LPFS Demo @ SC06
The Lake Pontchartrain Forecasting System
This page describes the LPFS demo done at SC06 (Tampa, Fl; Nov 2006) where multiple surge models were deployed in an event driven manner to the various machines of the LONI network
LPFS [info] is a collaboration which joins CCT with the Institute for Marine Sciences at UNC and the US Army Corps of Engineers using the computational resources of the Louisiana Optical Network Initiative (LONI). LPFS is also connected to the SCOOP project and the CCT Katrina Visualization Project.
This project is creating an automated wind speed and water level forecast system for Lake Pontchartrain on distributed LONI resources to help with planning flood gate closures. Forecasts are generated following each tropical storm advisory update issued by the National Hurricane Center.
The general procedure is to create an ensemble of multiple ADCIRC storm surge simulations based on the consensus storm forecast from the National Hurricane Center and perturbations to this forecast. Winds are generated using a synthetic wind model (based on the Holland wind model).
Brett and Archit testing the system before the LPFS demonstration.
The ensemble produces forecasts of predicted water level and wind speed along the southern shore of Lake Pontchartrain near the mouths of the drainage canals, and these forecasts are dissimintated via web pages within two hours of the initial advisory.
The LONI network topology across the State Louisiana is shown here. (see larger image)
The resources provided by LONI are "on demand" and event driven. The event driving the need for resources is a hurricane or strong tropical disturbance entering the Gulf of Mexico. The threat that the storm is posing on New Orleans determines the alert level (yellow, orange, or red), and this determines the level of priority the system gets on LONI's resources. At the highest threat level, red, LPFS is given dedicated access to the LONI resources.
The specific LONI machines provided by LONI for the demonstration were: Bluedawg (LaTech), Ducky (Tulane), and Zeke (ULL).
A 3D rendering by Werner Benger of the southern shore of Lake Pontchartrain showing the 3 canals of interest.
Alert Levels
| RED | Severe threat from storm. System in red alert status, highest priority for resource usage. |
| ORANGE | Severe threat from storm. System in red alert status, highest priority for resource usage. |
| YELLOW | Potential threat from storm. System in yellow alert status, resources on alert for potential use. |
| GREEN | Low threat from storm. System in green alert status, standard priority for resource usage. |
Demo System Description
A system diagram of the distributed LPFS system demonstrated at SC06.
For the demonstration, an LPFS run was initiated via the portal interface. With the alert level at GREEN, the action initiating the LPFS run changed the alert level to YELLOW.
A screenshot of the portal interface for the alert level, which is currently set at ORANGE.
During the analysis of the NHC consensus track and its 4 variants, the alert level was adjusted if any of the 5 tracks came within 570 nm (ORANGE) or 270 nm (RED) of New Orleans. The tracks were analyzed independantly of the portal, and the alert level was changed using an the XMLRPC interface provided for by the portal interface.
During run time, each simulation reported a status back to a daemon monitoring each run. Based on these status reports, another portal interface show the current status of each run on the different resources.
Alerts and Notifications
The portal interface also provides an XMLRPC interface to send alerts and notifications to different groups. For example, the instant message sequence illustrated below is sent to LPFS Admins during system start up.
A screenshot of the IM notifications received once a track is processed.
Results
As each simulation in the ensemble completes, the output files for each of the 3 canals are updated and the proper personnel are notified via the alert mechanisms described above. Results are also sent to the LSU SCOOP Archive for applications that may want to use this data at a later point in time.
Documents
Participants
Dr. Gabrielle Allen (CCT/CS), Brett Estrade, Archit Kulesthra, Shree Balasubramanian, Prathyusha Akunuri, Ian Kelly
|
Print
this Page
Refer
this page to a friend
