OSSEs for Pacific Predictability Josh Hacker, NCAR

1
OSSEs for Pacific PredictabilityJosh Hacker,
NCAR

• Contributors J. Anderson, R. Atlas, S. Benjamin,
  D. Emmitt, R. Frehlich, G. Hakim, J. Hansen, T.
  Hamill, R. Morss, C. Snyder, J. Whitaker

2
Large shops and small shops

• Some questions clearly require a (nearly)
  complete operational data stream and
  state-of-the-art assimilation system
• Some questions can be addressed independently
• Is there room for both?

3
Operational questions

• Value of existing and proposed observations to
  analysis and forecast skill
• Impact of observations in the context of all
  other observations
• Cost associated with ingestion, QC, and
  assimilation of an additional observation

4
Operational OSSE themes

• Must be carefully designed
• Multiple models (a credibility issue)
• Total observation error is difficult to estimate
• Interpretation can be done collaboratively

5
Example Observation error

• Rawinsonde in center of grid cell
• Large variations in sampling error
• Dominant component of total observation error in
  high turbulence regions
• Very accurate observations in low turbulence
  regions

Courtesy R. Frehlich
6
Small shops as an interpreter

• Only need to deal with output and know experiment
  details
• Interpretation of large-shop OSSEs
• Eliminating complications and using simpler
  models to aid interpretation

7
Hierarchical OSSEs

• Simpler models can be a useful complement to
  larger, more complex systems
• More accessible to smaller shops
• If questions are posed carefully, many results
  can be extrapolated to full systems
• Can rule out observations with simpler models
  (caveats)

8
OSSEs to develop paradigms

• Data assimilation methodology
• Approaches to understanding model error
• Approaches to understanding model phenomenology
• A (near) perfect-model OSSE makes these things
  far more tractable and serve as a test-bed

9
Examples from A Community White Paper

• State estimation adaptive observing strategies
  for different forecast objectives
• Model error proposing frameworks for quantifying
  it
• Error dynamics understanding the interaction
  between observation networks and phenomenological
  error growth
• Observing network design basic information
  content of classes of observations in the context
  of different DA systems

10
State estimation adaptive observing strategies
for different forecast objectives
Rocket Buoy System
COSMIC
Aerosonde
11
Observing network design
sample case 500 hPa geopotential
5500 m contour is thickened Black dots show
pressure ob locations
Full CDAS (120,000 obs)
EnSRF 1895 (214 surface pressure obs)
RMS 39.8 m
Optimal Interpolation 1895 (214 surface pressure
obs)
RMS 82.4 m
Courtesy Hamill/Whitaker
12
An incomplete laundry list

• Projection of observations on gravity or spurious
  modes
• Testing a variety of (new?) metrics
• Observations to impact societal benefit
• Disparate and similar observing and model scales
• Understanding scale interactions in models