A Distributed, Real-time Hurricane Wind Analysis System

Principle Investigator: Mark Powell
Co-Investigators:
Nirva Morisseau-Leroy
Sonia Otero
Sam Houston
Nick Carrasco
George Soukup

Sponsor:

NOAA High Performance Computing and Communications Program
National Institute for Building Sciences
HRD

Awards: NOAA Tech 2000 Conference Best JAVA Implementation

1. Statement of Problem

Response to a hurricane disaster depends on obtaining accurate and timely information on the magnitude and geographical distribution of the damage caused by the event. This information will determine which communities were most devastated and require immediate attention; it will also assist decision-making associated with the recovery process. Advances in computing and communications have made it possible to obtain tropical cyclone wind observations in near real-time (Burpee, et al., 1994, Griffin, 1992, Powell et al., 1996, 1998). However, scientists involved in hurricane forecasting and research have few tools that enable real-time examination and analysis of these data. Hurricane wind fields are determined subjectively based on the specialist's interpretation of flight-level reconnaissance data, satellite observations, pressure-wind relationships and available surface data. These fields are represented by text portions of the NHC forecast product as radii (from the storm center) of 34 kt, 50 kt, and hurricane force winds in four compass quadrants relative to north. Emergency managers require accurate and fine scale wind field information to input to geographic information systems (GIS) and damage models. Emergency managers have access to only coarse scale information on the wind field through the wind radii information provided in NHC forecasts and advisories, and are considering the use of parametric wind models to provide more detailed wind field coverage. These wind models are no substitute for real data and tend to oversimplify hurricane structure by not adequately accounting for asymmetries in the wind field caused by convection, land-sea friction differences, intensification, and environmental wind shear. Furthermore such models may contradict the operational forecast information which could lead to confusion among emergency managers. Numerical forecast models are initialized from dynamically constrained analyses of some of the same observations mentioned above. However, these fields are not available in real time, do not have the most recent observations due to early cut off times, lack mesoscale detail due to the dynamic constraint, and do not typically contain high resolution information from the reconnaissance and research aircraft.

2. Proposed Solution

We propose a real-time system for examining, synthesizing and objectively analyzing meteorological observations in hurricanes, using a common framework for wind exposure, measurement height, and averaging time. This system, called "H*WIND", will allow scientists to interact with the observations, perform quality control, and select from a menu of several graphical products depicting meteorological fields for storm diagnosis and forecast guidance. The HRD approach to hurricane wind analysis has evolved from a series of peer-reviewed, scientific publications analyzing landfalls of major hurricanes from 1979-1998 (Powell and Houston, 1996, 1998, Powell et al., 1998). Since 1994, HRD wind analyses have been conducted on an experimental basis to create real time hurricane wind field guidance for forecasters at the National Hurricane Center. These analyses are placed on the web after 24 h (www.aoml.noaa.gov/hrd see storm atlas) and have become the standard for assessing winds from new remote sensing platforms (e.g. RADARSAT, QUIKSAT) and parametric wind models used by the insurance industry. In addition we have had discussions with other operational and research organizations about using HRD wind fields on an experimental basis for initializing some numerical forecast models and to provide forcing for storm surge prediction models.

For prelandfall cases, the proposed system would create products designed to be useful for disaster planning. For example, wind analyses could be combined with the track forecast and error statistics to project "swath" fields depicting potential damage along an envelope of possible storm tracks. For hurricane landfall cases, the system would provide a menu of products designed specifically for real-time emergency management damage assessment applications. These fields will be formatted to be readily input to GIS and damage modeling software such as FEMA's HAZUS.

All available wind measurements gathered by reconnaissance aircraft, satellite remote sensors, airborne Doppler radar, ships, buoys, coastal and inland automatic weather stations will be automatically downloaded, preprocessed, and stored using a modern object-relational database located on a dedicated server. One or more additional machines will be dedicated as an objective analysis server and an application server. The application server will use distributed object technology to communicate actions and events between the clients and the database and analysis servers. All data and analyses will be archived on the database server and JAVA applets and IDL will allow graphical and gridded products to be created dynamically from the analysis archive and delivered to the client "on the fly". Scientific users would interact with the data and archive through a hardware-independent Web browser client or by a JAVA workstation client. Emergency managers would interact with current products based on operational or poststorm analyses through a Web interface. For scenario studies, users could construct products searching the archive by year, storm name, storm category, or geographic area.


Projects indexed by Scientist
Projects indexed by Subject
Return to Homepage