Optimization of ecosystem model parameters through assimilating eddy covariance flux data with an ensemble Kalman filter

Process-based terrestrial ecosystem models have been widely used to simulate carbon cycle, climate and ecosystem interactions. Some parameters used in biological functions often change seasonally and inter-annually. In this study, sequential data assimilation with an ensemble Kalman filter is designed to optimize the key parameters of the Boreal Ecosystem Producitivity Simulator (BEPS) model, taking into account the errors in the input, parameters and observation. The parameters adjusted through data assimilation include foliage clumping index (Cf), slope of stomatal conductance to the net photosynthetic rate (m), maximum photosynthetic carboxylation rate (Vcmax) and electron transport rate (Jmax) at reference temperature of 25°C, multiplier to the soil organic matter decomposition rates (Kr). The fluxes of CO2 (separated into gross primary production (GPP) and ecosystem respiration (RE)) and water vapor measured using the eddy covariance technique at the BOREAS/BERMS Old Aspen site, Canada during 1997–2004 are used for the optimization. Parameters are optimized at a daily time step and presented as 10-day averages. The results show that the parameters varied significantly at seasonal and inter-annual scales. Photosynthetic capacity (Vcmax, Jmax) usually increased rapidly at the leaf expansion stage and reached a plateau in the early summer, then followed an abrupt decrease when foliage senescence occurred. The multiplier Kr to soil respiration coefficients were reduced to 0.5 in wintertime; however it increased rapidly in the spring and reached about 1.0 in summertime. The intensity of soil respiration may be related to the metabolic responses of the microbial communities and the availability of labile substances in summer and winter. From leaf expanding in the spring to senescing in the autumn, Cf presented declining trend from 0.88 to 0.78 with slight variation; m increased from 5 and approached to an approximately stable value of 8 since early summer. With optimized parameters, the estimates of GPP, RE, net ecosystem production and water vapor fluxes were significantly improved compared with the measurements at daily and annual time steps. With eddy covariance fluxes, data assimilation with an ensemble Kalman filter can successfully retrieve the seasonal and inter-annual variations of parameters related to photosynthesis and respiration of this boreal ecosystem site.