نوع مقاله : مقاله مستخرج از رساله دکتری
موضوعات
عنوان مقاله English
نویسندگان English
Extended Abstract
Introduction
Dust is a characteristic of dry areas or any area that is located near the source of dust generation and benefits from its consequences by being located in the path of dust systems. Dust storms are one of the natural hazards that can have adverse effects on human health, agriculture and the environment. Iran is one of the arid regions of the world that is affected by severe dust storm events. The primary source of dust storms in western Iran often originates from Iraq, Syria and Saudi Arabia
Methodology
In this study, hourly dust data (horizontal visibility and current weather codes), average temperature, precipitation, humidity, and wind speed and direction data for 1987-2023 were used. Clustering was performed using the (FCM) method. In order to accurately predict future dust storms, RBF, MLP, ANFIS, and SVR neural network models were used, and to select the appropriate model, the accuracy criteria of Mean Square Error (MSE), Root Mean Square Error (RMSE), and correlation coefficient (R) were used. After selecting the appropriate model, the forecast of changes in dusty days for the coming years (2024-2040) was made and the output data was zoned.
Results and Discussion
Analysis based on the FCM method showed that the critical situation cluster includes Abadan and Ahvaz stations. The results of the evaluation and testing indicators of models for dust prediction showed that the best performance for dust estimation based on less error and more correlation is the RBF neural network model. The results of simulation of the observed and predicted values of dust storms of the studied stations for the test data by RBF showed that the curves of the observed and predicted values are close to each other; therefore, this model was able to model and predict days with dust storms with acceptable accuracy using the input variables. The forecast of dust conditions in the years 2024-2040 showed the highest dust frequency for Abadan, Ahvaz and Masjed-e-Suleiman stations, which are equal to 204, 185 and 144 days, respectively.
Conclusion
Paying attention to the dangers of dust storms, identifying the areas affected by them, and realistically predicting them will improve the quality of life and increase human health, and can be effective in environmental planning and optimal and desirable protection of natural resources. The results of this study showed that artificial neural network methods have higher superiority and accuracy and can extract hidden relationships between inputs and outputs.
کلیدواژهها English