Simulation Results of RECO. In this study, we chose RECO observations from the FLUNET 2015 dataset (Pastorello et al, 2020) as a standard, combining meteorological, remote sensing, and soil data, and used a convolutional neural network (CNN) to simulate RECO. According to different input parameters, six CNNs were established. During the experiment, each CNN model reached a stable state after 100,000 iterations with high accuracies in both the training and testing phases, indicating that the model did not overfit during the training process (Table 1).
The experimental results showed that the training accuracy R² of CNN-ORIGIN (CNN1) was 0.979, CC was 0.988, RMSE was 0.098, and SDE was 0.093. When introducing SPEI at different time scales, respectively, it was found that the introduction of SPEI12 (December SPEI on 12-month time scale) did not improve the performance of the model, i.e., the simulation accuracy of CNN2 was not improved compared to that of CNN1. However, SPEI24 (December SPEI on 24-month time scale) and SPEI36 (December SPEI on 36-month time scale) improved the model significantly. The training accuracy of CNN3 was 0.986 for R², 0.991 for CC, 0.085 for RMSE, and 0.074 for SDE; the training accuracy of CNN4 was 0.987 for R², 0.992 for CC, 0.078 for RMSE, and 0.072 for SDE, all of which were improved compared to that of CNN1. This shows that there is no significant effect of water conditions in the current year on RECO, while the effect of water conditions in the previous 1–2 years is more obvious, indicating that there is a significant lagged effect on the response of RECO to water condition changes, with a time scale of 1–2 years. The effect of the model was optimized when SPEI12 and SPEI24 were added jointly, i.e., CNN5—which had the highest accuracy and included the water conditions of the current year and the previous year and could reflect the lagged effect of RECO on water condition changes. Therefore, in this study, CNN5 was taken as the optimal model CNN-SPEIopt, and the results of this model were subject to subsequent analysis.
Table 1 CNN training results.
CNN Model
|
|
CNN1
|
CNN2
|
CNN3
|
CNN4
|
CNN5
|
CNN6
|
n
|
Input
|
|
Basic Data
|
Basic Data, SPEI12
|
Basic Data, SPEI24
|
Basic Data, SPEI36
|
Basic Data,
SPEI12, SPEI24
|
Basic Data, SPEI12, SPEI24, SPEI36
|
|
Training
|
R²
|
0.979
|
0.972
|
0.986
|
0.987
|
0.990
|
0.983
|
1000
|
|
CC
|
0.988
|
0.984
|
0.991
|
0.992
|
0.994
|
0.987
|
|
|
RMSE
|
0.098
|
0.113
|
0.085
|
0.078
|
0.068
|
0.103
|
|
|
SDE
|
0.093
|
0.105
|
0.074
|
0.072
|
0.064
|
0.083
|
|
Testing
|
R²
|
0.979
|
0.974
|
0.986
|
0.988
|
0.989
|
0.982
|
300
|
|
CC
|
0.988
|
0.985
|
0.991
|
0.993
|
0.994
|
0.986
|
|
|
RMSE
|
0.096
|
0.114
|
0.084
|
0.079
|
0.070
|
0.103
|
|
|
SDE
|
0.091
|
0.106
|
0.074
|
0.072
|
0.066
|
0.085
|
|
Note: Basic data contains: MAT (mean annual temperature), MAP (mean annual precipitation), PAR (photosynthetically active radiation), FAPAR (fraction of absorbed photosynthetically active radiation), LAL (leaf area index), NDVI (normalized difference vegetation index), NPP and SOC (soil organic carbon) from 0 to 30 cm; R²: coefficient of determination, CC: Lin's concordance correlation coefficient, RMSE: root mean square error, SDE: standard deviation of error; n: number of samples.
The simulation results of CNN-SPEIopt show that the mean value of global multi-year average RECO from 2000 to 2018 is about 0.87 kg·C·m-2·y-1, and the maximum value is about 3.75 kg·C·m-2·y-1. The high-value areas are mainly distributed in tropical rainforest areas, followed by mid-latitude and mid-high latitude forest-covered areas. The lowest RECO values are found in areas covered with grass and shrub and high latitude areas in the northern hemisphere. Fig. 1(b) shows the global spatial distribution of the standard deviation (STDEV) of RECO from 2000 to 2018, which reflects the magnitude of interannual fluctuations of RECO. The mean value of global RECO STDEV is about 0.22 kg·C·m-2·y-1, and the maximum value is 1.87 kg·C·m-2·y-1.
