Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Roles of Climate, Vegetation and Soil in Regulating the Spatial Variations in Ecosystem Carbon Dioxide Fluxes in the Northern Hemisphere

  • Zhi Chen,

    Affiliations Synthesis Research Center of Chinese Ecosystem Research Network, Key Laboratory of Ecosystem Network Observation and Modeling, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing, China, University of Chinese Academy of Sciences, Beijing, China

  • Guirui Yu ,

    yugr@igsnrr.ac.cn

    Affiliation Synthesis Research Center of Chinese Ecosystem Research Network, Key Laboratory of Ecosystem Network Observation and Modeling, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing, China

  • Jianping Ge,

    Affiliation College of Life Sciences, Beijing Normal University, Beijing, China

  • Qiufeng Wang,

    Affiliation Synthesis Research Center of Chinese Ecosystem Research Network, Key Laboratory of Ecosystem Network Observation and Modeling, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing, China

  • Xianjin Zhu,

    Affiliation Synthesis Research Center of Chinese Ecosystem Research Network, Key Laboratory of Ecosystem Network Observation and Modeling, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing, China

  • Zhiwei Xu

    Affiliations Synthesis Research Center of Chinese Ecosystem Research Network, Key Laboratory of Ecosystem Network Observation and Modeling, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing, China, University of Chinese Academy of Sciences, Beijing, China

Roles of Climate, Vegetation and Soil in Regulating the Spatial Variations in Ecosystem Carbon Dioxide Fluxes in the Northern Hemisphere

  • Zhi Chen, 
  • Guirui Yu, 
  • Jianping Ge, 
  • Qiufeng Wang, 
  • Xianjin Zhu, 
  • Zhiwei Xu
PLOS
x

Abstract

Climate, vegetation, and soil characteristics play important roles in regulating the spatial variation in carbon dioxide fluxes, but their relative influence is still uncertain. In this study, we compiled data from 241 eddy covariance flux sites in the Northern Hemisphere and used Classification and Regression Trees and Redundancy Analysis to assess how climate, vegetation, and soil affect the spatial variations in three carbon dioxide fluxes (annual gross primary production (AGPP), annual ecosystem respiration (ARE), and annual net ecosystem production (ANEP)). Our results showed that the spatial variations in AGPP, ARE, and ANEP were significantly related to the climate and vegetation factors (correlation coefficients, R = 0.22 to 0.69, P < 0.01) while they were not related to the soil factors (R = -0.11 to 0.14, P > 0.05) in the Northern Hemisphere. The climate and vegetation together explained 60 % and 58 % of the spatial variations in AGPP and ARE, respectively. Climate factors (mean annual temperature and precipitation) could account for 45 - 47 % of the spatial variations in AGPP and ARE, but the climate constraint on the vegetation index explained approximately 75 %. Our findings suggest that climate factors affect the spatial variations in AGPP and ARE mainly by regulating vegetation properties, while soil factors exert a minor effect. To more accurately assess global carbon balance and predict ecosystem responses to climate change, these discrepant roles of climate, vegetation, and soil are required to be fully considered in the future land surface models. Moreover, our results showed that climate and vegetation factors failed to capture the spatial variation in ANEP and suggest that to reveal the underlying mechanism for variation in ANEP, taking into account the effects of other factors (such as climate change and disturbances) is necessary.

Introduction

Terrestrial ecosystems play important roles in modulating the atmospheric carbon dioxide concentration and mitigating global warming [1]. The net carbon exchange between terrestrial ecosystems and the atmosphere is approximately 15–21 Pg yr-1 [2]. This carbon exchange is 2–3 times greater than the annual anthropogenic carbon emissions [1]. Moreover, ecosystem carbon dioxide exchanges are highly variable across space [3,4]. For example, evergreen plantations exhibit high carbon uptake, while drought or disturbed ecosystems produce large carbon emissions [5]. Hence, to reduce the uncertainties in estimated carbon fluxes of terrestrial ecosystems in global carbon cycling, a better understanding of the mechanisms and processes underlying the spatial variations in ecosystem carbon dioxide fluxes is required [6,7].

Climatic pattern is one principal control over the spatial variations in ecosystem carbon dioxide fluxes [8,9]. Study on global forests indicates that the spatial patterns of the mean annual temperature (MAT) and the mean annual precipitation (MAP) regulate the spatial variations in the annual net ecosystem production (ANEP) and its two component fluxes, i.e., the annual gross primary production (AGPP) and the annual ecosystem respiration (ARE) [10]. ANEP is largely associated with the MAT at mid-to-high latitudes and associated with dryness at mid-to-low latitudes [11]. AGPP and ARE are positively related with MAT and water availability across European forests [12,13]. In Asia, these climate-carbon fluxes relationships are relatively stronger because of the broader range in climate zones [1417].

Differences in vegetation characteristics, i.e., the leaf area index (LAI) and the length of carbon uptake period (CUP), also affect the spatial patterns of ecosystem carbon dioxide fluxes [18,19]. As the maximum LAI increases, the values of AGPP and ARE increase linearly, and ANEP increases exponentially across Asian ecosystems [14]. Variation in the CUP account for a large portion of the spatial variation in ANEP across a continental gradient of deciduous forest, evergreen forest, grass and crop ecosystems [20,21].

In addition to climate and vegetation, soil condition is another potential factor affecting the spatial pattern of ecosystem carbon dioxide fluxes. The spatial variation in soil respiration is found to be associated with the soil organic carbon (SOC) content [22,23] because SOC is the substrate for soil respiration [24] and meanwhile affects the soil respiration rate (measured at a given reference temperature) [24,25]. However, the role of soil conditions on the variations in AGPP, ARE and ANEP is not well documented. The interrelationships of climate, vegetation and soil and their effects on the spatial variations in AGPP, ARE and ANEP are much uncertain because few studies have considered all three factors simultaneously.

