Estimation of net carbon exchange using eddy covariance CO2 flux observations and a stochastic model

In this paper we use a stochastic model to estimate annual net carbon exchange (NCE) from eddy covariance data taken from various sites. The stochastic model is comprised of a signal and a noise component. The signal component is characterized using a semiparametric model relating CO2 flux to light, temperature and time fitted to the eddy covariance observations. The noise component is characterized from the resultant model residuals using empirical cumulative probability distributions. The estimates for NCE are then derived from multiple runs of the joint signal-noise model within a Monte Carlo framework. This model-based approach to estimating NCE is evaluated using synthetic data and found to give a reasonable partitioning of the signal and noise in these data. Building on this, we derive estimates of NCE from observed annual eddy covariance data sets for various sites. For the six sites analyzed, the noise to signal ratio for the annual NCE estimates lies between 5 and 17% highlighting the potential value of eddy covariance observations for this application.