توسعه مدل GA-ANFIS به‌منظور پیش‌بینی غلظت آلاینده 〖PM〗_10 در یک بازه زمانی بلندمدت مطالعه موردی: شهر تهران

ABSTRACT
Among the types of airborne particles, particles with a diameter of less than 10 microns have many adverse effects on human health. Meteorological parameters and the movement of a large volume of vehicles are considered the most important modulating factors in the distribution and concentration of atmospheric pollutants. In this study, in order to predict the concentration of PM-10 pollutant during a long-term interval in Tehran city, GA-ANFIS hybrid model was used. Wind speed, wind direction, temperature, relative humidity and traffic volume were considered as inputs and pollutant concentration PM_10 as the output of the model. The results of the calculation of the performance indicators showed that the combined GA-ANFIS model provides a better framework than the ANFIS model in predicting the pollutant concentration PM_10. In order to evaluate the spatio-temporal patterns of PM_10 pollutant concentration and to identify hot and cold spots in Tehran city, local Moran's statistic and Ard-J Gettys statistic were calculated. The results showed that there is a high level of clustering of PM_10 pollutant in Tehran (with 95% confidence level). The clusters of PM_10 have divided the city into two northern and southern parts so that most of the cold spots in the northern half and the hot spots in the south have spread to the center of the city
Extend Abstract
Introduction
Among the types of air particles, particles with a diameter of less than 10 microns have many adverse effects on human health. The main sources of emissions of this pollutant are fuel combustion, industrial processes, agriculture, cars and diesel machines do not have the necessary standards. Meteorological parameters and high volume displacement of vehicles are the most important modulating factors in the distribution and accumulation of air pollutants. Tehran city is the capital of Iran, it is the most populous cities in Iran over 8.69 million of population. Due to the special geographical location of the city, in the northern part it has a temperate and mountainous climate and in low-lying areas it is semi-arid. The metropolis of Tehran, as the second most populous capital in the Middle East and the twenty-fourth most populous city in the world, is one of the most polluted capitals in the world. Tehran is ranked 12th among 26 metropolises in terms of ambient   levels. The two main causes of the spread of this pollutant in Tehran are mobile sources (vehicles) and resident sources of pollution such as construction workshops and refineries around Tehran.
 
methods
Climatic parameters such as temperature, wind speed, wind direction and relative humidity have an important effect on the formation, transportation, accumulation and deposition of atmospheric pollutants. The role of wind speed as an important factor in the distribution of pollutants in the city of Tehran on a regional and local scale is undeniable. Due to the significant effect that high-speed winds have on the dispersion of pollutants and, consequently, on the concentration of pollutants in the atmosphere, as a result, chemical reactions are reduced. Regarding the role of temperature parameter in the process of changes in the concentration of air pollutants, it can be noted that in general, high temperature is associated with unfavorable air quality. The trend of relative humidity changes is in line with the pattern of temperature changes so that the maximum and minimum relative humidity are in accordance with the minimum and maximum temperatures. The next very influential factor in the process of changes in the concentration of pollutants is urban road traffic. The high number of vehicles in the central streets of the city and the subsequent increase in the volume and duration of traffic has led to an increase and accumulation of concentrations of pollutants in the streets and central areas of Tehran.  Recently, intelligent and data-driven methods have become increasingly well known for forecasting of air pollution. In the current investigation, an approach was presented for the training ANFIS by using GA based on a population algorithm. At first, Genetic algorithm was implemented for optimal consequent parameters of ANFS. Then, these optimal values was applied for training ANFIS model. At the next step, for spatial modeling of  concentration level during in the studied period, hotspot analysis in GIS was used to investigate the spatial changes of  concentration and to identify hot and cold spots. In this study, modeling of   by GA-ANFIS was done on three dataset such as air pollution data, meteorological, traffic volume. Meteorological parameter include the hourly wind speed (m/s), wind direction (), temperature () and relative humidity (%) were taken the weather measurement stations in Tehran. The air quality data, which comprises the hourly concentration of  (ppb). Concentration of pollutant () were obtained from the Air Quality Control Company of Tehran. Traffic data was obtained from Tehran Traffic Control Company. The time interval of traffic data and meteorological data were considered the same. These data, covering the period from 2010 to 2020, were acquired from 21 air monitoring stations. In this study, MATLAB 2020a programming language was used to implement ANFIS and hybrid of GA-ANFIS were proposed for  concentration prediction and four indices were used for evaluate the results of the predication by ANFIS and GA-ANFIS models. ANFIS is a combination of neuronal networks and fuzzy system that proposed by Jang and Sun. To reach the optimized output, the ANFIS can coupled with hybrid learning methods like Imperialist Competitive algorithm (ICA), PSO or GA. Genetic Algorithms (GA) is a type of evolutionary heuristic search algorithm based on genetic science. GA provides a random search that was used to solve optimization problems. It can be noted that GA-ANFIS algorithm can successfully improve the performance of the ANFIS model. In order to assessment of the patterns of spatial changes of   concentration in the study period at the seasons of the year, a GIS modeling was applied to produce the simulation maps at different seasons during a study period. Spatial correlation of case study overall pollution concentrations was measured using global Moran’s I and Getis-Ord general G indices.In order to identify hot and cold spots on maps, G statistics was used by Getis.
 
Result and discussion
The results of the global spatial autocorrelation analyses found high clustering levels for  in the case study, which are significantly different from random at the 95% confidence level. In the monthly results, the critical values for   pollutant occurred in the June to August (due to dust storms originating outside of Tehran). In general, there was no sensible similarity in the seasonal results. As the results of spatial autocorrelation analysis showed,  Clusters separated the city into southwest parts, as most of cold spots were situated in the north and west. With the increase in traffic volume at the beginning of the fall season due to the reopening of universities and schools, we are witnessing polluted critical days in Tehran. Starting of rainfall in late of winter, air pollution decreases and wind speed up to 4 m / s are ineffective in reducing pollution concentrations. The highest emission of particulate matter less than 10 microns is due to natural resources and due to the prevailing wind direction in Tehran from southwest to northeast, the high concentration of this pollutant in the districts of 18, 9, 10 and 15 at the entrances of Tehran, it seems logical. The lowest  concentration was observed in the northeastern areas of the city.
 
Funding
There is no funding support.
 
Authors’ Contribution
All of the authors approved thecontent of the manuscript and agreed on all aspects of the work.
 
Conflict of Interest
Authors declared no conflict of interest.
 
Acknowledgments
We are grateful to all the scientific consultants of this paper.