Given the reported controls of climate and vegetation on the carbon fluxes, we speculate that climate, vegetation and soil affect the spatial variations in AGPP, ARE and ANEP to different degrees. AGPP, ARE and ANEP are likely more strongly associated with climate and vegetation than soil factors. Furthermore, the spatial pattern of climate factors would shape the variations in AGPP, ARE and ANEP mainly by means of regulating vegetation properties.

Currently, the regional analysis of how climate, vegetation and soil determine the spatial variations in carbon dioxide fluxes becomes available that attributes to the intensive progresses made in the ecosystem carbon flux measurements. Over the past two decades, eddy-covariance flux measurements have been extensively recorded more than 400 sites worldwide [5]. Moreover, active measurements at numerous sites have been ongoing for more than a decade [5]. These valuable data underlie the solid basis for exploring the underlying drivers for the spatial variations in AGPP, ARE and ANEP. Classification and Regression Trees (CART) analysis and Redundancy Analysis (RDA) provide a suitable statistical approach to identify the critical factors and further quantify the effect size for each critical factor [2628]. Given ecological data are often complex, unbalanced, and contain missing data, moreover, relationships between explanatory and response variables are likely nonlinear and involve high-order interactions, the commonly used statistical modeling techniques often fail to find meaningful ecological patterns from such data [27]. CART can deal with nonlinear relationships, high-order interactions, and missing values, and represent easily interpretable results [26,27], which has been an ideally statistical technique used to investigate the controlling factors for complex and unbalanced ecological data, such as aboveground biomass, latent heat, and sensible heat and so on [29,30].

In this study, we aim to simultaneously consider climate, vegetation and soil factors and use CART and RDA methods to (1) identify critical factors that drive variations in AGPP, ARE and ANEP in the Northern Hemisphere, (2) quantify the contribution of each critical factor, and (3) elucidate the relationships among climate, vegetation and soil in controlling the spatial variations in AGPP, ARE and ANEP.

Materials and Methods

Data sources

The datasets analyzed in this study included (1) the climate factors of mean annual temperature (MAT, °C), mean annual precipitation (MAP, mm), and mean annual solar radiation (MAR, W m-2); (2) the vegetation factors of the mean maximum enhanced vegetation index (EVImax) and mean annual enhanced vegetation index (EVImean); (3) the soil factors of soil organic carbon content at the depth of 0–30 cm (SOC30, %) and soil organic carbon content at the depth of 30–100 cm (SOC100, %); and (4) the ecosystem carbon dioxide fluxes of the mean annual gross primary production (AGPP, g C m-2 yr-1), mean annual ecosystem respiration (ARE, g C m-2 yr-1), and mean annual net ecosystem production (ANEP, g C m-2 yr-1).

Flux data.

Published ecosystem carbon dioxide fluxes data (GPP, RE, and NEP) measured by the eddy covariance technique over the past two decades (1990–2010) in the Northern Hemisphere were compiled. The published data were filtered as follows. First, the carbon dioxide fluxes data must be processed and corrected by the authors at each site, including the procedures of three-dimensional coordinate rotation, WPL correction, storage flux calculation, outlier filter, nighttime flux correction, gap filling and flux partitioning [3135]. In the datasets, differences among sites mainly existed in the terms of the friction velocity (u*) threshold, the gap-filling and partitioning method. The u* threshold was defined as the u* value where temperature-normalized nighttime NEE start to drop off as u* decreases or where the night-time flux reached 95% of the average flux within the higher u* classes [35,36]. The u* threshold varied among ecosystems because it depended on the local topography, vegetation and weather. In this study the u* threshold fell within the general range from 0.1 to 0.4 m/s [35]. As gap-filling methods, the Mean Diurnal Variation (MDV), Look-Up Tables (LookUp), Marginal Distribution Sampling (MDS), Nonlinear Regression (NLR) and Artificial Neural Network (ANN) approaches were applied by researchers. Among these different gap filling approaches, there was no significant difference by comparison [31,34,37]. The effect of gap filling on the annual sum of NEE was reported to fall within a range of ± 25 g C m-2 yr-1 [37]. For flux partitioning, the nighttime data-based (NB) method [35] that respiration measured at night were extrapolated to the daytime based on the responses of respiration to air or soil temperature were overwhelmingly applied in the collected studies. Detailed information on each site was shown in S2 and S3 Tables.

Second, carbon fluxes data were required to be continuous for more than one year. For some sites where fluxes were difficult to measure during winter, we only considered sites where the winter CO2 flux was known to represent a negligible proportion of the annual flux [38] or the winter flux was extrapolated from the short-term non-growing-season or growing-season measurements [39].

Additionally, studied sites must be undisturbed by fire, logging or other serious processes in the previous ten years to eliminate the impacts of recent disturbance on the carbon dioxide fluxes. Croplands and grasslands are generally under management. We only selected croplands and grasslands with typical native plants that have long cultivation histories and constant cultivation systems and management practices.

A total of 861 site-year records of carbon dioxide fluxes from 241 sites were ultimately included in our analysis. These studied sites were distributed across Asia (67 sites), Europe-Africa (91 sites), and North-South America (83 sites) (Fig 1 and S2 Table). The sites consisted of evergreen broadleaf forest (13 sites), evergreen needleleaf forest (64 sites), deciduous broadleaf forest (25 sites), deciduous needleleaf forest (5 sites), mixed forest (12 sites), grassland (60 sites), cropland (34 sites), and wetland (28 sites) ecosystems. For a spatial analysis, we calculated the multi-year mean annual GPP, RE, and NEP (AGPP, ARE, and ANEP) for each site.

thumbnail
Fig 1. Spatial distribution of eddy covariance flux sites used in this study.

