Under the Global Covenant of Mayors (GCoM) initiative, cities present their action plans committing to mitigate greenhouse gas (GHG) emissions and/or adapt to climate change. One concrete objective consists in setting a reduction target, by which cities commit to reduce their baseline GHG emissions for a chosen target year. In monitoring their emissions, cities report inventories for any arbitrary year(s), making available only discrete readings in what can be considered a very sparse yearly time series. Examining the performance of the cities for the target years of 2020 and 2030, the actual measurements are usually not available. Therefore, a machine learning methodology is proposed to predict the GHG emissions inventories for each city on their target year, enabling the assessment of the cities' performance inside a common reporting framework. Using the reported inventories, the methodology identifies a model for each city, minimizing the error for the last known reported value. As a result, the proposed method allows predicting GHG emissions for cities from their yearly inventories, controlling the uncertainty associated to the estimations and extracting reliable information that can be updated as soon as new emissions inventories become available.

FRANCO DE LOS RIOS Camilo;  MELICA Giulia;  TREVILLE Aldo;  BALDI Marta Giulia;  PISONI Enrico;  BERTOLDI Paolo;  THIEL Christian; 

2022-09-05

ELSEVIER

JRC128979

2210-6707 (online),   

https://www.sciencedirect.com/science/article/pii/S2210670722004279,    https://publications.jrc.ec.europa.eu/repository/handle/JRC128979,   

10.1016/j.scs.2022.104114 (online),