Search

Under the Copernicus programme, an operational CO2 Monitoring Verification and Support system (CO2MVS) is being developed and will exploit data from future satellites monitoring the distribution of CO2 within the atmosphere. Methods for estimating CO2 emissions from significant local emitters (hotspots; i.e. cities or power plants) can greatly benefit from the availability of such satellite images that display the atmospheric plumes of CO2. Indeed, local emissions are strongly correlated to the size, shape, and concentration distribution of the corresponding plume, which is a visible consequence of the emission. The estimation of emissions from a given source can therefore directly benefit from the detection of its associated plumes in the satellite image. In this study, we address the problem of plume segmentation (i.e. the problem of finding all pixels in an image that constitute a city or power plant plume). This represents a significant challenge, as the signal from CO2 plumes induced by emissions from cities or power plants is inherently difficult to detect, since it rarely exceeds values of a few parts per million (ppm) and is perturbed by variable regional CO2 background signals and observation errors. To address this key issue, we investigate the potential of deep learning methods and in particular convolutional neural networks to learn to distinguish plume-specific spatial features from background or instrument features. Specifically, a U-Net algorithm, an image-to-image convolutional neural network with a state-of-the-art encoder, is used to transform an XCO2 field into an image representing the positions of the targeted plume. Our models are trained on hourly 1 km simulated XCO2 fields in the regions of Paris, Berlin, and several power plants in Germany. Each field represents the plume of the hotspot, with the background consisting of the signal of anthropogenic and biogenic CO2 surface fluxes near to or far from the targeted source and the simulated satellite observation errors. The performance of the deep learning method is thereafter evaluated and compared with a plume segmentation technique based on thresholding in two contexts, namely (1) where the model is trained and tested on data from the same region and (2) where the model is trained and tested in two different regions. In both contexts, our method outperforms the usual segmentation technique based on thresholding and demonstrates its ability to generalise in various cases, with respect to city plumes, power plant plumes, and areas with multiple plumes. Although less accurate than in the first context, the ability of the algorithm to extrapolate on new geographical data is conclusive, paving the way to a promising universal segmentation model trained on a well-chosen sample of power plants and cities and able to detect the majority of the plumes from all of them. Finally, the highly accurate results for segmentation suggest the significant potential of convolutional neural networks for estimating local emissions from spaceborne imagery.

Cities are responsible for more than half of all greenhouse gas emissions. While many cities have committed to emission reduction trajectories, many lack the infrastructure to develop their emissions budgets. The measurement of CO2 plumes from cities by satellite imagery, coupled with atmospheric inversion methods, could allow quantifying direct CO2 emissions from cities, or at least detecting trends in their evolution.OCO-3, with its Snapshot Area Maps (SAMs) mode, is the first instrument to provide 2D (≈80km*80km) images of the total CO2 column at high resolution (≈2km*2km). In particular, these SAMs target atmospheric plumes of CO2 from cities and powerplants, with the goal of quantifying their emissions. Methods to estimate these emissions must be reliable and fast to process all available images (several thousands for OCO-3), whose number will increase with the CO2M and GeoCarb missions. The inversion methods by direct flux calculation (Integrated Mass Enhancement, Cross-Sectional and Source Pixel) or with a Gaussian plume model require little computation time. This thesis aims to evaluate the accuracy of these CO2 plume inversion methods and to study the favorable cases in terms of target and observation condition. This is done in a theoretical framework (atmospheric transport simulations) and by applying the methods to acquired SAMs.We quantify and analyze the different sources of error of these methods in detail using satellite pseudo-images of plumes, first over Paris and then over 31 cities in the world. The error of these methods is mainly due to errors in the estimation of the background concentration (XCO2 concentration that does not come from the city emissions) and in the estimation of the effective wind that carried the plume. We show, with a decision tree learning method, the sensitivity of the error on the emission estimate to the variability of the wind direction in the PBL and to the city's emission budget. The set of pseudo-images for which the emissions are large (>2.1ktCO2/h) and the wind direction variability low (2.1ktCO2/h) et la variabilité de la direction du vent faible (

Dataset and weights release for preprint "Segmentation of XCO2 images with deep learning: application to synthetic plumes from cities and power plants" Joffrey Dumont Le Brazidec et al.