Variance-based sensitivity analysis of BIOME-BGC for gross and net primary production

•We identified the uncertainty in BIOME-BGC (a process-based simulator) input parameters in the form of probability distribution functions.•We identified the uncertainty in BIOME-BGC simulated gross and net primary production.•We reported the significant control of the input parameter FRC:LC (fine root carbon to leaf carbon) on simulated gross and net primary production. Parameterization and calibration of a process-based simulator (PBS) is a major challenge when simulating gross and net primary production (GPP and NPP). The large number of parameters makes the calibration computationally expensive and is complicated by the dependence of several parameters on other parameters. Calibration can be simplified by first identifying those parameters for which GPP and NPP are most sensitive. For an appropriate application of a PBS, a sensitivity analysis is an essential step. Sensitivity analysis based on local derivatives (i.e., one-at-a-time analysis) does not examine the PBS behaviour over the whole parameter space. This study therefore implements a variance-based sensitivity analysis (VBSA) addressing the full range of PBS input. A VBSA is also independent of non-linearity in a PBS. This paper performs a VBSA of the process-based simulator BIOME-BGC for GPP and NPP output in a Douglas-fir stand at the Speulderbos forest site, The Netherlands. The results show that GPP and NPP are highly sensitive to the following parameters: fraction of leaf nitrogen in Rubisco, the ratio of fine root carbon to leaf carbon, the ratio of carbon to nitrogen in leaf and fine root, the leaf and fine root turnover, the water interception coefficient and soil depth. GPP and NPP are particularly sensitive to the ratio of fine root carbon to leaf carbon that is responsible for leaf area index development. The study concludes that a VBSA analysis provides a reliable and useful approach for a sensitivity analysis of process-based simulators with a complicated structure in the parameters.