LPFS 2009 Scope of Work Rick Luettich Jason Fleming Robert Weaver February 19, 2009
Jan 04, 2016
LPFS 2009 Scope of Work
Rick Luettich
Jason Fleming
Robert Weaver
February 19, 2009
LPFS Overall Goals
To provide reliable, timely, accurate guidance for
storm surge in southern Louisiana To provide accurate estimates of high water marks
and areas of inundation immediately after landfall
LPFS Overview
Automated software system for storm surge guidance in SE LA; written around ADCIRC
Uses NHC Official Forecast Generates 5-storm ensemble (consensus, 20%
stronger, 20% slower, veer right, veer left) Post process results for south shore of Lake
Pontchartrain and West Bank (Harvey Canal) Autogenerated output includes hydrograph and
wind speed plots
2008 Season Retrospective
Hurricane Zeus: Preseason Drill
Created and ran 20
forecast tracks Made Google maps of
each Provided hydrographs
of LP and West Bank Summary spreadsheet
for coastal parishes
Hurricane Zeus: Preseason Drill
Type and scope of output broader than anticipated Sapphire (ERDC supercomputer) reboot Drill held 13 May; COE pleased with the output
we provided (too much information in hindsight) Hurricane Zeus drill was picked up by media
Humid City blog: “Corps Cutting it Close at
Floodgates This Year” Baton Rouge Advocate “Corps Tests System of Storm
Protection”
Hurricane Dolly
Hurricane Dolly
7/20: Dolly formed in Western Caribbean; Corps'
internal forecaster Bill Frederick indicated Dolly
not a threat to MVD; LPFS activated at UNC 7/21: Corps Directive: Only activate LPFS on
Sapphire when Frederick forecasts MVD impacts 7/21: Dolly made 1st landfall in Mexico; we were
added to Frederick's forecast email list
Hurricane Dolly
7/22 Frederick forecast MVD impacts of 3 ft surge
west of Morgan City (outside target area) Preparations made to run on Sapphire Final landfall 7/23 on South Padre (TX) as Cat 1 LPFS ran at UNC from 7/20 to 7/24 and reliably
produced surge guidance for 15 advisories (no
impact predicted for target area)
Tropical Storm Edouard
Tropical Storm Edouard, 3 August
Tropical Storm Edouard, 3 August
TS Edouard forms just off the coast of LA We activated at ERDC (in standard queue) with
backup at UNC (Lake Pontchartrain only) We request dedicated queue on ERDC machine ERDC computer experienced power fluctuations
causing network failure Bill Fredericks forecasts 2-3 ft surge in coastal LA
Tropical Storm Edouard, 4 August
Request for dedicated queue on ERDC computer
temporarily denied pending higher authorization;
subsequently granted LPFS ran from 8/3 to 8/4, producing surge
guidance for 5 advisories at ERDC and 8 at UNC;
worst case predictions were ½ foot surge, winds 8-
10 kts TS Edouard made landfall near Port Aurthur, TX
Tropical Storm Fay
Tropical Storm Fay, 15 August
Tropical Storm Fay, 15 August
TS Fay forms over Dominican Republic LPFS activated on ERDC computer, as well as
LONI computer as backup, and UNC computer (as
tertiary backup) Dedicated queue requested on ERDC computer,
and granted LPFS running on all platforms – no significant
impact predicted
Tropical Storm Fay, 17 August
Tropical Storm Fay
8/17: Bill Fredericks forecasted no MVD impacts,
will continue to monitor; Captain Royston
indicated no further updates are necessary 8/18: Request made for surge guidance based on
the latest track from the GFS model instead of
official forecast; 1st landfall in Florida Keys 8/18: We begin the “ALT Fay” series of guidance
using manually developed forecast tracks based on
guidance from meteorological models
Tropical Storm Fay
GFS based guidance:
no significant impact 8/19 2nd landfall near
Naples FL ERDC opens up
additional computer 8/20 Surge guidance
requested based on
BAMM model instead
Tropical Storm Fay, Met Guidance
Tropical Storm ALT Fay, 20 August
Tropical Storm Fay, August 21
3rd Landfall, near Daytona Guidance based on BAMM model indicated 3-6 ft
surge along eastern coast of SE LA Guidance requested for NGFDL track with fwd
speed of 8 mph and 80 kt max wind; results
indicated 5 ft surge on south shore of Lake Pont. Sapphire (ERDC machine) down for emergency
maintenance; UNC machine down all September
Tropical Storm Fay, 22 August
Sapphire rebooted, removed from service for
emergency maintenance, then rebooted again Request made for guidance based on 6 day run of
NGFDL using 60kt and 80kt max winds Lake Pontchartrain runs available quickly, West
Bank runs require much more time Contour plot of surge guidance used to estimate
surge in IHNC Fay re-emerged into Gulf
Tropical Storm Fay, 22 August
Tropical Storm Fay, 23 August
4th (and final) Landfall near Carrabelle on FL
panhandle (dissipated inland following day) Last sets of official results indicated no significant
impacts in target area LPFS ran from 8/15 to 8/23 and provided guidance
for 19 advisories from ERDC; 35 advisories at
UNC (Lake Pontchartrain only) Also produced custom guidance for 5 tracks from
3 advisories based on met. model guidance
Hurricane Gustav
Hurricane Gustav, 25 August
Formed west of the
Windward islands and
rapidly strengthened
from disturbance to
hurricane Bill Frederick issued
wait-and-see forecast
Hurricane Gustav 26 August
We request LPFS Activation at ERDC at 0800;
