Experimental P-Surge 2.0: Probabilistic Hurricane Storm Surge (with tide)

Overview | Product Description | Education

P-Surge stands for Probabilistic Hurricane Storm Surge. It is an ensemble model based on the SLOSH model, a storm surge model in use by the National Weather Service to help forecast surge from hurricanes. P-Surge takes the hurricane forecast from the National Hurricane Center (NHC) and uses error statistics from past hurricane forecasts to create many statistically probable storms. P-Surge then computes the possible storm surge from each of these and combines them together to get a better idea of the probable storm surge within the next 80hours.

Creation of Possible Storms

The P-Surge model takes the National Hurricane Center's (NHC) official forecast and varies it based on the historic forecast error. An example of NHC's official forecast for Katrina 2005:

The model varies it based on cross track errors:

as well as errors in size, forward speed and intensity:

To do so, the model uses the error statistics for NHC's forecasts in 2008:

Error Type 0-h 12-h 24-h 36-h 48-h 72-h 96-h 120-h
Cross-track (nm) 3.1 15.5 26.7 35.4 45.6 66.5 103.0 133.3
Along-track (nm) 3.6 18.0 31.1 44.3 56.3 83.0 113.6 152.0
Intensity (kt) 2.4 7.8 11.4 13.6 14.8 16.0 17.5 21.0

The error statistics for the along-track (or forward speed) and cross-track errors are 85% of the five year mean errors from 2003-2007 inside a box defined by 10-45N, 60-100W. The 85% accounts for the fact that forecast tracks of storms with winds of 50 knots or higher are more accurate. The reason for the box was to sample storms near the rawinsonde network.

The intensity errors are the five year mean errors for storms with winds of 50 knots or higher in the entire Atlantic basin.

Each permutation is then assigned a weight based on its likelihood. The storm surge model (SLOSH) is then run for each permuation and the results are combined to create probabilities based on those weights.

Above Ground vs Above Datum

The original P-Surge model computed water level above the model's datum, or zero point. The new model also computes water level above ground. The differences between these two products is illustrated below. The current model uses a datum of NAVD-88, which is approximated as Mean Sea Level in the below images.