A regional inverse modelling system implemented as fully automated reproducible self-documenting workflow

The ICOS research infrastructure collects, maintains and supports a large range of high quality and long-term greenhouse gas observations in its FAIR data repository Carbon Portal, together with inventory, modelled surface fluxes and ancillary data. All of these receive unique persistent identifiers as key pre-requisite for achieving reproducible scientific workflows.

The Lund University Modular Inversion Algorithm (LUMIA) has been developed at the Lund University hosting department of Carbon Portal for inverse modelling of carbon dioxide and (isotope resolved) methane. The model uses as input observational and prior flux data like those mentioned before. The overarching goal of this so called top-down inverse method is to provide optimized emission fields. Next to common and scientific interest, the method also can inform policy makers on the progress in reaching the goals of the 2015 Paris agreement, through independent estimates of emissions and natural fluxes. For this purpose an open, flexible, transparent, modular and reproducible system like presented here is needed.

Here we present a self-documenting, fully reproducible workflow for our inverse carbon model system LUMIA. The workflow design is based on frameworks developed through EU initiatives like Copernicus CAMS and CoCO2.

The design goals include using the system as demonstrator of the general principle of the use of inverse modelling tools based on simplified and reduced datasets, to perform network design for observational systems based on real and synthetic data sets, up to full blown regional inversions based on all available data, including an analysis of uncertainties of the optimized fluxes.