To validate the CNN-SPEIopt simulation results, we compared them with six commonly used process models. The mean value of RECO simulated by CNN-SPEIopt is close to most models, but its standard deviation is much higher than others. It can be speculated that although the traditional process models can simulate RECO, the interannual fluctuations of RECO may be significantly underestimated by about 71% to 88%.
To reveal whether these models can accurately reflect the interannual fluctuations of RECO, this study collected 173 observations from 30 sites of AmeriFlux (https://ameriflux.lbl.gov) and ChinaFlux (http://www.cnern.org.cn) to validate each model. First, by comparing the 173 observations with the simulated values of the models, we found that CNN-SPEIopt had the highest accuracy and the smallest simulation error, with an R² of 0.535 and an RMSE of 0.444. While the R² of the other models ranged from 0.002 to 0.468 and an RMSE from 0.475 to 0.856 (Fig. 2). Second, we calculated the standard deviation of the multi-year observations for each site and the standard deviation of the models’ simulations and conducted a correlation analysis between the standard deviation of the observations and the standard deviation of the simulations (Fig. 3). The results show that the Pearson correlation coefficient between CNN-SPEIopt and site observations is 0.475, which is significantly correlated at the p < 0.01 level. Moreover, the Pearson correlation coefficients between other process models and site observations are very low and some models even show a negative correlation. This indicates that among these models, only CNN-SPEIopt can capture the interannual fluctuations of RECO because it is more consistent with the observed interannual fluctuations of RECO. The standard deviations of the other models’ simulations are generally low and inconsistent with the site observations, indicating that these models would substantially underestimate the interannual fluctuations of RECO.
Spatial Patterns and Changing Trends of RECO and NPP. The CNN-SPEIopt shows that the global average RECO has a significant increasing trend (p < 0.01) with climate change. Spatially, 54.92% of the global RECO shows an increasing trend, while 45.08% shows a decreasing trend. This indicates that the global RECO is dominated by growth.
The global average NPP also shows a significant increasing trend (p < 0.01), but the growth rate of NPP (slope of 0.154) is higher than that of RECO (slope of 0.142). In terms of spatial distribution, NPP shows a significant increasing trend in most of the global regions, except for the Amazon region. Globally, 73.35% of the NPP shows an increasing trend and 26.65% shows a decreasing trend. This indicates that although the growth trends of global average RECO and NPP are similar, the proportion of increasing NPP is higher than that of increasing RECO, and the carbon sink potential of the global terrestrial ecosystem generally shows an increasing trend.
The land area of the northern extra-tropics (30°N~90°N) accounts for about 60% of the world, with extensive forest cover and high carbon sink potential. On the one hand, influenced by the warming and humidification of climate and the effect of CO2 fertilization, the growth period of boreal vegetation is prolonged and vegetation productivity is increasing (Bellassen and Luyssaert, 2014), which enhances terrestrial carbon sink capacity. On the other hand, the boreal permafrost contains a large amount of soil carbon, and global warming will accelerate the permafrost melt and release CO2, which also risks enhancing terrestrial carbon emissions (Koven et al., 2011). The CNN-SPEIopt results show that the average RECO in the northern extra-tropics is significantly increased (p < 0.01). The spatial distribution of RECO is dominated by growth, with 55.69% of RECO showing an increasing trend and 44.31% shows a decreasing trend.
Meanwhile, MODIS NPP also shows a significant increasing trend (p < 0.01) in the northern extra-tropics, however, the growth rate of NPP (slope of 0.152) is much higher than that of RECO (slope of 0.107). In terms of spatial distribution, 80.12% of NPP is increasing, while 19.88% of NPP is decreasing. The increasing proportion of NPP is much larger than that of RECO. In other words, the NPP in the northern extra-tropics shows a substantial increase, while RECO does not increase as significantly. The carbon sink potential of this region has a gradual increase trend dominated by the growth of NPP.
The tropics (30°N~30°S), which covers about 36% of the global land area, contains the largest tropical rainforests and the highest vegetation productivity in the world, as well as the largest carbon sink. CNN-SPEIopt shows that RECO in this region has a significant increasing trend (p < 0.01). In terms of spatial distribution, 53.53% of RECO shows an increasing trend, while 46.47% shows a decreasing trend.
NPP in the tropics also shows a significant increasing trend (p < 0.01) and the growth rate of NPP is close to that of RECO (slope of NPP is 0.122 and RECO is 0.114). Spatially, 62.05% of NPP shows an increasing trend and 37.95% shows a decreasing trend. The percentage of increasing RECO in the tropics is close to that in the northern extra-tropics, however, the percentage of increasing NPP is much lower than that in the northern extra-tropics. Although the increasing area of NPP in the tropics is larger than that of RECO, the excess area is less than 10%. The advantage of carbon sink growth in the tropics is weaker than that in the northern extra-tropics where the area of increasing NPP is about 25% more than that of RECO.