Red circles indicated the locations of eddy covariance flux sites in the Northern Hemisphere. The base map was the 2008 MODIS (Moderate-resolution Imaging Spectroradiometer) land-cover product (MCD12C1) at a spatial resolution of 0.05 degree plotted by the North Pole Azimuthal Equidistant projection tool in ArcGIS 10.0 software.

https://doi.org/10.1371/journal.pone.0125265.g001

Climate data.

The climate variables (MAT and MAP) during the period of flux observation were simultaneously collected. For nine sites where miss temperature data and six sites where miss precipitation data, we used the neighboring meteorological station observations from the global surface summary of daily data produced by the National Climatic Data Center (NCDC) (ftp://ftp.ncdc.noaa.gov/pub/data/gsod/) to fill. The MAR data were uniformly obtained from the Climate Research Unit (CRU05) Monthly Climate Database (http://daac.ornl.gov/ISLSCPII/guides/cru_monthly_mean_xdeg.html) provided by the International Satellite Land Surface Climatology Project (ISLSCP).

Vegetation data.

The EVI data at a spatial resolution of 250 m and a temporal resolution of 16 days from 2000 to 2010 was extracted from the MODIS data product (MOD13Q1) available from the Oak Ridge National Laboratory’s Distributed Active Archive Center (http://daac.ornl.gov/MODIS/). We extracted regions consisting of 81 pixels at a 1 km radius centered on the flux tower. The EVI value quality of each pixel was examined using the pixel reliability flags included in the product. We filtered out poor-quality pixels with pixel reliability flags of 0 or 1. If more than a quarter of pixels were with high quality, we averaged the 16-day EVI value; otherwise, we treated the data point as missing and used the multi-year mean EVI. We calculated the multi-year mean annual EVI (EVImean) and the mean maximum EVI (EVImax) at each site.

Soil data.

The soil organic carbon contents at the depths of 0–30 cm (SOC30) and 30–100 cm (SOC100) were extracted from the Harmonized World Soil Database (version 1.2) [40] (http://webarchive.iiasa.ac.at/Research/LUC/External-World-soil-database/HTML/HWSD_Data.html?sb=4). We also compiled the SOC data from the published literatures to assess the robustness of the abstracted SOC data. The result indicated that the SOC from the Harmonized World Soil Database exhibited a good agreement with the SOC measured at the flux tower sites (R2 = 0.62, n = 17, P = 0.002).

Data analysis

Distribution and correlation analysis.

The ‘car’ and ‘corrgram’ packages in the R software environment were applied to analyze the statistical distributions and correlations of climate, vegetation, soil factors, and carbon dioxide fluxes. The ‘ScatterplotMatrix’ function was used to draw histograms of each factor and carbon dioxide fluxes. Linear regression was applied to analyze the correlations between each pair of variables.

Identify the critical factors.

Classification and Regression Tree (CART) analysis was employed to identify the critical factors determining the spatial variations in AGPP, ARE and ANEP. CART operates by splitting the data into mutually exclusive subgroups (nodes), and each of the subgroups as homogeneous as possible [26]. During this process, a binary splitting procedure is applied. Beginning with the split of the root (or parent node) that contains all of the objects into two nodes (or child nodes), the splitting procedure is subsequently applied separately and repeatedly to each child node. The tree grows as the splitting process continues, and it is then pruned to a reasonable size [27]. In this study, the ‘rpart’ package in the R software environment was used to perform the CART analysis. First, the ‘rpart’ function was used to establish a formula that included all of the explanatory and response variables. Then, the parameters that dictated the splitting process were set using the ‘rpart.control’ function, in which cp was a complexity parameter. The main function of cp was to reduce the computing time by pruning splits that were not important. Initially, the cp value was set to 0.005 to construct a tree. However, the initial tree had such a large number of splits that it was difficult to interpret and produced a high prediction error. Thus, the ‘prune’ function was used to prune the tree. The x-error served as a criterion for obtaining the optimal tree. When the x-error reached to the minimum value, the tree was pruned and the main environmental factors of the optimal tree were simultaneously screened. The x-error was found to be minimized, followed by a slow increase as the number of splits increased. Therefore, when x-error was minimum the corresponding cp value was used to prune the tree to obtain the final optimal model.

Quantifying the contributions.

Redundancy Analysis (RDA) was conducted to quantify the contributions of critical explanatory variables to the response variables. RDA is presented as an eigen-analysis of covariance matrices. It assesses the explanatory power of each defined variable by parsing out other terms as constraints to calculate its proportion of total variance [28]. In the present study, the ‘vegan’ package containing the ‘rda’ function in the R software environment was applied to perform the RDA analysis. All critical factors identified by CART were initially included in the model setup to calculate the total explanatory power (the proportion of the constrained inertia to the total inertia). Subsequently, each critical factor was alternately parsed out as constraint to analyze the individual and mixed effects. Finally, the ‘venneuler’ function was used to visualize the proportional contributions.

Plotting.

The site distribution was plotted using ArcGIS 10.0 software. All of the other figures were plotted using the R-freedom software (version 3.0.0).

