Atmospheric lidars are crucial remote sensing tools in aerosol andclimate research. A pulsed time-of-flight lidar continuously samplesvertical atmospheric profiles across day and night, yielding aspatiotemporal atmospheric map. However, lidar analysis ischallenged by low signal-to-noise ratios, sunlight interference, andthe need for frequent calibration.To address these challenges, our work begins by developing theAtmospheric Lidar Data Augmentation (ALiDAn) framework tosimulate spatiotemporal and multiwavelength lidar data underdiverse atmospheric and system conditions. Using ALiDAn's data,we identify limitations and flaws in standard processing methods.To address these, we develop a Maximum Likelihood Estimation(MLE) approach tailored to lidar's photon-counting Poissonstatistics. This significantly improves signal analysis, especiallyduring the daytime, allowing for more frequent and accurateretrievals.With this enhanced signal estimation, we also reveal unexpectedflaws in the standard calibration approach. Therefore, we propose anovel learning-based lidar calibration that uses spatiotemporalmeteorological and lidar data. This approach helps address lowsignal-to-noise ratios, keeps calibration more updated than thecurrent operational method, and achieves higher accuracy onsimulated data. These advancements enable continuous, reliableaerosol retrievals and offer insights and tools to guide futureadvancement in lidar analysis.This talk presents several key advancements from my PhD research,conducted under the supervision of Prof. Yoav Y. Schechner at theAndrew and Erna Viterbi Faculty of Electrical & ComputerEngineering, the Technion.
