Global sensitivity analysis of a modified CENTURY model for simulating impacts of harvesting fine woody biomass for bioenergy

Modeling the long-term effects of intensive forest biomass harvesting scenarios over time, including the more complete removal of tree tops and branches, is a scientific and policy need. Yet, due to our incomplete understanding about complex forest ecosystems, model simulations are to various degrees uncertain. In this study we first modified a well-evaluated and widely used ecosystem model - CENTURY 4.5 - to model management scenarios that retain various sizes and quantities of small-diameter woody material after intensive biomass harvests. Second, we used a global sensitivity analysis approach to evaluate the sensitivity of nine model outputs to 55 parameters, grouped into 17 factors. The values of the parameters were generated with a normal distribution and sampled with the extended Fourier amplitude sensitivity test. Our analysis indicated that within a harvest rotation, the model output sensitivity varied over years in response to different factors. The model was most sensitive to factors consisting of temperature effects on potential production as well as N deposition and non-symbiotic N fixation. In response to the uncertain parameter values, the model simulation revealed that outputs of net N mineralization rates in slow and passive soil organic matter pools had the highest uncertainties. However, due to the very low fraction of the N supplied from these two pools, forest production and other simulations were not strongly affected, ending with overall variations less than 6%. Ultimately, this study exhibits a novel approach in modeling the effects of harvesting fine woody debris for bioenergy on long-term ecosystem C and N cycles, and illustrates that sensitivity testing the most uncertain parameters is crucial for minimizing model uncertainty.