Reconstructing Carbon Monoxide Reanalysis with Machine Learning

TL;DR

This study investigates machine learning methods to predict monthly-mean total column of Carbon Monoxide re-analysis from a control model simulation and finds a promising approach to compensate for data losses.

Abstract

The Copernicus Atmospheric Monitoring Service provides reanalysis products for atmospheric composition by combining model simulations with satellite observations. The quality of these products depends strongly on the availability of the observational data, which can vary over time as new satellite instruments become available or are discontinued, such as Carbon Monoxide (CO) observations of the Measurements Of Pollution In The Troposphere (MOPITT) satellite in early 2025. Machine learning offers a promising approach to compensate for such data losses by learning systematic discrepancies between model configurations. In this study, we investigate machine learning methods to predict monthly-mean total column of Carbon Monoxide re-analysis from a control model simulation.

Figures (15)

• Figure 1: The global means timeseries of CR CO (Input), EAC4 CO (Target), ML and $\text{ML}_{\text{ano}}$ CO (prediction) for the whole dataset period. For each year the ML model is trained on the previous (max.) 5 years.
• Figure 2: Top row: Monthly $R^2$ scores and RMSE between the different datasets over the full period. Bottom row: Mean monthly $R^2$ scores and RMSE, highlighting seasonal variations.
• Figure 3: This figure shows the scores for each grid point over the yeas from 2004-2024. The left column shows the ML scores and the right column shows the CR scores.
• Figure 4: This figure shows the spatial error for ML prediction and CR for 2024. The NN is trained on years 2019--2023
• Figure 5: .