Continuing the MLS water vapor record with OMPS LP using neural networks

Abstract. Stratospheric water vapor (SWV) plays an important role in atmospheric chemistry, dynamics, and radiative forcing. Satellite measurements by the Aura Microwave Limb Sounder (MLS), SciSat-1 Atmospheric Chemistry Experiment Fourier Transform Spectrometer (ACE-FTS), and Stratospheric Aerosol and Gas Experiment III on the International Space Station (SAGE III/ISS) have provided key constraints on SWV for the past decades. MLS provides the best geographical coverage among these instruments, but it approaches the end of its life cycle in the coming years, which will result in a data desert for satellite-based SWV measurements given that ACE-FTS and SAGE III/ISS only measure at a few dozen geolocations per day. The Ozone Mapping and Profiler Suite Limb Profiler (OMPS LP) is flying aboard the Suomi National Polar-orbiting Partnership (SNPP) and NOAA-21 satellites and is planned for additional platforms in the coming years. While not designed to measure SWV, it shows weak sensitivity to it, particularly in the wake of the Hunga eruption's significant injection of water vapor into the stratosphere. By utilizing the frequent co-locations between OMPS LP and MLS measurements, we developed a neural network-based retrieval algorithm to estimate SWV from SNPP OMPS LP radiances between 11.5 and 40.5 km. We find that the LP SWV profiles typically agree with MLS within 5 %, and agreement with ACE-FTS and SAGE III/ISS is typically within 10 %. We show that the SNPP-trained model is applicable to NOAA-21 OMPS LP without retraining, though minor differences in radiances between the instruments results in a ∼ 5 % bias under most conditions. Our results suggest that OMPS LP can continue the global water vapor record in the lower stratosphere into the 2030s, though continued independent measurements by satellite and balloon-borne instruments will be key to verifying the stability of our approach for quantifying decadal-scale SWV variability.