Bayesian forecasting and dynamic models (2nd ed.)
Bayesian forecasting and dynamic models (2nd ed.)
| Hi-index | 0.00 |
The parameters of ecosystem models are conventionally optimized through nonsequential inversion methods, which treat observations as a whole and lack the flexibility to investigate possible temporal evolution of the model parameters. This research developed a smoothed ensemble Kalman filter (SEnKF) to assess to what extent the parameters and state variables of an ecosystem model can be simultaneously optimized through the assimilation of eddy flux observations. The performance of the SEnKF was demonstrated in one case study: the assimilation of measurements of carbon exchange between a mixed forest and the atmosphere at Niwot Ridge Forest (Colorado, USA) from 2000 to 2004 into a carbon flux partition model. Our analyses demonstrated that some model parameters, such as light use efficiency and respiration coefficients, were highly constrained by eddy flux data at daily to seasonal time scales. Light use efficiency was strongly seasonal. Model predictions based on parameters modified by the SEnKF were much improved, compared to predictions made without progressive data assimilation. The SEnKF reduced the variance of state variables that is caused by uncertainties of parameters and driving variables. The analysis of net ecosystem exchange of carbon between the forest and the atmosphere was improved.