Applying Machine Learning Techniques to Airborne Science Planning and Real-Time Column Mole Fraction Estimates for Integrated Path Differential Absorption Lidar

Spectral Sensor Solutions LLC (S3) and our subcontractor, Texas Tech University, are pleased to submit this proposal in response to the NASA SBIR Phase I solicitation, subtopic S5.03 Accelerating NASA Science and Engineering through the Application of Artificial Intelligence. In Phase I we propose to apply modern machine learning (ML) methods to enable real-time flagging of integrated path differential absorption (IPDA) lidar measurements using ancillary data collected inflight. In addition, during Phase I, we propose to leverage significant improvements in research forecast models to build tools to extract data relevant to same-day flight planning tools and inputs required for low-uncertainty, inflight, preliminary data analysis of IPDA measurements. In Phase II we will use ML methods to score the quality of research forecast inputs to preliminary column mole fraction (CMF) retrievals. The IPDA lidars used in the Atmospheric Carbon Transport - America NASA Earth Venture Suborbital Program, the Multi-functional Fiber Laser Lidar (MFLL) and the High Altitude Lidar Observatory (HALO), will provide the primary data for these retrievals. Ultimately, the goal is to demonstrate the feasibility of using high-resolution rapid refresh (HRRR) models to enable improved NASA airborne science mission planning and to demonstrate application of the derived thermodynamic variables to near-real-time CMF retrievals with low uncertainty. Tools developed in this SBIR will provide a significant improvement for future airborne science missions in general through enhanced tools for flight planning, resulting in maximization of science return per flight hour. Additionally, science campaigns using active remote gas monitoring systems will be significantly enhanced through the ability to provide low uncertainty retrievals of column mole fractions for the gas of interest in near-real-time from the remote measurements versus the weeks- or months-long lag currently available for these systems.