Assimilation of GPS Radio Occultation Refractivity Data from CHAMP and SACC Missions over High South - PowerPoint PPT Presentation

Loading...

PPT – Assimilation of GPS Radio Occultation Refractivity Data from CHAMP and SACC Missions over High South PowerPoint presentation | free to view - id: 1abdbf-ZDc1Z



Loading


The Adobe Flash plugin is needed to view this content

Get the plugin now

View by Category
About This Presentation
Title:

Assimilation of GPS Radio Occultation Refractivity Data from CHAMP and SACC Missions over High South

Description:

Assimilation of GPS Radio Occultation Refractivity Data. from CHAMP and SAC-C ... cloud-tracked winds ~ from geostationary satellites ... – PowerPoint PPT presentation

Number of Views:86
Avg rating:3.0/5.0
Slides: 29
Provided by: graupelI
Category:

less

Write a Comment
User Comments (0)
Transcript and Presenter's Notes

Title: Assimilation of GPS Radio Occultation Refractivity Data from CHAMP and SACC Missions over High South


1
Assimilation of GPS Radio Occultation
Refractivity Data from CHAMP and SAC-C Missions
over High Southern Latitudes with MM5 4DVAR
TAE-KWON WEE AND YING-HWA KUO University
Corporation for Atmospheric Research, Boulder,
Colorado DAVID H. BROMWICH AND ANDREW J.
MONAGHAN Byrd Polar Research Center, The Ohio
State University, Columbus, Ohio
2008 Monthly Weather Review, 136, 2923-2944
2
Outline
  • Introduction
  • The CHAMP and SAC-C refractivity
  • Experiment framework
  • Data types used for assimilation
  • 4DVAR assimilation method and experiments
  • Results
  • Assimilation of GPS RO data without other
    observation types
  • Results for first 12-h assimilations
  • Results for extended assimilation periods
  • Summary and conclusions

3
Introduction
  • High southern latitudes cyclonic storms Play a
    role in poleward energy transport
  • in the Ross Sea in December 2001, shutting down
    aircraft operation nearly 1 week at McMurdo
    Station
  • AMPS (Antarctic Mesoscale Prediction System)
  • Operated at NCAR
  • Polar MM5 (Bromwich et al. 2001)
  • requires global analyses for its initial
    conditions from AVN of NCEP
  • the forecast skill of AMPS is strongly tied to
    the quality of the AVN analyses (Bromwich et al.
    2003)
  • poor quality of initial conditions was
    responsible for the unusually poor forecasts of
    AMPS (Monaghan et al. 2003)
  • comparing global analyses of ECMWF, NCEP UKMO,
    found a improvement in the accuracy of forecasts
    (Simmons 2002) in SH. However, the largest
    discrepancy between these global analyses still
    appeared over the Southern Ocean and
    Antarctic.(Pendlebury et al. 2003, comparing
    ECMWF, NCEP, UKMO, GASP JMA)
  • FIG. 1. suggests that information obtained from
    the observing systems for the period analyzed was
    insufficient to fully characterize the Antarctic
    atmosphere and that novel observational platforms
    should be made available and utilized.
  • ? GPS radio occultation (RO) offers one such
    opportunity.

4
Introduction
FIG. 1. The differences between NCEP and ECMWF
analyses (NCEP minus ECMWF) averaged for 15 Nov
200115 Jan 2002. Zonal mean of (a) geopotential
height (gpm) and (b) temperature (K). Contour
intervals are 10 gpm and 0.5 K. Heavy solid and
dashed contours correspond to zero and negative
values, respectively.
5
Introduction
  • GPS limb soundings
  • This concept has been successfully demonstrated
    with the GPS/MET (Ware et al. 1996)
  • Adventage of GPS RO data (Kursinski et al.
    1996)
  • high accuracy
  • high vertical resolution
  • all-weather capability
  • independent estimates of pressure as a function
    of height
  • 3D global coverage
  • calibration-free operation
  • GPS/MET ? CHAMP ? SAC-C ? COSMIC
  • data retrieval chain can be used for data
    assimilation
  • phase amplitude ? bending angle ? refractivity
    ? retrived T, P, Q.
  • to assimilate the GPS RO data, the refractivity
    and bending angle are likely candidates for
    practical applications (Kuo et al. 2000)

