بررسي روابط بين عناصر آب و‌ هوايي و نوسانات آلودگي هوا (مورد: شهر مشهد)

Investigating the Relationship among Climate Factors and Air Pollution Fluctuation (Case Study of the City of Mashhad)

هدف اصلی تحقیق استفاده از مدلی است كه بتواند بین عناصر اقلیمی و آلودگی هوا ارتباط برقرار كند. بدین منظوراز سه مدل متفاوت شبكه عصبی احتمالی، مدل رگرسیون خطی و مدل پرسپترون چندلایه استفاده شد. برای این تحقیق از آمار یك ساله‌ی اداره‌ی حفاظت محیط زیست مشهد استفاده شد. این ‌آمار مربوط به آلاینده های هوا شامل (CO- NO- O3- SO2) و آمار هواشناسی شامل پارامترهای اقلیمی (رطوبت نسبی، درجه حرارت، جهت‌ باد و سرعت باد)‌ می‌باشد. داده‌ های آلودگی هوا از تعداد 11 ایستگاه آلوده سنجی جمع‌آوری شده است. این داده‌ها به صورت ساعتی بوده و سپس از آنها میانگین گرفته شد.پس‌ازدرون‌یابی ‌فاصله ‌معكوس ‌وزندار وتحلیل داده‌ها به منظور پیش‌بینی روابط داده‌ها با استفاده از مدل شبكه عصبی داده‌ها به دسته‌های آموزشی(70%)، ارزیابی(15%)‌ و تست(15%) طبقه بندی شدند. در این تحقیق برای تحلیل، از دسته‌ی داده‌های آموزشی استفاده شد. نتایج نشان داد‌ كه میزان میانگین مربعات خطا(MSE) و میانگین مطلق خطا (MAE) در مدل شبكه عصبی احتمالی‌ پایین‌تر بوده و نتایج نشان داده است كه مدل شبكه عصبی احتمالی، توانسته است رابطه منطقی بین آلودگی هوا و پارامترهای هواشناسی برقرار كند. از بین عناصر اقلیمی تاثیرگذار بر منواكسید كربن، رطوبت نسبی در ساعت 12:30و جهت باد بیشترین اثر را داشته‌اند، همچنین عواملی اقلیمی تاثیرگذار بر غلظت دی‌اكسید گوگرد رطوبت نسبی در ساعت 6:30 و درجه حرارت مطلق‌ بوده است.

1. Introduction During the past decade, many efforts for investigating the relationship among air pollution and meteorological factors have been undertaken and some statistical methods have been proposed. Population increase and technological advancement were among the factors affecting creation and extension of air pollution in big cities, forcing large cities all over the world to contemplate measures to curb and prevent the spread of pollutants. Air pollution is emblematic of a wide range of impacts on biological, physical and economic systems. In recent years, air quality has been debated as a pivotal factor in materialization of the quality of life in urban areas, specifically in densely populated and industrial areas. In the present research, in order to evaluate air pollution in the city of Mashhad, eleven air-monitoring stations were used and several parameters for air pollution were studied. The main purpose of the research is to use a model which can establish a link between weather and pollutant factors as well as to identify climate factors affecting air pollution in the city of Mashhad and predict the density of pollutant gases using neural network and linear regression. 2. Study Area Mashhad Metropolis, the capital of Khorasan Razavi province spanning over a land area of 204 square kilometers, located in northeast of Iran at a longitude of 59 degrees and 15 minutes and a latitude of 35 degrees and 43 minutes up to 38 degrees and 8 minutes in the vicinity of Abrood, Kashafrood, is flanked by Hezarmasjed and Binalood mountains. It has a cold and arid climate and has its own specific climate and regional specifications. The preponderance of the city has a cold and air climate, with some cold and semi-arid regions and a small cold and wet section situated in the loftiest heights of Hezarmasjed and Binalood mountains. Overall, the city of Mashhad has a changing climate, inclining towards cold and arid. 3. Material and Methods Data and information regarding pollutants in the city of Mashhad for a one-year period was obtained from Khorasan Razavi’s environment organization. The data is collected from eleven air-monitoring stations on an hourly basis. The average for each day has been used as the data for that day. Then meteorological data, including wind speed; wind direction; rain; air pressure; relative humidity; average temperature and maximum and minimum temperature were obtained from meteorological organization. After collecting the required data and information, first the average for the data concerning air pollutants, gathered on an hourly basis, was calculated. Then using ArcGis, climate parameters and pollution information were merged, using merge button, to form a coherent daily output. The next step involved using inverse distance weighting interpolation, which is one of the methods used in meteorological and geographical studies. In this method, for the purpose of prediction in locations where data is not collected, the data gathered in the vicinity of the intended location is used. In the next stage, the correlation among meteorological data and air pollutants was calculated. Finally, Neuorosolutions was used in order to predict non-linear factors and teach the neural network. 4. Conclusion The findings of the study reveal that the potential neural network, compared to regression and multilayer perceptron, was able to establish a link between meteorological parameters and the density of pollutants (NO-O3-CO-SO2) in the city of Mashhad. The aim of the present study is to use a model which can establish a link between climate factors and air pollutants and identify the climate factors which affect air pollution. The result of sensitivity analysis graph indicates that among the climate factors affecting the pollutant of CO, relative humidity at the hour (12:30) and the direction of wind are the most influential. It was also revealed that relative humidity at the hour (6:30) and maximum absolute temperature are the climate factors affecting the density of sulphur dioxide. As for the graphs depicted in the study, it was revealed that the potential neural network, compared to the other two models, was able to establish a more logical relation among air pollutants and climate factors. In the potential neural network for carbon monoxide, the Mean Square Error (MSE) was calculated to be 66.61 and in regression model it was 76.92. Similarly, the Mean Absolute Error (MAE) was 3.005 and 3.12 respectively. In the neural network model for sulphur dioxide, MSE was calculated to be 14.12 and MAE was 2.15; while in the regression model these measurements were 14.003, -0.18 and 2.13 respectively.