NOAAs National Weather Service Probabilistic Storm Surge PSurge
NOAA’s National Weather Service: Probabilistic Storm Surge (P-Surge) Arthur Taylor, Bob Glahn, Wilson Shaffer MDL / OST November 30, 2005
Introduction NHC begins operational SLOSH runs 24 hours before landfall. • Provides a storm surge estimate for non-evacuation applications. Problem: Surges are based on a single NHC forecast track and associated parameters. • When provided accurate input, SLOSH results are within 20% of high water marks. • Track and intensity prediction errors are the largest cause of errors in SLOSH surge forecasts and can overwhelm the SLOSH results. Probabilistic Storm Surge 2005
Probabilistic Storm Surge Methodology An ensemble of SLOSH runs is created based on NHC’s official advisory and historic errors. • Create a SLOSH input track based on the advisory. • Create a set of SLOSH input tracks by varying the various input parameters based on historic errors. • Assign a probability to each SLOSH input track based on the likelihood of that track. • Run the SLOSH model on all the input tracks, and join the results together to compute the probability of surge exceeding various thresholds. Probabilistic Storm Surge 2005
SLOSH’s Input Track Location • Can get from NHC’s advisory Forward Speed • Can compute from NHC’s advisory. Radius of Maximum Winds (Rmax) • Not given in NHC’s advisory due to lack of skill in forecasting changes in Rmax. Pressure • Can only get the current value (no forecast values) from NHC’s advisory. Probabilistic Storm Surge 2005
SLOSH’s Rmax and Pressure Since NHC’s advisory does not provide Rmax, or forecast Pressure, we need to compute them. • The SLOSH parametric wind model relates Rmax, Pressure, and Maximum Wind Speed (Vmax). Given any two, the third can be computed. • Vmax is provided in NHC’s advisory. • Since the current Pressure is provided, one can estimate the current Rmax. • We assume that Rmax remains constant, then calculate the resulting Pressures. Probabilistic Storm Surge 2005
Example: Katrina Advisory 23 Probabilistic Storm Surge 2005
Varying Katrina’s Tracks • 1. 645 standard deviations (sd) to left and right, is equivalent to 90% of storms • 0. 67 sd to left and right would be average error • Spacing based on size of the storm • Calculations are done when 34 knot winds or greater are in the SLOSH basin Probabilistic Storm Surge 2005
Varying the Other Parameters: Size: Small (30%), Medium (40%), Large (30%) Speed: Fast (30%), Medium (40%), Slow (30%) Intensity: Strong (30%), Medium (40%), Weak (30%) Probabilistic Storm Surge 2005
Determine Which Basins to Run We try all SLOSH input tracks in all operational basins: • For each basin, eliminate tracks which never forecast tropical storm force winds. • Remove basins where all the tracks were eliminated. • Treat eliminated tracks as if they generated no surge in a basin. Probabilistic Storm Surge 2005
Calculate probability of exceeding X feet To calculate probability of exceeding X feet, we look at each cell in each SLOSH run’s envelope. • If that value exceeds X, we add the weight associated with that SLOSH run to the total. • Otherwise we don’t increase the total. • The total weight is considered the probability of exceeding X feet. • We are examining the need to calibrate the probabilities. Probabilistic Storm Surge 2005
Katrina Adv 23 Probability > 5 ft (approx. 24 hr before landfall) Probabilistic Storm Surge 2005
Arlene Adv 10 Probability > 5 ft (approx. 24 hr before landfall) Probabilistic Storm Surge 2005
Potential Products Product Types: • Probability of storm surge > X feet at any time during the run. • Probability of storm surge > X feet from time T 0 to T 1. Formats: • GRIB 2 (WMO’s GRIdded Binary) with multiple choices of X, and multiple time slices. • Images in the form of. png files • GIS data in the form of. shp files Dissemination Methods: • Use the National Digital Guidance Database (NDGD). • Put images / data on the NHC web / ftp site. Display Methods: • Improve the SLOSH display program to display / animate GRIB 2 • Web browser for the. png images, GIS for the. shp files Plans: FY 06 Experimental Products, FY 07 Operational Products Probabilistic Storm Surge 2005
Calibration To produce better probability forecasts, we can calibrate the method. • If we forecast 50% chance of exceeding X feet, does it actually exceed X feet 50% of the time? For all calculated probabilities (in 10% bands), find the actual relative frequency of occurrence. • For observations, we can use SLOSH’s best track analysis. • Use all historic storms making landfall over the last 10 years. • Since a single basin doesn’t have a large number of historic cases, we work on a single uniform grid • The uniform grid also has the advantage that each cell is the same size, so it can be weighted equally. Probabilistic Storm Surge 2005
Preliminary Calibration Results Combining: • Bonnie 98 • Floyd 99 • Isabel 03 • Charley 04 (FL) Probabilistic Storm Surge 2005
Preliminary Calibration Results Combining: • Bonnie 98 • Floyd 99 • Isabel 03 • Charley 04 (FL) Probabilistic Storm Surge 2005
- Slides: 16