6
Introduction
  • Refractivity
  • The simulated atmospheric Refractivity data
    (OSSEs) are useful for recovering atmospheric
    structure, leading to positive impacts on NWP.
    (Zou et al. 1995)
  • The assimilation of refractivity data from a
    COSMIC-like constellation could improve the
    quality of Antarctic regional analyses and
    forecasts. (Wee Kuo 2004)
  • Healy et al. (2005) assimilated CHAMP
    refractivity data with a UKMO 3DVAR scheme and
    reported reduction of temperature error in the
    upper troposphere and lower stratosphere, most
    notable in the Southern Hemisphere.
  • Bending angle
  • The assimilation of GPS/MET bending angles using
    SSI slightly improves temperature, especially in
    the Southern Hemisphere. (Zou et al. 2000)
  • The assimilation of CHAMP bending angle data
    along with many other types of observations
    slightly reduces analysis and forecast errors in
    the tropics and in the Southern Hemisphere. (Zou
    et al. 2004)
  • Temperature error in the upper troposphere and
    lower stratosphere of Antarctica reduced by
    employing a 1D bending angle observation operator
    in the ECMWF 4DVAR system (Healy Thépaut 2006)
  • The main purpose of this study is to assess the
    impact of CHAMP and SAC-C GPS RO refractivity
    data on analyses and short-term forecasts over
    the Antarctic area.

7
The CHAMP and SAC-C refractivity
For neutral atmosphere
  • pairs of retrieved temperature (collocated CHAMP
    and SAC-C soundings occurring within 30 min and
    200 km for a period of about 2 months) profiles
    agree to less than 0.5 K in standard deviation
    and to within 0.1K in the mean between 5- and
    20-km altitude. (Hajj et al. 2004)
  • Global CHAMP soundings comparison with ECMWF
    analyses and radiosonde observations, the mean
    difference is less than 0.4 K in retrieved
    temperature (with standard deviation of
    difference is 1 K at 10 km and 2 K at 30 km) and
    less than 0.5 in refractivity, respectively. (
    Wickert et al. 2004)
  • In this study
  • The estimated error of CHAMP SAC-C GPS RO
    refractivity, standard deviation, was generally
    0.30.5 for 525 km.
  • A low-pass filter has been applied to the
    refractivity profiles to prevent potential
    aliasing on small scales.
  • Thinning the refractivity data down to an amount
    close to the vertical resolution of our
    assimilation model to prevents GPS RO data from
    contributing unduly to the observational cost
    function by outnumbering other data.
  • The observational error is assumed to be
    vertically uncorrelated, in common with other
    types of data used in this study.

8
The CHAMP and SAC-C refractivity
FIG. 2. Geographic distribution of CHAMP (open
circles) and SAC-C (filled squares) GPS RO
soundings for the period of 0000 UTC 90000 UTC
20 Dec 2001. The shaded area, including the
Antarctic continent, denotes the region used for
forecast verification. The lower histogram
represents the number of GPS RO soundings
available for each 12-h period.
9
Experiment framework a. Data types used for
assimilation
  • BUFR archive of the NCEPNCAR reanalysis project
  • geopotential height, temperature, horizontal
    wind, and specific humidity from radiosonde
    pibal observations
  • surface reports of pressure, temperature,
    horizontal wind, and specific humidity
  • cloud-tracked winds from geostationary
    satellites
  • aircraft observations of wind and temperature
  • ATOVS ( Advanced Television and Infrared
    Observation Satellite (TIROS) Operational
    Vertical Sounder ) virtual temperature
  • soundings from NOAA polar-orbiting satellites.
  • Additional satellite data
  • temperature and moisture soundings from MODIS
  • rainfall rate, total liquid water, precipitable
    water vapor, and surface wind speed from SSM/I
    (only used over the ocean)
  • surface wind vectors from QuikSCAT (only used
    over the ocean)

10
Experiment framework b. 4DVAR assimilation
method and experiments
  • The assimilation system used here is based on
    the MM5 adjoint model (Zou et al. 1995).
  • full-physics nonhydrostatic limitedarea model
    with a grid spacing of 90 km, 31 sigma levels,
    and a 10-hPa model top
  • The nonlinear MM5 forecast model and its adjoint
    share a suite of physical packages
  • simple-ice grid-resolvable precipitation
  • subgrid-scale precipitation
  • bulk aerodynamic planetary boundary layer
  • simple radiation scheme
  • Some new features have been implemented in MM5
    4DVAR
  • various additional observation types are
    assimilated
  • A weak constraint of a digital filter (Wee and
    Kuo 2004) is included to reduce the inadvertent
    effect of fast oscillations.
  • A combined 4DVAR method that provides not only
    an optimal initial condition but also an estimate
    of model error (by using the variational
    continuous assimilation (VCA) technique) is
    implemented.

