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

Summary

This paper extends gap-filling methodology evaluation to challenging managed ecosystems and methane fluxes, assessing random forest regression against established approaches. Findings indicate that RFR is a competent alternative to standard algorithms, particularly superior for filling longer gaps (>30 days) in CO₂ and for other trace gases, whilst marginal distribution sampling remains preferred for short gaps (<12 days) in CO₂. Importantly, RFR reliably filled cumulative fluxes over gaps exceeding three months whilst preserving key environmental-flux relationships, and performed effectively using globally available reanalysis data when measured drivers were unavailable.

UK applicability

The findings are directly applicable to UK grassland and pasture monitoring networks, as European managed pastures were explicitly included in the study sites. UK researchers conducting eddy covariance measurements on livestock farms and permanent grasslands can adopt RFR approaches for handling data gaps more reliably than current standard methods.

Key measures

Gap-filling algorithm performance for CO₂, H₂O, energy, and CH₄ fluxes; cumulative flux accuracy for gaps exceeding 3 months; preservation of environment-flux response relationships

Outcomes reported

The study evaluated random forest regression (RFR) and marginal distribution sampling (MDS) approaches for filling missing data in eddy covariance time-series across carbon dioxide, water, energy, and methane fluxes. Testing occurred in European managed pastures, Southeast Asian converted peatlands, and North American drylands.

Topic tags