Tuesday, 17 April 2012
Heritage Ballroom (Sawgrass Marriott)
During the 2010 Northern Hemisphere tropical cyclone (TC) season, the Naval Research Laboratory ran an experimental NOGAPS ensemble for the purpose of TC track prediction. The ensemble consisted of 21 members, each run at T319L42 resolution, identical to that of the contemporary operational deterministic NOGAPS forecast. Here we present an assessment of the quality of the NOGAPS ensemble-based TC track predictions, both deterministic and probabilistic. Two forecast assessment issues will receive special focus. The first issue arises from the fact that all ensemble members do not simultaneously dissipate a predicted TC at the same forecast lead time. Hence, the number of ensemble members with a position forecast for a particular TC often changes with forecast lead time. We will address the implications of changing ensemble membership for interpretation and assessment of ensemble-based TC position forecasts, particularly the ensemble mean. The second focus area concerns fitting a continuous forecast probability distribution to a discrete ensemble of TC forecast positions. There are a variety of plausible choices for modeling the TC position distribution; for instance, circles or ellipses could be used to define contours of equal probability. We will use probabilistic forecast verification to assess the relative performance of fitted TC position distributions under different fitting methodologies.
![](images/flag.png)
![](images/blue_ribbon.png)