%0 Journal Article %T Modeling and Prediction of Dust in Western Iran %J Physical Geography Research %I University of Tehran %Z 2008-630X %A Sobhani, Behrooz %A Safarian Zengir, Vahid %A Faizollahzadeh, Sina %D 2020 %\ 03/20/2020 %V 52 %N 1 %P 17-35 %! Modeling and Prediction of Dust in Western Iran %K simulation %K Hazard %K RBF and ANFIS models %K Iran's dusty areas %K Statistical analysis %R 10.22059/jphgr.2020.284389.1007408 %X Introduction Dust in hazardous areas anywhere in the world is harmful for human societiesand life organisms.  Dust Mineral Aerosols can significantly affect Earth's climate (Zhiyuan et al., 2019: 3). The prevalence of dust storms is devastating human health and agricultural activities in Central Asia (Tiangang et al., 2019: 16). Dust plays an important role in socio-economic development, but on the other hand, such supply can have a negative impact on the environment of the forest (Narayan et al, 2019: 4). According to the previous studies, the importance of dust and the resulting hazard can show that the dust parameter is important for natural hazards. According to the studies, the existing methods for studying the dust that has been done so far have been general and have not adequately addressed the subject. Dust in the areas under its control anywhere in the world has had a risk for various parts of life. Dust storms have also been growing in recent years (Mohammad Khan, 2017: 495). Dust phenomenon due to recent droughts caused adverse biological effects and damages in agriculture, industry and transportation in the provinces of Khuzestan and other neighboring areas (Darvishi et al., 2017: 1). Today, dust is one of the common phenomena and is one of the major environmental problems in arid and semi-arid areas (Hejazi Zadeh et al., 2018: 108). The purpose of this study is to analyze the dust data first to address this issue and then, using ANFIS and RBF models, to make a modeling comparison. Finally, the results can predict for a better view of the dust situation for the future, in dusty regions of Iran. Material and methods In this study, after analyzing 29-year-old dust data for 28 stations the regions afflicted by the phenomena in Iran, they were first analyzed and then normalized. After normalizing the dust data using two new and powerful applied models for modeling and forecasting in climateology, the ANFIS and RBF models were modeled. Then, the two models were compared for accurate prediction of the future, and after training the dust data, they were predicted for the coming years. Finally, using the TOPSIS multivariate decision making model, regions are more involved with the priority hazardous dust hazard phenomenon and by utilizing ArcGIS software output data. RBF neural network model Neural networks with radial base function are widely used for nonparametric multi-dimensional functions through a limited set of training information. Radial neural networks with a fast and comprehensive learning are very interesting and efficient, and they pay particular attention to it, Hartman et al. (1990). Gyrosy, Pogni, as well as Hartmann and Kepler, in the 1990s proved that radial-basis grid networks are very powerful approximation devices, so that by having a sufficient number of hidden neurons, they can be able to approximate each function accurately with every degree. These networks are often compared to the neural network back propagation error. The main architecture of the RBF consists of a two-tier network (Khanjani et al., 2016). ANFIS Neural Network Model In this step, it is possible to model and predict dust in the study area using the ANFIS comparative neuro-fuzzy network model (Ansari, 2010: 29). In this study, the phenomenon of dust in a series of time of 276 months (23 × 12 276) was considered in two ANFIS and RBF neural networks models in each station. In a time series consisting of n examples x_1, 〖x〗 _ (2), ..., x_n is the next value of relation (6) of its previous value (Asghari Oskouei, 2002: 75).     The fuzzy system is based on the "conditional-result" logical rules that, using the concept of linguistic variables and fuzzy decision making process depicts the space of input variables on the space of the output variables. The combination of fuzzy systems based on logical rules, and artificial neural network methods can enable the extraction of knowledge from numerical data. It has led to the introduction of a comparative neural system inference. A sogeven fuzzy system was presented with three inputs, one output and two laws and an equivalent ANFIS system. This system has two inputs x and y and one f output. Proximity to Ideal Mode (TOPSIS) Huang and Yun proposed TOPSIS in 1981. In this method, m options (A1, A2, ..., Am) were evaluated with n indices (C1, C2, ..., Cn) (Momeni, 2008). Solving this problem with this method was carried out in the following steps (Makvandi et al., 1391; Law and Order, 2014). Results and discussion The zoning of dust phenomena in dusty regions of Iran using TOPSIS The results of the implementation of the Topsis model, using the degree of importance of the criteria derived from the entropy method, indicate that, in terms of dust intensity, places more and less dusty for the next 14 years in dusty areas Iran, three stations of Abadan, Masjed Soleyman and Ahvaz were exposed to dust (1, 95%, and 81%), respectively, for simulated years. The northern stations of the study area including Khoy, Boroujen and Ahar showed a lower dust intensity with percentages (0.1, 0.4 and 0.6), respectively. According to TOPSIS model, south west and west of Iran were exposed to dust for simulated years. Conclusion According to comparisons of ANFIS and RBF neural network models, the two models were trained to predict dust. The results obtained from the training of the ANFIS neural network model at best, the RMSE value was 11.67 and the R2 value was 0.5879. But the results obtained from the training of the RBF neural network model, at best, were RMSE equal to 2.19 and the R2 value was 0.9854. By comparing these two models, it was finally concluded that the performance of the RBF neural network model was better. According to the modeling and the results obtained from the comparison of the models, the accuracy and reliability of the RBF neural network model was confirmed for prediction, then it was used modelling in this study. %U https://jphgr.ut.ac.ir/article_75677_704218a5e220866ff0ea0c117f732224.pdf