Results

Distributions and correlations of climate, vegetation, soil factors and carbon dioxide fluxes

The sites analyzed in this study reflected the high spatial variability in ecosystems in the Northern Hemisphere. The MAT varied from -12 to 29°C and the MAP varied from 145 to 3485 mm. AGPP ranged from 75 to 3760 g C m-2 yr-1, ARE ranged from 103 to 3805 g C m-2 yr-1, and ANEP varied from carbon emissions (-236 g C m-2 yr-1) to high carbon uptake (857 g C m-2 yr-1) (Fig 2). Based on the available dataset in this study, AGPP was 1258 ± 667 g C m-2 yr-1 (mean ± std). ARE and ANEP were 1052 ± 598 g C m-2 yr-1 and 205 ± 208 g C m-2 yr-1, respectively, in the Northern Hemisphere (Fig 2).

thumbnail
Fig 2. Scatterplot matrix of climate, vegetation, soil factors and ecosystem carbon dioxide fluxes.

The diagonal panels were histograms of each factor and carbon dioxide flux. The lower panels were linear regression for each pair of factor and carbon dioxide flux. MAT: mean annual temperature (°C); MAP: mean annual precipitation (mm); MAR: mean annual solar radiation (W m-2); EVImax: mean maximum enhanced vegetation index; EVImean: mean annual enhanced vegetation index; SOC30: soil organic carbon content at the depth of 0–30 cm (%); SOC100: soil organic carbon content at the depth of 30–100 cm (%); AGPP: mean annual gross primary production (g C m-2 yr-1); ARE: mean annual ecosystem respiration (g C m-2 yr-1); ANEP: mean annual net ecosystem production (g C m-2 yr-1).

https://doi.org/10.1371/journal.pone.0125265.g002

Correlations among the spatial variations in the climate, vegetation, soil factors and ecosystem carbon dioxide fluxes were shown in Fig 2. AGPP, ARE and ANEP were significantly and positively related with the MAT and MAP (correlation coefficients (R) = 0.22 ~ 0.64, P < 0.01), but they were not related with the MAR (R = -0.02 ~ 0.06, P > 0.05). AGPP, ARE and ANEP showed positive correlations with the vegetation index (R = 0.19 ~ 0.69, P < 0.01), but showed no significant correlations with the soil factors (R = -0.11 ~ 0.14, P > 0.05). The correlations in different ecosystem types showed that AGPP, ARE, and ANEP were similarly related to the climate and vegetation factors (P < 0.01) but they were not related to the soil factors (P > 0.05) in the forest, cropland, grassland and wetland ecosystems (S1 Table). However, compared to forests and grasslands where carbon fluxes were related to both of the climate and vegetation factors, carbon fluxes tended to be more closely related to the vegetation factors in croplands and more closely related to the climate factors in wetlands (S1 Table).

Critical factors for determining the spatial variations in carbon dioxide fluxes

The CART results showed that the climate and vegetation factors were the main factors for differentiating and grouping AGPP, ARE and ANEP in the Northern Hemisphere. Soil factors only had a minor effect on the spatial variations in AGPP, ARE and ANEP (Fig 3).

thumbnail
Fig 3. Classification and Regression Trees analysis of the controlling factors for AGPP(A), ARE(B), ANEP(C).

Left part: Classification and regression trees for AGPP(A), ARE(B), ANEP(C); Right part: The n_split, x_error and cp value used in tree building for AGPP(A), ARE(B), ANEP(C).

https://doi.org/10.1371/journal.pone.0125265.g003

For AGPP, the optimal tree was obtained when the x-error was the minimum value of 0.536 and the corresponding cp value was 0.0125 (Fig 3A). Five critical factors for AGPP were identified, i.e., MAP, EVImean, MAT, MAR and SOC100. About 73% of the AGPP samples (n = 175) were separated by the MAP, EVImean and MAT factors (Fig 3A). The optimal tree for ARE was obtained when the x-error was the minimum value of 0.522 and the corresponding cp value was 0.0243 (Fig 3B). Four critical factors, i.e., MAP, EVImean, MAT and SOC30, were identified for ARE. Approximately 96% of the ARE samples (n = 231) were separated by the MAP, EVImean and MAT factors (Fig 3B). For ANEP, the optimal tree was obtained when the x-error was the minimum value of 0.967 and the corresponding cp value was 0.0294 (Fig 3C). Compared with AGPP and ARE, less number of split occurred for ANEP. Only the MAP and EVImax were identified as the critical factors for ANEP (Fig 3C).

Contributions of critical factors to the spatial variations in carbon dioxide fluxes

The RDA results showed that the MAT, MAP, and EVImean together explained 60% and 58% of the spatial variations in AGPP and ARE in the Northern Hemisphere, respectively (Figs 4A and 4B). EVImean exhibited a large direct effect on the spatial variations in AGPP and ARE (13.3% and 13.1%, respectively), which were higher than the individual direct effects of the MAT and MAP. When the regulation of MAT and MAP on EVImean was included, the MAT and MAP together explained 45–47% of the spatial variations in AGPP and ARE, which accounted for three quarters of the total explained variation (Figs 4A and 4B). However, the climate constraint on the vegetation index explained approximately 75%, indicating that the effects of climate on the spatial variations in AGPP and ARE are mainly realized by means of their regulation on vegetation properties. The MAP and EVImax together only explained 10.1% of the spatial variation in ANEP in the Northern Hemisphere, which was much less than the corresponding explanation for AGPP and ARE (Fig 4C).

thumbnail
Fig 4. Contributions of critical factor to the spatial variations in AGPP(A), ARE(B), ANEP(C).

