Robert Dumais

1
HIGH RESOLUTION METEOROLOGY FOR THE U.S. ARMY
Robert Dumais James Cogan
U.S. Army Research Laboratory
Computational and Information Systems
Directorate Boundary Layer
Meteorology Branch AHPCRC
Workshop on Mesoscale and CFD Modeling
for Military Applications Jackson State
University

May 25, 2004
2
MISSION
To conduct scientific research to characterize
Army-scale boundary layer atmospheres using
observation-based numerical modeling, and to
apply and evaluate meteorological models and
observational methods for planning and operations
for the net-centric Army of the Future Forces
era.
3
MISSION
TECHNOLOGY FOR REAL-TIME OPERATIONS
(1) THAT CAN BE EXECUTED DIRECTLY ON LOWER
ECHELON TACTICAL SYSTEMS TO PRODUCE ENHANCED
METEOROLOGICAL GUIDANCE (SOME COMPROMISE BETWEEN
PHYSICAL ACCURACY COMPUTATIONAL EFFICIENCY)
(2) THAT CAN LEVERAGE HPC VIA
REACH-BACK MODE TO PROVIDE UNPRECEDENTED HIGH
RESOLUTION DETAIL AND ACCURACY FOR SPECIFIC
TACTICAL ENVIRONMENTAL PROBLEMS
4
Boundary Layer Meteorology for the Army

• GOAL
• - Provide the capability to describe and predict
  the atmosphere at meso-gamma and micro scales in
  the boundary layer for use in applications for
  the Future Forces and Future Combat System.
• BARRIERS
• - Inadequate ability to describe and predict
  boundary layer phenomena at smaller scales in
  close to real time, especially in urban and
  complex terrain.
• - Lack of real time or near real time method to
  properly assimilate data for model initialization
  and post-processing at smaller scales (for
  example, lack of proper 3DVAR constraint criteria
  at mesogamma), especially for non-traditional and
  non-met data. Also, improved physical
  turbulence parameterizations needed for sub-km.
• APPROACH
• Combine models at different scales that allow
  best of predictive (mesoscale) and diagnostic
  (microscale) methods.
• Combine small scale model output and
  observations.
• - Cooperative research with other agencies and
  academia in areas such as data assimilation
  LES/CFD modeling (NCAR, PSU, ASU, CSU,
  Alaska-Fairbanks, Clark-Atlanta, UND, OSU, etc.)
  
• - In-house development to emphasize Army specific
  areas not adequately addressed elsewhere.

5
HIGH RESOLUTION METEOROLOGY FOR THE ARMY
WHY IS IT IMPORTANT?

• ? Supporting Future Forces/Future Combat System
  requirements for faster, smaller and lighter Army
  of the future.
• Accurate, timely weather information at high
  resolutions for mission execution and planning
  developing the Weather Running Estimate.
• Using battlefield weather to see first,
  understand first.
• Supporting small unit tasks and purposes, and
  synchronizing assaults into a decisive end-state.
• Rapidly receiving and disseminating local
  weather information to assess rapidly changing
  battlefield environmental conditions.
• Real-time warning and dissemination to protect
  the forces against Chem/Bio/Nuclear/Radiological
  (CBNR) hazards.
• Collaborative and decisive decision aids use as
  a combat multiplier.
• Optimize use of battlefield electro-optical and
  acoustic systems, and allow for accurate
  deployment of wide area, long duration obscurants.

6
Small Scale Met Modeling and Nowcasting
Small Scale for Models refers to those models
having a horizontal grid spacing running from
about 2.5 km down to about 10 m. The smallest
scale models (i.e., a few hundred meters or less)
frequently have a vertical extent from the
surface layer up through all or part of the
boundary layer. Nowcasting (Weather Running
Estimate) may be defined as analyses and very
short term forecasts of combine model output with observations in a
timely manner (for IMETS, MMS-P, Future AF/ARMY
JET, DCGS-A)
7
Some Components of Small Scale Met Modeling

• MULTISCALE MODELING STRATEGIES FOR MESO to MICRO
  SCALES IN THE BOUNDARY LAYER
  -
  Non-hydrostatic Numerical Weather Prediction
  (NWP) modeling
  - Diagnostic
  microscale models for the Planetary Boundary
  Layer (PBL)
  
• - Coupling of NWP diagnostic models
  
  - Ensemble modeling of mesoscale NWP models
• - Focus on smaller areas
  meso/microscale resolutions in space and time
• - Evaluate and determine how to include
  and portray uncertainty in met products
  
• - Use to drive transport in dispersion
  models, and for signature propagation models.

