Spatial source attribution of eddy covariance flux data by inversion optimization

Since eddy covariance (EC) flux towers are typically mounted within structured landscapes, interpreting EC flux data is complicated due to spatial heterogeneity, which may exhibit sources and sinks simultaneously. This complexity makes it challenging to understand mechanisms and controls determining flux budgets for the individual land cover types that make up the entire ecosystem. Therefore, it complicates the scaling of flux results in space and/or time, or comparing EC fluxes under different environmental conditions.

We present a novel tool to decompose blended flux data from EC towers into individual components emitted by different land cover types within the tower’s footprint. The tool has two key components: 1) an exceptionally efficient algorithm that solves the steady-state transport equation, and 2) a linear optimizer to solve the inversion problem. This design allows for the analysis of years of continuous EC data on a typical desktop computer in a short time, with output consisting of half-hourly flux data for each land cover type individually.

The approach is entirely data-driven and can be applied to the fluxes of energy and scalars such as methane, N₂O, or CO₂. The model takes as input a land cover map containing the footprint and the standard output from the raw eddy data processing software, EddyPro. The accuracy of the flux attribution tool was validated using two EC towers in close proximity, sharing the same ecosystem and meteorological conditions, but with different land cover structures in the footprint. The agreement between the inversion results for each of the towers proves its applicability for a wide range of research questions.