Constraints on ecosystem carbon and water flux : estimates in a temperate Australian evergreen forest

Abstract

Land-atmosphere carbon dioxide (CO2) exchange is the least constrained component of the global carbon cycle, yet it is driving most of its inter-annual variability. Seasonal and interannual variations in weather conditions affect biological activity and resulting CO2 exchanges, but the relative effects of phenology and climate on carbon cycling are not well understood. I used four years of eddy covariance data from a eucalypt woodland located near Sydney, South-East Australia, to better constrain carbon and water fluxes from this forest type. At our site, I observed a seasonal pattern of net ecosystem exchange (NEE) that contrasted with other flux tower sites in eucalypt forests. While similar Australian sites acted as a sink of carbon all year, especially in summer, our site behaved as a net sink of carbon in winter and a net source of carbon in summer. This pattern was caused by ecosystem respiration (Reco) driving the seasonal course of NEE, as the seasonal variability in Reco was bigger than that of gross primary production (GPP). GPP was limited by stomatal closure at high vapour pressure deficit in summer, but remained high in winter, while Reco was high in summer, and lower in winter. Leaf area index (LAI) varied seasonally, increasing rapidly mid-summer to reach a maximum in late summer, then decreased until the next year. LAI was a good predictor of canopy photosynthetic capacity (PC). The Community Atmosphere Biosphere Land Exchange (CABLE) land surface model was able to reproduce the seasonal variation in forest NEE but did not entirely capture canopy PC variability. Leaf demography, which is not accounted for in the model, may partly explain the mismatch between observed and simulated PC and should be further investigated. Our estimate of allocation of net primary productivity (NPP) to leaf growth was dynamic seasonally, which contrasts with the CABLE model assumption of a constant allocation factor in the evergreen broadleaf forest biome. Improved representation of dynamic allocation may further improve carbon cycle predictions in evergreen broadleaf forests. A semi-mechanistic model of heterotrophic respiration, the Dual Arrhenius Michaelis Menten model (DAMM), reproduced seasonal variations of Rsoil and Reco as a function of temperature and soil moisture. Daily to seasonal patterns of soil CO2 efflux were similar to those of Reco, but hourly dynamics were different, as Rsoil remained nearly constant overnight while Reco decreased. While decreasing air temperatures overnight may explain decreasing above-ground respiration, advection could also play a role, leading to a systematic data bias. Additional continuous, high frequency measurements of Reco components such as leaf respiration, stem respiration and soil respiration would improve mechanistic understanding of nighttime and daytime Reco. While weather variation was the major control of fluxes, the canopy phenology (leaf area index variations and leaf demography) also played an important role and needs to be incorporated in land surface models.

Date of Award	2019
Original language	English