Differences in environmental factors
Environmental factors showed value differences between forest types, while the significance of differences differed among variables, which were both found with corrected values and original measurements (Fig. 1).
The differences in environmental factors between naturally regenerating forests (NF) and planted forests (PF) in China. The environmental factors include three annual climatic factors (a–c), three seasonal temperature factors (d–f), three seasonal precipitation factors (g–i), three biotic factors (j–l), and two soil factors (m,n). Three annual climatic factors include mean annual air temperature (MAT, a), mean annual precipitation (MAP, b), and aridity index (AI, c) defined as the ratio of MAP to annual potential evapotranspiration. Three seasonal temperature factors include the temperature of the warmest month (Tw, d), the temperature of the coldest month (Tc, e), temperature annual range (TR, f). Three seasonal precipitation factors include precipitation of the wettest month (Pw, g), precipitation of the driest month (Pd, h), and precipitation seasonality (Ps, i) defined as the standard deviation of monthly precipitation during the measuring year. Three biological factors include the mean annual leaf area index (LAI, j), the maximum leaf area index (MLAI, k), and stand age (SA, l). Two soil factors include soil organic carbon content (SOC, m) and soil total nitrogen content (STN, n). The differences are tested for each variable with one-way analysis of variance (ANOVA), where * and ** indicate significant differences between forest types at significance levels of α = 0.05 and α = 0.01, respectively. The corrected values are mean values during 2003–2019 after correcting the original measurements with the interannual trend (See methods), which are listed in each panel, while original measurements are mean values during the measuring period of each ecosystem, which are not shown in each panel.
For annual climatic factors, the significant difference between NF and PF only appeared in MAT (Fig. 1a). The mean MAT of NF was 10.50 ± 7.81 °C, which was significantly lower than that of PF (15.65 ± 6.23 °C) (p < 0.05). However, the MAP and aridity index (AI, defined as the ratio of MAP to annual potential evapotranspiration) showed no significant difference between forest types (Fig. 1b,c), although their values differed.
For seasonal climatic factors, only the temperature of the warmest month (Tw) showed a significant difference between NF and PF (p < 0.05) (Fig. 1d), while other factors, including the seasonal temperature factors and the seasonal precipitation factors, showed no significant differences between NF and PF (Fig. 1e–i).
However, the differences in biological factors showed divergent significance among factors (Fig. 1j–l). Mean annual leaf area index (LAI) showed no significant difference between NF and PF (Fig. 1j), while the maximum leaf area index (MLAI) and stand age (SA) significantly differed between forest types (Fig. 1k–l). The MLAI of NF (4.83 ± 1.49 m2 m−2) was significantly higher than that of PF (p < 0.05) (Fig. 1k). NF also had a significantly higher SA (116.72 ± 84.14 years) than PF (p < 0.01), which was 25.65 ± 28.48 years (Fig. 1l).
Soil factors showed significant differences between forest types, with a higher content appearing in NF (Fig. 1m,n). The soil organic carbon content (SOC) of NF was 1645.28 ± 949.18 gC m−2, which was significantly higher (p < 0.05) than that of PF (929.25 ± 686.67 gC m−2). The soil total nitrogen content (STN) of NF, which was 121.31 ± 70.60 gN m−2, was also significantly higher (p < 0.05) than that of PF (75.05 ± 49.49 gN m−2).
Differences in NEP values
Carbon flux values differed between forest types but their differences were not statistically significant, which were both true with corrected values and original measurements (Fig. 2).
The differences in carbon flux values between naturally regenerating forests (NF) and planted forests (PF) in China. The carbon fluxes include net ecosystem productivity (NEP, a), gross primary productivity (GPP, b), and ecosystem respiration (ER, c). The differences are tested for each flux with one-way analysis of variance (ANOVA) and analysis of covariance (ANCOVA) by fixing other variables as a covariant. All results show no significant difference between NF and PF (p > 0.05). The corrected values and original measurements are same to those in Fig. 1.
The mean NEP of NF was 326.08 ± 240.87 gC m−2 year−1, which was lower than that of PF (444.81 ± 308.76 gC m−2 year−1), while the results from one-way analysis of variance (ANOVA) indicated that NEP values did not significantly differ between forest types (F = 1.6, p > 0.05) (Fig. 2a). Even considering the significant effects of MAT on NEP, the results from analysis of covariance (ANCOVA) by fixing MAT as a covariant also suggested that no significant difference was found in NEP values between forest types (F = 2.09, p > 0.05). Fixing other variables as a covariant also found no significant differences in NEP values between NF and PF.
