Water in All Phases (WIAP) Analysis

Goals:

The long-term goal of the project is to offer quarter-hour, 5km-resolution analyses of water species in all phases (vapor, clouds, liquid, ice, and soil moisture) over the CONUS.  The analysis will use state-of-the-art data in real-time, including radar, satellite, GPS, aircraft, and surface reports.  The intended purpose is to improve model initialization.  This task will require the analysis to be run on the new FSL HPTi supercluster computer (Jet).  We will progress toward this goal in logical steps beginning at 13km, followed by 10km 1 hr (current) and then to smaller scales and shorter cycle intervals.
 

Status:

Currently the analysis runs every 3 hours starting at 00ut.  The domain is as pictured below and the resolution is 13 km.  A 10-km version is under test and runs once daily.  The plan is to reduce the memory requirements of the software to allow it to run as frequently and at a high a resolution as possible.  While this work continues on the preliminary test platform (IBM 397) our group is porting the software and configuring it to run on the HPTi.  We understand that our goals can only be met on the HPTi.

Current HPTi issues are:

• Interfacing output files to the web (meeting to be held this week with FD)
• Automating weekly builds on Jet (now tested weekly)
• Running the analysis in cron on Jet (problems with cron, John Smart is working details with George Carr)
• Casting the code in parallel form (Not a big item yet, first priority is to get the code to run on one node)
• Debugging the software on jet.  Unfortunately we are seeing runtime errors not emulated on other machines running the same domain.  These appear unique to the Jet compiler/architecture.

• Running the analysis at 15 minute intervals (only 30 minute updates have been achieved to date)
• Integrating GPS data. Now available in real-time. Testing on the ROC grid initially.
• Debugging problems running GOES derived TPW in the large-scale analysis.  This presents no problems to ROC, but was found to have problems in the larger domain grids such as WIAP.
• Integrating the balance package

Domain:

The following parameters were chosen to maximize the continental US in a rectangle.

• True at 40 degrees north latitude
• Pole longitude is 100 degrees west
• Tangent at 40 degrees north latitude
• Lower left coordinate 23.9N 120.45W

• 469 x 285 points

• 937 x 567 points  (= 531279 points, if you include 21 levels, =11.1 million points per 3D field)

Test Data (netCDF files of various fields, updated every 3 hours):