Investigation of ice water content retrieval from active satellite data

In the course of this thesis work different aspects of the retrieval of ice water content from satellite data were investigated. In particular, an algorithm to retrieve the ice water content of clouds from CALIPSO lidar data has been developed with the aim of investigating which parts of a lidar IWC retrieval might be most challenging and which parameters would have the largest influence on the retrieval results. The data retrieved was then used to simulate a CloudSat radar signal in order to investigate the validity of the assumed McFarquhar-Heymsfield particle size distribution in different regions of a cloud. The ice water content retrieval was performed using a forward model algorithm that computed a lidar signal taking into account the effects of multiple scattering. For each altitude level the input parameters into this algorithm, which depend on the ice water content, were varied until the computed signal corresponded to the measured CALIPSO signal, thus determining the ice water content at one altitude. This step was repeated for each altitude, proceeding from the level closest to the instrument to the lowest altitude and using the results obtained for higher altitudes at each level (``onion peeling'' approach). The ice water content profiles retrieved from several CALIPSO profiles were then used to compute an overall particle size distribution according to McFarquhar and Heymsfield for one CloudSat footprint. This overall particle size distribution was then used to compute the radar reflectivity CloudSat would measure according to Mie theory. This simulated signal was compared to the signal actually measured by CloudSat in order to investigate the validity of the assumed particle size distribution. Regarding the IWC retrieval algorithm it was found that the effects of multiple scattering could well be accounted for with the retrieval method proposed. Additionally, it could be concluded that the lidar ratio has a very strong influence on the retrieval results and an accurate parameterization for this variable is essential for the quality of the retrieval results. From the CloudSat signal simulation and the subsequent comparison with the measured CloudSat signal it was found that the assumption of a constant lidar ratio for tropical clouds could not be made for CALIPSO lidar data. Additionally, it was found that the McFarquhar-Heymsfield particle size distribution tends to overestimate the amount of large particles at the cloud top. For lower regions of the cloud it was found that the PSD tends to underestimate the amount of large particles, which was attributed to an observed shattering effect for one of the instruments used in the particle size measurements. It was concluded that the particle size distribution should be parameterized according to the position in the cloud and that a better parameterization of the lidar ratio would help to draw more accurate conclusions about the PSD validity.Validerat; 20101217 (root