Contributions of MAT, MAP, and EVImean to the spatial variations in AGPP(A) and ARE(B); Contributions of MAP and EVImax to the spatial variation in ANEP (C).

https://doi.org/10.1371/journal.pone.0125265.g004

Discussion

Climate factors strongly affect ecosystem production and respiration by determining heat and water availability. Lieth [41] proposes that two climate indices, i.e., temperature and precipitation, serve as key predictors of plant production and thereby provides a world map of terrestrial primary production based on MAT and MAP alone. The strong regulation of climate on the geospatial pattern of AGPP and ARE was similarly observed in this study. The MAT and MAP accounted for 45–47% of the spatial variations in AGPP and ARE in the Northern Hemisphere. This result is strongly supported by regional-scale studies within the Northern Hemisphere, which reveal that temperature and precipitation are two important controls for determining the spatial variations in AGPP and ARE in Europe [13], North America [12], China [16] and Asia [14,15,17].

Noteworthily, we found that the climate influenced AGPP and ARE mainly by means of regulating vegetation properties. Approximately three quarters of the climate explanatory power achieved from their constraint on the vegetation index. The strong dependencies of AGPP and ARE on the climate factors revealed in the previous studies [1217] are likely because climate pattern determine the geographical distribution of vegetation types, and these various vegetation types represent varying vegetation indices and leaf span that directly affect ecosystem carbon dioxide exchanges [14,42,43]. Compared to forests and grasslands, the vegetation index exerts larger effect on the carbon fluxes than climate in croplands, which is likely associated that croplands mostly concentrate on the temperate zones and more importantly, croplands are generally under long-term intensive managements such as fertilization, irrigation, and cultivation, which would weaken the direct climate regulation on ecosystem carbon exchange to some extent [17].

In addition to temperature and precipitation, solar radiation is another important climate factor. Our results demonstrated that the spatial pattern of MAR was not the principal factor for determining the spatial variations in AGPP, ARE and ANEP. It is probably associated with the relatively small geospatial variation in solar radiation because the spatial variation coefficient of the MAR (0.21) was much lower than that of MAT (0.76) and MAP (0.61). Additionally, solar radiation may affect carbon dioxide fluxes mainly by constraining the temperature [44]. A close positive correlation between solar radiation and temperature was demonstrated in this study.

The role of the spatial variation in SOC in regulating AGPP, ARE and ANEP, however, is not as well documented as those climate and vegetation factors. Our results demonstrated that the variation in SOC had a minor effect on the spatial variations in AGPP, ARE and ANEP. This minor regulation is partially attributed to the complex geospatial variation in SOC because SOC is the net balance of inputs from plant production and outputs from decomposition and leaching [45]. SOC content could be low in high productivity ecosystems if these ecosystems concurrently experience intensive losses by decomposition and leaching. Moreover, SOC is less sensitive to changes in climate compared to vegetation properties. SOC keeps constant along the precipitation gradient [46], and it was weakly correlated with variations in MAT and MAP. The limited influence of SOC on the spatial variations in AGPP, ARE and ANEP may also be associated with the potential influence of other soil properties, such as soil N and P content [47,48]. In the forest study, nutrient availability (soil type, nutrient concentrations, pH, C/N ratio, nitrogen supply and mineralization) is the key regulator of net production, which accounts for 19% of the spatial variation in ANEP [49]. However, these findings are limited to forests and based on analysis that qualitatively divides the studied sites into high or low soil nutrient levels. The direct associations between the spatial variations in carbon dioxide fluxes with soil N and P contents and availabilities still require further corroboration.

In this study, the MAT, MAP, and EVI together explained 58–60% of the spatial variations in AGPP and ARE and 10% of ANEP. These values indicated that there were considerable variations not captured by the evaluated climate and vegetation factors, particularly for the ANEP. The spatial patterns of ecosystem carbon dioxide fluxes are likely simultaneously affected by other factors, for example (1) Changes in temperature and precipitation, and nitrogen deposition. Magnani et al. [50] demonstrates that the nitrogen deposition influences the spatial patterns of ecosystem carbon fluxes. Piao et al. [51] indicates that temperature change in the recent past is an important driver for current and future forest carbon balances; (2) Disturbances. Modeling studies show that changes in vegetation properties and succession stages under disturbances can weaken the climate dependency of carbon fluxes [52]. Yuan et al. [53] points out disturbances alter ecosystem age structure and vegetation composition that cause substantial impacts on the ecosystem carbon cycle.

Based on the above illustration, we may develop a simple biogeographical ecological framework for the spatial variations in AGPP, ARE and ANEP in the Northern Hemisphere (Fig 5). This framework reveals that (1) temperature and precipitation patterns together determine the geographical distribution of vegetation types. (2) these various vegetation types represent varying physiognomic properties of vegetation (e.g., leaf area index and length of carbon uptake period) that directly affect the carbon dioxide exchanges and thus shape the spatial pattern of AGPP, ARE and ANEP. (3) the geographic pattern of soil organic carbon content has only a minor and secondary effect on the spatial variations in AGPP, ARE and ANEP. (4) there are considerable variations not captured by the evaluated climate and vegetation factors because of the impacts of climate change and disturbances and so on. Our results indicate that the effects of climate, vegetation and soil on the spatial variations in AGPP, ARE and ANEP are in different magnitudes. Climate factors influence the spatial variations in AGPP and ARE via their regulations on vegetation properties, while soil factors only exert a minor effect. Take into account these discrepant roles of climate, vegetation and soil would be helpful to accurately assess the global carbon balance and predict the ecosystem responses to environmental changes.

thumbnail
Fig 5. Biogeographic-ecological framework for the spatial variations in AGPP, ARE and ANEP.

