การพัฒนาอัลกอริทึมและแบบจำลองคาดการณ์ความเข้มข้น ของก๊าชไนโตรเจนไดออกไซด์ด้วยข้อมูลดาวเทียม ในประเทศไทย

Abstract

Nitrogen dioxide (NO2) is one of air pollution that currently affects in Thailand. The aims of this study are separated into two objectives. Firstly, this research aim to develop an algorithm to analyze the distributing pattern and the related factor that led to the generation of NO2 in Thailand during January 2019 to December 2020 by weekly average. Secondly, this research develops a model for 7 days forecasting the concentration of NO2 in the Central and the Eastern region of Thailand. The algorithm uses data from ground station, and satellite data from Sentinel-5P, and analyzes by applying 9 types of regression equation model. The analysis of spatial distribution and factors related to NO2 concentration uses NO2 data along with meteorological data such as temperature, humidity, and windspeed, and uses Pearson correlation coefficient to analyze. The model for 7 days forecasting the concentration of NO2 will create two types of the model. The first type is analyzed from ground station, and the second type is analyzed from algorithm using data from the first objective, metrology data, and forecasting metrology data by using multiple linear regression. The result of the research reveals that, the average of NO2 concentration in 2019 was higher than NO2 concentration in 2020, and the value was 10.95 ppb and 10.67 ppb respectively. The maximum total annual average of NO2 concentration, seasonal comparison, occurred during the dry season (between November and April) in January 2019 (18.37 ppb), while the minimum average occurred during the wet season (between May and October) in June 2020 (8.05 ppb). The highest average NO2 concentration, regional comparison, was found in the Central region during the fourth week of January 2019 (40.69 ppb), while the lowest average concentration was found in the Northern region during the first week of September 2019 (2.84 ppb). Moreover, the factor that correlated with NO2 concentration were relative humidity and wind speed. In addition, the suitable model for estimating the concentration of NO2 is the Cubic model with coefficient of determination (R2) at 0.72 and the overall accuracy of this model is 70.29 %. Additionally, the model for 7 days forecasting the concentration of NO2 analyzed from ground station had a lower error than the model 7 days forecasting that analyzed from cubic model, with RMSE values of 9.34 and 15.04 respectively. The forecasting model analyzed from ground station has higher overall accuracy than the forecasting model analyzed from cubic model, with values of 72.97 % and 59.05 % respectively.