A hybrid optimal estimation and machine learning approach to predict atmospheric composition

Atmospheric measurement techniques 2026 Open access

Read full text
Authors: Frank Werner, K. W. Bowman, Seungwon Lee, Joshua L. Laughner, Vivienne H. Payne, James McDuffie
DOI: 10.5194/amt-19-3095-2026
Cited by: 2
References: 54

Abstract

We present a HYbrid REtrieval Framework (HYREF) that predicts subcolumn carbon monoxide (CO) from Cross-track Infrared Sounder (CrIS) observations by emulating TRopospheric Ozone and its Precursors from Earth System Sounding (TROPESS) optimal estimation (OE) retrievals. While OE provides physically rigorous results for only a small subset of CrIS data due to radiative transfer costs, HYREF combines machine learning (ML) with OE training to deliver full coverage and high-resolution predictions for every valid sample. Beyond CO, HYREF reproduces retrieval diagnostics, namely column averaging kernels, degrees of freedom, and errors, enabling robust comparisons, model evaluation, and data assimilation. Benchmarking shows excellent skill (correlation >0.99, bias

Fields of Study & Subjects

Air Quality Monitoring and Forecasting

Publication details

Journal / source
Atmospheric measurement techniques
Publisher
Copernicus GmbH
Volume
19
Issue
9
Pages
3095–3109
Publication date
11 May 2026
Language
EN