11
Experiment framework b. 4DVAR assimilation
method and experiments
FIG. 3. Experiment timeline. The assimilation
experiments (GPS and No GPS) are conducted, with
a 12-h window, for 48-h periods. The 4DVAR
experiments consist of 19 cases that begin at
12-h intervals from 0000 UTC 9 Dec 2001. They are
performed with a cycling mode in which a 4DVAR
analysis is used as the background field for the
next cycle (upper-right inset). The initial
background field, for each case, is provided with
a 6-h MM5 forecast that has been initialized with
the AVN analysis. For each of the experiments,
cycles, and cases, a 5-day forecast is made
starting from the 4DVAR analysis. The forecast is
also made for the cold start run that takes its
initial condition directly from the AVN analysis.
12
Results a. Assimilation of GPS RO data without
other observation types
Analysis increments (AIs) analysis-background
(wind field) hydrostatic and geostrophic
adjustment processes as a response to the nearby
heating
(H L) coincide with the divergence and
convergence regions of midlevel outflow
FIG. 4. Analysis increments in the experiment
when only GPS RO data (marked with circled
crosses) are assimilated, valid at 1200 UTC 11
Dec 2001 (a) temperature and wind vector at s
0.54, which is about 545 hPa, with contour
interval of 0.2 K (b) surface pressure and wind
vector at the lowest model level with contour
interval 0.3 hPa zero value omitted.
13
Results a. Assimilation of GPS RO data without
other observation types
Sharp the tropopause increase horizontal ?T
FIG. 5. Vertical cross section along the heavy
line XY in Fig. 4. (a) Background potential
temperature (solid, contoured at every 10 K) and
wind speed (dashed, contoured at every 5 m s-1).
Shaded areas denote where the background wind
speed exceeds 20 m s1. (b) Analysis increment in
temperature (heavy solid and dashed, contoured
every 0.3 K, zero contour suppressed). Vertical
lines in (b) designate the locations of GPS RO
soundings.
14
Results b. Results for first 12-h assimilations
much larger than what is expected from the
contribution of GPS RO to the total observational
cost function changes in the background due to
the assimilation of GPS RO can stimulate the
reduction of observational misfit in other
observation types suggesting that the
assimilation of refractivity tends to change
moisture more easily unless limited by the
absolute amount of moisture
FIG. 6. Sensitivity of 4DVAR assimilation to GPS
RO data (see text for the definition), averaged
over the verification domain in Fig. 2 in
temperature (T), specific humidity (Q), pressure
(P), and vector wind (W).
15
Results b. Results for first 12-h assimilations
FIG. 7. Difference of forecast errors, GPS
runs-No GPS runs, with forecast time (15 days,
from left to right). The forecast error is
defined as the RMS difference against the
verifying ECMWF analysis. (top) Temperature (K)
(middle) geopotential height (gpm) (bottom)
vector wind (m s1). Horizontal bars represent one
standard deviation envelope among 19 cases.
Negative value means the error reduction due to
the assimilation of GPS RO refractivity.
16
Results c. Results for extended assimilation
periods
FIG. 8. Time series of RMS forecast error from
the second to fourth analysis cycles (a) 500-hPa
geopotential height (gpm), (b) 500-hPa
temperature (K), and (c) 850-hPa specific
humidity (g kg-1). For cold start runs, errors
for the first 12-h periods are not presented.
Note that the actual time difference between the
consecutive cycles is 12 h even though they are
24 h apart in the plots. This is only for
graphical convenience.
17
Results c. Results for extended assimilation
periods
FIG. 9. Time series of mean bias forecast error
from the second to fourth analysis cycles (a)
500-hPa geopotential height (gpm), (b) 500-hPa
temperature (K), and (c) 850-hPa specific
humidity (g kg-1). Note that the actual time
difference between the consecutive cycles is 12 h
even though they are spaced 24 h apart in the
plots. This is only for graphical convenience.
18
Results c. Results for extended assimilation
periods
FIG. 10. The 66-h forecasts of MSLP valid at 0600
UTC 14 Dec 2001, contoured at every 4 hPa (a)
cold start run initialized with NCEP AVN
analysis, (b) No GPS run, and (c) GPS run. The
forecasts of No GPS run and GPS run are made
after 48-h assimilations. Verifying ECMWF
analyses, dashed contours, are overlain.
19
Results c. Results for extended assimilation
periods
FIG. 11. Time series of 500-hPa error-increment
correlation coefficients of (a) geopotential
height and (b) temperature. The coefficients
represent the correlation between two departures
of No GPS run, one from the verifying ECMWF
analysis (error) and another from GPS run (GPS
ROinduced increment). Note that the actual time
difference between the consecutive cycles is 12 h
even though they are spaced 24 h apart in the
plots for graphical convenience.
20
Results c. Results for extended assimilation
periods
and sa and s GPS are the standard deviations of
the No GPS run from the verifying analysis and
GPS run.
FIG. 12. The 500-hPa temperature differences in
No GPS run-GPS run (contours with 0.3-K interval,
negative dashed, and zero omitted) and in No GPS
run-ECMWF analysis (color shaded) (a) analyses
valid at 0000 UTC 12 Dec 2001 after 48-h
assimilations (b) 24-h forecasts from the
analyses.
21
Results c. Results for extended assimilation
periods
FIG. 13. RMS forecast errors in refractivity ()
verified against GPS RO data available for each
24-h period.
22
Results c. Results for extended assimilation
periods
FIG. 14. RMS forecast errors verified against GPS
RO retrievals available for each 24-h period (a)
temperature (K), (b) specific humidity (g kg-1),
and (c) pressure (hPa).
23
Summary and conclusions
  • First 12-h assimilations
  • The impact for the first cycles was marginally
    positive or nearly neutral, and it varied greatly
    from one case to another.
  • The largest differences tended to be near the
    occultations that occurred in regions with strong
    baroclinicity.
  • The differences were highly oriented toward the
    baroclinically unstable areas where numerical
    models tend to have large, rapidly growing errors
    associated with the baroclinic development
  • The large case-to-case variation is attributed
    to the low density of GPS RO data and a strong
    dependency of the data impact on the location of
    GPS RO soundings relative to synoptic activity.
  • Because of the low observational density of
    current CHAMP and SAC-C GPS RO missions, an
    immediate and large impact in the analyses and
    very short-term forecasts was difficult to obtain.

