The Lake Pontchartrain Forecasting System

Brett D. Estrade

Credits

• Center for Computation and Technology at LSU

– Shree Balasubramanian

– Archit Kulshrestha

– Prathyusha Akunuri

– Ian Kelley

– Chirag Dekate

– Jon MacLaren

– Gabrielle Allen

– Ed Seidel

• Louisiana Optical Network Initiative (LONI)

– Charlie McMahon

– Brian Ropers-Huilman

– LONI HPC Ops Staff

• Institute of Marine Sciences at UNC

– Jason Fleming

– Crystal Fulcher

– Rick Luettich

• A collaboration of CCT, UNC, and the US Army Corp of Engineers (USACE) in New Orleans

• Is a system used by USACE to help determine storm surge at 3 New Orleans' drainage canals– 17th Street Canal (-90.12566 29.98837)

– Orleans Ave. Canal (-90.10094 29.99637)

– London Ave. Canal (-90.06807 29.98851)

• Produces a forecast of what the storm surge will be at the heads of 3 canals given an NHC consensus track forecast and 4 variations of it

LPFS ...

A View of the Canalscompliments of CCT's Viz Group

N

The USACE's Dilemma• 3 major canals used to rid the

city of water during heavy rains will back up and provide a conduit for lake water to flow into the city during tropical events as evidenced by Katrina

• The current solution to close the canals during a tropical event requires a 6 hour effort and the use of tall cranes, which are by affected high winds

• Gating the canals reduces each canal's effectiveness by a great amount because it requires the use of pumps to move the water out

• A timely decision must be made wrt high winds and a reduction in ability to rid the city of rain water

wwltv.com's “17th Street Canal” Online Cam

How Does LPFS Help?

• LPFS provides easy to read graphs for each of the 3 locations showing for each of the 5 tracks:– Storm surge (i.e., elevation above sea level)– Wind speeds

• Due to the availability of dedicated LONI resources

during high and severe alerts, results are provided in a timely manner – within 2 hours of an NHC track forecast

LPFS Input – Hurricane Tracks

1.NHC Consensus2.20% Faster Winds3.20% Slower Winds4.Veer Right of Cone5.Veer Left of Cone

Picture of local news coverage at approximately the same time as the consensus track to the left

LONI Resources

LPFS System Overview

• LPFS is a distributed ensemble of 5 parallel ADCIRC simulations running on the Louisiana Optical Network Initiative's (LONI) state wide resources

• LPFS may run in a continuous way for the duration of a particular storm once it is initiated

• The status of the LPFS simulations are tracked in real time via a web portal interface

• Automated alerts are sent out at key points in the process and during specified events via various means – email, sms, and instant messenger

• Distributed LPFS is implemented such that no components need to reside on the same machine; additionally, redundant components could easily be used to arbitrarily increase the fault tolerance of the system. This allows us to determine the reliability of this system with some confidence.

ADCIRC – Storm Surge

• A finite element coastal ocean model

• Uses Generalized Wave Continuity Equation (GWCE) to solve for elevations about the geode (sea level)

• Uses Momentum Equation to solve for the u and v components of current velocities

• Uses Continuous Galerkin Finite Element Methods (node based)

• Single Program Multiple Data (SPMD) & uses MPI; scaling is linear up to some point where communication overhead dominates (~3-5000 nodes/cpu)

• Domain and input data decomposed prior to program execution

ADCIRC Website:

http://www.adcirc.org

LPFS General Work Flow

LPFS Distributed Implementation

LPFS Components

• ADCIRC

• Ensemble track generation

• Resource allocation and simulation distribution utilities

• Storm track analysis tools and automatic alert assessment

• A daemon that monitors the status of the simulations

• Portal interface to manage alerts, track the status of each running simulation, and distribute simulation results and product

A General LPFS Scenario

• A tropical event is detected well outside of New Orleans; an LPFS admin sets the current alert level from Green to Yellow via a portal interface;

– Alerts go out via email, instant messenger, and SMS to all personnel

– LPFS admins coordinate LONI resources and begin preparing to activate the LPFS system.

• NHC forecast places the track of the hurricane within 540 nm of New Orleans , and the current alert level is set from Yellow to Orange automatically

– Alerts go out via email, instant messenger, and SMS to all personnel

– LPFS admins begin running the system regularly for each new NHC advisory using LONI resources under the PRIORITY queue;

– USACE personnel are notified of new storm surge forecasts via E/IM/SMS within 2 hours of each NHC advisory

• NHC forecast places the tracks of the hurricane with 270 nm of New Orleans,and the current alert level is set from Orange to Red automatically

– Alerts go out via E/IM/SMS to all personnel

– LPFS now utilizes LONI resources in a preemptive, dedicated mode (checkpt'd jobs are killed)

– USACE personnel are notified of results as they are readied

LPFS Alerts and Notifications

• The color coded alert system indicates various states of readiness and resource allocation on LONI resources

• Green

– Low threat from storm. System in green alert status, standard priority for resource usage.

• Yellow

– Potential threat from storm. System in yellow alert status, resources on alert for potential use.

• Orange

– High threat from storm. System in orange alert status, enhanced priority for resource usage.

• Red

– Severe threat from storm. System in red alert status, highest priority for resource usage.

LPFS Alerts and Notifications

Instant Messaging

E-mail

SMS

LPFS Portal Interface

http://panther.hpc.lsu.edu:50080/gridsphere/gridsphere

LPFS Output – Wind & Elevation

The Future of LPFS

• First and foremost, we must continue to improve LPFS' usefulness and dependability in order to aid USACE and the people of New Orleans;

• Improve the system's ease of use and product delivery mechanisms

• Allow for the selection of other point of interest with in the simulation domain

• Develop further the alert/notification scheme further

• Create a more distributed, asynchronous, and data driven system– i.e., be able to distribute an arbitrary number

ensembles with an arbitrary number of tracks per ensemble

Conclusion

• LPFS' first priority is to serve the needs of USACE and their efforts to protect the New Orleans Metro Area from the Lake's storm surge during a hurricane

• LPFS is quickly maturing into a positive example of State, Federal, and Academic institutions cooperating in the name of public safety

• The distributed LPFS is becoming a good framework to begin investigating many interesting distributed, data driven problems