Characterizing Regime-Based Flow Uncertainty for Source Term Estimation Applications

Source term estimation remains a considerable challenge, not least of which is due to the uncertainty in analyzed and forecast wind fields that advect the constituent. The goal of this work is to develop a regime-based quantification of horizontal wind field uncertainty utilizing a global ensemble numerical weather prediction model. In this case, we are utilizing the Global Ensemble Forecast System Reforecast (GEFSR). We will test several unsupervised machine learning methods for clustering horizontal flow fields and the forecast uncertainty in these flow fields for different vertical levels and for different forecast times for regions across the globe. We will test results of our model trained on GEFSR with numerically created HYSPLIT plumes created from analysis fields of the High Resolution Rapid Refresh as well as field experiment data plume releases such as CAPTEX. The ultimate goal of this research is to develop an applications to inform decision makers conducting source term estimation on the confidence of the forecast trajectory from a suspected constituent to drive decisions on sensor emplacement and utilizing resources to increase forecast confidence, such as conducing high resolution model runs. Another application is regime-based confidence in field campaign sensor emplacement plans.