D' T' Wahl

1
D. T. Wahl

• Comparing NOGAPS and COAMPSTM
• with In Situ Data

2
Outline

• NWP
• Primitive Equations
• Sigma Pressure Levels
• Hydrostatic vs. Non-hydrostatic
• Satellite Data Processing
• Boundary Conditions
• Steps to obtain Numerical Forecast
• Product Distribution
• Model Output Statistics
• Model Tendencies
• Cruise Data
• Discussion
• Conclusion

3
Short Review of NWP Fundamentals
4
Primitive Equations
Courtesy of COMET http//www.meted.ucar.edu/
5
NOGAPS - Sigma Pressure Levels
Courtesy of COMET http//www.meted.ucar.edu/
6
Hydrostatic vs. Non-Hydrostatic

• Hydrostatic (NOGAPS)
• Assumes hydrostatic equilibrium
• (PGF Gravity)
• Good for global and synoptic features
• Non-Hydrostatic (COAMPS)
• Define a reference state and the perturbations
  from that state
• Good for mesoscale modeling as they account for
  vertical motion and accelerations, rather than
  inferring them from continuity.

7
NOGAPS Satellite Data Processing

• Atmospheric Variational Data Assimilation System
  (NAVDAS) for converting satellite data into model
  input.
• NAVDAS comprises the data quality control and
  analysis elements of the new NOGAPS data
  assimilation system. It is a three-dimensional
  Variational (3DVAR) analysis scheme.
• NAVDAS is a process by which satellite
  observations are converted to usable parameters
  and produces data from the surface to .1 mb.
• Within the NAVDAS both satellite and conventional
  data are further checked for quality and for
  consistency with neighboring observations and the
  model short-term forecast
• This process of quality control is referred to as
  buddy checking.

8
COAMPS Satellite Data Processing

• Multivariate Optimal Interpolation System (MVOI)
  with a six-hour data assimilation cycle.
• When quality checking data, a first guess of
  the initial conditions is taken from either a
  previous COAMPS forecast or the NOGAPS forecast
  and is interpolated onto the COAMPS grids.
• These first guess fields are then updated with
  real data. It is important to point out that
  this system of quality control depends on using
  prior models that are of good quality (MetEd,
  2006).
• Conventional data are subjected to quality
  control by the same process as NOGAPS and NOGAPS
  quality control checking system

9
Boundary Conditions

• Characteristic of all models
• Lower boundary
• Terrain height
• NOGAPS uses ½ degree Global Land One-kilometer
  Base Elevation (GLOBE)
• Ocean surface
• Land roughness
• Upper boundary
• Where to stop modeling (4 to 7 mb)
• Characteristic of regional models
• Lateral boundary conditions
• Supplied through coupling with a global or
  another regional model

10
(No Transcript)
11
Steps to Obtaining a Numerical Forecast

• Data collection
• Quality control
• Data Assimilation
• Forecast Integration
• Post-Processing of Forecast Fields
• Distribution of the Product

12
Numerical Weather Prediction Model Process
DATA COLLECTION
QUALITY CONTROL
ANALYSIS
FORECAST MODELS
POST- PROCESSING
VERIFICATION
13
Data Collection

• Observations are assembled from
• ship and buoy reports
• surface stations
• aircraft
• rawinsonde upper air soundings
• satellite derived products
• etc.
• Data cuts occur hourly -- observations from the
  previous hour are collected and prepared for
  input into the models.

14
Quality Control

• This is the step in which the validity of the
  individual observations is checked.
• Observations are screened for errors, redundancy,
  and consistency with the previous forecast.

15
Data Assimilation

• The purpose of data assimilation is to turn
  irregularly spaced observations into a regularly
  spaced grid of values from which the model can be
  run.
• The first guess is blended with incoming
  observations.

16
Data Assimilation

• Non-simple interpolation
• Use balance relationships to introduce dynamical
  consistency into the analysis.
• Scales of motion that the model cannot resolve
  are filtered out.
• Errors with each type of observation are known
  and weighted accordingly based on instrument use
  and historical accuracy.

17
Forecast Integration

• This is the step in which the model analysis is
  integrated forward in time.

18
Post-Processing

• Numerical filters and smoothers are applied to
  the raw numerical output to eliminate any high
  frequency noise in the model fields that is not
  removed by damping already in the model.
• This is where the forecasted model variables are
  interpolated from model coordinates to map
  coordinates or numerical guidance.

19
Product Distribution

• Generation of charts
• Dissemination of charts
• Output used by other models
• ocean wave models
• sea ice models
• ocean circulation models
• ocean thermodynamics models
• tropical cyclone models
• aircraft and ship-routing program

20
Model Output Statistics (MOS)

• This is a statistical approach to forecasting,
  which eliminates model biases and systematic
  errors.
• Based upon long term model performance statistics
  and climatological observations are kept (all
  stats, no physics).

21
FNMOC Model Tendencies
Models characteristics and tendencies are track
every week and as they change, they are posted
at https//www.fnmoc.navy.mil/PUBLIC/
22
NOGAPS Tendencies

• Complex lows are merged into one at the extended
  forecast periods.
• Developing oceanic lows are slightly
  under-forecast and are slow to deepen through 72
  hours.
• Mature oceanic lows are 2-3 mb over-forecast and
  are slow to fill.
• Surface winds associated with deepening (filling)
  lows are under-forecast (over-forecast).

23
COAMPS Tendencies

• Synoptically performs as well as other models on
  the mesoscale it frequently out performs other
  models.
• The strongest feature is its ability to capture
  localized winds and small scale effects.
• Does not do well over open ocean.

24
Cruise Data
25
Cruise Data Plot
26
1200 24 JAN 06
27
(No Transcript)
28
(No Transcript)
29
1800 24 JAN 06
30
(No Transcript)
31
(No Transcript)
32
0000 25 JAN 06
33
(No Transcript)
34
(No Transcript)
35
1200 25 JAN 06
36
(No Transcript)
37
(No Transcript)
38
Cruise Data Plot
39
Discussion

• Data sparse area.
• We had very little significant weather almost
  no clouds.
• Quality of satellite data is moisture (clouds)
  dependent
• (Remember back to Remote Sensing).
• 88 of observational input is derived from polar
  orbiting or geostationary satellites.
• 12 come from ship observations, dropsondes,
  pilot balloons, and rawinsondes.
• NOGAPS uses NAVDAS an improved translator of
  satellite data vice COAMPS MVOI.

40
Conclusion

• NOGAPS out performed COAMPS during our cruise
  (agrees with model tendencies).
• Both models over forecasted the winds by 5 to 20
  m/s.
• Both models forecasted the winds up 50 degree
  north that the rawinsonde indicated
• Both errors likely attributed to lack of
  atmospheric moisture (clouds).
• The trends of the model output was correct.

41
Questions?