Abstract
Tropical cyclone (TC) forecasts are subject to large amounts of uncertainty. The current operational method of estimating the uncertainty is via historical errors from the past five years of forecasts. Circles are drawn around each of the forecast points such that the actual position of the TC is within the circle 2/3 of the time. The edges of the cone are then connected to create the "cone of uncertainty"
There are drawbacks to the standard "cone of uncertainty". The first is that the cone is misunderstood by many. The second is that the cone remains the same size for all TCs, regardless of the forecast uncertainty or confidence. A forecast with very high confidence will have the same cone size as one with low confidence. The third drawback is that only one possible track can be depicted. Thus, it is possible that a path well away from the area drawn by the cone will be possible.
To overcome the limitation of a historical error- based "cone of uncertainty", we can use model ensembles. Model ensembles are multiple runs of operational forecast models with tweaks made to the initial conditions or the model physics. The ensembles create multiple track forecasts. Probabilistic forecasts can be generated from these forecasts, which can create an objective estimate of the forecast uncertainty based upon the current state of the atmosphere.
Results from the 2016 Atlantic and Pacific hurricane seasons indicate that an ensemble-based probability swath can be used to estimate the forecast uncertainty for TCs. For hurricanes Darby, Lester, Madeline, and Matthew, data indicate that this ensemble-based probability swath is a suitable replacement for the "cone of uncertainty". These four hurricanes remained within the 20 percent swath 71 percent of the time for a 5-day period, which is slightly more than the desired 2/3 of the time depicted by the standard error cone.
