International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004
the scan and can be calculated from the along-track distance on
the ground and the satellite velocity after Lorenz (1985). For
MISR, the time difference for different cross-track positions
within a block or scene can be assumed as constant. Northerly
winds lead to an underestimation of the heights and the along-
track wind component has to be added to the y-parallax, while
southerly winds result in overestimation of cloud-top heights.
The motion vectors retrieved from Meteosat-6 and Meteosat-7
data were resampled to the ATSR2 or MISR grid and the cross-
and along-track wind components calculated. Using the time
difference between the acquisition of the two views, the along-
track wind components were converted into CTH corrections.
3. VALIDATION
As each observation and each retrieval method have its own
characteristics, much can be learned from a intercomparison of
the results. The objective is to document the relative
performance of the different observations of cloud top height,
and where possible to understand this. In collaboration with
Eumetsat and the Rutherford Appleton Laboratory (RAL), two
case studies were analyzed with spaceborne observations from
ATSR2, MISR and Meteosat-6/-7 and ground-based
observations by the Chilbolton radar and radiosondes. The
results are summarized in Table 2; the retrieval methodologies
(next to our stereo processing) are described in detail in
Tjemkes et al. (2002).
Date ATSR2 MISR Radar Radiosonde
28/06/2000 235: 02 24+0.2 |2.48+0.03/2.5+0.2
13/06/2001 4.5 € 0.3 4.4+0.2 14.27 + 0.08 14.6 + 0.2
Table 2. CTH results (in km above sea-level) from ATSR2,
MISR, radar and radiosondes for two cases (20 June
2000 and 13 June 2001) over Chilbolton, UK.
Based on the validation experience from Cloudmap2, a new
comparison study of cloud height assignment methods has been
started by Eumetsat in November 2003. Next to the multi-view
photogrammetric retrieval from MISR and AATSR, the cloud
height will be determined by optimal estimation from AATSR
(by RAL), CO; slicing from MODIS and Oxygen A-band from
MERIS (by Free University Berlin). The comparison of these
different height products with the operational Eumetsat AMV
(Atmospheric Motion Vector) product will allow to evaluate the
strengths and weaknesses of each individual height assignment
method within the AMV production chain. Results from Phase
| of the study are reported in Fischer et al. (2004).
4. COMBINATION WITH GROUND-BASED STEREO
DATA
For the acquisition of ground-based stereo images of clouds, a
camera imager system has been developed (Figure 2). The
system, data acquisition and data processing is described in
detail in Seiz et al. (2002). The imager system was part of the
MAP-Special Observation Period (SOP) composite observing
system at the Rhine Valley, Switzerland, in autumn 1999 and
was later installed at the Ziirich-Kloten airport in September
2001 and April 2002. The system was used for the retrieval of
cloud-bottom heights, which were then used for the validation
of stereo cloud-top heights of vertically thin clouds (e.g. cirrus,
contrails) and for the combination of cloud-bottom boundaries
with satellite-derived cloud-top boundaries for a 3D
representation of the current cloud field. This second
application is explained in the next section.
*
Figure 2. Ground-based imager system (Skycam).
4.1 Combination of Satellite- and Ground-based Datasets
For assimilation experiments with a very high-resolution
version of the operational NWP model at MeteoSwiss, several
Skycam measurements series were performed at Zurich-Kloten
airport in April 2002, in coincidence with satellite (ASTER,
MISR) overpasses. In collaboration with the German Aerospace
Center (DLR), a 3D cloud data set was then derived from the
combination of MISR stereo cloud-top heights, MODIS
microphysical data and Skycam stereo cloud-bottom heights.
On a 50 m x 50 m grid, the cloud geometry was defined by the
cloud mask and the cloud-bottom heights from Skycam and by
the cloud-top heights from MISR. According to the effective
radius reached close to the cloud top and the total optical
thickness (both from MODIS MODO6 data), a constant liquid
water gain and a droplet number density could be derived for
each cloudy column. Thus, an internal profile of microphysics
of a vertical resolution of 50 m was achieved (Figure 3). More
details about the 3D cloud boundary extraction and subsequent
application in radiative transfer simulations at DLR can be
found in Seiz (2003) and in the Cloudmap2 Final Report,
D13/D14 (Cloudmap2, 2004).
T-Ronge (af
Q
Rr T
Y [km] 0
X [km]
Figure 3. 3D cloud water distribution, derived from 3D cloud
boundary data (MISR, ground-based camera system)
and MODIS information, 19 April 2002 (image
courtesy: Tobias Zinner, DLR).
1158
A+ pe FM QC £4 -——-* As
mj *n f.
=
n
EJ yt
