3.1 Stability and productivity in planted and natural forest
Natural forests showed greater temporal stability in AGB than planted forests over the last four decades (T value=5.19, P<0.0001) (Fig. 2a). Despite the low diversity of planted forest (often with a single tree species), the AGB stocks of planted forest (0.028±0.027 Tg ha-1) was similar to that of natural forest (0.029±0.028 Tg ha-1); however, the productivity of planted forest was much higher compared with natural forest (T value=4.65, P<0.001) (Fig. 2b). AGB stability significantly decreased at a rate of -0.013 yr-1 (T value=-9.06, df=4893, P<0.0001) in natural forest and -0.011 yr-1 (T value=-3.53, df=546, P<0.0001) in planted forests (Fig. 3a). There was a significant decrease in productivity (T value=-4.76, df=4893, P<0.0001) in natural forest, with a slope of -0.0065 Mg ha-1 year-1 per calendar year; productivity in planted forest did not change significantly (Fig. 3b).
3.2 Environmental and forest structure predictors of productivity and AGB stability in planted and natural forests
In natural forest, annual precipitation (Z value=3.47, P=0.00053) and latitude (Z value=2.04, P=0.042) had a negative effect on AGB stability in the conditional average models after accounting for spatial autocorrelation (Fig. 4a). AGB stability was higher in natural forest plots with older stand age and higher altitude (Fig. 4a). However, richness (Z value=1.95, P=0.055) and stand age (Z value=1.71, P=0.087) were not significantly related to AGB stability in planted forest in the conditional average models (Fig. 4a). Based on the GAMM model, there was a significant non-linear relationship of tree density (F value=6.12, P <0.0001) and stand age (F value=8.64, P <0.0001) with stability in natural forest (Fig. S3 a and c). And, there was no significant association between richness and stability in natural forest (F value=0.97, P =0.34) and planted forest (F value=3.81, P =0.053) (Fig. S3b) in GAMM model. In addition, density (F value=0.26, P =0.61) and stand age (F value=2.80, P =0.11) did not significantly affect stability in planted forest (Fig. S3 a and c)
The conditional average models for natural forest productivity change included precipitation, stand age, latitude, soil depth, and tree density (Fig. 4b). Annual precipitation (Z value=3.73, P=0.0002), soil depth (Z value=5.51, P<0.0001), and latitude (Z value=5.68, P<0.0001) played a significant positive role in productivity in natural forest (Fig. 4b). Productivity decreased with Stand age was in natural forest (Z value=5.87, P<0.0001) (Fig. 4b). No significant effect of tree density on productivity in natural forest was observed (Z value=0.16, P=0.87) (Fig. 4b). The change in productivity in planted forest was most strongly affected by stand age, tree density, richness, and altitude. Productivity increased significantly with richness (Z value=4.42, P<0.0001) and tree density (Z value=3.38 P=0.00072) (Fig. 4b), and decreased significantly with stand age (Z value=3.71, P=0.00021) (Fig. 4b). Productivity of planted forests was higher at higher altitudes (Z value=2.25, P=0.025) (Fig. 3b). Based on the GAMM models, the productivity of natural (F value=25.95, P <0.0001) and planted forest (F value=4.49, P =0.011) increased logarithmically with tree density (Fig. S2 d) but displayed contrasting patterns with tree density. Productivity also increased near-logarithmically with stand age (F value=5.80, P <0.0001) in natural forest (Fig. S3 f), and stand age had negative and linear effects on the present-day productivity of planted forest (F value =-1.96, P =0.050). Richness played a positive role in productivity in planted forest (F value=17.72, P <0.0001) but not in natural forest (F value=0.64, P =0.43) (Fig. S3 e). In general, the results of the GAMM models and GLS models were consistent.