Journal of Spatial Analysis

---

Analysis and Comparing Satellite Products and Simulated
 Of AOD in West Iran (2000-2018)
 
Kaveh Mohammadpour, Ph.D. Student in Climatology, Kharazmi University of Tehran
Mohammad Saligheh, Associate Professor in Climatology, Kharazmi University of Tehran
Ali Darvishi Bloorani, Assistant Professor in RS & GIS, Tehran University
Tayeb Raziei, Assistant Professor in Climatology, SCWMRI, Iran
 
Introduction
Dust are the main type of aerosols that affects directly and indirectly radiation budget. In addition, those affect the temperature change, cloud formation, convection, and precipitation. In recent years, the increase of different sensors and models has made possible to research the dust. The most important studies about dust analysis has been considered of Aerosol Optical Depth (AOD) as the most key parameter, which are based on the use of remote sensing technique and global models for analyzing the behavior and dynamics of dust in recent two decades. To achieve this, it has used of MODIS and MACC to study and identify the behavior of dust in the last two decades over west Iran.
 
Materials and methods
Areas in this study are Ilam, Kermanshah, Kurdistan, Lorestan and Hamedan provinces. The area has studied of two data series such as: first is MACC data with a spatial precision of 14 km2 and a 3-hour time scale; and other one is MODIS sensor production on the Terra satellite with a 10-square-kilometers resolution. In order to analyze the dust in the area in the period 2000 to 2018, statistical methods and simulation has used of the AOD parameter in MACC and MODIS. Before any processing, the data regraded to 0.2 × 0.2 degrees in order to compare the data. Then, the average daily AOD formed in a 22 × 23 matrix with 560 pixels that presented with 3653 × 560 for MACC during 2003 to 2012 and 6489 × 560 for MODIS during 2000-2018. Average of daily AOD obtained of MACC and MODIS calculated using of statistical equations. Then, the spatial distribution of AOD during the dusty months for synoptic stations and total province surface extracted using of R packages during the daily time series of the periods. Finally, the spatial distribution of the obtained AOD interpolated using the kriging function.
 
Results and Discussion
The average annual AOD obtained from Deep Blue algorithm from MODIS was less than MACC in all of the interested stations, except for Hamedan and Khorramabad stations, and provinces surfaces.
Correlation of AOD between MODIS and MACC shown that the correlations is high between model and sensor data (R2 = 59). In addition, the spatial correlation map shows 0.38 to 0.76 in which indicates a significant relationship between the MACC and MODIS pixels and the relationship is more in the western provinces of the area than the northeast of the region (Hamedan). The monthly comparison of the mean of AOD of the sensor and the model in the whole the area shows a highest correlation between the AOD in February and October.
The interpolation of the spatial distribution in the decade of the study (2003-2012) in MACC showed that the spatial variations of AOD is decreasing from the south of Ilam to the north of Kurdistan and reached the lowest level in the north of Kurdistan province. In general, the findings of annual and seasonal spatial distribution (dry period) of dust showed that MACC overestimated AOD compared to MODIS in the area. Nonetheless, the dust pattern in both of the sensor and the model increased from south to north. Although, the dust pattern is more regular in the sensor than the model. The spatial distribution of dust in Ilam, Kermanshah, and Kurdistan provinces in MODIS and MACC shows that dust in the southern point of the Ilam province has the highest concentration and the lowest is observed in the northeast of Kurdistan province. This spatial distribution of dust showed that dust in western provinces of the area follow latitudinal trend , in which is influenced by the high topography of Kermanshah and Kurdistan provinces and the proximity of Ilam province to dust sources in the distribution of dust intensity.
 
Conclusion
The results showed that there was a significant correlation between the sensor and the model and the coefficient was more than 0.4 in all months on the area. The findings of the annual amount of dust in MODIS showed that the amount of dust in the years 2000 to 2009 has increased in whole areas and from 2009 onwards, this annual trend has been reduced by 2018. MACC findings also showed that the AOD has been growing up in the period, although AOD amount have had a steep slope by 2010, but since 2010, dust has a steady slope. Therefore, West Iran has experienced two active (before 2010) and inactive (after 2010) periods in dust during an 18-years period on the area. The findings of MODIS and MACC in the study area indicate that the monthly distribution of dust from April to August has the highest concentration. In general, the annual and seasonal spatial distribution (months with the highest AOD) of dust indicates that the intensity of AOD in MACC was higher than MODIS in the area. Although the sensor and model has a roughly similar pattern and increases from south to north, but the trend in MODIS is more regular than MACC.
 
Keywords: Aerosol Optical Depth (AOD), MACC, MODIS, West Iran
 
 
 
 

---

Application of multivariate techniques in-line with spatial regionalization of AOD over Iran

Introduction
Models, satellites and terrestrial datasets have been used to detect and characterize aerosol. Nontheless, micoscale classification using remote sensing parameters considers as a deficiency. Thus, regionalizion and modeling aerosol without regard to political boundaries or a specific stations over Iran  demonstrates the spatial distribution of simple AOD structures.
Materials and methods
 Present study attempted to simulate and detect homogeneous areaes of aerosol in Iran using AOD (areosol optical depth) datast at 550 nm across Iran. Among the eigen techniques, principal component analysis (PCA) is the most applicable and controversial classification applied as multivariate analysis approach. In the line of the target, PCA, S-Mode separate the AOD subgroups with similar correlations. In the mode, m time series apply to each n station or grid points as a variable in the analysis, which is the territory of the region or geographical area. Mathematically, if the input data column in the Z matrix is applied as mathematical variables and the Z matrix has n points in the time series and m is the time step, then in the Zs decomposition has 3654×9985. In addition, the scree test and North's rule were used to cut-off the principal components and to select the number of appropriate special vectors to be kept.
Results and Discussion
For the study purpose, 85 percentaile of loadings were used to determine AOD areas over Iran. Using the method, the spatial patterns of Iran's aerosolshave been divided into six subregions, which are the major centers affected by the AOD. These major AOD hotspots affect by AOD extermes that are originated from aerosol surrounding sources. So that, the geographical location of sources areas have caused the northeastern atmosphere of Iran to be influenced by severe storms originating from the Karakum Desert. The same is correct concerning the East and Southeast regions. While, the intensification and transfer of aerosol from the Sistan plain to the south is increased AOD load over southeast Iran. Moreover, this study revealed a set points associated with distinguishing spatial differences between the west-northwest and southwest regions as well as central region that have not addressed in previous studies because of focus on ground-based observations. Also, the method illustrated that formation of the identified regions are a function of the volume, growth, and spread of aerosol particles resulting from the source regions in the Middle East. Finally, the classification techniques converting dynamic phenomenon such as aerosol into simpler structures presented a interpretable understanding of the geographical distribution of the phenomenon.
Conclusion
The present study identified the spatial patterns of AOD hotspots into six distinct regions including northeast, west-northwest, southeast, southwest, central and eastern Iran affected by the aerosol as well as major centers or high gradient areas. In addition, the present study not only supported by previous studies, but also it  make sense a regionalization that was neglected by former studies, whileseperated  the boundaries of the AOD areas without considering  provincial boundaries. Overall, the classification techniques, PCA, simplified a dynamic phenomenon such as aerosol into a simpler and illustrated geographical and interpretable understanding of the spatial distribution of the phenomenon.

Keywords: Aerosol Optical Depth (AOD), Multivariate Techniques, Regionalization, Iran