The application of optimal estimation retrieval to lidar observations

Optimal estimation retrieval is a nonlinear regression scheme to determine the conditions statistically most-likely to produce a given measurement, weighted against any a priori knowledge. The technique is applied to three problems within the field of lidar data analysis. A retrieval of the aerosol backscatter and either the extinction or lidar ratio from two-channel Raman lidar data is developed using the lidar equations as a forward model. It produces profiles consistent with existing techniques at a resolution of 10-1000 m and uncertainty of 5-20%, dependent on the quality of data. It is effective even when applied to noisy, daytime data but performs poorly in the presence of cloud. Two of the most significant sources of uncertainty in that retrieval are the nonlinearity of the detectors and the instrument's calibration (known as the dead time and overlap function). Attempts to retrieve a nonlinear correction from a pair of lidar profiles, one attenuated by a neutral density filter, are not successful as uncertainties in the forward model eliminate any information content in the measurements. The technique of Whiteman et al. [1992] is found to be the most accurate. More successful is a retrieval of the overlap function of a Raman channel using a forward model combining an idealised extinction profile and an adaptation of the equations presented in Halldórsson and Langerholc [1978]. After refinement, the retrieval is shown to be at least as accurate, and often superior to, existing methods of calibration from routine measurements, presenting uncertainties of 5-15%. These techniques are then applied to observations of ash over southern England from the Eyjafjallajökull eruption of April 2010. Lidar ratios of 50-60 sr were observed when the plume first appeared, which reduced to 20-30 sr after several days within the planetary boundary layer, indicating an alteration of the particles over time.This thesis is not currently available in ORA