An Integrated Profiling Technique IPT and its accuracy assessment

1
An Integrated Profiling Technique (IPT) and its
accuracy assessment
Ulrich Löhnert Meteorological Institute
University of Munich
2
Ground-based atmospheric profiling

• Microwave remote sensing allows the retrieval of
  high-quality temperature, humidity and
  microphysical cloud parameters
• long-term times series of cloud profiles are
  necessary for the evaluation of NWP and climate
  models
• more accurate knowledge of cloud microphysics are
  needed in order to better understand the
  cloud-radiation interaction
• continuous measurements of microwave profilers
  may have the potential to partially substitute a
  number of stations in an existing radiosonde
  network
• ground based remote sensors can perform important
  benchmark measurements for satellite retrieval
  validation

KNMI, April 27, 2004
3
IPT-Objective
Get more accurate by
combining as much instruments as you can
difficult need to specialize in many instruments
do it in a physically consistent way
even more difficult need to calculate the
forward models !
KNMI, April 27, 2004
4
Main instruments at the Cabauw site
Cloud radar (KNMI/GKSS) Advantage accurate
determination of cloud vertical
structure Disadvantage direct determination of
microphysical cloud parameters highly erroneous
Microwave radiometer MICCY (Bonn) Advantage
accurate LWP, temperature and humidity profile
Disadvantage very limited vertical resolution
concerning liquid water
Laser ceilometer (KNMI) Advantage accurate
determination of cloud base Disadvantage high
absorption within liquid cloud
KNMI, April 27, 2004
5
Microwave profiler MICCY
MICCYs absorption characteristics are suited for
determining T, q, (and LWC) profiles
complex underdetermined inversion problem
Simplification linearization and discretization

• still an ambiguous problem
• KJ is ill-conditioned

further constraints required
with
KNMI, April 27, 2004
6
IPT scheme (off-line)
19 MICCY brightness temperatures (TB)
dBZ profiles
lidar cloud base

• optimized profiles of
• Temperature (T)
• humidity (q)
• LWC

Bayesian Retrieval
ground measurements of T, p, q
a priori LWC profile (model climatology)
a priori T und q profiles (nearest-by radiosonde)
measurement-consistent with respect to error
covariances
KNMI, April 27, 2004
7
IPT-equations
y measurement vector (TB, dBZ, Tgr, qgr)
Forward model F K Jacobi-matrix
(dF/dx) xa a priori profile Se error
covariance matrix Sa a priori covariance
IPT equations (optimal estimation) derivable
from Bayesian probability theory condition
Gaussian distributed parameters

• solution is not exact, but rather probability
  density
• a priori profile stabilizes the solution
• iterative approach guarantees physical
  consistency

KNMI, April 27, 2004
8
Error covariance matrix Se

• 0.5 K random calibration error main
  diagonal components
• forward model error of RTO included in Se
  (actual error is
  unknown and is thus interpreted as the covariance
  between the two standard microwave absorption
  models of Rosenkranz 1998 and Liebe 1993)
  main and off-diagonal components

no correlation

• 3 dBZ random error main diagonal components
• inclusion of error covariances due to a standard
  attenuation correction for water vapor and liquid
  water main and off diagonal components
  