2. HIGH RESOLUTION DATA FUSION AND DATA
ASSIMILATION - Fuse local observational
data with NWP background fields to
generate best high resolution picture over small
areas of interest - Analyses generated
using established methods (e.g., successive
corrections) and developing methods such as
those of Askelson (UND) - New sources of
non-conventional meteorological and surface data
may be available (satellite, deployable
met towers, profilers, UAVs, UGVs,
mini-lidars, etc.) - Work with external
researchers to develop investigate state of the
art methods (e.g., 3DVAR/4DVAR Ensemble
KF) for future Army application
8
MM5 as NWP component of multi-scale model system
9
INITIAL REMOTE SENSING INGEST
NMSU CARSAME
LAND/SEA SURFACE TEMPERATURE (7 DAY COMPOSITES)
FROM NOAA 16 PASSES
2 PASSES PER DAY/ ABOUT 1 KM RESOLUTION/ NO CLOUD
PIXELS
WORK HAS INITIATED TO PUT DATA INTO GRIB FORMAT,
FOR INPUT INTO MM5 PREGRID MODULE
10
3DWF Simulation of OKC Wind (z 9m)
dxdy10.66m, dz3m Grid points 129x129x129
Bank One Tower
Initialized with lidar VAD profile and sonic
anemometer data (1530 UTC 9 July 2003)
11
Development of ARL DAT

• Left Panel - View/Display
• GIS View
• Source Site Selection
• Deposition Field Display
• Right Panel Control Widgets
• Map Scale
• Source Information
• Interactive Map Location
• Time of Deposition Field Overlay
• Current Time
• Data Status
• Current Weather Status

12
Value-Added Studies Forecast Resolution
Leverage AHPCRC Resources for enhancing NWP
capability for this project
13
HPC MICROSCALE MODELING OF COMPLEX TERRAINS
Objective Leveraging heavily off of ARL HPC
and internal ABLE facilities, develop distributed
memory numerical modeling strategies for complex
terrain environments, focusing on sub-km scale
resolutions. Participate in variational and/or
ensemble model data assimilation research efforts
(such as at CG/AR and AHPCRC) that can be used to
provide superior mesomicro short range NWP
forecasts and boundary conditions for urban-scale
models, in addition to allowing for a way of
objectively determining uncertainties and optimal
meteorological and/or chemical-biological sensor
placement on the battlefield.
14
UW-NMS ON ARL HPC (IBM SP)

• Domain Afghanistan (500 x 384 _at_ 3km horizontal
  resolution, 35 vertical levels)
• Depicted
• Freezing Level (Blue)
• Cloud Liquid Water Content (Yellow)
• Surface Winds (Light Blue) Note not full
  resolution

Most of the optimization was accomplished through
loop unrolling and parallelizing the independent
loops. Some things were done for optimizing the
microphysics such as changing the array sizes
15
FUTURE EFFORT WITH CG/AR Ensemble Data
Assimilation

• Develop and Assess Capability of the Maximum
  Likelihood Ensemble Filter (Zupanski, 2003) Data
  Assimilation approach on the meso-micro scales
• suited for highly nonlinear operators
• - no adjoints
• - no assumptions about
  observation correlations
• - perturb analysis rather
  than individual obs
• - optimize non-quadratic
  cost function, using a
• superior Hessian
  preconditioner
• - allows for both an
  estimate of uncertainty and for
• the most likely
  deterministic solution

16
Conclusion
Branch work already has had and is having an
impact on Army systems (e.g., MMS-P,
IMETS). Cooperative research important part of
branch technical program (e.g., CSU, ASU, UND,
UTEP). Looking to advance technology via in-house
and external cooperative efforts (e.g., WRF
committees, university seminars, visits to/from
universities and other organizations). HPC
resources have application to much of the work of
the branch. Not really a conclusion, but a
continuation.