On Modeling of Atmospheric Visibility Classification Forecast with Nonlinear Support Vector Machine

Based on the consecutive high temporal resolution data observed at the special automatic weather stations ROSA on Beijing airport highway during 2006-2007, the modeling of atmospheric visibility classification forecast model with the nonlinear support vector machine method was discussed and evaluated in this paper. The evaluation result shows that the performance of the forecast model by the support vector machine method was good. 40% of atmospheric visibility classification forecast is consistent with the observed data; more than 90% of the forecast classification errors is within one level (including equality). Moreover, in the future 3-48 h forecast the atmospheric visibility performed stably. The perfect forecast results verify that the support vector machine method has strong capability of processing the nonlinear relationship between atmospheric visibility and meteorological factors.