The mean GPP value of NF was 1448.41 ± 471.13 gC m−2 year−1, which was slightly lower than that of PF (1500.68 ± 582.84 gC m−2 year−1) (Fig. 2b), while ANOVA results indicated that the GPP value of NF showed no significant difference from that of PF (F = 0.07, p > 0.05) (Fig. 2b). Even considering the significant effects of MAT on GPP, ANCOVA results obtained by fixing MAT as a covariant also suggested that GPP values did not significantly differ between forest types (F = 1.52, p > 0.05). Fixing other variables as a covariant also drew a similar result.
The mean ER value of NF (1087.08 ± 473.42 gC m−2 year−1) was slightly lower than that of PF (1095.77 ± 349.83 gC m−2 year−1). ANOVA results also indicated that ER values did not significantly differ between forest types (F = 0.00, p > 0.05) (Fig. 2c). Even considering the significant effects of MAT on ER, ANCOVA results obtained by fixing MAT as a covariant also suggested that ER values did not significantly differ between forest types (F = 0.01, p > 0.05). Fixing other variables as a covariant also drew a similar result.
Therefore, NF showed a lower NEP resulting from the lower GPP than PF, while their differences were not statistically significant (Fig. 2).
Differences in NEP latitudinal patterns
Carbon fluxes showed divergent latitudinal patterns between NF and PF, while their latitudinal patterns varied among carbon fluxes, which were both found with corrected values and original measurements (Fig. 3).
The latitudinal patterns of carbon fluxes over Chinese naturally regenerating forests (NF) and planted forests (PF). The carbon fluxes include net ecosystem productivity (NEP, a,b), gross primary productivity (GPP, c,d), and ecosystem respiration (ER, e,f). Each panel is drawn with the corrected values (blue points) and original measurements (grey points), respectively. The blue and black lines represent the regression lines calculated from the corrected values and original measurements, respectively, with their regression statistics listed in blue and black letters. Only the regression slope (Sl) and R2 of each regression are listed. The grey lines represent the regressions between carbon fluxes added by random errors and latitude. Only significant (p < 0.05) regression lines are drew in each panel.
NEP showed no significant latitudinal pattern among NFs (Fig. 3a), while that of PF exhibited a significant decreasing latitudinal pattern (Fig. 3b). With increasing latitude, the NEP of NF showed no significant spatial variation. However, the increasing latitude caused the NEP of PF to significantly decrease. Each unit increase in latitude led to a 23.69 gC m−2 year−1 decrease in NEP, with an R2 of 0.31.
The GPP of NF and PF both exhibited significant decreasing latitudinal patterns with similar decreasing rates (Fig. 3c,d). With increasing latitude, the GPP of NF significantly decreased. Each unit increase in latitude led to a 37.75 gC m−2 year−1 decrease in GPP, with an R2 of 0.54. The increasing latitude significantly decreased the GPP of PF. Each unit increase in latitude decreased the GPP of PF at a rate of 53.68 gC m−2 year−1, with an R2 of 0.43. Although the decreasing rates differed in values between NF and PF, their differences were not statistically significant (F = 0.71, p > 0.05).
The ER of NF showed a significant decreasing latitudinal pattern (Fig. 3e), while that of PF exhibited no significant latitudinal pattern (Fig. 3f). The increasing latitude caused the ER of NF to significantly decrease. Each unit increase in latitude led to a 28.71 gC m−2 year−1 decrease in ER, with an R2 of 0.31. However, the increasing latitude contributed little to the ER spatial variation of PF (p > 0.05).
In addition, the latitudinal patterns of carbon fluxes and their differences between forest types were also obtained with the original measurements (Fig. 3, grey points). The latitudinal patterns of random error adding carbon fluxes were comparable to those of our corrected carbon fluxes (Fig. 3), which confirmed that the latitudinal patterns of carbon fluxes and their differences between forest types would not be affected by the uncertainties in generating the corrected carbon fluxes.
Therefore, among NFs, the similar decreasing latitudinal patterns of GPP and ER meant that NEP showed no significant latitudinal pattern, while the significant decreasing latitudinal pattern of GPP and no significant latitudinal pattern of ER caused NEP to show a decreasing latitudinal pattern among PFs.
