Beekmans, Christoph: 3-D Cloud Morphology and Evolution Derived from Hemispheric Stereo Cameras. - Bonn, 2020. - Dissertation, Rheinische Friedrich-Wilhelms-Universität Bonn.
Online-Ausgabe in bonndoc: https://nbn-resolving.org/urn:nbn:de:hbz:5-57512
@phdthesis{handle:20.500.11811/8283,
urn: https://nbn-resolving.org/urn:nbn:de:hbz:5-57512,
author = {{Christoph Beekmans}},
title = {3-D Cloud Morphology and Evolution Derived from Hemispheric Stereo Cameras},
school = {Rheinische Friedrich-Wilhelms-Universität Bonn},
year = 2020,
month = feb,

volume = 88,
note = {Clouds play a key role in the Earth-atmosphere system as they reflect incoming solar radiation back to space, while absorbing and emitting longwave radiation. A significant challenge for observation and modeling pose cumulus clouds due to their relatively small size that can reach several hundreds up to a few thousand meters, their often complex 3-D shapes and highly dynamic life-cycle. Common instruments employed to study clouds include cloud radars, lidar-ceilometers, (microwave-)radiometers, but also satellite and airborne observations (in-situ and remote), all of which lack either sufficient sensitivity or a spatial or temporal resolution for a comprehensive observation. This thesis investigates the feasibility of a ground-based network of hemispheric stereo cameras to retrieve detailed 3-D cloud geometries, which are needed for validation of simulated cloud fields and parametrization in numerical models. Such camera systems, which offer a hemispheric field of view and a temporal resolution in the range of seconds and less, have the potential to fill the remaining gap of cloud observations to a considerable degree and allow to derive critical information about size, morphology, spatial distribution and life-cycle of individual clouds and the local cloud field.
The technical basis for the 3-D cloud morphology retrieval is the stereo reconstruction: a cloud is synchronously recorded by a pair of cameras, which are separated by a few hundred meters, so that mutually visible areas of the cloud can be reconstructed via triangulation. Location and orientation of each camera system was obtained from a satellite-navigation system, detected stars in night sky images and mutually visible cloud features in the images. The image point correspondences required for 3-D triangulation were provided primarily by a dense stereo matching algorithm that allows to reconstruct an object with high degree of spatial completeness, which can improve subsequent analysis.
The experimental setup in the vicinity of the Jülich Observatory for Cloud Evolution (JOYCE) included a pair of hemispheric sky cameras; it was later extended by another pair to reconstruct clouds from different view perspectives and both were separated by several kilometers. A comparison of the cloud base height (CBH) at zenith obtained from the stereo cameras and a lidar-ceilometer showed a typical bias of mostly below 2% of the lidar-derived CBH, but also a few occasions between 3-5%. Typical standard deviations of the
differences ranged between 50 m (1.5 % of CBH) for altocumulus clouds and between 7% (123 m) and 10% (165 m) for cumulus and strato-cumulus clouds. A comparison of the estimated 3-D cumulus boundary at near-zenith to the sensed 2-D reflectivity profiles from a 35-GHz cloud radar revealed typical differences between 35 - 81 m. For clouds at larger distances (> 2 km) both signals can deviate significantly, which can in part be explained by a lower reconstruction accuracy for the low-contrast areas of a cloud base, but also
with the insufficient sensitivity of the cloud radar if the cloud condensate is dominated by very small droplets or diluted with environmental air.
For sequences of stereo images, the 3-D cloud reconstructions from the stereo analysis can be combined with the motion and tracking information from an optical flow routine in order to derive 3-D motion and deformation vectors of clouds. This allowed to estimate atmospheric motion in case of cloud layers with an accuracy of 1 ms-1 in velocity and 7° to 10° in direction. The fine-grained motion data was also used to detect and quantify cloud motion patterns of individual cumuli, such as deformations under vertical wind-shear.
The potential of the proposed method lies in an extended analysis of life-cycle and morphology of cumulus clouds. This is illustrated in two show cases where developing cumulus clouds were reconstructed from two different view perspectives. In the first case study, a moving cloud was tracked and analyzed, while being subject to vertical wind shear. The highly tilted cloud body was captured and its vertical profile was quantified to obtain measures like vertically resolved diameter or tilting angle. The second case study shows a
life-cycle analysis of a developing cumulus, including a time-series of relevant geometric aspects, such as perimeter, vertically projected area, diameter, thickness and further derived statistics like cloud aspect ratio or perimeter scaling. The analysis confirms some aspects of cloud evolution, such as the pulse-like formation of cumulus and indicates that cloud aspect ratio (size vs height) can be described by a power-law functional relationship for an individual life-cycle.},

url = {https://hdl.handle.net/20.500.11811/8283}
}

Die folgenden Nutzungsbestimmungen sind mit dieser Ressource verbunden:

InCopyright