  
• forward model error of dBZ-LWC relation included
  main diagonal components

KNMI, April 27, 2004
9
Covariance matrix Sa (off-line)
Height
Height
temporal interpolation
Profile
Profiel
Operational radiosonde t to 12h in 30km
distance to Cabauw
Operational radiosonde t to in 30km distance to
Cabauw
Height
Main diagonal components (Toper,i Tcab,i)2 Off
diagonal components (Toper,i Tcab,i)(Toper,j
Tcab,j)
Radiosonde at Cabauw t to 6h
Profile
KNMI, April 27, 2004
10
Example Cloud Classification
KNMI, April 27, 2004
11
IPT retrieval May 19, 2003
KNMI, April 27, 2004
12
IPT retrieval May 19, 2003
KNMI, April 27, 2004
13
IPT retrieval May 19, 2003
KNMI, April 27, 2004
14
IPT evaluation T q (BBC1)
(IPT-DeBilt RS, interpol.)
KNMI, April 27, 2004
15
IPT evaluation LWC (BBC1)
KNMI, April 27, 2004
16
Comparison with Cabauw radiosondes (BBC2)
with clouds
KNMI, April 27, 2004
Thanks to A. Schomburg, U-Bonn
17
Comparison with Cabauw radiosondes (BBC2)
with clouds
KNMI, April 27, 2004
Thanks to A. Schomburg, U-Bonn
18
Comparison with Cabauw radiosondes (BBC2)
without clouds
KNMI, April 27, 2004
Thanks to A. Schomburg, U-Bonn
19
Comparison with Cabauw radiosondes (BBC2)
without clouds
KNMI, April 27, 2004
Thanks to A. Schomburg, U-Bonn
20
Comparison Cabauw radiosondes with LM (BBC2)
with clouds
KNMI, April 27, 2004
Thanks to A. Schomburg, U-Bonn
21
Comparison Cabauw radiosondes with LM (BBC2)
with clouds
KNMI, April 27, 2004
Thanks to A. Schomburg, U-Bonn
22
Summary

• IPT is a method for deriving measurement-consisten
  t profiles of T, q, and LWC using a combination
  of remote and in-situ sensors in an optimal sense
• Together with the cloud classification scheme,
  IPT can be applied automatically
• DT lt 1K, Dq lt 1 gm-3 overall
  DLWC 20
• Temperature and humidity profiles derivable
  during cloudy cloud-free cases!
• Microwave absorption uncertainties present a
  serious limitation

23
Outlook A sophisticated IPT-evaluation using
RACMO output

• Apply IPT to RACMO model output and verify the
  results with respect to the original model
  parameters
• No uncertainties due to absorption model
• Detailed and exact calculation of (error)
  covariance matrices is possible
• Consider accuracy issue w.r.t. a priori
  radiosonde
• distance
• Required working steps prior to IPT application
• Simulate MICCY brightness temperatures from model
  parameters
• Extract ceilometer radar parameters
• Determine a priori profiles of T, q, LWC
• Calculate (error) covariance matrices Sa, Se

KNMI, April 27, 2004
24
Outlook

• Preparations are on-going for continuous IPT
  application to the combined data set available at
  Cabauw from August 2001 until today
• Future extensions will include
• an algorithm merger product with other
  complimentary retrievals (TUD, Reading )
• Ice-phase and mixed-phase micro-physics
• N, reff
• forward modelling of solar and infrared radiation
  ?
• wind profilers ??

25
Available IPT data

• updated BBC1 cloud classification scheme and T,
  q, LWC retrievals available on the BBC data base
• bbc.knmi.nl/BBC1/cabauw/microwaves/ubonn/daily/le
  vel2av/
• (CLIWA-NET format)
• Refer to Readme file or me (uloeh_at_uni-bonn.de)
• BBC2 IPT data will follow in a matter of weeks
  from now broadcast mail will be sent to all

26
Questions concerning RACMO output

• How ist LWP calculated with respect to cloud
  cover?
• Is LWC an output parameter?
• At what time intervals are fresh analyses
  available?
• What are the typical forecast times, which are
  openly available?

27
Specific points

• has been previously carried out by Erik van
  Meijgaard
• Phase1 IPT without radar data
• Phase2 Use some kind of noisy Z-LWC relation
  to simulate a realistic Z-LWC dependency from
  RACMO
• 3a. Off-line IPT (as done momentarily)
• Simulated radiosonde as T q a priori
• use temporally averaged T q profiles
  assuming a 6h (12h) radiosonde time
  difference from a grid point corresponding to
  DeBilt
• Sensitivity study concerning radiosonde time
  interval and spatial distance
• LWC a priori use existing 1D-model data?
• 3b. On-line IPT (as to be done in future)
• RACMO analysis/forecast as T, q, LWC a priori

KNMI, April 27, 2004
28
Specific points
3b. On-line IPT (as to be done in future)
next available RACMO analysis/forecast as T, q,
LWC a priori (knowledge when forecasts are
available is crucial) 4a. Calculate covariances
as previosly shown 4b. Calculate covariance
matrices for T, q and LWC by evaluating
differences between existing radiosonde ascents
and previously run forecasts
KNMI, April 27, 2004