Differences in the environmental effects on NEP spatial variations
Environmental factors, including the annual climatic factors, seasonal temperature factors, seasonal precipitation factors, biological factors, and soil factors, exerted divergent effects on the spatial variations of NEP and its components, which also differed between forest types (Table 1). No factor was found to affect that the spatial variation of NEP among NFs, while most annual and seasonal climatic factors were found to affect that among PFs. The spatial variations of GPP and ER among NFs were both affected by most annual and seasonal climatic factors and LAI, while those among PFs were primarily shaped by most annual and seasonal climatic factors. Though LAI showed no significant effect on GPP and ER spatial variations among PFs, SA exerted a significant negative effect. In addition, the spatial variations of soil variables contributed little to the spatial variations of carbon fluxes. Therefore, among NFs, most annual and seasonal climatic factors and LAI were found to affect GPP and ER spatial variations, while no factor was found to significantly influent the NEP spatial variation. However, among PFs, most annual and seasonal climatic factors were found to affect the spatial variations of NEP and its components, while LAI showed no significant effect. Using the original measurements also generated the similar correlation coefficients (Supplementary Table S1).
Given the high correlations among annual climatic factors and seasonal climatic factors (Supplementary Table S2), the partial correlation analysis was applied to determine which factors should be employed to reveal the mechanisms underlying the spatial variations of NEP. Partial correlation analysis showed that MAT and MAP exerted the most important roles in spatial variations of NEP and its components (Table 2). After controlling MAT (or MAP), other factors seldom showed significant correlation with carbon fluxes, especially fixing MAT (Table 2). In addition, MAT and MAP exerted similar effects on the spatial variations of NEP and its components (Table 1). Using the original measurements also generated the similar partial correlation coefficients (Supplementary Table S3). Therefore, we only presented the effects of MAT on carbon flux spatial variations and their differences between forest types in detail.
The increasing MAT increased carbon fluxes, while the increasing rates differed between forest types (Fig. 4). The increasing MAT contributed little to the NEP spatial variation of NF but raised the NEP of PF (Fig. 4a,b). Each unit increase in MAT caused the NEP of PF to increase at a rate of 27.77 gC m−2 year−1, with an R2 of 0.31 (Fig. 4b). The increasing MAT significantly raised GPP in NF and PF (Fig. 4c,d). For NF, each unit increase in MAT increased GPP at a rate of 43.76 gC m−2 year−1, with an R2 of 0.49 (Fig. 4c), while each unit increase in MAT increased the GPP of PF at a rate of 69.18 gC m−2 year−1, with an R2 of 0.57 (Fig. 4d). The GPP increasing rates did not significantly differ between NF and PF (F = 1.52, p > 0.05). The increasing MAT also raised ER in both NF and PF (Fig. 4e,f), whose increasing rates were 38.97 gC m−2 year−1 (Fig. 4e) and 36.79 gC m−2 year−1 (Fig. 4f), respectively, while their differences were not statistically significant (F = 0.01, p > 0.05). In addition, using the original measurements also generated the similar spatial variations and their differences between forest types (Fig. 4). Furthermore, the random error adding carbon fluxes responded similarly to those of our correcting carbon fluxes (Fig. 4), indicating that the effects of MAT on carbon fluxes would not be affected by the uncertainties in our correcting carbon fluxes. Therefore, the similar responses of GPP and ER to MAT made MAT contribute little to NEP spatial variations among NFs, while GPP and ER showed divergent response rates to MAT, which made NEP increase with MAT among PFs.
The effects of mean annual air temperature (MAT) on the spatial variations of carbon fluxes over Chinese naturally regenerating forests (NF) and planted forests (PF). The carbon fluxes include net ecosystem productivity (NEP, a,b), gross primary productivity (GPP, c,d), and ecosystem respiration (ER, e,f). Each panel is drawn with the corrected values (blue points) and original measurements (grey points), respectively. The blue and black lines represent the regression lines calculated from the corrected values and original measurements, respectively, with their regression statistics listed in blue and black letters. Only the regression slope (Sl) and R2 of each regression are listed. The grey lines represent the regressions between carbon fluxes added by random errors and latitude. Only significant (p < 0.05) regression lines are drew in each panel.
Differences in the spatial couplings between carbon fluxes
The spatial couplings between carbon fluxes differed between NF and PF (Fig. 5).
