The Antarctic Mesoscale Prediction System

1
The Antarctic Mesoscale Prediction System
David H. Bromwich, Andy Monaghan Polar
Meteorology Group Byrd Polar Research Center The
Ohio State University Columbus, Ohio, USA Jordan
G. Powers Mesoscale and Microscale Meteorology
Group National Center for Atmospheric
Research Boulder, Colorado, USA 20 September,
2006
2
AMPS Overview
3

• AMPS The Antarctic Mesoscale Prediction System
• Nested NWP system dedicated to supporting
• the operations of the U.S. Antarctic Program
• and International programs.
• Employs Polar MM5, a limited-area atmospheric
  model adapted for high-latitude applications by
  Ohio State Universitys Polar Meteorology Group
  at the Byrd Polar Research Center.
• Project began in October 2000 as a small
  collaboration between 2 U.S. institutions, and
  since has grown into an international program.
• Other nations are developing similar systems
  (Aust., Italy)

4
Current AMPS domains (60-km, 20-km, 6.7-km,
2.2-km)
60-km
6.7-km (Ross Island)
20-km
2.2-km (Ross Island)
6.7-km (South Pole)
6.7-km (Ant. Penin.)
5
60-km
20-km
60-km
6.7-km (South Pole)
6.7-km (Ant. Penin.)
6
AMPS Accomplishments
7
Rescue of the Magdalena Oldendorff in June 2002

• German supply vessel became trapped in thickening
  sea ice with 107 aboard
• South Africa sent the S.A. Agulhas to rescue the
  passengers.
• AMPS forecasts were used by the South African
  Weather Service (SAWS) for Agulhas southward
  journey
• AMPS indicated a brief window of favorable
  weather during which the passengers were
  airlifted to safety

Oldendorff and Argentinas Almirante Irizar
X
AMPS Sea Level Pressure and precipitation (green)
8
AMPS NWP for S.A. Agulhas in 2005/2006 Austral
Summer

• South Africas S.A. Agulhas makes an annual
  relief voyage to Antarctica
• AMPS provides tailored forecast products to SAWS
• 20-km domain over SANAE
• 15-km domain over ship track
• AMPS visibility and wind forecasts have proven
  especially valuable
• AMPS products are freely available to the
  international community

Materials courtesy I. Hunter, SAWS press release,
2 February 2006
9
AMPS Validation Work
10
Validation AMPS Forecasts of Clouds
LEFT AMPS 00z 13 Dec forecasted integrated
cloud liquid water (blue gray shades) and cloud
ice (white shades).
RIGHT Low resolution IR satellite composites
from the UW-AMRC archive
Materials courtesy R. Fogt and D. Bromwich, BPRC
11
Using Self Organizing Maps (SOMs) to compare the
frequency of observed and forecast circulation
features in AMPS
Right A 5x4 SOM array is employed to
characterize the sea level pressure over the Ross
Sea (inset below) into 20 distinct patterns.
Then the 060 hour forecasts are compared to the
analyses (000) to determine modeled circulation
biases in AMPS/
AMPS Frequency of Occurrence (000 060)
Materials courtesy J. Cassano, CIRES, U.
Colorado, USA
12
Analyzing Mountain Waves in the complex terrain
near McMurdo Station during intense wind storms

• Downslope windstorm on lee slopes from
  large-amplitude mountain waves
• Hydraulic jump downstream as flow adjusts to
  ambient conditions
• Incorrect flow pattern sets up a deflection zone
  over the McMurdo area
• Hydraulic jump located at interface of deflection
  zone, which is upstream of McMurdo, causing wind
  speeds to be too low there

McMurdo
0000 UTC 16 May 2004 sea-level pressure (hPa) and
surface streamlines.
Steinhoff et al. (2006)
13
Evaluating AMPS 500-hPa Performance (12-36 hr
forecasts)

• High correlations for temperature, geopotential
  height, and wind speed.
• Biases are typically small, and either /- (not
  systematic)
• Lower skill in dewpoint forecasts may be due to
  low absolute humidity at 500 hPa, which makes
  small errors look big.

Geopotential Hgt
Temperature
Dewpoint Temp
Wind Speed
Bromwich et al. (2005)
14
Evaluating the AMPS 10-km vs 3.3-km domain
(24-36 hr forecasts)
Observedthick gray 3.3-km solid black 10-km
dashed black
Ratio of 3.3-km standard deviation to 10-km
standard deviation (normalized)
Spectral Density vs frequency for wind speed at 2
AWS sites near McMurdo
Conclusion The 3.3-km domain is more sensitive
and more accurate than the 10-km domain
Bromwich et al. (2005)
15
Evaluating AMPS Performance in data-sparse
regions with COSMIC GPS radio occultation
soundings

• When GPS satellites communicate, their signals
  travel through slices of earths atmosphere.
• The atmosphere refracts the GPS signals a small
  amount depending temperature, humidity, and
  pressure.
• Thus, the refractivity residuals from these
  communications can be converted to soundings of
  temperature and humidity.
• Conversely, as shown here, modeled temperature
  and humidity fields can be converted to
  refractivity and compared to COSMIC soundings.