The width of arrows indicates the strength of effects. Plus sign (+) indicates the positive effects. Plus and minus sign (+/-) indicates the positive or negative effects. MAT: mean annual temperature; MAP: mean annual precipitation; MAR: mean annual solar radiation; GSL: length of growing season; LAI: leaf area index; SOC: soil organic carbon; AGPP: mean annual gross primary production; ARE: mean annual ecosystem respiration; ANEP: mean annual net ecosystem production. (1) Temperature and precipitation patterns together determine the geographical distribution of vegetation types; (2) These various vegetation types represent varying physiognomic properties of vegetation (e.g., leaf area index and length of carbon uptake period) that directly affect the carbon dioxide exchanges and thus shape the spatial patterns of AGPP, ARE and ANEP; (3) Geographic pattern of soil organic carbon content has only a minor and secondary effect on the spatial variations in AGPP, ARE and ANEP; (4) there are considerable variations not captured by the evaluated climate and vegetation factors because of the impacts of climate change and disturbances and so on.

https://doi.org/10.1371/journal.pone.0125265.g005

In this study, several sources of uncertainty also existed. Errors in the eddy covariance measurements were likely caused by the influence of complex topography, atmospheric stability and frequency of data acquisition [5456]. Some systematic errors were probably introduced during the processes of outlier removal, gap filling and flux partitioning. Moreover, given the incompletely uniform study periods of the analyzed flux sites, several uncertainties were likely caused by this uneven and limited monitoring time. Finally, large variation in ANEP was not captured by the variations in the climate and vegetation factors. Hence, integrative consideration of the effects of climate change, disturbances and other factors on ANEP is highly needed in the future studies.

Supporting Information

S1 Table. Correlation coefficients between the carbon fluxes with climate, vegetation and soil factors in the forest, cropland, grassland and wetland ecosystems.

MAT: mean annual temperature (°C); MAP: mean annual precipitation (mm); MAR: mean annual solar radiation (W m-2); EVImax: mean maximum enhanced vegetation index; EVImean: mean annual enhanced vegetation index; SOC30: soil organic carbon content at the depth of 0–30 cm (%); SOC100: soil organic carbon content at the depth of 30–100 cm (%); AGPP: mean annual gross primary production (g C m-2 yr-1); ARE: mean annual ecosystem respiration (g C m-2 yr-1); ANEP: mean annual net ecosystem production (g C m-2 yr-1). ** indicates significant correlation at the 0.01 level (two-tailed). * indicates significant correlation at the 0.05 level (two-tailed).

https://doi.org/10.1371/journal.pone.0125265.s001

(DOC)

S2 Table. Sites characteristics of this study.

EBF: evergreen broadleaf forest; ENF: evergreen needle forest; DBF: deciduous broadleaf forest; DNF: deciduous needle forest; MF: mixed forest.

https://doi.org/10.1371/journal.pone.0125265.s002

(DOC)

S3 Table. Sites measurement systems and data processing approaches.

OPEC: Open-path eddy covariance; CPEC: Closed-path eddy covariance; MDV: Mean Diurnal Variation; LookUp: Look-up table; MDS: Marginal Distribution Sampling; NLR: Nonlinear Regression; ANN: Artificial Neural Network; NB: nighttime data-based estimate by respiration equation; DB: daytime data-based estimate by light response equation;-: unspecified.

https://doi.org/10.1371/journal.pone.0125265.s003

(DOC)

Acknowledgments

We acknowledge all of the researchers who contributed to the eddy covariance flux measurements for providing valuable data.

Author Contributions

Conceived and designed the experiments: GY. Performed the experiments: ZC. Analyzed the data: ZC. Contributed reagents/materials/analysis tools: ZC. Wrote the paper: JG QW XZ ZX.

