Tuesday, 25 January 2011: 2:00 PM
613/614 (Washington State Convention Center)
An accurate quantification of uncertainty is of vital importance for a reliable risk assessment of hazards attributable to or influenced by meteorology, including (but not limited to) hurricane landfall, wintertime roadway conditioning, and the transport of harmful airborne contaminants. This last application is the focus of this study. Ensembles are a common way to quantify the uncertainty in meteorological variables, but ensembles may not accurately reflect the actual uncertainty (the variance of the forecast errors). The authors have previously introduced the Linear Variance Calibration (LVC) technique as a way to calibrate ensembles to more reliably predict uncertainty. The LVC calculates a linear regression between ensemble variance and error variance. The statistical correlation of this calibration is high for low-level winds, which are an important field for atmospheric transport & dispersion (AT&D) forecasts, but the impact of the calibration on the resulting application (here, AT&D forecasts) needs to be evaluated as well.
We require a large sample to properly determine the impact of wind variance calibration on the reliability of AT&D uncertainty predictions. To that end, we have created a joint meteorology-dispersion testbed to produce this large sample. This testbed produces not only regular dispersion forecasts based on a meteorological ensemble forecast, but also an analysis of meteorological conditions that is used to create a baseline dispersion simulation. In the absence of a long-lived regional dispersion field experiment, this baseline provides a pseudo-truth against which we can measure the effectiveness of dispersion forecasts with and without wind variance calibration. Different variations of the LVC are also tested. All forecasts are compared to the baseline simulation to determine the accuracy and reliability of the resulting concentration and dosage forecasts.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner