HFIP Real-Time Experimental Forecast System for 2011

  • About
  • Schedule and Presentations


April 14, 2011


SSMC2 14316
1325 East West Hwy
Silver Spring, MD


We need to have a discussion of the target system we want to run next summer that we will run on t-jet as part of the demo system and stream 1.5.

In general the system this summer should consist of a global ensemble and a regional ensemble (each could have more than one core) with the number of members and resolution TBD. Beyond that we need to see what will fit, what will be ready to go by next summer and what makes sense. As close as possible this should be a stream 1.5 activity though that isn't a requirement.

For the format of the meeting I suggest that the various organizations planning or proposing to run in real-time present their proposal to the group for discussion. This presentation should include detail of the model, initialization and planned post processing/products. They will also need to present detailed estimates of the computing resource requirement (# of cores and run time), when the components are proposed to be run and if the core requirements change during the run (for example if the requirement is different for initialization vs the forecast).

We also plan a presentation on the t-jet resource, how the real-time runs will be managed and options for expanding the HFIP computing resource in Boulder. We will also give an opportunity for NHC to indicate their preferences for what they would like to see run on the HFIP computing system and for products from that system.

There will be ample time for discussion and a goal of the workshop will be a reasonably detailed plan on what will be run, how it will be managed and when the various components will be run (for example, what happens when there is more than one storm, what happens when there are none).

A rough schedule for the meeting is shown below.

I will present a preliminary proposal for the makeup of the real-time t-jet system and during general discussion that plan will be modified and tuned.

The goal of the meeting is to come to agreement on the 2011 real-time system recognizing that there may need to be some later modifications as new information becomes available.

Agenda and Presentations

Thursday April 14, 2011

  8:30 am    Welcome, overall objectives/goals for t-jet (Bob Gall)
  9:00 am    Goals of workshop/a tentative proposal (Bob Gall)
  9:30 am    MMM Proposal for real-time runs in 2011 hurricane
                   season (Ryan Torn)
10:00 am    The t-jet system/ real-time management, options
                   for enlarging the HFIP computing system.
                   (Craig Tierney)

10:30 am    Break

11.00 am    ESRL Proposal for real-time runs in 2011 hurricane
                   season (Jeff Whittaker)
11:30 am    EMC Proposal for real-time runs in 2011 hurricane
                   season (beyond the operational models) (Vijay)

12:00 noon Lunch

  1:00 pm    AOML Proposal for real-time runs in 2011 hurricane
                   season (Frank Marks)
  1:30 pm    NRL Proposal for real-time runs in 2011 hurricane
                   season (Alex Reinecke)
  2:00 pm    PSU Proposal for real-time runs in 2011 hurricane
                   season (Fuqing Zhang/Yonghui Weng)

                  GFDL Proposal for real-time runs in 2011 hurricane
                   season (Tim Marchok)

  2:30 pm    Begin general discussion of HFIP 2011 real-time
                   system (Gall-Lead)

                   - Start with Gall preliminary proposal and iterate
                   - Consider:
                         o Systems that are likely to be part of the
                             2012 experimental forecast system
                         o Cost (we need firm estimates of runtimes
                             and core requirements of individual
                             proposed components)
                         o Scheduling
                               What do we do when there are multiple
                               Which models should run first if they have
                                 to be run sequentially
                   - Which components will be stream 1.5 candidates
                         o Scheduling retro runs and analysis

  3:00 pm    Break

  3:30 pm    Continue discussion
  4:30 pm    Finalize tentative plan

  5:00 pm    Adjourn