Evaluation of a novel approach to partitioning respiration and photosynthesis using eddy covariance, wavelets and conditional sampling

Abstract

The eddy covariance (EC) technique remains a cornerstone for direct, continuous monitoring of greenhouse gases fluxes, particularly for carbon dioxide (CO₂). Traditionally, EC-derived net ecosystem exchange (NEE) is partitioned into gross primary productivity (GPP) and ecosystem respiration (R_eco) using model-based approaches. Here, we present a novel, fully empirical partitioning method that applies conditional sampling to wavelet-decomposed signals, isolating positive and negative contributions of the wavelet co-spectrum of vertical wind velocity and CO₂ dry molar fraction, conditioned by the water vapour flux. This method was evaluated across two French ICOS sites, a mixed forest (FR-Fon) and a cropland (FR-Gri), over multiple years.

The approach is grounded in the hypothesis that wavelet decomposition enables separation of oppositely signed turbulent structures across scales, a claim supported by co-spectral analysis. The resulting flux components exhibited distinct frequency signatures under neutral and unstable atmospheric conditions, though not under stable stratification.

Daily partitioned fluxes derived from this method aligned well with GPP and R_eco estimates from established nighttime- and daytime-based partitioning, with inter-method differences smaller than those observed between the conventional approaches themselves. Conceptually the method approximates net photosynthesis and offered improved coherence with site-specific ecological and management dynamics, capturing events such as growing season, harvest, and manure application at FR-Gri, more reliably than standard methods. It also avoided spurious GPP estimates common error in the night-time approach. Moreover, the diel R_eco cycle revealed a bimodal pattern, suggestive of combined influences from solar radiation and soil temperature, in contrast to the predominantly single temperature-driven dynamics inferred by conventional models.

Our findings demonstrate that wavelet-based conditional sampling offers a promising alternative for CO₂ flux partitioning, one that is entirely empirical, calibration-free, and grounded in the physical co-emission dynamics and transport from surface to the atmosphere.