The Sensitivity of a Real-Time Four-Dimensional Data Assimilation Procedure to Weather Research and Forecast Model Simulations: A Case Study

1
The Sensitivity of a Real-Time Four-Dimensional
Data Assimilation Procedure to Weather Research
and Forecast Model Simulations A Case Study

• Hsiao-ming Hsu and Yubao Liu
• NCAR/RAP

2

• Motivations

• WRF real-case initialization schemes
• SI interpolation from other models
• 3DVAR hopeful, but 3D, simplified balance
• 4DVAR bright future
• Hereby, we look into a method to initialize WRF
  with a four-dimensional dynamically and
  physically consistent analysis, which
  incorporates all available synoptic and asynoptic
  observations.
• NCAR/ATEC MM5-based RT-FDDA system provides this
  kind of analysis to initialize WRF forecast.

3
NCAR/ATEC RTFDDA

• Built around MM5 (Jennifer et al. 2001, Liu et
  al. 2002)
• Continuous observation nudging (Stauffer and
  Seaman 1994)
• Multi-grids (1 km fine meshes)
• 3 hourly-cycling
• Operated at 5 ATEC ranges and support several
  special tasks (CO-fire, Olympics)

Coldstart
t
FDDA
Forecasts
Day N
Day 0
4
(No Transcript)
5
CASE
During Year-2002 Winter Olympics at SLC, RTFDDA
was operational for 2 months. There was a snow
storm event during March 13. A pair of contrast
experiments of 12-hr WRF forecasts with different
initial conditions were conducted, started at
00Z, March 13.
6
Experiment Design

• EXP1- Cold start WRF
• WRF initial condition was generated by
  re-analysis of ETA forecast with available
  observations at 00Z, March 13.
• EXP2 - Warm start WRF
• WRF initial condition was obtained from the
  RTFDDA analysis which had been running
  continuously from a cold start 84 hours ago.
• EXP3 - Warm start MM5
• Same as EXP2, but with MM5
• (from op-RTFDDA).

7
Domain configuration

• 82 x 70
• dx 36 km
• 36 layers
• 12 levels in 1 km AGL
• Coarse mesh only

8
Hourly Precipitation of 1 12 Forecasts
Cold start
Warm start
9
Subjective verification of 1 hour precipitation
at 3-h forecast
A
A
C
C
B
B
Warm start
Cold start
C
A
B
C
B
OBS IR
OBS Radar
10
(No Transcript)
11
WRF Forecast at 03Z
RTFDDA forecast at 03Z
12
Summary

• Significant differences were observed between the
  cold start and the warm start WRF forecasts.
• The warm-start WRF run compares more favorable
  to observations.
• The Warm start WRF results are very similar to
  those from RTFDDA (MM5) during the first few
  fours of forecasts.
• It is evident that reasonable benefit of reduced
  dynamical and cloud/precipitation spin-up
  during first few hours can be obtained by
  interfacing MM5 RTFDDA process to WRF
  initialization.

13
Future Work

• Comparison study on higher resolutions and severe
  weather cases.
• Ingesting RTFDDA cloud/precipitation analyses
  into warm start WRF.
• Implement warm start WRF in the same
  operational environment of RTFDDA MM5
• Quantitative verification of warm-start WRF
  against various observations for a longer-term
  parallel tests with MM5.