Global Validation of a Process-Based Model on Vegetation Gross Primary Production Using Eddy Covariance Observations

Dan Liu; Wenwen Cai; Jiangzhou Xia; Wenjie Dong; Guangsheng Zhou; Yang Chen; Haicheng Zhang; Wenping Yuan

doi:10.1371/journal.pone.0110407

Abstract

Gross Primary Production (GPP) is the largest flux in the global carbon cycle. However, large uncertainties in current global estimations persist. In this study, we examined the performance of a process-based model (Integrated BIosphere Simulator, IBIS) at 62 eddy covariance sites around the world. Our results indicated that the IBIS model explained 60% of the observed variation in daily GPP at all validation sites. Comparison with a satellite-based vegetation model (Eddy Covariance-Light Use Efficiency, EC-LUE) revealed that the IBIS simulations yielded comparable GPP results as the EC-LUE model. Global mean GPP estimated by the IBIS model was 107.50±1.37 Pg C year⁻¹ (mean value ± standard deviation) across the vegetated area for the period 2000–2006, consistent with the results of the EC-LUE model (109.39±1.48 Pg C year⁻¹). To evaluate the uncertainty introduced by the parameter V_cmax, which represents the maximum photosynthetic capacity, we inversed V_cmax using Markov Chain-Monte Carlo (MCMC) procedures. Using the inversed V_cmax values, the simulated global GPP increased by 16.5 Pg C year⁻¹, indicating that IBIS model is sensitive to V_cmax, and large uncertainty exists in model parameterization.

Citation: Liu D, Cai W, Xia J, Dong W, Zhou G, Chen Y, et al. (2014) Global Validation of a Process-Based Model on Vegetation Gross Primary Production Using Eddy Covariance Observations. PLoS ONE 9(11): e110407. https://doi.org/10.1371/journal.pone.0110407

Editor: Christopher Carcaillet, Ecole Pratique des Hautes Etudes, France

Received: February 7, 2014; Accepted: September 19, 2014; Published: November 6, 2014

Copyright: © 2014 Liu et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Funding: This study was supported by the National Science Foundation for Excellent Young Scholars of China (41322005), the National High Technology Research and Development Program of China (863 Program) (2013AA122003), National Natural Science Foundation of China (41201078), Program for New Century Excellent Talents in University (NCET-322 12-0060) and the Fundamental Research Funds for the Central Universities. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: The authors have declared that no competing interests exist.

Introduction

Terrestrial gross primary production (GPP) is the largest carbon flux in terrestrial ecosystems, and it is approximately 20 times larger than the amount of carbon introduced from anthropogenic sources [1]. Thus, even small fluctuations in GPP can cause large changes in the airborne fraction of carbon and subsequently influence future climate change [2]. Vegetation also contributes to human welfare by providing food, fiber and energy [3]–[4]. Therefore, regular monitoring and reliable estimation of global terrestrial GPP is important for improving our understanding of the global carbon cycle, accurately predicting future climate, and ensuring the long-term sustainability of terrestrial ecosystem services.

Ecosystem models serve as a backbone for evaluating large-scale and global GPP. Two categories of ecosystem models are widely used: process-based and satellite-based. Satellite-based models are driven by remotely sensed data and provide simple means of estimating GPP [5]; however, they are limited in their ability to model mechanisms. Process-based models typically exhibit detailed expressions of terrestrial processes, such as photosynthesis, respiration, phenology, and hydrological cycle. Therefore, process-based models play important roles in investigating the mechanisms underlying current biases in estimated ecosystem production [6], predicting the future conditions of the terrestrial carbon cycle, and exploring its feedback to climate change [7].

Numerous attempts have been made to develop and improve process-based models. However, a recent study from the North American Carbon Project (NACP) showed that current models perform poorly and difference between observations and simulations far exceed the observational uncertainty [8]. Model parameter uncertainty is a key source limiting the accuracy of process-based models. Knorr and Heimann analyzed the uncertainties of process-based models [9]. They found that parameter uncertainties could explain much of the large variance among models and that the largest uncertainties arose from plant photosynthesis, respiration and soil water storage.

