Stable gap-filling for longer eddy covariance data gaps: A globally validated machine-learning approach for carbon dioxide, water, and energy fluxes

Continuous time-series of CO2, water, and energy fluxes are useful for evaluating the impacts climate-change management on ecosystems. The eddy covariance (EC) technique can provide continuous, direct measurements ecosystem fluxes, but to achieve this gaps in data must be filled. Research-standard methods gap-filling have tended focus CO2 temperate forests relatively short less than two weeks. A method applicable other capable filling longer is needed. To address challenge, we propose a novel approach, Random Forest Robust (RFR). RFR accommodate wide range gap sizes, multiple flux types (i.e. water fluxes). We configured using either three (RFR3) or ten (RFR10) driving variables. was tested globally latent heat (LE), sensible (H) from 94 suitable FLUXNET2015 sites by artificial (from 1 30 days length) benchmarked against standard marginal distribution sampling (MDS) method. In general, improved MDS's R2 15% 30% reduced uncertainty 70%. RFR's improvements H LE were more twice improvement observed fluxes. Unlike MDS, performed well gaps; example, 30-day dropped 4% relative 1-day gaps, while MDS 21%. Our results indicate that timeseries. Such continuous measurements, with low bias, enhance our understanding ecosystems globally.