Location of COSMIC Soundings over Southern Ocean
from 50-60o S.
Courtesy of K. Manning, NCAR, USA
16
Evaluating AMPS Performance in data-sparse
regions with COSMIC GPS radio occultation
soundings

• When GPS satellites communicate, their signals
  travel through slices of earths atmosphere.
• The atmosphere refracts the GPS signals a small
  amount depending temperature, humidity, and
  pressure.
• Thus, the refractivity residuals from these
  communications can be converted to soundings of
  temperature and humidity.
• Conversely, as shown here, modeled temperature
  and humidity fields can be converted to
  refractivity and compared to COSMIC soundings.

Above Comparison of WRF and MM5 refractivity
statistics versus COSMIC soundings from 50-60oS
for the 60-km AMPS domain for a 22-forecast
period during austral winter, 2006. Both MM5 and
WRF are currently being run side-by-side in AMPS
in order to compare the skill of WRF versus MM5
as it undergoes development. This figure
indicates that that WRF is comparable to MM5 over
the Southern Ocean, and that the lowest skill in
both models occurs in the lower troposphere.
This suggests attention should be focused on
resolving marine stratus and the treatment of the
PBL within the models.
Location of COSMIC Soundings over Southern Ocean
from 50-60o S.
Courtesy of K. Manning, NCAR, USA
17
Ongoing AMPS Development
18
Developing Polar WRF to replace Polar MM5 in
AMPS (testing done over Greenland)
Run Correlation Bias WRF -
Swiss Camp 0.865 0.47 MM5 Swiss Camp
0.938 4.32 WRF - Summit 0.863
1.83 MM5 Summit 0.896 2.34
WRF simulations with Hall-Thompson microphysics,
YSU PBL, Noah LSM, Dudhia shortwave and RRTM
longwave radiation, and AVN-driven boundary
conditions MM5 simulations with Eta PBL, Reisner
microphysics and ECMWF-driven boundary conditions
Bromwich, Hines, and Bai, 2006
19
Variational and Ensemble Data Assimilation in AMPS

• Motivation
• Antarctica is a challenge for Data Assimilation.
• Implementation of 3D-Var in AMPS.
• Research advanced DA techniques.
• Investigate data impact (GPS, AWS, AIRS, MODIS).
• Modeling Systems
• MM5 (currently testing WRF ARW).
• WRF-Var (3D- in 2005, 4D- in 2006).
• Ensemble Square Root Filter (EnSRF).

00 UTC 15th May 2004 (/-2hrs) Thinned AIRS obs
(above), MODIS AMV obs (below)
Materials courtesy D. Barker, NCAR
20
Assimilating MODIS cloud-track winds into AMPS
ms-1
ms-1
34
34

25
25
L
L
Sfc Winds (ms-1) SLP (hPa)
L
MOD1
STD
Standard AMPS forecast
AMPS forecast with MODIS Winds

• Use of MODIS winds dramatically increases number
  of upper-air observations over an otherwise
  data-sparse continent
• Preliminary results show a statistically
  significant positive impact

Materials courtesy J. Powers, NCAR
21
MM5 4D-Var Assimilation of GPS Radio Occultation
data over Antarctica
1200 UTC 11 66H
Wee et al. (2006)
Wee et al. (2006) assimilated 50 GPS RO
soundings from CHAMP and SAC-C over the Antarctic
over a two-day period and demonstrated positive
impact on the prediction of an intense Antarctic
cyclone out to 3 days. The GPS RO soundings
account for 3 of the data used. The
recently-launched COSMIC constellation will
record many daily RO soundings over Antarctic for
years to come. Solid contour Predicted Sea Level
Pressure by MM5 using different initial
conditions. Dashed lines are verifying ECMWF
analysis.
22
AMPS forecast archive used for science
23
Climate of the McMurdo region based on 1-year of
AMPS 3.3-km forecasts

• Key Points
• First high-resolution model climatology of the
  rugged McMurdo region.
• Time-mean vortices occur in the lee of Ross
  Island, perhaps a factor in the high incidence of
  mesoscale cyclogenesis noted in this area.
• The first-ever detailed precipitation maps of the
  region.
• Orographic precipitation maxima occur on the
  southerly slopes of Ross Island and mountains to
  the southwest.
• A precipitation-shadow effect for the McMurdo Dry
  Valleys.

Annual Mean Fields
Monaghan et al. (2005)
24
(No Transcript)
25
Websites

• Realtime live action server for AMPS
• http//www.mmm.ucar.edu/rt/mm5/amps/
• http//polarmet.mps.ohio-state.edu/PolarMet/antarc
  ticnwp.html
• AMPS archives based at NCAR
• Multiple domains, 3 or 1 hourly forecast data
• AMPS-related publications by Bromwichs group
• http//polarmet.mps.ohio-state.edu/PolarMet/abstra
  cts.html
• 2006 Antarctic Meteorological Observation,
  Modeling, and Forecasting Workshop
• http//www.mmm.ucar.edu/events/antarctic06/index.p
  hp

26
Antarctic Southern Ocean Regional Reanalysis

• This is an idea that I am developing for the NSF
  IPY call with a probable submission date of
  February 2007.
• The spatial resolution will likely be 20 km for
  the modern satellite era and 60 km before that.
  (Bromwich)