Estimating mean carbon turnover time of three forest ecosystems in South Korea using data-model fusion

Mean Carbon Turnover Time (MCTT) determines carbon (C) storage capacity of forest ecosystems. Recently, researches estimating MCTTs of forest ecosystems using data-model fusion are emerging. Data-model fusion is a statistical approach constraining model parameters to minimize deviations between data and model, thereby producing more reliable simulation results such as ecosystem C stocks and ecosystem C effluxes. This study used data-model fusion to estimate MCTTs of evergreen coniferous forests (EC), deciduous broad-leaved forests (DB), and mixed forests (Mix) in South Korea. The national forest inventory (NFI) data including ecosystem C stocks and the Forest Biomass and Dead organic matter Carbon model (FBDC) were adopted to conduct data-model fusion. Model parameters of the FBDC were constrained, and the FBDC estimated ecosystem C stocks, ecosystem C effluxes, and MCTTs of EC, DB, and Mix using the constrained model parameters. Most of the model parameters were well constrained, except for decay constants for dead wood, fallen leaves, and fallen branches. The estimated ecosystems C stocks showed a consistency with the observed ecosystem C stocks data in NFI, resulting in the coefficients of determination (R²) of 0.93, 0.85, and 0.81 for EC, DB, and Mix, respectively. The estimated mean ecosystem C stock (Mg C ha^-1) of EC (156.27) was smaller than that of DB (184.86) and Mix (188.10). The estimated ecosystem C effluxes (Mg C ha^-1 year^-1) of EC, DB, and Mix were 4.55, 5.43, and 5.48, respectively. However, there were only marginal differences among the estimated MCTTs (years) of EC (33.72), DB (33.13), and Mix (33.80). These results indicate that three forest ecosystems in South Korea might have similar C storage capacity even though ecosystem C stock of EC is relatively smaller. Moreover, the C storage capacity of EC will be underestimated when we assume equivalent ecosystem C effluxes for all the three forest ecosystems without considering the estimated ecosystem C effluxes. We suggest using data-model fusion for estimating MCTT when observations in ecosystem C effluxes are limited.

Funding sources: Korea Forest Service (Korea Forestry Promotion Institute; 2018110C10-2020-BB01), the Ministry of Land, Infrastructure, and Transport (Korea Agency for Infrastructure Technology Advancement; 20UMRG-B158194-01)