References

  1. 1. Le Quéré C, Peters GP, Andres RJ, Andrew RM, Boden T, Ciais P, et al. Global carbon budget 2013. Earth System Science Data 2014; 6: 235–263.
  2. 2. Jung M, Reichstein M, Margolis HA, Cescatti A, Richardson AD, Arain MA, et al. Global patterns of land-atmosphere fluxes of carbon dioxide, latent heat, and sensible heat derived from eddy covariance, satellite, and meteorological observations. Journal of Geophysical Research 2011; 116. pmid:24307747
  3. 3. Valentini R, Matteucci G, Dolman AJ, Schulze ED, Rebmann C, Moors EJ, et al. Respiration as the main determinant of carbon balance in European forests. Nature 2000; 404: 861–865. pmid:10786790
  4. 4. Baldocchi D. Turner review No.15. 'Breathing' of the terrestrial biosphere: lessons learned from a global network of carbon dioxide flux measurement systems. Australian Journal of Botany 2008; 56: 1–26.
  5. 5. Baldocchi D. Measuring fluxes of trace gases and energy between ecosystems and the atmosphere—the state and future of the eddy covariance method. Global Change Biology 2014; 20: 3600–3609. pmid:24890749
  6. 6. Schimel DS, House JI, Hibbard KA, Bousquet P, Ciais P, Peylin P, et al. Recent patterns and mechanisms of carbon exchange by terrestrial ecosystems. Nature 2001; 414: 169–172. pmid:11700548
  7. 7. Wofsy SC, Harriss RC. The North American Carbon Program (NACP). Report of the NACP Committee of the US Interagency Carbon Cycle Science Program. US Global Change Research Program, Washington, DC; 2002. pp. 1–56.
  8. 8. Xia JZ, Liu SG, Liang SL, Chen Y, Xu WF, Yuan WP. Spatio-temporal patterns and climate variables controlling of biomass carbon stock of global grassland ecosystems from 1982 to 2006. Remote Sensing 2014; 6: 1783–1802.
  9. 9. Yuan WP, Luo YQ, Richardson AD, Oren R, Luyssaert S, Janssens IA, et al. Latitudinal patterns of magnitude and interannual variability in net ecosystem exchange regulated by biological and environmental variables. Global Change Biology 2009; 15: 2905–2920.
  10. 10. Wang XC, Wang CK, Yu GR. The spatial and temporal pattern of global forest carbon exchange based on eddy covariance. Science in China Series D: Earth Sciences 2008; 38: 1092–1102.
  11. 11. Yi CX, Ricciuto D, Li R, Wolbeck J, Xu XY, Nilsson M, et al. Climate control of terrestrial carbon exchange across biomes and continents. Environmental Research Letters 2010; 5: 034007.
  12. 12. Law BE, Falge E, Gu L, Baldocchi DD, Bakwin P, Berbigier P, et al. Environmental controls over carbon dioxide and water vapor exchange of terrestrial vegetation. Agricultural and Forest Meteorology 2002; 113: 97–120.
  13. 13. Reichstein M, Papale D, Valentini R, Aubinet M, Bernhofer C, Knohl A, et al. Determinants of terrestrial ecosystem carbon balance inferred from European eddy covariance flux sites. Geophysical Research Letters 2007; 34: L01402.
  14. 14. Kato T, Tang YH. Spatial variability and major controlling factors of CO2 sink strength in Asian terrestrial ecosystems: evidence from eddy covariance data. Global Change Biology 2008; 14: 2333–2348.
  15. 15. Hirata R, Saigusa N, Yamamoto S, Ohtani Y, Ide R, Asanuma J, et al. Spatial distribution of carbon balance in forest ecosystems across East Asia. Agricultural and Forest Meteorology 2008; 148: 761–775.
  16. 16. Yu GR, Zhu XJ, Fu YL, He HL, Wang QF, Wen XF, et al. Spatial pattern and climate drivers of carbon fluxes in terrestrial ecosystems of China. Global Change Biology 2013; 19: 798–810. pmid:23504837
  17. 17. Chen Z, Yu GR, Ge JP, Sun XM, Hirano T, Saigusa N, et al. Temperature and precipitation control of the spatial variation of terrestrial ecosystem carbon exchange in the Asian region. Agricultural and Forest Meteorology 2013; 182–183: 266–276.
  18. 18. Paster J, Post WM. Linear regressions do not predict the transient responses of Eastern North American forests to CO2—induced climate change. Climatic Change 1993; 23: 111–119.
  19. 19. Potter CS, Randerson JT, Field CB, Matson PA, Vitousek PM, Mooney HA, et al. Terrestrial ecosystem production: a process model based on global satellite and surface data. Global Biogeochemical Cycles 1993; 7: 811–841.
  20. 20. Baldocchi D, Falge E, Gu LH, Olson R, Hollinger D, Running S, et al. FLUXNET: a new tool to study the temporal and spatial variability of ecosystem-scale carbon dioxide, water vapor, and energy flux densities. Bulletin of the American Meteorological Society 2001; 82: 2415–2434.
  21. 21. Churkina G, Schimel D, Braswell BH, Xiao XM. Spatial analysis of growing season length control over net ecosystem exchange. Global Change Biology 2005; 11: 1777–1787.
  22. 22. Gough CM, Seiler JR. The influence of environmental, soil carbon, root and stand characteristics on soil CO2 efflux in loblolly pine (Pinus taeda L.) plantations located on the South Carolina Coastal Plain. Forest Ecology and Management 2004; 191: 353–363.
  23. 23. Rodeghiero M, Cescatti A. Main determinants of forest soil respiration along an elevation/temperature gradient in the Italian Alps. Global Change Biology 2005; 11: 1024–1041.
  24. 24. Yu GR, Zheng ZM, Wang QF, Fu YL, Zhuang J, Sun XM, et al. Spatiotemporal Pattern of Soil Respiration of Terrestrial Ecosystems in China: The Development of a Geostatistical Model and Its Simulation. Environmental Science and Technology 2010; 44: 6074–6080. pmid:20704202
  25. 25. Zheng ZM, Yu GR, Sun XM, Li SG, Wang YS, Wang YH, et al. Spatio-temporal variability of soil respiration of forest ecosystems in China: influencing factors and evaluation model. Environmental Management 2010; 46: 633–642. pmid:20571797
  26. 26. Breiman L, Friedman JH, Olshen RA, Stone CJ. Classification and Regression Trees. Wadsworth, Belmont, California, USA; 1984.
  27. 27. De’ath G, Fabricius KE. Classification and regression trees: a powerful yet simple technique for ecological data analysis. Ecology 2000; 81: 3178–3192.
  28. 28. Oksanen J. Multivariate Analysis of Ecological Communities in R: Vegan Tutorial. R package version. 2013.
