HYCOM Brief: Moving the Model from Science to Operations
Implementing HYCOM as an Operational Model at NAVO Frank L. Bub Naval Oceanographic Office Ocean Modeling Technical Lead HYCOM Consortium Meeting 24-26 April 2007 Approved for Public Release Distribution Unlimited NAVOCEANO HYCOM Presentation 24APR07 Slide 1 Outline 1. OCEAN MODEL APPLICATIONS 2. INSTALLATION AT NAVOCEANO 3. THE TRANSITION PROCESS ABSTRACT: A discussion of the practical aspects of modeling like schedules, computer asset requirements, data flows (in, internal and out), NAVOCEANO and MSRC infrastructure, model products (construction, delivery, archiving), policy hurdles, the movement from R&D through system installation to operational in real-time, etc. That is, how NAVOCEANO takes the work of science and turns it into production. NAVOCEANO HYCOM Presentation 24APR07 Slide 2 1. OCEAN MODEL APPLICATIONS Navy business - fight and win wars
Ocean knowledge - a force multiplier Products for Navy customers ASW Sound speed Fronts and eddies Deep water to coastal regions MIW Currents Coastal to deep water NSW Currents, waves Nearshore, beaches, estuaries Fleet Operations amphibious landings Nearshore currents Surf zone NAVOCEANO HYCOM Presentation 24APR07 Public web images Slide 3 ASW Forecasting Sound Speed Structure Surface duct Sound trapping wave guide (mixed layer, sonic layer) Frequency trapped relative to inverse of thickness Cylindrical spreading Sound loss due to reflection (roughness-waves) and leakage to deep
Structure of sound speed Vertical wobbles internal tides change surface duct thickness Variations change ensonification Diurnal surface warming changes duct form NAVOCEANO HYCOM Presentation 24APR07 Bob Helber, Charlie Barron (NRL) personal communication Slide 4 Forecast Requirements 3D ocean eddies and fronts change horizontal structure Tau Location and structure affect sound propagation Movement (advection) means changes that affect ASW 00 Forecast Periods 2 days synoptic or real time (model run finishes and delivered at about +18 hours) 3 days tactical or tomorrows fleet placement 7 days strategic or plan the next few days NAVOCEANO HYCOM Presentation 24APR07 Tau 72 Global NCOM - East Coast
Slide 5 Aside: Model Resolution Questions What improvements does 1/12 or 1/25 degree bring us over 1/8 G-NCOM? Should we run global 1/25 or a 1/12 resolution with nested high resolution domains (1/30, 1/60) in areas of interest? What resolution is required to resolve Navy applications? What does higher spatial resolution give us scientifically If 5 grid-points define a feature, is a 12 nm eddy relevant to ASW? Can we actually get the high-res physics right or is parameterization all that really works now? Same question for time Does a 1-hour time step mean anything outside strongly tidal regions? Balance science with capabilities A 2X increase in 3D resolution means 8X the storage space and 10X the computational requirements Can we move the additional data around, store it, will we use it again? NAVOCEANO HYCOM Presentation 24APR07 Slide 6 Assessing Model Skill is Important
Need to be able to show how the model is degraded over scales of time and space How much better than MODAS is GNCOM? Real-time data assimilation expected Adjust model start field to reality of observations Provides indicators of uncertainty Ensemble approach perturbations indicate variability Correlation length scales How to observe and set model resolution Required by Navy to determine Fleet risks for ocean-related actions NAVOCEANO HYCOM Presentation 24APR07 Bob Helber, Charlie Barron personal communication Slide 7 2. HYCOM INSTALLATION AT NAVOCEANO Fit into NAVO Infrastructure Use available data sources Constrained by MSRC computational power
Comply with NAVO IT support Deal with NAVO/MSRC I/O constraints Web-based deliveries Fleet constraints and formats NAVOCEANO HYCOM Presentation 24APR07 Slide 8 NAVOCEANO Models Data-Flow Wiring Diagram (Simplified Version) DATA COLLECTION Data Sources External Flow RTDHS SATELLITE DATA PROCESSING SST Altimetry Color ADFC MSRC UNCLASSIFIED MSRC UNCLASSIFIED OCEAN MODELS OCEAN MODELS NLOM NLOM NCOM NCOM
1,130 7.6 6.8 7.6 CPU-hours AVBL CNMOC AVBL CNMOC CLAS 22,131 19,258 13,619 3,320 2,889 2,043 69,888 67,968 43,008 10,483 10,195 6,451 U U C 55,008
8,251 180,864 27,130 PROJECTED NAVOCEANO MSRC OCEAN MODEL REQUIREMENTS Estimated FY07 requirements 7,095 gigaflops or ~81% of projected capacity 14000 12000 INFRASTRUCTURE 10000 ASSIMILATION * A gigaflop is defined as a billion (10^9) Floating Point Operations. This is calculated by multiplying the speed of a processor (CPU) times the number of CPUs used, times the wall clock time in seconds, to determine model cycles required. This is multiplied by 4 flops/cycle. GFLOPS COAMPS WAVES & SURF 8000 ICE