EOC elects to hold off dedicated queue request Bill Frederick forecast is northern Gulf coast We make request for increase from 160 CPUs to
640 CPUs on Sapphire to speed West Bank results 2100 Official request made to activate LPFS on
Sapphire with “more power”
Hurricane Gustav 27 August Media and private citizens requests for
information made directly to LPFS Team at UNC,
appealing for release of results All inquiries forwarded to COE; LPFS results to
be made available only to Heath Jones and Nancy LPFS Activated on Sapphire using standard queue,
dedicated queue access scheduled to begin on 28th Need for 640 CPUs reiterated to MSRC MSRC requests justification for more CPUs;
eventually grants 640 CPUs for Gustav only
Hurricane Gustav 28 August
LPFS running in dedicated queue on 640
processors Contour plots of High Water Marks in Google
Earth format provided in addition to hydrographs Clarification requested for storm parameters:
pressure, windspeed, Rmax, lat/lon at landfall, etc Ensemble redefined for custom Rmax and custom
tracks (e.g., veer east by 1 degree only)
Hurricane Gustav, 29 August
Hurricane Gustav 29 August
3 of 4 LPFS storm tracks failed on Sapphire for
the 5am EDT advisory cycle LPFS guidance indicates 4ft/40kts in Lake
Pontchartrain and 10 ft/45kts on West Bank Bill Frederick issues estimates of 14-20ft on
eastern shore and 4-9 ft on northern Gulf coast
Hurricane Gustav 29 August
Hurricane Gustav 29 August
Hurricane Gustav 29 August
Hurricane Gustav 29 August
Hurricane Gustav 30 August
Overnight advisory cycle failed on Sapphire (node
failure) Additional custom adjustments to Rmax for
ensemble storms were requested 5 additional output stations were requested beyond
the original 2 (Lake Pontchartrain and Harvey
Canal) for a total of 7 Had to rewrite backend post processing code
Hurricane Gustav 30 August
Hurricane Gustav 31 August
Run failures on Sapphire became more and more
frequent ADCIRC runs performed on LONI hardware used
to advise Corps LONI runs used much larger grid, were available
much less frequently
Hurricane Gustav, Landfall
Shape file of high water marks were requested,
looking at hindcast First time LPFS results have been been used by
emergency responders after landfall LPFS ran from 8/25 to 9/2 and produced guidance
from 11 advisories with 2 output stations, and 3
advisories for expanded list of 7 output stations
September 2nd
Hurricane Ike
Hurricane Ike
Formed September 1 List of output stations expanded to 23 Additional outputs of wind model requested at 0,
3, 6, and 9 hours Request made for adjustments to hydrographs and
contour plots 640 CPUs requested on Sapphire and Jade
(approved)
Hurricane Ike
Problems continued on Sapphire Post processing code for 23 stations pushed the
limits of our plotting code Ike trended west, out of the target zone for the
LPFS grid Full sl15 grid was run on LONI to provide output
for whole LA coast Outputs consisted of contour plots of High Water
Marks
2008 Season Summary
Very active season for the Gulf Broadest participation in the surge guidance
process to date Tremendous effort by all involved to overcome
challenges and provide best guidance New technical requirements evolved as the season
went on
2009 Changes and Enhancements
Reliability
Flexibility
Coverage
Timeliness
Accuracy
Reliability
Limit the number of in-season changes any change may introduce/reveal bugs (incl. hardware) in-season changes make rigorous testing more difficult LPFS2009 will be more flexible reducing the need for
late changes Implement structured testing
run forecasts in test mode test mode carefully designed to cover all possible
situations
Flexibility
2008 experience indicated the need for swappable bathymetry input files (mesh) reconfigurable storm ensemble seamless modification of station output
Changes will be made such that ADCIRC input files easily swapped number and character of storms in ensemble may be
dynamically modified, advisory-by-advisory number and location of output stations dynamically
modifiable
Coverage
Geography Beyond Lake Pontchartrain, beyond West Bank 2009 will fully cover Louisiana coastline
Time span Early stages
What if only model guidance is available? 2009 will cover pre-official forecast period
Late stages After storm has made landfall, hurricane forecast uncertainty
disappears, surge guidance accuracy is maximized 2009 will advise after landfall for emergency responders
Timeliness
Fast turnaround of results will require greater
processing power 640 CPUs on sapphire produced a result in 1 hour
(per storm in the ensemble) Therefore, 3-storm ensemble will require 1920
CPUs to produce results in one hour We will request a dedicated reservation of 1920
CPUs for the 2009 season
Accuracy
Pleased with the accuracy of our surge guidance in
2008 season Detailed comparison of measured data and surge
guidance is still ongoing Enhanced accuracy in 2009 season via
use of most up-to-date bathymetry transition to asymmetric vortex wind model
Conclusions
Broad participation has provided a tremendous boost to the guidance process
The efforts, frustrations and successes of the 2008 season have informed the development of LPFS
The 2009 season will build on past experiences to provide the most reliable, timely and accurate surge guidance available