Dust storm ensemble forecast experiments in East Asia

The ensemble Kalman filter (EnKF), as a unified approach to both data assimilation and ensembleforecasting problems, is used to investigate the performance of dust storm ensemble forecasting targetinga dust episode in the East Asia during 23--30 May 2007. The errors in the input wind field, dust emissionintensity, and dry deposition velocity are among important model uncertainties and are considered in themodel error perturbations. These model errors are not assumed to have zero-means. The model error meansrepresenting the model bias are estimated as part of the data assimilation process. Observations from aLIDAR network are assimilated to generate the initial ensembles and correct the model biases. The ensembleforecast skills are evaluated against the observations and a benchmark/control forecast, which is a simplemodel run without assimilation of any observations. Another ensemble forecast experiment is also performedwithout the model bias correction in order to examine the impact of the bias correction. Results show thatthe ensemble-mean, as deterministic forecasts have substantial improvement over the control forecasts andcorrectly captures the major dust arrival and cessation timing at each observation site. However, theforecast skill decreases as the forecast lead time increases. Bias correction further improved the forecastsin down wind areas. The forecasts within 24 hours are most improved and better than those without the biascorrection. The examination of the ensemble forecast skills using the Brier scores and the relative operatingcharacteristic curves and areas indicates that the ensemble forecasting system has useful forecast skills.