The southern extra-tropics (30°S to 90°S) has a small land area covered by vegetation, accounting for only about 4% of the globe except for the Antarctic continent. The RECO in this region still shows an increasing trend, but with a lower significance (p < 0.1) than other regions. Spatially, 56.28% of RECO shows an increasing trend, while 43.72% of RECO shows a decreasing trend.
The NPP in the southern extra-tropics shows an increasing trend, with a slightly lower significance (p < 0.05) than other regions. Spatially, 71.90% of NPP shows an increasing trend, while 28.10% shows a decreasing trend. The percentage of increasing NPP in the southern extra-tropics is larger than that of RECO—about 15% more. However, because of the small land area of this region, it has less impact on the global carbon sink.
In addition, we calculated the proportions of spatial grids where RECO and NPP changed in the same and opposite directions both globally and regionally (Fig. 5c and Table 3). The areas where NPP increased indicated that the vegetation in these areas is in good condition. The areas where NPP increased and RECO decreased are obvious carbon sinks with the greatest carbon sink capacity. The areas where NPP decreased implied vegetation death or reduced productivity. A decrease in NPP and an increase in RECO means that these areas are at risk of becoming carbon sources. Presently, the results show that the global carbon sink potential of terrestrial ecosystems is increasing, as nearly half of the areas where NPP and RECO are increasing simultaneously (41.08%) and the proportion of NPP increasing and RECO decreasing is also higher (32.28%). The proportions of the other two scenarios are lower, 13.91% are NPP decreasing and RECO increasing, and 12.73% are both NPP and RECO decreasing.
The northern extra-tropics has 44.87% of areas where both NPP and RECO increased and 35.57% of areas where NPP increased and RECO decreased. Both are lower in the tropics, with 35.06% of the area where both NPP and RECO increased and 27.27% of the area where NPP increased and RECO decreased. The proportion of area with increasing carbon sink potential is much higher in the northern extra-tropics than in the tropics.
On the contrary, in the northern extra-tropics, 10.93% of areas have decreasing NPP and increasing RECO, and 8.62% has a decrease in both. These regions are mainly concentrated on the west coast of the United States, eastern North America, central Europe, and central and southwestern Russia. In the tropics, 18.49% of areas have decreasing NPP and increasing RECO, and 19.17% of areas have a decrease in both. These two are the highest in the tropics. This means that the regions with the highest carbon source risk are most concentrated in the tropics, mainly in the southern tip of North America, Amazonia, eastern Brazil, central Africa, eastern India, southeastern China, and Southeast Asia.
In summary, the growth trends of global RECO and NPP show imbalances. First, there is an imbalance between the growth rate of RECO and NPP; moreover, the growth rate of NPP is larger than that of RECO, and the proportion of NPP increasing area is also higher than that of RECO, which contributes to maintaining the carbon sink of the global terrestrial ecosystem in an increasing trend. Second, the carbon sink potential of different regions is unbalanced. In the northern extra-tropics, the growth rate of NPP is significantly larger than that of RECO, and the proportion of NPP increasing area is also the highest. Furthermore, in the tropics, the growth rates of RECO and NPP are markedly close, and the proportion of NPP increasing area is much lower than that in the northern extra-tropics.
Table 2 Increasing or decreasing proportions of global and regional RECO and NPP.
|
|
Global
|
Northern Extra-tropics
|
Tropics
|
Southern Extra-tropics
|
|
|
Increase
|
Decrease
|
Increase
|
Decrease
|
Increase
|
Decrease
|
Increase
|
Decrease
|
RECO
|
|
54.92%
|
45.08%
|
55.69%
|
44.31%
|
53.53%
|
46.47%
|
56.28%
|
43.72%
|
|
p < 0.1
|
13.01%
|
8.37%
|
13.12%
|
7.88%
|
12.72%
|
8.99%
|
13.90%
|
9.60%
|
NPP
|
|
73.35%
|
26.65%
|
80.12%
|
19.88%
|
62.05%
|
37.95%
|
71.90%
|
28.10%
|
|
p < 0.1
|
30.13%
|
6.06%
|
30.88%
|
2.23%
|
29.93%
|
12.74%
|
21.38%
|
4.49%
|
Table 3 Proportions of global and regional RECO and NPP in the same and opposite change directions.
