Tanya L. Otte and Robert C. Gilliam

1
A Nudging Strategy for Mesobeta-Scale WRF
Simulations Suitable for Retrospective Air
Quality ModelingPreliminary Results

• Tanya L. Otte and Robert C. Gilliam
• NOAA Air Resources Laboratory, Research Triangle
  Park, NC (In partnership with U.S. EPA National
  Exposure Research Laboratory)
• 6th Annual CMAS Conference
• Chapel Hill, NC
• 2 October 2007

2
What is Nudging?

• Formally, Newtonian relaxation
• Method of dynamically relaxing model toward
  observed state
• Includes non-physical forcing term in prognostic
  equations
• Uses difference between model and best estimate
  of observation in space and time
• Used throughout retrospective simulations (i.e.,
  a dynamic analysis) to keep meteorology as
  close to observed as possible
• Extends run length of usable meteorology for AQ
  modeling
• One method of FDDA in MM5 and WRF
• Typically applied to
• Horizontal wind components (U and V)
• Temperature (T) above the PBL
• Water vapor mixing ratio (Q) above the PBL

3
Motivation

• Current nudging strategies for MM5 and WRF
  typically based on four papers
• Stauffer and Seaman, MWR, 1990
• Initial MM5 nudging paper both analysis and obs
  nudging
• Stauffer, Seaman, and Binkowski, MWR, 1991
• Nudging in PBLOK for wind, restrict for mass and
  moisture
• Stauffer and Seaman, JAM, 1994
• Multiscale approach to phase from analysis to obs
  nudgingas horizontal grid spacing decreases
• Seaman, Stauffer, and Lario-Gibbs, JAM, 1995
• Define nudging coefficients for multiscale
  approach
• Use 36/12/4 km nesting for air quality modeling

4
Motivation (continued)

• Seaman et al. (1995) used input analyses derived
  from 12-h, 2.5-deg 3D analyses and 3-h, 2.5-deg
  surface analyses
• Analysis to observations done in MM5 with
  RAWINS toward conventional surface observations
  and rawinsondes
• Current archived analyses are often as fine as
  3-h, 12-km (e.g., NAM218) for 3D and surface
• Analyses include combination of conventional and
  remote-sensed observations
• Can use RAWINS with MM5 only 3DVar in WRF (for
  now)
• Is nudging strategy (coefficients and multiscale
  approach) in Seaman et al. (1995) too restrictive
  given todays data availability?

5
WRFv2.2 Setup
Input Analyses NARR (32-km) grid, 3-hourly
Explicit Microphysics WSM 6
Convective Parameterization Kain-Fritsch
PBL Model ACM2
Land-Surface Model Pleim-Xiu LSM
Radiation RRTM LW Dudhia SW
Not in released code to be included in WRFv3
6
WRF Domain
12 km, 290 x 251 x 34 layers
7
Sensitivity Overview
12 UTC 4 Aug 00 UTC 25 Aug 2006 Four 5.5-day
overlapping run segments

• Low Typical nudging coefficients for 12-km
• 1.0 x 10-4 s-1 for U,V,T 1.0 x 10-5 s-1 for Q
• From Seaman et al., JAM, 1995 for analysis
  nudging with obs nudging _at_12-km
• Std Typical nudging coefficients for 36-km
• 3.0 x 10-4 s-1 for U,V,T,Q
• 1-h e-folding time for physical processes
• High Nudging coefficients with half e-folding
  time of Std
• 5.5 x 10-4 s-1 for U,V,T,Q
• 0.5-h e-folding time for physical processes
• StdPBL Same as Std, but with nudging toward
  temperature and moisture in PBL
• HighPBL Same as High, but with nudging toward
  temperature and moisture in PBL

8
Preliminary Analysis

• Show mean error (bias) from 20-day runs
• Keep it simple, for now
• MAE and RMSE do not change bottom line
• Verify against 700 NWS surface stations
• Includes T, Q, wind, surface pressure
• Included in analyses (i.e., nudged)
• Includes urban, suburban, and rural sites
• Verify against 67 CASTNET observations
• Includes T, RH, wind, SW radiation
• Independent of analyses (i.e., not considered in
  nudging)
• Largely rural sites

9
2-m Temperature vs. CASTNET
Daily
By Day in WRF Run
10
2-m Dew Point vs. CASTNET
Daily
By Day in WRF Run
11
SW Radiation vs. CASTNET
Daily
By Day in WRF Run
12
10-m Wind Speed vs. CASTNET
Daily
By Day in WRF Run
13
Diurnal 2-m Temperature
Nudged
Not Nudged
14
Diurnal 2-m Mixing Ratio
Nudged
Not Nudged
2-m Q for CASTNET not included because CASTNET
does not report surface pressure, which is
required to convert RH to Q.
15
Diurnal 10-m Wind Speed
Nudged
Not Nudged
16
Diurnal 10-m Wind Direction
Nudged
Not Nudged
17
Summary

• Preliminary results suggest
• Nudging coefficients for wind, temperature, and
  moisture can be increased over Seaman et al.
  (1995), i.e., Low values, if not obs nudging
• Stronger nudging, i.e., High values, reduces
  bias in 10-m wind speed, but has little impact on
  2-m temperature and dew point.
• Nudging toward moisture may still need to be
  weaker than towards temperature and wind.
• Nudging toward temperature and moisture in the
  PBL increases the bias for 2-m temperature, 2-m
  dew point, 10-m wind speed, and shortwave
  radiation at CASTNET sites (not nudged).
• As expected, statistics vs. NWS observations are
  better than statistics vs. CASTNET sites, which
  are independent of the analyses.
• Model behavior with and without nudging in the
  PBL is vastly different during stable (nighttime)
  regime than convective (daytime) regime

18
Next Steps

• Further evaluation with current runs
• Evaluation against upper-air observations, PBL
  heights, precipitation, etc.
• Additional methods
• Repeat sensitivities with NAM-218 (12-km) input
• Vary nudging strength by variable
• Use observation nudging with analysis nudging
• Test with analysis (e.g., RAWINS) in WRF

19
Looking Ahead2-m Temp. using NARR and NAM218 vs.
CASTNET
Uses 32-km NARR analyses initialized 12 UTC 4 Aug
MAE and RMSE K
Uses 12-km NAM218 analyses initialized 00 UTC 20
Jul
20
Acknowledgments

• Jonathan Pleim (NOAA)
• Lara Reynolds (CSC)

Disclaimer The research presented here was
performed under the Memorandum of Understanding
between the U.S. Environmental Protection Agency
(EPA) and the U.S. Department of Commerce's
National Oceanic and Atmospheric Administration
(NOAA) and under agreement number DW13921548.
This work constitutes a contribution to the NOAA
Air Quality Program. Although it has been
reviewed by EPA and NOAA and approved for
publication, it does not necessarily reflect
their policies or views.