This article discusses a novel framework developed for Eddy Covariance CO2 flux gap filling—a process used to fill in data gaps in long-term series of CO2 flux measurements. Developed by Ugolini et al., the framework focuses on the combination of machine learning and time series decomposition to fill data gaps.
The framework begins with the application of linear interpolation for small gaps in the data. Requirements for larger gaps are met by utilizing advanced machine learning techniques such as Random Forests. Finally, a series of time series decompositions bring the noise out of the data.
The MDPI (Multidisciplinary Digital Publishing Institute) is a scholarly open-access publisher specializing in the digital publication of peer-reviewed journals and books. Founded in 1996, this Geneva-based research-driven company offers its services in more than 150 thematic areas, excluding the physical sciences. All articles published by MDPI are immediately available worldwide under open access.
Ugolini et al. is a team of computer scientists and researchers at the National Research Council (CNR) of Italy’s Institute of Information Science and Technologies (ISTI). Their research focuses on artificial intelligence, machine learning, and data analysis. In particular, they are studying ways to fill in data gaps in long-term measurements through the use of advanced techniques based on machine learning and time series decomposition. Ugolini et al.’s research paper on Eddy Covariance CO2 Flux Gap Filling highlights the importance of filling gaps in long-term series of CO2 flux measurements in order to better understand greenhouse gas emissions and their role in the global climate.
