Title: The LAPS hot start Initializing mesoscale forecast models with active cloud and precipitation proces
1The LAPS hot startInitializing mesoscale
forecast models with active cloud and
precipitation processes
- Paul Schultz
- NOAA Forecast Systems Laboratory
- Local Analysis and Prediction Branch
2Developers of the LAPS hot-start techniques
- John McGinley, branch chief, variational methods
- Paul Schultz, project manager, modeler
- Brent Shaw, modeler
- Steve Albers, cloud analysis, temp/wind analysis
- Dan Birkenheuer, humidity analysis
- John Smart, everything
3Goals
- Address NWP spin up problem
- Explicit short-range (0-6 h) QPFs and cloud
forecasts - Focus on a local modeling capability
- Must be computationally inexpensive
- Exploit all locally-available meteorological data
- High-resolution grids
- Robust data ingest, QC, and fusion
- Develop a flexible solution for easy technology
transfer - Hardware/OS independence
- Choice of mesoscale model
4LAPS II Three-Dimensional Cloud Analysis
METAR
METAR
METAR
5Cloud typing
Cumulus vertical motions
6LAPS II Dynamic Balance Adjustment
( ) b are background quantities () are
solution increments from background ( ) are
observation differences from background
7LAPS II Dynamic Balance Adjustment
FH FL
8Example first forecast hour, 5-min frames
hot start loop
9First real-time implementation
- Real-time diabatically initialized MM5 runs since
Fall 2000 - MM5v3-4 with minor modifications
- 125 x 105 x 34 domain, 10-km grid spacing
- K-F cumulus parameterization
- Schultz explicit microphysics
- Four 24-h forecasts per day with hourly output
- Displayed on AWIPS (FSL and BOU NWS) and WWW
- http//laps.fsl.noaa.gov/cgi/laps_fcst_products.cg
i - Typically available 1.5 h after cycle time
10Quantitative Assessment
- Comparison of parallel model runs using three
kinds of initialization (hot, warm, cold)
otherwise identical - Objective verification of model performance using
hot start vs. other initialization methods - Approximately 40 forecast cycles during Jan 2001
- Gridded comparisons using LAPS analysis as truth
- Computed various threat scores, RMSE, etc.
11Model Initialization Comparisons
Time-n Time
Cold start
MM5 Forecast
no LAPS analysis interpolate from larger-scale
model
Eta
Warm start
LAPS Analyses
pre-forecast nudging to a series of LAPS
analyses sometimes called dynamic initialization
MM5 Nudging
MM5 Forecast
Hot start
MM5 Forecast
diabatic initialization using the balanced LAPS
analysis
LAPS II
Dynamically balanced, Cloud-consistent LAPS
Eta LBC for all runs
12Results of Initialization Comparisons
13Results of Initialization Comparison
14Summary of first results
- Software is very reliable
- Hot start works well for winter precipitation
systems - Several forecast-hours of added value
- What about summertime?
- Verification not complete
- Encouraging cases
- Some known problems
15Example 21 June 2001/0600 UTC Run
MM5 00 hr Forecast, Valid 21/0600 UTC
GOES IRNOWRAD, Valid 21/0600 UTC
16Example 21 June 2001/0600 UTC Run
MM5 01 hr Forecast, Valid 21/0700 UTC
GOES IRNOWRAD, Valid 21/0700 UTC
17Example 21 June 2001/0600 UTC Run
MM5 02 hr Forecast, Valid 21/0800 UTC
GOES IRNOWRAD, Valid 21/0800 UTC
18Example 21 June 2001/0600 UTC Run
MM5 03 hr Forecast, Valid 21/0900 UTC
GOES IRNOWRAD, Valid 21/0900 UTC
19Example 21 June 2001/0600 UTC Run
MM5 04 hr Forecast, Valid 21/1000 UTC
GOES IRNOWRAD, Valid 21/1000 UTC
20Example 21 June 2001/0600 UTC Run
MM5 05 hr Forecast, Valid 21/1100 UTC
GOES IRNOWRAD, Valid 21/1100 UTC
21Convection -- the next problem
- 10-km grid is too coarse except for biggest
storms - too fine for conventional convective
parameterizations - storms start late, produce excessive precip
- subgrid-size cumuli alias to resolvable scale
- 4 km? 3 km?
- Improvements will come, QPF bias will persist
- Parameterization for subgrid cumuli probably
required down to 250 m - Real life get a computer, configure the model,
solve problems as they arise
22Intriguing aspects of the D/FW TMU project
- Texas mesonet
- surface obs in every country
- wind profilers
- CRAFT project
- real-time access to full-volume, full-resolution
WSR-88D radars - Internet II (Abilene)
- Includes Lubbock, D/FW, but nothing south of that
yet - Local or remote implementation?
23Related Presentations at NWP conference
- Dynamic Balance
- WAF/NWP P1.11, McGinley and Smart
- Cloud Analysis
- WAF/NWP JP2.5A, Schultz and Albers
- Operational Evaluation/Case Studies
- WAF/NWP 3.9, Shaw et al.
- WAF/NWP P2.3, Birkenheuer et al.
- MP P4.21, Szoke and Shaw
Available in pdf format from laps.fsl.noaa.gov