The spatial couplings between carbon fluxes over Chinese naturally regenerating forests (NF) and planted forests (PF). The spatial couplings include the coupling between gross primary productivity (GPP) and ecosystem respiration (ER) (a,b) and that between GPP and net ecosystem productivity (NEP) (c,d). Each panel is drawn with the corrected values (blue points) and original measurements (pink points), respectively. The blue and black lines represent the regression lines calculated from the corrected values and original measurements, respectively, with their regression statistics listed in blue and black letters. Only the regression slope (Sl) and R2 of each regression are listed. The grey lines represent the regressions between carbon fluxes added by random errors and latitude. Only significant (p < 0.05) regression lines are drew in each panel.
GPP and ER were spatially coupled in both NF and PF, while the responses of ER to GPP differed between NF and PF (Fig. 5a,b). With increasing GPP, the ER of NF linearly increased at a rate of 0.86 gC m−2 year−1, with an R2 of 0.73 (Fig. 5a). The increasing GPP also caused the ER of NF to linearly increase, with an increasing rate of 0.53 gC m−2 year−1 and an R2 of 0.78 (Fig. 5b). The ER increasing rate of NF was significantly higher than that of PF (F = 5.78, p < 0.05).
However, GPP was only spatially coupled with NEP in PF (Fig. 5c,d). The increasing GPP contributed little to the NEP spatial variation of NF, whereas the increasing GPP caused NEP to linearly increase at a rate of 0.47 gC m−2 year−1 in PF, with an R2 of 0.68 (Fig. 5d).
In addition, the spatial couplings of carbon fluxes and their differences between forest types were also obtained with the original measurements (Fig. 5, pink symbols) and the error adding carbon fluxes (Fig. 5).
Comprehensive understanding on differences in NEP spatial patterns
Based on NEP latitudinal patterns, effects of environmental factors on NEP spatial variations, and the couplings between carbon fluxes, we found that NEP spatial patterns sourced from the spatial patterns of GPP and ER, which resulted from the increasing MAT induced by the decreasing latitude (Figs. 3, 4, 5).
The increasing latitude induced a decreasing MAT (Supplementary Table S2), which showed no significant difference between NF and PF. With increasing latitude, MAT showed a significant decreasing trend at a rate of 0.70 °C, with an R2 of 0.68 (Eq. (1)).
$${text{MAT}} = 35.77 – 0.70;{text{Latitude}},;{text{R}}^{2} = 0.68,;{text{n}} = 35,;{text{p}} < 0.01$$
(1)
The increasing MAT exerted a positive effect on the spatial variation of GPP (Fig. 4), which also showed no significant difference between NF and PF. The increasing MAT raised GPP at a rate of 52.37 gC m−2 year−1, with an R2 of 0.48 (Eq. (2)).
$${text{GPP}} = 779.6 + 52.37;{text{MAT}},;{text{R}}^{2} = 0.48,;{text{n}} = 30,;{text{p}} < 0.01$$
(2)
The increasing GPP exerted a positive effect on ER spatial variation, while its effects significantly differed between NF and PF (Fig. 5). ER spatial variation was jointly shaped by the forest type, GPP, and their interaction, with an R2 of 0.72 (Eq. (3)). NF showed a lower intercept and a higher increasing rate, while PF had a higher intercept and a lower increasing rate.
$${text{ER}} = 302.03 – 460.31;{text{Forest}} + 0.53;{text{GPP}} + 0.33;{text{Forest}} times {text{GPP}},;{text{R}}^{2} = 0.72,;{text{n}} = 30,;{text{p}} < 0.01$$
(3)
where Forest represents the forest type with NF of 1 and PF of 0, while Forest × GPP reflects the interaction between forest type and GPP, which is set to GPP at NF and 0 at PF.
With the joint effects of GPP and ER, NEP spatial variation was also primarily shaped by the spatial variation of GPP, while the effects of GPP significantly differed between NF and PF (Fig. 5). NEP spatial variation was jointly shaped by the forest type, GPP, and their interaction, with an R2 of 0.42 (Eq. (4)).
$${text{NEP}} = – 283.42 + 417.52{text{Forest}} + 0.47{text{GPP}} – 0.33{text{Forest}} times {text{GPP}},;{text{R}}^{2} = 0.42,;{text{n}} = 30,;{text{p}} < 0.01$$
(4)
Using the original measurements, we also got the similar equations (Supplementary Text S1).
Therefore, the increasing latitude decreased MAT thus GPP, which did not significantly differ between NF and PF. However, the differences in SA and soil variables made the portion of GPP allocated to ER thus NEP significantly differ, which made NEP show divergent spatial patterns between NF and PF.
Source: Ecology - nature.com