|
Global
|
Northern Extra-tropics
|
Tropics
|
Southern Extra-tropics
|
NPP+RECO+
|
41.08%
|
44.87%
|
35.06%
|
41.34%
|
NPP+RECO-
|
32.28%
|
35.57%
|
27.27%
|
30.56%
|
NPP-RECO+
|
13.91%
|
10.93%
|
18.49%
|
14.94%
|
NPP-RECO-
|
12.73%
|
8.62%
|
19.17%
|
13.15%
|
Global and Regional Carbon Sink Capacities and Potential Risk Assessment. In this study, the difference between the NPP growth rate and RECO growth rate at each spatial grids was calculated (Fig. 6) to quantitatively assess the global and regional capacities and potential risk of carbon sink. The areas dominated by NPP growth are carbon sink potential areas (the area where NPP grows faster is larger than the area where RECO grows faster), and the areas dominated by RECO growth are carbon source risk areas (the area where NPP grows faster is smaller than the area where RECO grows faster). As shown in Fig. 6, 65.59% of the global areas have a faster NPP growth rate, and 34.41% are areas with a faster RECO growth. This indicates that the global terrestrial ecosystem has a certain capacity of carbon sink dominated by NPP growth.
In the northern extra-tropics, 69.33% of the areas have a faster NPP growth rate and 30.67% of the areas have a faster RECO growth rate. The northern extra-tropics is clearly dominated by NPP growth, and the capacity of the carbon sink is gradually increasing. The regions where the NPP growth rate is significantly higher than that of RECO are concentrated in northeastern China and most regions of Europe and North America. The regions where the RECO growth rate is significantly higher than the NPP growth rate are concentrated in central and southwestern Russia and central and eastern North America. The vegetation distributed in the northern extra-tropics is extensive and rich in types, such as deciduous broadleaf forest (DBF), evergreen coniferous forest (ENF), grassland (GRA), and savanna (SAV), all of which have high carbon sink capacities (Fig. 7a), implying that the northern extra-tropics has substantial carbon sink capacity.
In the tropics, the distribution of carbon sink potential areas is considerably uneven. On the one hand, in the Amazon and African rainforest areas, the growth rate of RECO is significantly higher than that of NPP, which has a greater risk of carbon source. In addition, the RECO growth rate is also higher in eastern India, southern China, and Southeast Asia. On the other hand, central South America, central Africa, some regions of China, and India also have obvious carbon sink areas, i.e., the NPP growth rate is higher than that of RECO. Specifically, 59.67% of the tropical areas are experiencing faster growth rates of NPP and 40.33% are experiencing faster growth rates of RECO. Although the tropics are still dominated by NPP growth, the dominance is less pronounced than that of the northern extra-tropics. The carbon cycle balance in the tropics has been severely disturbed in recent years, mainly due to the destruction of tropical rainforests and the fact that their carbon sink capacity has become saturated. Fig. 7a shows that the evergreen broadleaf forest (EBF), which is widely distributed in the tropics, has a smaller carbon sink capacity and the strongest capacity to release carbon by respiration among the many vegetation types. Several high-risk carbon source regions such as Amazonia, central African, and Southeast Asia, all of which are widely distributed in tropical forests, now have significantly reduced carbon sink capacity. This could cause a break in the global terrestrial ecosystem carbon cycle and threaten the entire planetary ecosystem.
In the southern extra-tropics, 62.19% have a faster NPP growth rate and 37.81% have a faster RECO growth rate. The carbon sink capacity is slightly more optimistic than that of the tropics. However, this region has less vegetation area, mainly grassland and shrubs, and contributes less to the global terrestrial ecosystem.
In addition, this study assessed the capacities of several major economies regarding their contributions to the global terrestrial ecosystem carbon sink, including the United States, China, the European region (except Russia), Brazil, and India (Fig. 7b). Among these countries and regions, China has the largest carbon sink capacity, followed by the U.S. and the European region, and Brazil and India have smaller carbon sinks. The United States and the European region have high forest cover and rich vegetation carbon stocks. However, as forests in these two countries are generally older and most ecosystems are highly mature, the growth potential of carbon sink is not prominent and is in a more stable state. Brazil, a country with a large tropical rainforest area, is no longer outstanding for its carbon sink, which is caused by the decrease in tropical rainforest area and the saturation of carbon sink capacity. Most notably, China and India both have large-scale greening areas subject to anthropogenic regulation, and together dominate the global trend of vegetation greening, however, their carbon sink capacities are vastly different. The expansion of greening in China is attributed to the national policy of returning farmland to forest and grassland, of which forests contribute 42%, thus enhancing the carbon sink capacity. In contrast, 82% of greening in India comes from farmland, with forest greening accounting for only 4.4% (Chen et al., 2019). Furthermore, the respiration rate of farmland is relatively high and contributes minorly to the carbon sink, thus India has a weaker carbon sink capacity.