Gap-filling carbon dioxide, water, energy, and methane fluxes in challenging ecosystems: Comparing between methods, drivers, and gap-lengths

Summary

This paper evaluates gap-filling methodologies for eddy covariance flux measurements in challenging ecosystems where data gaps are common due to instrument failure or management disturbances. Random forest regression proved more reliable than standard approaches for filling longer gaps (> 30 days) and for methane and energy fluxes, whilst marginal distribution sampling remained preferable for short CO₂ gaps. The findings are particularly relevant for agri-ecosystems and peatland conversions, where maintaining complete flux time series is essential for accurate cumulative carbon accounting.

UK applicability

The study's findings on European managed pastures directly apply to UK grassland monitoring and greenhouse gas quantification programmes. The RFR approach and recommendations for gap-filling methodology could enhance UK eddy covariance networks and improve reliability of UK agricultural emissions data under the UK's net-zero and emissions reporting obligations.

Key measures

Gap-filling accuracy and reliability for CO₂, water, energy, and methane fluxes; performance across gap lengths (< 12 days, 12–30 days, > 30 days, > 3 months); ecosystem types assessed: European managed pastures, Southeast Asian converted peatlands, North American drylands

Outcomes reported

The study evaluated random forest regression (RFR) and marginal distribution sampling (MDS) approaches for filling missing data in eddy covariance time series measuring carbon dioxide, water, energy, and methane fluxes across managed pastures, converted peatlands, and drylands. Performance was assessed across different gap lengths and flux types to determine optimal gap-filling methodologies for challenging ecosystems.

Topic tags