The eddy covariance method is the standard way flux towers measure how much CO₂ an ecosystem absorbs or releases. A sonic anemometer and a fast gas analyzer sample the vertical wind speed w and CO₂ concentration c many times per second. Each is split into a mean and a turbulent fluctuation (Reynolds decomposition):
w = w̄ + w′ c = c̄ + c′
Flux: F_c = ρ · mean(w′ · c′) over the averaging window
Turbulent eddies that carry air upward with a different CO₂ content than eddies carrying air downward produce a nonzero correlation between w′ and c′ — that correlation, not the mean wind, is the flux. A photosynthesizing canopy pulls CO₂ out of the air moving past it, so updrafts tend to run CO₂-depleted and downdrafts CO₂-enriched: w′ and c′ become negatively correlated, giving a negative F_c (net uptake). A respiring canopy at night does the opposite.
- Photosynthesis ⇄ Respiration — sets which process dominates, i.e. the true sign and strength of the w′–c′ correlation the turbulence carries.
- Turbulence intensity — bigger, gustier eddies (visualised as the moving spheres above the canopy) move more air and more CO₂ per unit time.
- Averaging window — eddy covariance needs enough samples (real towers use ~30 minutes at 10-20 Hz) for the random turbulent noise to cancel out and the true covariance to emerge; a short window is visibly noisier here, exactly as in a real dataset.
- Sensor height — a sensor mounted higher above the canopy integrates over a larger "footprint" of ground, which damps small-scale noise a little, just as real flux-footprint theory predicts.
The cumulative NEE (Net Ecosystem Exchange) panel integrates the live flux estimate over simulated time and converts µmol CO₂ m⁻² s⁻¹ into grams of carbon per square metre — the same quantity flux-tower networks (FLUXNET, AmeriFlux, ICOS) report to track how much carbon forests and fields pull out of the atmosphere.