, respectively.
کلیدواژهها
عنوان مقاله [English]
Identifying Dust Storm-Prone Areas Using Google Earth Engine Data and Classified Variable Data Mining Methods: A Case Study of Yazd Province, Iran
نویسندگان [English]
-
Mohamad Kazemi
1
-
Atefeh Jafarpoor
2
-
Reza Naderi
3
1
Hormozgan University
2
Tarbiat modares university
3
Researcher Soil Conservation and Watershed Management Research Department, Isfahan Agricultural and Natural Resources
چکیده [English]
As complex climatic events, dust storms could be managed by considering their nature and attributes. Therefore, this study sought to investigate interactions between the aerosol optical depth index and climatic land surface characteristics using data mining and zoning techniques in dust-prone regions of the Yazd province, Iran. To this end, the required data was collected from several climatic products of the University of Idaho and Modis Sensor for the 2000–2017 period using Google Earth Engine.
Moreover, the image of the maximum dust was processed using AOD Modis and ENVI 5.1 software. Then, the underlying correlation between the variables was identified through various data mining techniques. In addition, the ROC curve was used for cross-validation, and different metrics were applied to assess the model, including Square Root of Error, Absolute Normalized Error, Classification Error, Absolute Error, and Crucial Class Fraction Ratio. Finally, the best data mining approach was used to determine the location and zoning of dust-prone regions.
The findings of the study indicated that the decision tree outperformed the Bayesian theory with 89.53% accuracy and that it performed better than the nearest neighbor with an accuracy of 61.3% and 81.31%, respectively. As for the validation of the models, the decision tree methods, nearest neighbor search, and Bayesian network theory showed 74.21%, 64.39%, and 55.42% values, respectively.
Moreover, in regions with crucial harvest and dust ranges, wind speed and soil surface moisture were found to have the most significant role. On the other hand, the zoning of dust-prone regions revealed that 888,067067 Km2 of areas were covered with the highest concentrations of dust, most of which were located in the central and eastern parts of Yazd province, with the AOD values being 0.465, 0.309, 0.162, and 0.065 for the ranges of 0-0.036, 0.036-0.072, 0.072-0.107, and 0.107-0.3, respectively.
کلیدواژهها [English]
-
Dust
-
Optical Depth
-
Climatic Variables
-
Data Mining Models
-
Machine Learning
)