AUTOMATIC METHANE PLUMES DETECTION IN TIME SERIES OF SENTINEL-5P L1B IMAGES

Reducing methane emissions is essential to tackle climate change. In this paper, we introduce MetFluxNet, a deep learning model for methane plume source rate estimation. This model is trained on a new synthetic dataset designed to avoid network overfit. MetFluxNet can accurately estimate low source rates even in the case of heterogeneous backgrounds. To demonstrate its reliability for real-world plume estimation, we validated its predictions on real plumes with known source rates.

CO₂ radiocarbon content (Δ¹⁴CO₂) is a unique tracer that helps to accurately quantify anthropogenic CO₂ emitted into the atmosphere. Δ¹⁴CO₂ measured in airborne flask samples is used to distinguish fossil versus biogenic CO₂ sources. Mid-Atlantic US CO₂ variability is found to be driven by the biosphere. Errors in modeled fossil fuel CO₂ are evaluated using Δ¹⁴CO₂ airborne data as an avenue to improving future regional models of atmospheric CO₂ transport.

This study addresses uncertainties in atmospheric models by analyzing CO₂ dynamics in a complex urban environment characterized by a dense population and tropical vegetation. High-accuracy sensors were deployed, and the Weather Research and Forecasting model coupled with Chemistry (WRF-Chem) was utilized to simulate CO₂ transport, capturing variations and assessing contributions from both anthropogenic and biogenic sources.

We study the capacity of XCO₂ spaceborne imagery to estimate urban CO₂ emissions with synthetic data. We define automatic and standard methods and objective criteria for image selection. The wind variability and urban emission budget guide the emission estimation error. Images with low wind variability and high urban emissions account for 47 % of images and give a bias in the emission estimation of −7 % and a spread of 56 %. Other images give a bias of −31 % and a spread of 99 %.

We characterize the variability in CO and CO₂ emissions over Mexico City from long-term time-resolved Fourier transform infrared spectroscopy solar absorption and surface measurements from 2013 to 2021.  Using the average intraday CO growth rate from total columns, the average CO / CO₂ ratio and TROPOMI data, we estimate the interannual variability in the CO and CO₂ anthropogenic emissions of Mexico City, highlighting the effect of an unprecedented drop in activity due to the COVID-19 lockdown.

We have designed and deployed a mid-cost medium-precision CO₂ sensor monitoring network in Paris since July 2020. The data are automatically calibrated by a newly implemented data processing system. The accuracies of the mid-cost instruments vary from 1.0 to 2.4 ppm for hourly afternoon measurements. Our model–data analyses highlight prospects for integrating mid-cost instrument data with high-precision measurements to improve fine-scale CO₂ emission quantification in urban areas.

Description of the network for measuring greenhouse gas concentrations in the Paris region and analysis of eight years of continuous monitoring. 

We show that MethaneAIR, a precursor to the MethaneSAT satellite, demonstrates accurate point source quantification during controlled release experiments and regional observations in 2021 and 2022. Results from our two independent quantification methods suggest the accuracy of our sensor and algorithms is better than 25 % for sources emitting 200 kg h⁻¹ or more. Insights from these measurements help establish the capabilities of MethaneSAT and MethaneAIR.

A consistent estimation of CO₂ emissions is complicated due to the scarcity of CO₂ observations.  In this study, we showcase the potential to improve the CO₂ emissions estimations from the NO₂ concentrations based on the NO₂-to-CO₂ ratio, which should be constant for a source co-emitting NO₂ and CO₂, by comparing satellite observations with atmospheric chemistry and transport model simulations for NO₂ and CO₂. Furthermore, we demonstrate the significance of the chemistry in NO₂ simulations.

This study quantifies urban CO₂ emissions via an atmospheric inversion for the Paris metropolitan area over a 6-year period from 2016 to 2021. Results show a long-term decreasing trend of about 2 % ± 0.6 % per year in the annual CO₂ emissions over Paris. We conclude that our current capacity can deliver near-real-time CO₂ emission estimates at the city scale in under a month, and the results agree within 10 % with independent estimates from multiple city-scale inventories.

Monitoring of CO₂ emissions is key to the development of reduction policies. Local emissions, from cities or power plants, may be estimated from CO₂ plumes detected in satellite images. CO₂ plumes generally have a weak signal and are partially concealed by highly variable background concentrations and instrument errors, which hampers their detection. To address this problem, we propose and apply deep learning methods to detect the contour of a plume in simulated CO₂ satellite images.

In support of the global stocktake of the Paris Agreement on climate change, we proposed a method for reconciling the results of global atmospheric inversions with data from UNFCCC national greenhouse gas inventories (NGHGIs). Here, based on a new global harmonized database that we compiled from the UNFCCC NGHGIs and a comprehensive framework presented in this study to process the results of inversions, we compared their results of carbon dioxide (CO₂), methane (CH₄), and nitrous oxide (N₂O).

This study presents a simple atmospheric inversion modeling framework for the localization and quantification of unknown CH₄ and CO₂ emissions from point sources based on near-surface mobile concentration measurements and a Gaussian plume dispersion model. It is applied for the estimate of a series of brief controlled releases of CH₄ and CO₂ with a wide range of rates during the TOTAL TADI-2018 experiment. Results indicate a ~10 %–40 % average error on the estimate of the release rates.