In temperate and boreal ecosystems, phenological transitions (particularly the timing of spring onset and autumn senescence) are thought to represent a major control on spatial and temporal variation in forest carbon sequestration, as these transitions effectively determine growing season length. To investigate these patterns, we analyzed 153 site-years of data from the FLUXNET ‘La Thuile’ database. Eddy covariance measurements of surface-atmosphere exchanges of carbon and water from 21 research sites at latitudes from 36°N to 67°N were used in the synthesis.
We defined a range of phenological indicators based on the first (spring) and last (autumn) dates of (1) C source/sink transitions (‘carbon uptake period’); (2) measurable photosynthetic uptake (‘physiologically active canopy period’); (3) relative thresholds for latent heat (evapotranspiration) flux; (4) phenological thresholds derived from a range of remote sensing products; and (5) a climatological metric based on the date where soil temperature equals mean annual air temperature. We then tested whether site-level flux anomalies were significantly correlated with phenological anomalies across these metrics, and whether the slopes of these relationships (representing the sensitivity to phenological variation) differed between deciduous broadleaf (DBF) and evergreen needleleaf (ENF) forests.
Results/Conclusions

Across sites, growing season length was tightly correlated with annual GPP (gross primary productivity) for most phenological metrics, and the slope of this relationship (~6 gC/m2/ d) did not differ between evergreen and deciduous sites. By comparison, within sites, interannual phenological anomalies were less strongly correlated with GPP anomalies; a one-day anomaly in growing season length increased deciduous annual GPP by ~3 gC/m2/d, and evergreen GPP by only ~2 gC/m2/d. GPP was consistently more sensitive than Reco (ecosystem respiration) to spatial or interannual variation in phenology; as a result, both earlier springs, and later autumns, tended to result in increased C sequestration. Phenological metrics derived from fluxes themselves, rather than remote sensing, were better correlated with spatial and temporal patterns of flux variation.
In relation to both within- and across-site variation in phenology and fluxes, results obtained depended on the phenological metric used, i.e. definition of “start” and “end” of growing season, emphasizing the need for improved understanding of relationships between these different metrics and ecosystem processes. Furthermore, differences in flux-phenology relationships in the context of spatial and temporal variation in phenology raise questions about using results from either short-term or space-for-time studies to anticipate responses to future climate change.