Suppost vector machine regression applied to MODIS data for PM10 concentaration analysis

The density of absorbable particulate matter less than 10um termed as PM₁₀ is one the most important contamination index for air quality monitoring. This article presented a new PM₁₀ concentration analysis approach based on a quick atomospher correction (QUAC) model and support vector machines reggression (SVR). The deriviation of six MODIS bands before and after QUAC model is calculated as indicating features to atomospher matters. Several regression models including liner, logarithmic, quadratic, power and SVR are compared in term of the statistical correlation between the derivation values and groud measured concentration of PM₁₀. The experimental result shows SVR outperforms than the other regression models.