Ground- and Satellite-based Multi-view Photogrammetric Determination of 3D Cloud Geometry
Buch in deiner Nähe kaufen
Beschreibung
In this doctoral thesis, the possibilities of ground- and satellite-based multi-view measurements of clouds with modern photogrammetric methods are examined, with the objective to derive cloudbase/-top heights and motion. These parameters are important for a better description of clouds for nowcasting, numerical weather prediction and climate research. The presented work is part of the EU projects Cloudmap and Cloudmap2.
For the ground-based data acquisition, a new ground-based multi-camera system was developed. The camera system consists of at least two commercial digital CCD cameras, with a horizontal distance of about 500-1000 m, each connected to a laptop computer for camera control and image storage and to a radio clock for high-precision time synchronization. The stereo-photogrammetric method to calculate height and motion of the cloud-base included the precise determination of the interior and exterior orientation parameters of the cameras which was carried out with an in-house close-range photogrammetric testfield and an on-site orientation with GPS and stars. The cameras were installed at Mels, Switzerland, in October 1999 during the Mesoscale Alpine Programme (MAP) and at Zürich-Kloten Airport, Switzerland, in September 2001 and April 2002, in coincidence with other cloud measurement instruments (i.e. ceilometer, lidar, IR camera and soundings) and satellite overpasses of ERS-2 and EOS-Terra.
To calculate cloud-base and cloud-top height automatically, different image processing steps, including image preprocessing, feature extraction, image matching, blunder detection and further postprocessing, were applied to the multi-view images. Existing photogrammetric methods were thereby tested on the various cloud images and adapted when necessary to the specific problems encountered with clouds. First, a Wallis filter was used for radiometric equalization of the images and contrast enhancement. Suitable matching features were then selected with the Förstner or Harris operator for points and with the Canny operator for edges, respectively. If a cloud mask was available, it was used to thin out the feature set before matching to include only cloud features. The hierarchical matching approach with the Multi-Photo Geometrically Constrained (MPGC) LSM software, developed at our Institute, was successfully applied to the cloud images. For blunder detection, the matching results were quality-controlled with absolute and relative tests on the LSM statistics. The matching results illustrated that the MPGC LSM method can be used for the task of cloud matching, if applied with the necessary preprocessing, match point approximation and quality control strategies. It was further shown how a third camera can facilitate the ground-based cloud-base height retrieval through the additional geometric constraints.
Seven ground-based cases were analyzed in detail with the cloud-adapted matching algorithm. Thereby, different matching strategies were tested, including use of original images versus preprocessed images, use of more than two cameras and sequence-based analysis. As validation, the results were compared with semi-automatically measured points and with several other instrument data like radiosondes, ceilometers, lidars and IR cameras. Comparisons with these data have shown a good correspondence for the analyzed cases. The stereo camera system was able to retrieve accurate height values of cloud features for most cloud situations with relatively small standard errors (i.e. from a few meters for low clouds to about 100-150 m for high clouds at 10-12 km altitude). For most clouds, these stereoscopically matched features are a good proxy for the bottom boundaries of the clouds because most clouds become optically thick within a few hundred meters from their boundaries.
For the satellite part, multi-view images from MISR (on board EOS Terra), ASTER (on board EOS Terra) and ATSR2 (on board ERS-2) were used. As stereo image pairs from polar-orbiting satellites are never perfectly synchronous (i.e. time delay of some seconds between the image acquisition from the different viewing angles), the height error of the cloud-top heights, introduced by the along-track motion component, was corrected with cloud-top winds extracted from Meteosat-6 5-minute/10-minute rapid scan and Meteosat-7 30-min data. For MISR, with nine viewing angles, this height correction is only needed when two camera views or three symmetric views are taken. With at least three images from non-symmetric cameras, it is possible to directly separate the along-track parallax (due to cloud height) from the along-track wind contribution (due to cloud motion).
Four satellite-based cases with coincident stereo measurements of ASTER, MISR and ATSR2 and Meteosat-6/-7 image sequences were treated in detail. The results were compared to other operational cloud-top height and motion products as well as to radiosonde and ground-based cloud radar data. Thereby, the stereo cloud-top heights proved to be in very good correspondence (i.e. within 200-300 m) with the radar and radiosonde measurements, with the advantage that they depend only on basic geometric relationships of observations of cloud features from at least two different viewing angles and on the texture of these cloud features, while other cloud height estimation methods are dependent on knowledge of additional atmospheric parameters, like cloud emissivity, ambient temperature or lapse rate. As an interesting matching validation option, it was shown that, by chance, the cloud motion error for the MISR AN-AF and ASTER stereo cloud-top heights is approximately the same, independent of the actual cloud height and cloud motion. Therefore, it was possible to evaluate the accuracy of MISR AN-AF matching for one coincident ASTER-MISR case over Zürich-Kloten, independent of artifacts due to the subsequent wind correction.
Finally, three case studies of coincident ground- and satellite-based retrieval of cloud-base/cloudtop height and motion are presented. The case studies illustrated validation of satellite-based cloud-top height retrievals for vertically thin clouds with ground-based imagers. The described 3D cloud geometry data sets will be further used in the project Cloudmap2 for modelling and visualization studies. Together with the experiences from other ongoing cloud research projects on the assimilation of existing and new in-situ, ground-based and satellite-based cloud data into numerical weather prediction (NWP) models, it will significantly improve our understanding of cloud parametrization and representation in NWP and global climate models.
Autor*in
Gabriela Seiz
Themen in »Ground- and Satellite-based Multi-view Photogrammetric Determination of 3D Cloud Geometry«
Anwendungen der Photogrammetrie Wolkenbeobachtung Wolkenklassifikation