Sensitivity Analysis of Biome-BGCMuSo for Gross and Net Primary Productivity of Typical Forests in China

15 Background: Process-based models are widely used to simulate forest productivity, but complex 16 parameterization and calibration challenge the application and development of these models. Sensitivity 17 analysis of numerous parameters is an essential step in model calibration and carbon flux simulation. 18 However, parameters are not dependent on each other, and the results of sensitivity analysis usually vary 19 due to different forest types and regions. Hence, global and representative sensitivity analysis would provide reliable information for simple calibration. Methods: To determine the contributions of input parameters to gross primary productivity (GPP) and net primary productivity (NPP), regression analysis and extended Fourier amplitude sensitivity testing 23 (EFAST) were conducted for Biome-BGCMuSo to calculate the sensitivity index of the parameters at four observation sites under climate gradient from ChinaFLUX. 25 Results: Generally, GPP and NPP were highly sensitive to C:N leaf (C:N of leaves), W int (canopy water 26 interception coefficient), k (canopy light extinction coefficient), FLNR (fraction of leaf N in Rubisco), 27 MR pern (maintenance respiration in kg C/day per kg of tissue N), VPD f (vapor pressure deficit complete 28 conductance reduction), and SLA1 (canopy average specific leaf area in phenological phase 1) at all 29 observation sites. Various sensitive parameters occurred at four observation sites within different climate 30 zones. GPP and NPP were particularly sensitive to FLNR, SLA1 and W int , and C:N leaf in temperate, 31 alpine and subtropical zones, respectively. 32 Conclusions: The results indicated that sensitivity parameters of China's forest ecosystems change with 33 climate gradient. We found that parameter calibration should be performed according to plant functional 34 type (PFT), and more attention needs to be paid to the differences in climate and environment. These 35 findings contribute to determining the target parameters in field experiments and model calibration.


102
The daily GPP and NPP values were calculated by inverse operation of GEE and NEE. to September) during the plant growth period is 437.5 mm, accounting for 82% of the annual precipitation 124 (Li et al. 2019

176
The model simulation is divided into two stages. The first is spinup simulation, which starts with very 177 low initial levels of soil C and N and runs until it reaches a steady state under a given climate to estimate             Seven key sensitive parameters for GPP and NPP were summarized from the two sensitivity analysis 335 methods: C:Nleaf, Wint, k, FLNR, MRpern, VPDf, and SLA1 (Figure 7 and Figure 8). The sensitive 336 parameters selected by regression analysis were similar to those from EFAST, and the key sensitive 337 parameters differed significantly under various climate environments and PFTs. FLNR was the key 338 sensitive parameter in temperate zone, and C:Nleaf was the key sensitive parameter in subtropical zones.

339
VPDf was the key sensitive parameter for DBF, and GR was the key sensitive parameter for ENF at the 340 CBS site. At the DHS site, C:Nleaf and k were the key sensitive parameters of EBF, and C:Nfr was the key 341 sensitivity parameter of ENF. In addition, C:Nfr was the sensitive parameter only for ENF-type GPP at the 342 DHS site, while GR was the sensitive parameter only for ENF-type NPP at the CBS site. The key sensitive parameters extracted according to different sensitivity analysis methods have a high 352 degree of similarity. It is therefore necessary to perform a series of steps for key sensitivity parameters to 353 simplify the process from sensitivity analysis to parameter correction. Here, we refer to these steps as the 354 "workflow". The workflow for the application of key parameters is shown in Figure 9. The colored 355 parameters should be of importance in sensitivity analysis and modeling. We described and discussed the 356 workflow and the choices that the user must make in each step to provide practical guidelines to support 357 the user in the sensitivity analysis-based simulation correction of the parameters.

364
The results described in sections 3.2 and 3.3 indicate that the sensitivity of the climate environment 365 to productivity simulation is greater than the difference in PFTs. Specifically, the most common sensitivity 366 parameter of different PFTs in the temperate zone was FLNR, followed by MRpern. In the alpine zone, the 367 common sensitive parameters of different PFTs were SLA1 and Wint. In addition, C:Nleaf was the key 368 sensitive parameter for different PFTs in the subtropical zone, followed by MRpern. Next was the selection 369 of key sensitive parameters for different PFTs. In the temperate zone, DBF was sensitive to VPDf, and 370 ENF was sensitive to GR. In the subtropical zone, EBF was sensitive to k, and ENF was sensitive to C:Nfr.

371
In addition, GR and C:Nfr were the key sensitive parameters of NPP and GPP, respectively, excluding the 372 common key sensitive parameters of NPP and GPP. As the importance of the parameters gradually 373 decreases with the deepening of the workflow, the disturbance effect of the following parameters on the 374 simulation results gradually decreases. In the application workflow for key parameters, if some key 375 sensitive parameters (such as GR and C:Nfr) have been added before, the corresponding steps can be 376 omitted in the following process.

377
The uncertainty discussed in Section 3.1 is intended to assess the uncertainty of the results of a

416
The consistency of the parameter range is guaranteed to a certain extent.