Evaluating Microphysical Parameterization Schemes for Use in Hurricane Environments

1
Evaluating Microphysical Parameterization Schemes
for Use in Hurricane Environments
Robert Rogers NOAA/AOML Hurricane Research
Division Collaborators Michael Black and Robert
Black NOAA/AOML Hurricane Research Division Shuyi
Chen Rosenstiel School of Marine and Atmospheric
Science/U. of Miami
2
Official yearly average 48-h forecast errors and
trend lines over period shown below for Atlantic
basin. (from Rappaport et al., NOAA Research
Initiative, 2003)
30
500
450
25
400
20
350
300
15
Error (kt)
250
Error (nm)
200
10
150
5
100
50
0
0
1975
1977
1979
1981
1983
1985
1987
1989
1991
1993
1995
1997
1999
2001
1961
1964
1967
1970
1973
1976
1979
1982
1985
1988
1991
1994
1997
2000
Year
Year
Track errors
Intensity errors
3
OUTLINE

1. Motivation
2. Methodology
3. Results
4. Summary and future work

4
Motivation

• Tropical cyclone intensity governed by magnitude
  and distribution of latent heat release
• Latent heat release influenced by hydrometeor
  production, conversion, fallout -- all processes
  that must be parameterized
• The development of a technique for evaluating
  models against in situ and remotely-sensed data
  in hurricane environments can point to possible
  deficiencies in parameterization schemes
• These comparisons can ultimately lead to
  improvements in the models

5
Reflectivity comparison for Floyd observations
and simulation
P-3 lower fuselage radar composite (240 x 240 km)
at 4255 m altitude between 2300 and 2330 UTC 13
September
Simulated reflectivity (250 x 250 km) at 600 hPa
at 2300 UTC 13 September
6
METHODOLOGY
7
Methodology

• Compare high-resolution simulations of tropical
  cyclones with observations from multitude of
  storms
• Stratify regions of observations and simulations
  into eyewall, rainband, and stratiform regions
  using objective technique (Rogers et al. 2004)
• Identify biases in microphysical scheme based on
  comparisons

8
Process diagram for microphysical scheme (Lin et
al. 1983)
9
Storms used in VI database
Black et al. (1994) also analyzed these data
Both NOAA WP-3D aircraft flew concurrently
10
Track and intensity plots for Floyd and Bonnie
simulations
pressure (hPa)
time after initialization (hr)
Intensity
Track
11
Evaluation techniques

• Interpolate model output, radar data to common
  cylindrical grid
• (r,?,z) dimensions of 165 x 72 x 50
• Separate datasets into eyewall, rainband, and
  stratiform regions
• Statistical comparison methodology
• Profiles
• Distributions
• Contoured frequency by Altitude Diagrams (CFADs
  Yuter and Houze (1994)
• Correlations

12
Eyewall, rainband, and stratiform regions
height (km)
radius (km)
13
Examples from Hurricane Olivia (VI) and Floyd
(MM5)
240
eyewall
rainband
eyewall
A
A
B
distance (km)
120
stratiform
B
stratiform
rainband
0
120
240
0
rainband
rainband
stratiform
stratiform
stratiform
eyewall
eyewall
B
A
B
A
P-3 Radar observations of Hurricane Olivia
MM5 simulation of Hurricane Floyd
14
Examples from Hurricane Gustav (VI) and Bonnie
(MM5)
240
stratiform
B
B
stratiform
stratiform
rainband
eyewall
A
distance (km)
eyewall
distance (km)
120
A
0
120
240
0
rainband
rainband
stratiform
eyewall
eyewall
stratiform
stratiform
stratiform
B
A
B
A
Radar observations of Hurricane Gustav
MM5 simulation of Hurricane Bonnie
15
RESULTS
16
Characteristics of regions in VI and model data
17
Profiles of mean vertical motion and reflectivity
Vertical motion
Reflectivity
18
Contoured Frequency by Altitude Diagram (CFAD)
(Yuter and Houze 1994)
19
CFADs of vertical velocity
Multi-case radar
Simulations
20
CFADs of reflectivity
Multi-case radar
Simulations
21
Scatter plots of vertical motion and ice
concentration
Bonnie flight-level observations
Bonnie simulation
22
Mean eyewall reflectivity (shaded) stratified by
vertical motion bins
height (km)
Multi-case radar
w (m/s)
Simulations
height (km)
w (m/s)
23
Tests with different schemes
24
Intensity plots for different microphysical
schemes
pressure (hPa)
time after initialization (hr)
25
CFADs for various schemes
26
SUMMARY AND FUTURE WORK
27
Summary of comparisons between simulations and
observations

• Profiles of simulated vertical motion much weaker
  and reflectivity much higher
• Reflectivity decreases much more rapidly with
  height, and over a deeper layer, above melting
  level in observations
• Distributions of w narrower, dBZ broader in model
• Modal and maximum w weaker in model, dBZ is
  higher in model
• Distribution of vertical motion narrows with
  height in upper troposphere in model, broadens
  with height in observations
• Correlation between vertical motion and
  hydrometeor mixing ratio/reflectivity much
  stronger in model

28
Possible biases in microphysical scheme

• Water loading too high in model fall speed
  errors
• Weak model w, high model dBZ
• High model correlation between w and dBZ
• Slow decrease in model dBZ above melting level
• Narrowing w distribution with height in model in
  upper-levels, broadening distribution in
  observations
• Rain and frozen particle production too strong,
  and/or conversion terms too weak
• Weak model w, slow decrease in dBZ
• High graupel values lead to high dBZ below
  melting level
• Size distributions not representative of tropical
  cyclone environments
• Large variability across systems, even within
  systems

29
Average size distributions from P-3 aircraft in
Hurricane Emily (1987) for eyewall and stratiform
regions (4 km altitude)
30
Future work

• Use evaluation techniques to test modifications
  to scheme, different schemes (e.g.,
  double-moment, spectral schemes)
• Use evaluation techniques to test different
  models (e.g., operational hurricane forecast
  models)
• Collect data and perform simulations of
  additional cases (especially weak systems) to
  broaden spectrum of storm structures, increase
  robustness of statistics
• Incorporate different data sources into
  observational database (e.g., TRMM PR
  reflectivity, NASA EDOP)

31
Comparisons of Z-M relations for (a) ice and (b)
rain from simulations and from observed hurricane
cases
32
Profiles of mean reflectivity (dBZ) for all
regions using different Z-M relations.