Application of an alternative method to derive reliable estimates of nighttime respiration from eddy covariance measurements in moderately complex topography

Even moderately complex topography can lead to significant horizontal and vertical advection and a consequent underestimation of nocturnal CO₂ effluxes derived from eddy covariance measurements on a single tower. The standard approach to select nighttime eddy flux data uses a threshold in friction velocity to exclude periods when advection is important but this is problematic in situations where turbulence is intermittent. van Gorsel et al. [van Gorsel, E., Leuning, R., Cleugh, H.A., Keith, H., Suni, T., 2007. Nocturnal carbon efflux: reconciliation of eddy covariance and chamber measurements using an alternative to the u^*-threshold filtering technique. Tellus 59B, 397-403] have developed an alternative method that estimates the CO₂ flux from the maximum of the eddy flux plus change in storage term in the period after sunset when stable stratification develops. During this time the advection terms do not contribute significantly to the mass balance of the air layer below the eddy flux instruments at the Tumbarumba flux station (SE Australia). Advection dominates only later in the night, following the development of large horizontal and vertical gradients of CO₂. As net nighttime and daytime fluxes are often of similar magnitude but opposite in sign, underestimation of respiration can lead to large errors in annual ecosystem carbon budgets. Nocturnal respiration rates obtained using the new approach are systematically higher compared to previously published results that used a u^*-threshold filter, and are in excellent agreement with independent chamber measurements. Revised net ecosystem exchange (NEE) calculations are therefore substantially reduced compared to those previously published (differing by 5.6 ± 0.5 t C ha^-1 year^-1 based on a 34-month period) and compare well with results from an ecosystem model (differing by -0.2 ± 0.7 t C ha^-1 year^-1 based on the same time period). Due to a drought period in 2003 and a subsequent insect attack NEE at Tumbarumba has been highly variable. Within 5 years the forest changed from being a carbon sink of -7.5 t C ha^-1 year^-1 to a source of +2.3 t C ha^-1 year^-1. This high variability demonstrates the importance of long-term, high-resolution time series when interpreting ecosystem data. The study further confirms the importance of using multiple measurement techniques and modelling to improve confidence in the mutually constrained NEE values.