  29. 29. Jung M, Reichstein M, Bondeau A. Towards global empirical upscaling of FLUXNET eddy covariance observations: validation of a model tree ensemble approach using a biosphere model. Biogeosciences 2009; 6: 2001–2013.
  30. 30. Sun J, Cheng GW, Li WP. Meta-analysis of relationships between environmental factors and aboveground biomass in the alpine grassland on the Tibetan Plateau. Biogeosciences 2013; 10: 1707–1715.
  31. 31. Falge E, Baldocchi D, Olson R, Anthoni P, Aubinet M, Bernhofer C, et al. Gap filling strategies for defensible annual sums of net ecosystem exchange. Agricultural and Forest Meteorology 2001; 107: 43–69.
  32. 32. Gilmanov TG, Verma SB, Sims PL, Meyers TP, Bradford JA, Burba GG, et al. Gross primary production and light response parameters of four southern plains ecosystems estimated using long-term CO2-flux tower measurements. Global Biogeochemical Cycles 2003; 17: 1071.
  33. 33. Lasslop G, Reichstein M, Papale D, Richardson AD, Arneth A, Barr A, et al. Separation of net ecosystem exchange into assimilation and respiration using a light response curve approach: critical issues and global evaluation. Global Change Biology 2010; 16: 187–208.
  34. 34. Papale D, Reichstein M, Aubinet M, Canfora E, Bernhofer C, Kutsch W, et al. Towards a standardized processing of Net Ecosystem Exchange measured with eddy covariance technique: algorithms and uncertainty estimation. Biogeosciences 2006; 3: 571–583.
  35. 35. Reichstein M, Falge E, Baldocchi D, Papale D, Aubinet M, Berbigier P, et al. On the separation of net ecosystem exchange into assimilation and ecosystem respiration: review and improved algorithm. Global Change Biology 2005; 11: 1424–1439.
  36. 36. Massman WJ, Lee X. Eddy covariance flux corrections and uncertainties in long-term studies of carbon and energy exchanges. Agricultural and Forest Meteorology 2002; 113: 121–144.
  37. 37. Moffat AM, Papale D, Reichstein M, Hollinger DY, Richardson AD, Barr AG, et al. Comprehensive comparison of gap-filling techniques for eddy covariance net carbon fluxes. Agricultural and Forest Meteorology 2007; 147: 209–232.
  38. 38. Matsuura Y, Kajimoto T, Nakai Y, Osawa A, Morishita T, Zyryanova OA, et al. Proceedings of the Seventh International Conference on Gobal Change: Connection to the Arctic (GCCA-7) International Arctic Research Center. University of Alaska Fairbanks 2007; 312 p: 258–261.
  39. 39. Arneth A, Kurbatova J, Kolle O, Shibistova OB, Lloyd J, Vygodskaya NN, et al. Comparative ecosystem–atmosphere exchange of energy and mass in a European Russian and a central Siberian bog II. Interseasonal and interannual variability of CO2 fluxes. Tellus Series B 2002; 54: 514–530.
  40. 40. FAO/IIASA/ISRIC/ISSCAS/JRC. Harmonized World Soil Database (version 1.2). FAO, Rome, Italy and IIASA, Laxenburg, Austria; 2012.
  41. 41. Lieth H. Primary production: terrestrial ecosystems. Human Ecology 1973; 1: 303–332.
  42. 42. Prentice IC, Cramer W, Harrison SP, Leemans R, Monserud RA, Solomon AM. A global biome model based on plant physiology and dominance, soil properties and climate. Journal of Biogeography 1992; 19: 117–134.
  43. 43. Prentice KC. Bioclimatic Distribution of Vegetation for General Circulation Model Studies. Journal of Geophysical Research 1990; 95: 11811–11830.
  44. 44. Cramer W, Kicklighter DW, Bondeau A, Moore B, Churkina G, Nemry B, et al. Comparing global models of terrestrial net primary productivity (NPP): overview and key results. Global Change Biology 1999; 5: 1–15.
  45. 45. Jobbagy EG, Jackson RB. The vertical distribution of soil organic carbon and its relation to climate and vegetation. Ecological Applications 2000; 10: 423–436.
  46. 46. Zhou XH, Talley M, Luo YQ. Biomass, litter, and soil respiration along a precipitation gradient in Southern Great Plains, USA. Ecosystems 2009; 12: 1369–1380.
  47. 47. LeBauer DS, Treseder KK. Nitrogen limitation of net primary productivity in terrestrial ecosystems is globally distributed. Ecology 2008; 89: 371–379. pmid:18409427
  48. 48. Vitousek PM, Porder S, Houlton BZ, Chadwick OA. Terrestrial phosphorus limitation: mechanisms, implications, and nitrogen-phosphorus interactions. Ecological Applications 2010; 20: 5–15. pmid:20349827
  49. 49. Fernández-Martínez M, Vicca S, Janssens IA, Sardans J, Luyssaert S, Campioli M, et al. Nutrient availability as the key regulator of global forest carbon balance. Nature Climate Change 2014; 4: 471–476.
  50. 50. Magnani F, Mencuccini M, Borghetti M, Berbigier P, Berninger F, Delzon S, et al. The human footprint in the carbon cycle of temperate and boreal forests. Nature 2007; 447: 848–850. pmid:17568744
  51. 51. Piao SL, Friedlingstein P, Ciais P, Peylin P, Zhu B, Reichstein M. Footprint of temperature changes in the temperate and boreal forest carbon balance. Geophysical Research Letters 2009; 36: L07404.
  52. 52. Thornton PE, Law BE, Gholz HL, Clark KL, Falge E, Ellsworth DS, et al. Modeling and measuring the effects of disturbance history and climate on carbon and water budgets in evergreen needleleaf forests. Agricultural and Forest Meteorology 2002; 113: 185–222.
  53. 53. Yuan WP, Liu SG, Dong WJ, Liang SL, Zhao SQ, Chen JM, et al. Differentiating moss from higher plants is critical in studying the carbon cycle of the boreal biome. Nature Communications 2014; 5: 4270. pmid:24967601
  54. 54. Massman WJ. A simple method for estimating frequency response corrections for eddy covariance systems. Agricultural and Forest Meteorology 2000; 104: 185–198.
  55. 55. Finnigan JJ, Clement R, Malhi Y, Leuning R, Cleugh HA. A re-evaluation of long-term flux measurement techniques part I: averaging and coordinate rotation. Boundary-Layer Meteorology 2003; 107: 1–48.
  56. 56. Aubinet M, Berbigier P, Bernhofer CH, Cescatti A, Feigenwinter C, Granier A, et al. Comparing CO2 storage and advection conditions at night at different carboeuroflux sites. Boundary-Layer Meteorology 2005; 116: 63–93.