Study On The Haze/Fog Detection Methods Based On MODIS Data

Haze/fog which reduces visibility and contains a large amount of pollutants has great influence on the quality of life and environment for residents. With the development of remote sensing technology, it is significant to use RS to monitor haze and pollution level. Base on the key indicators for distinguishing haze/fog, this paper uses MODIS aerosol product, atmospheric profile product and field observation data to create visibility and relative humidity Remote Sensing Model of the Xianlin region in Nanjing. Besides, various factors affecting the precision of the model was analysed, and then detect the haze/fog phenomenon of Nanjing and its severity level.Base on MODIS aerosol products, this paper uses aerosol scale height and power law model to establish visibility prediction model. The two methods have their own advantages and disadvantages. Overall, power law model results are more accurate. Four season correlation coefficients were0.754,0.745,0.636and0.761, and the prediction results are more accurate for the middle portion of visibility and have a certain bias for large and small values. Relative to the power law method, the trend of predicting results from aerosol scale height model is similar with the real value. But when some of the aerosol optical thickness (AOT) values are low, the forecast visibility values of aerosol scale height model sometimes are negative. This paper uses MODIS atmospheric profile products to establish two relative humidity estimation models, by use of dew point temperature and atmospheric precipitable water vapor separately. The average error between predictive and true value of the two models are13%and15%,error rote for haze are7%and8%separately, and for fog both are1%.Besides, ground meteorological data was used to analyze the models’ influence factor. The results show that air pressure and surface wind speed have influences on the models.According to the haze categories standards, haze pollution level was divided based on the visibility prediction model and relative humidity model. The results show power law model for visibility and relative humidity model using precipitable water vapor to distinguish haze pollution level is better. Accuracy of distinguish haze days from clean days can reach91.4%, slight haze days and mild haze were76.1%and56.4%. But for the moderate haze days and severe haze days, aerosol scale height model for visibility and the PWV model for relative humidity has better results.