TIDAL 6000 COASTAL REGIONAL GLOBAL 4000 MODAS 10% OF MSRC CAPACITY 2000 0 2005 2006 2007 2008 2009 2010 2011 CALENDAR YEAR / BY QUARTER NAVOCEANO HYCOM Presentation 24APR07 Slide 10 We Must be Mindful of Security Issues Data flow constraints
Navy access requirements Most fleet support is classified Totally open code NAVOCEANO HYCOM Presentation 24APR07 Slide 11 Strive for Standardized Input/Output MSRC upgrades every 2 years (IBM now, LINUX next?) Use ESMF Automatic jobstream runs Ocean model operators not 24/7 Operator intervention by exception Log events to ROAMER Inform operators / less experienced of progress /problems Maintain consistent look and feel Archive and deliver in a standardized format NetCDF 5D (properties, x-y-z, t) Regular lat-lon grid Properties T-S-U-V-E sound speed, density, etc. Save in tiles / blocks
Conduct model assessments Model to model comparisons Consistent model-observation comparisons NAVOCEANO HYCOM Presentation 24APR07 Slide 12 We Need to Meet a Navy Schedule Navy cycle at sea At 0800, the admirals brief At 2200, the plan for tomorrow goes out Daily hindcast a reanalyses Pick up SST/SSH analyses Assimilate the latest observations Revise start fields Forecast has to be available to meet the morning schedule Compress run times May have to cut some cool stuff 2-day, 3-day, 7-day products SWAFS Mediterranean Sea NAVOCEANO HYCOM Presentation 24APR07 Slide 13 Resources Are Always a Consideration
Limited MSRC cycles Competition from FNMOC, acoustics Delivery pipe communications can be very restricted Via WEB Old days JPEG graphics (150K/image) Now data for onboard computers Looking for standardized compression Public delivery via NOAA Coastal Data Development Center (CDDC) NAVOCEANO HYCOM Presentation 24APR07 Slide 14 3. TRANSITION SCIENCE TO OPERATIONS NRL Research does the science Peer-reviewed papers Validation test report (VTR - Navy / NOAA procedures) Milestone I, move from basic research (6.1 / 6.2) to 6.4 Install at NAVO with NRL running model
NAVO provides operational cycles NAVO ocean model operators are trained Fit into existing infrastructure Establish feedbacks for improvements, corrections Prepare users manuals, design documents Move to meet schedule, output, skill requirements Milestone II demonstrate that model can run operationally NAVOCEANO HYCOM Presentation 24APR07 Slide 15 The Transition Process - 2 NAVO takes over Becomes part of the operations job stream and schedule NAVO personnel troubleshoot, fix run problems Conduct an OPEVAL Demonstrate meets Navy requirements Determine strengths and weaknesses Milestone III declare model operational Continuous feedback of issues to NRL Example MLD problems with NCOM Warranty tail Use experience, emerging requirements for next cycle NAVOCEANO HYCOM Presentation 24APR07 Slide 16 In Summary
NAVO models are developed to meet Navy requirements ASW sound speed forecasts The operating system is constrained Available resources Timetables Security issues Communications and delivery There is a transition process Tested and declared operational for specific applications A series of AMOP* milestones are to be met * Administrative Modeling Oversight Panel NAVOCEANO HYCOM Presentation 24APR07 Slide 17
Designing Virtual Organizations for Citizen Science Wiggins & K. Crowston IFIP 8.2 OASIS workshop 15 December, 2009 Motivation Problem: The Tyranny of Scale in scientific research Geographic scale Temporal scale Data scale Solution: Redesigning scientific work eScience: leverages ICTs to...
Hazard Classification. Each physical or health hazard is a "hazard class" (e.g., Carcinogenicity is a hazard class). A "hazard class" may be sub-divided in the criteria into several "hazard categories" based on the degree of severity of the hazard.
Unit 4: Lesson 1. INTRODUCTION TO THE UNIT OF STUDY. In Unit 4 we are going to learn how to Make Every Contact Count by learning and practicing the skills which are needed to provide opportunistic brief advice and to...
Fixed Prosthodontics Gold crowns-full and partial veneer Inlays-gold and all ceramic Porcelain-all porcelain and PFM Fixed partial dentures Replace one or more missing teeth Restorative Dentistry Restorative dentistry covers the range from simple direct restorations to complex full mouth ...
Challenge is to spend time with each other when it suits and when it doesn't. Do you spend time with your siblings? ... Try to be the big sister for her little sister (the one she always wanted for herself)....
Best Practices for Descriptive Metadata for Web Archiving. Recommendations of the OCLC Research Library Partnership . Web Archiving Metadata Working Group . Alexis Antracoli, Kate Bowers, Jackie Dooley. oc.lc / wam
CS 447 Networks and Data Communication B Results of Midterm Exam (as of 4:15 PM, October 22, 2019) Average: 71.6 Highest: 100.0 Grades (in 100-point scale) These results do NOT include the extra credit from the programming project These results...
Ready to download the document? Go ahead and hit continue!