Enhancement of a Semi-empirical Model using Genetic Algorithm for Estimation of Near Surface Particulate Matter (PM10) Concentration in City of Tehran Using Satellite Images and Weather Data



Extended abstract
1- Introduction
Exposure to fine particulate matter with aerodynamic diameters less than 10 μm (PM10) has negative effects on human health and may induce respiratory problems, cardiovascular and lung diseases, and additional health problems. Both short-term and long-term exposures to PM10 have been linked to increased morbidity. The measurement of ground-level PM 10 concentration on a regular basis is therefore of great importance to epidemiological studies it also provides valuable information for an effective management and forecasting of air quality. Air quality monitoring networks have been established in many industrialized countries to take measurement of pollutant concentrations at different locations, on a daily or hourly basis.

Anumber of data fromvarious sourceswere collected for this research, including the historical air quality data sets, MODIS aerosol imagery, and ground-based meteorological measurements.
In this work, a semi-empirical model is coupled with the Genetic Algorithms (GA) to enhance the estimation of the particulate matter concentration in a local scale and at the satellite passing time. For this, the corrected values of Aerosol Optical Depth (AOD) retrieved from MODIS images are used. Hence the AODs were corrected for the effects of air humidity and the air mixing height. In semi-empirical method, using genetic algorithm, few weather data were introduced to the model and their effects were considered. Temperature and humidity caused improvement in assessment of PM concentration.

3– Discussion
Finally the method was evaluated for Tehran where an acceptable correlation of R2=0. 51 with RMSE of 28 was achieved. However, in the procedure of AOD correction, limitations such as insufficient number of ground-based weather stations and approximate determination of the air mixing height may cause some uncertainties in the proposed method.

4– Conclusion
This paper has proposed an effective semi-empirical model for the prediction of ground-level PM10 concentration (GL-[PM10]). Remotely sensed MODIS AOD, and ground-based measurements of surface temperature and surface relative humidity have been found to be highly significant in the prediction.
Although MODIS extracted AOD may contain valuable information about particulate matter but this parameter is highly affected by the atmospheric conditions where usually it is hard to acquire the weather data during many satellite overpasses.