Simulation and Sensitivity for Biomass ofFour Main Forest Types in Zhejiang Using TRIPLEX Model

Inventory data of growth and yield from 147 sample plots aged from 5 to 50 years of four common forest types in 21 counties of Zhejiangprovince was used to test the process-based model of TRIPLEX. The four main forest types are evergreen broad-leaved forest (EF), coniferous andbroad-leaved mixed forest (MF), Pinus massoniana forest (PF) and Cunninghamia lanceolatae forest (CF). Simulation values by TRIPLEX modelwith stand density, diameter at breast height(DBH), height(H), litterfall pool, aboveground and total biomass were compared with measured ones. Theresults showed that significant correlation (p<0.001) between the simulated and measured values of forest stands for four main forest types inZhejiang province was found and the coefficients of determination (r2) was 0.92 for density, 0.77 for DBH, 0.80 for H, 0.53 for litterfall pool, 0.92 for aboveground biomass and 0.91 for total biomass, with small errors of -0.32 for DBH, -0.59 for H, 1.15 for litterfall pool, -0.51 for the aboveground biomass and -2.64 for the total biomass, except for stand density (53.56), and low biases (2.7% for density, -3.3% for DBH, -8.6% for H, -1.0% for the aboveground biomass and -1.5% for the total biomass) except for litterfall pool (16.8%). The simulated and measured values had significant correlation (p<0.001) for the aboveground and total biomass for each forest type, PF had highest coefficient of determination within 0.95 and 0.94 respectively, and CF had lowest coefficient of determination. For the four main forest types, Aboveground and total biomass of tested forest types had negative correlation with temperature except CF, and positive relation with relative humidity, and no correlation with precipitation. Simulated values of biomass and NPP of four forest types in Zhejiang by TRIPLEX model were much closer to the surveyed ones than those by the CEVSA and CASA models. These results suggest that TRIPLEX model did not decrease the predictive accuracy at the low demand for parameters, which can simulate and predict biomass with complex conditions in climate and soil.