بررسي تأثير عناصراقليمي برآلودگي هواي شهر مشهد با استفاده ازمدل شبكه عصبي

Study of Climatic Elements Effect on Air Pollution of Mashhad City Using Neural Network Model

شهر مشهدازجمله شهرهاي آلوده ايران مي باشدكه عوامل متعددي درآلودگي آن نقش دارندازجمله عناصراقليمي كه دراين مقاله موردبررسي قرارگرفته است. به اين منظور دادههاي مربوط به آلودگي هوا (منوكسيد كربن، دي اكسيد نيتروژن، دي اكسيد گوگرد، ازن و ذرات معلق (PM10) به صورت روزانه وازسال 1384 تاسال 1390 ازاداره كل محيط زيست خراسان رضوي تهيه گرديدو براي بررسي تاثيرعناصراقليمي برآلودگي ازداده هاي هواشناسي شامل حداقل وحداكثردما، حداقل وحداكثررطوبت، جهت وسرعت بادازايستگاه سينوپتيك مشهداستفاده گرديد. دراين پژوهش از شبكه‌هاي عصبي با ساختارهاي مختلف همانند پرسپترون استفاده شد وسعي ­گرديد تا بهترين و كاراترين شبكه با تعيين مقدار خطاي آن انتخاب و مورد استفاده قرارگيرد. در نهايت جهت انتخاب مدل مناسب و بهينه از شاخص­هاي R2 و RMSE استفاده گرديد. نتايج برآوردها نشان داد روش رگرسيون هيچ برآورد درستي از روزهاي آلوده ندارد اما روش شبكه عصبي در 25 درصد موارد برآورد درست‌تري از روزهاي آلوده دارد.

The city of Mashhad is one of the polluted cities in Iran so that various factors Playing the role in its pollution such as climate elements that studied as one of these factors. For this purpose, the data related to the air pollution (carbon monoxide, nitrogen dioxide, sulfur dioxide, ozone and particulate matters (PM10)) was provided daily by environment department of Razavi Khorasan province from 2003 to 2009. To study the impact of climate elements on the pollution, the meteorology data including minimum and maximum temperature, minimum and maximum humidity, direction and speed of wind from the synoptic station of Mashhad was used. In this survey, neural networks with different architectures like Perceptron were used and it was tried to select the best and most efficient network and to use through determining its error value. Finally, in order to select a suitable and optimum model, R2 and RMSE indices were used. In this research, meteorology parameters and the data related to the day before pollution were considered as the input data and to polluted days as the output data. The results of evaluations showed that regression method provides no proper evaluation on polluted days but neural network method more accurate evaluation on polluted days for 25% of the cases. However, it seems that except climatic elements, the other important factors such as the number of vehicles and their traffic have an important impact on the air pollution of Mashhad.