Parameter estimation and uncertainty analysis of forest carbon dynamics: Ensemble Kalman filter method

Background/Question/Methods

Ecosystem carbon (C) sequestration plays a critical role in ecosystem responses to elevated atmospheric CO₂. To estimate the C sequestration precisely and efficiently, Ensemble Kalman Filter (EnKF) combining a series of measurements with a terrestrial ecosystem model was used in this study to optimize parameters and quantify the uncertainties with eight types of measurements. The data included plant biomass, foliage and root biomass, woody biomass, litterfall, microbial biomass, forest floor carbon, soil carbon, and soil respiration at Duke Forest FACE site over a 9-year (1996-2004) period.

Results/Conclusions Results/Conclusions Results/Conclusions

The test results of the recovery rate for linear, nonlinear, and EnKF indicated that EnKF was an optimum method for parameter estimation in forest ecosystem C dynamics model. The optimization results showed that 6 of 8 parameters were constrained well, except the transfer coefficients of metabolic litter and passive soil organic mater. The prediction of data using estimated parameters is in very good agreement with observations. The forecasting of C sequestration showed that there would be 17,800-18500 g C·m^-2 stored in the ecosystem by the year 2014, which was consistent with most of the forecast results estimated using other approaches (e.g., MCMC). The uncertainties of parameters quantified by EnKF decreased with sequential estimations and increased with the length of periods without new data available, and they were also affected by measurement errors. Therefore, EnKF, as one of the most efficient data-model assimilation approaches, is a powerful tool to forecast terrestrial C sequestration in responses to global change.