Lidar-Seeded Polarimeter Aerosol and Cloud Retrievals

Presenter: Snorre Stamnes
Organization: NASA LaRC
Co-Authors: Xiaomei Lu, Xu Liu, Brian Cairns, Andrzej Wasilewski, Karena Lai, Michael Jones, Eduard Chemyakin, Nan Chen

Abstract

This project develops an AI framework to enhance Earth and Planetary Science mission returns by combining the global coverage of passive polarimeter data with lidar's high-fidelity vertical profiling. The approach leverages PACE (SPEXone and HARP2 polarimeters) and EarthCARE (ATLID lidar), collocated to generate 100,000+ paired profiles by December 2025.

Using PyTorch, neural networks trained on these datasets will enable PACE polarimeters to detect clouds, retrieve cloud top height (<1 km uncertainty), and profile aerosol vertical distribution (CALIOP-like uncertainty at 1 km resolution), reducing dependence on model assumptions and calibration errors.

Deliverables include cloud and aerosol products with quantified uncertainties, results for the PACE Software Operations Team, and open-source AI code.

By fusing polarimetry's microphysical richness with lidar's vertical precision, this work maximizes science return from existing assets and builds a 3D cloud/aerosol validation framework for next-generation AI weather forecasting models.