EXCITED

The goal of the EXCITED project was to develop open source code for a machine learning model that can produce gridded CO2 flux estimates based on eddy-covariance measurements, satellite data and meteorological fields from reanalysis products. The open source software can be used to produce CO2 flux data sets at different spatial and temporal resolutions and for different regions and time periods. An important innovation is to not only use eddy-covariance data to train the machine learning model, but to also use gridded CO2 flux estimates from inverse models. The eddy covariance is able to locally capture the short variability (e.g., the diurnal cycle that is driven primarily be the availability of light), whereas the inversely derived CO2 flux estimates are robust on larger spatiotemporal scales.

Besides the main code development, other important contributions to the field resulting from this project include the development of an international workshop at the Lorentz Center where early career to senior researchers met and discussed on important challenges and methodological developments in the field of terrestrial carbon cycle research. Finally, an important achievement of the project is the development of the open source software “xarray-regrid”. This is an extension to the popular Python package “xarray”, which is used to retrieve and store gridded data. As the name already suggests, the extension that we developed be used for regridding datasets, a task that researchers frequently encounter. Since the functionality of xarray-regrid is so generic, it is useful to a large group of researchers.