Optimal selection of satellite XCO2 images over cities for urban CO2 emission monitoring using a global adaptive-mesh model

Abstract. There is a growing interest in estimating urban CO₂ emission from space-borne imagery of XCO₂. Emission estimation methods are already being tested and applied to actual or synthetic images. However, we still need automatic and standard methods, as well as objective criteria for selecting the images to be processed. This study shows the performance of an automated process for estimating urban emissions, standardised for all cities, using synthetic satellite images of XCO₂. We also use a decision tree learning method to define satellite image selection criteria.

We show that our method, based on a Gaussian plume model, has a success rate of 92 % when applied to our database of 9920 images covering 31 cities worldwide. Using our learning method, we show that the two main criteria guiding the error on the emission estimate are the wind direction's spatial variability and the targeted city's emission budget. Our learning method also allows us to separate images giving statistically accurate estimations from those giving erroneous estimations based on the two above mentioned criteria. Images for which the spatial variability of wind direction is low (less than 12°) and urban emissions high (greater than 12.1 ktCO₂/h) account for 47 % of images and have a bias on the emission estimation of -7 % of the emissions and a spread (IQR) of 56 %. Images with high spatial variability in wind direction or low urban emissions account for 53 % of the images and have a bias in the emission estimate of -31 % and a spread of 99 %.