A FourSeason Impact Study of Rawinsonde, GOES, and POES Data in the Eta Data Assimilation System. Pa

1
A Four-Season Impact Study of Rawinsonde, GOES,
and POES Data in the Eta Data Assimilation
System. Part I The Total Contribution

• Authors Tom Zapotocny, W. Paul Menzel, James
  Jung and James Nelson III

2
Motivation

• To assess the impact of rawinsonde, GOES and POES
  data on Eta forecasts during each season.
• To examine the importance of GOES versus POES
  data.

3
Background for this Paper

• An earlier paper by Zapotocny et al. looked at
  the sensitivity of all 34 data types used in the
  Eta Data Assimilation/Forecast System (EDAS) for
  a single winter case study.
• A follow up to that paper then examined the
  impact of a 10 data type subset to 11-day periods
  in 3 of the 4 seasons.

4
Study Design

• Eta model run for 15-day period during each of
  the four seasons, 2 times per day
• Oct. 25 Nov. 8, 2001
• Jan. 16 30, 2002
• April 12 26, 2002
• June 24 July 8, 2002
• Temperature, u and v wind direction, and relative
  humidity are examined at the 7 mandatory levels
• Mean SLP is also examined

5
Study Design cont.

• Control run uses all EDAS inputs
• Experimental runs deny data from either
  rawinsonde (raobs), GOES or POES
• Forecasts are examined per run out to the 48-h
  forecast

6
GOES Denials

• Wind
• Marine Infrared Cloud Drift Winds
• Marine Cloud-Top Water Vapor Winds
• Cloud Picture triplet winds
• Land clear-air 3-layer precipitable water
• Marine clear-air radiances

7
POES Denial

• Special Sensor Microwave Imager data
• Column total marine precipitable water
• Low level marine superobed winds
• High Resolution Infrared Radiation Sounder (HIRS)
  radiance
• Microwave Sounding Unit (MSU) Data
• Advanced MSU (AMSU) temperature and moisture

8
RAOB Denial

• Mass and Wind data is denied
• Mass constitutes temperature, moisture and
  pressure/height data

9
3 Dimensional Variable Analysis

• X- resultant series of analysis values
• Xb- Background value series
• B-background error covariance matrix
• R - observational error covariance matrix

• Yo - Observation series
• H - forward model
• Hbal - Forward model for the balance relationship
• Bbal - magnitude of error assigned to the balance
  relationship

Zapotocny et al. 2000
10
3D-VAR Sensitivity

• Small number of high quality targeted obs can
  have more influence than high number of error
  prone obs in a small forecast error region
• Accuracy of the forward model, H
• Accuracy of the model forecast used as the
  background

11
Available Data at 3 hr Increments
12
Available vs Usable Observations
13
Root Mean Square Forecast Impact
14
Time Averaged Forecast Impact
15
Time Averaged RMS FI at 24 hrs
16
Time Averaged RMS FI at 48 hrs
17
Fall Time Averaged FI after 24 hrs
18
Fall Time Averaged FI after 24 hrs
19
Winter Time Averaged FI after 24 hrs
20
Winter Time Averaged FI after 24 hrs
21
Spring Time Averaged FI after 24 hrs
22
Spring Time Averaged FI after 24 hrs
23
Summer Time Averaged FI after 24 hrs
24
Summer Time Averaged FI after 24 hrs
25
4 Season Summary Time Averaged FI after 24 hrs
26
4 Season Summary Time Averaged FI after 24 hrs
27
Forecast Impact of all 120 runs by data type
28
Median Distribution and 95 Confidence Limits for
Data Component by Pressure Level
29
Fall Time Averaged FI after 48 hrs
30
Fall Time Averaged FI after 48 hrs
31
Fall FI for 24 hr forecast from 0 Z to 12 Z
32
Fall FI for 24 hr forecast from 0 Z to 12 Z
33
Spring FI for 24 hr forecast from 0 Z to 12 Z
34
Spring FI for 24 hr forecast from 0 Z to 12 Z
35
Conclusions

• RAOBS contribute the most to positive FI for mean
  SLP and upper air temps
• GOES and ROAB have similar success on FI for
  upper-air winds

• RH is about equal for all 3 data types
• 24 hr FI is better than 48 hr FI for all seasons