24
Summary and conclusions
  • Assimilation continued over extended periods
  • Considerable positive impact of GPS RO data on
    the forecasts developed. It was observed in all
    parameters examined throughout the entire model
    atmosphere and forecast length out to 5 days and
    continually increased as the assimilation period
    was extended.
  • GPS RO data may be useful for early detection of
    the upper-level precursors for storm development
    (e.g., intensification of troughs and ridges, or
    enhancement of baroclinicity due to the increase
    of the thermal gradient)
  • The changes in the mass field also accompanied
    corresponding changes in the winds.
  • Consequently, the positive impact of GPS RO data
    was more evident in longer forecasts.
  • A correlation between the error and GPS
    ROinduced increment is introduced to demonstrate
    how the upscale growth of perturbations, attained
    by the assimilation of GPS RO data, results in a
    prominent positive impact in longer-range
    forecasts.
  • To complement NWP analyses and in situ
    observations, observed GPS RO refractivity data
    and their retrievals are used for the
    verification over the data-sparse Antarctic area.
    The verification not only showed a positive
    impact of GPS RO data, consistent with the
    results analyzed against the ECMWF analyses, but
    also confirmed their high value for NWP
    verification and monitoring.

25
End
26
APPENDIX
27
APPENDIX
FIG. A1. Observed MSLP (hPa) and wind speed (m
s-1) at Ferrell AWS (77.9S, 170.8E), near
McMurdo, December 2001.
28
APPENDIX
FIG. A2. Storm tracks for the 1217 Dec low in
the Ross Sea manually identified from satellite
imagery (black, circles) and from the 6-hourly
ECMWF analyses (gray, triangles). MSLP (hPa) is
also plotted for the ECMWF analyses.
About PowerShow.com