The maximum rate of carboxylation by the enzyme Rubisco (V_cmax) is fundamental in modeling photosynthesis [10]. Sensitivity analysis shows that the projections of ecosystem production are particularly sensitive to the fixed parameters associated with V_cmax [11]. Therefore, the parameterization scheme of V_cmax is essential for GPP simulation, and its impacts on model performance need to be tested.

Eddy covariance measurements recorded by the increasing number of eddy covariance (EC) towers provide a great opportunity for model validation and improvement. Concurrent measurements include carbon fluxes, latent heat, and sensible heat, as well as meteorological conditions such as air temperature and relative humidity, and provide unprecedented datasets for model validation and evaluations of parameter constraints [12]. The current network of EC sites covers a wide range of ecosystem types, thus it has the potential to significantly improve our understanding of the variation in GPP across time, space and biomes [13].

The goal of this study was to validate a process-based ecosystem model (Integrated BIosphere Simulator, IBIS) based on measurements from 62 EC sites [14]. The specific objectives were to (1) examine the performance of IBIS over several ecosystem types, (2) compare IBIS model performance with a satellite-based model (i.e., Eddy Covariance-Light Use Efficiency, EC-LUE), and (3) investigate the impacts of the parameter V_cmax on model performance.

Data and Methods

2.1 IBIS model and parameter inversion

The Integrated BIosphere Simulator (IBIS) is designed to integrate a variety of terrestrial ecosystem processes within a single, physically consistent modeling framework. It represents land surface processes, canopy physiology, vegetation phenology, long-term vegetation dynamics, and carbon and water cycling [14]. The photosynthesis module of the IBIS model is provided by the formulations of Farquhar [15].

C₃ photosynthesis and C₄ photosynthesis are expressed separately in the IBIS model. For C₃ plants, the gross photosynthesis rate per unit leaf area, (mol CO₂ m⁻² s⁻¹) is expressed as(1)where is light-limited rate of photosynthesis, represents the Rubisco-limited rate of photosynthesis, and is the photosynthesis limited by the inadequate rate of utilization of triose phosphate.

The light-limited rate of photosynthesis is given as(2)where is the intrinsic quantum efficiency of CO₂ uptake in C₃ plants (mol CO₂ mol⁻¹ quanta), is the flux density of photosynthetically active radiation absorbed by leaf (mol quanta m⁻² s⁻¹), is the concentration of CO₂ in the intercellular air spaces of the leaf (mol mol⁻¹), and is the compensation point for gross photosynthesis (mol mol⁻¹).

The Rubisco-limited rate of photosynthesis is calculated as(3)where is the maximum carboxylase capacity of Rubisco (mol CO₂ m⁻² s⁻¹) and and are the Michaelis-Menten coefficients (mol mol⁻¹) for CO₂ and O₂, respectively.

Under conditions of high intercellular CO₂ concentrations and high irradiance, photosynthesis is limited by the inadequate rate of utilization of triose phosphate. This limitation is expressed as(4)where is the rate of triose phosphate utilization.

Photosynthesis in C₄ plants is similarly modeled as the minimum of three potential capacities to fix carbon [16]. The gross photosynthesis rate is given by(5)where is the light-limited rate of photosynthesis, is the Rubisco-limited rate of photosynthesis and is the CO₂-limited rate of photosynthesis at low CO₂ concentrations.

The parameter V_cmax (mol CO₂ m⁻² s⁻¹) is very important for simulating the photosynthesis process. In the IBIS model, it is established as a constant that differs among plant functional types (PFTs) (Table 1). To validate the IBIS model and investigate the impact of the V_cmax parameter scheme on model performance, we conducted two simulations (IBIS and IBIS-Type) for each site. IBIS simulation used the default values of V_cmax (Table 1). For the IBIS-Type simulation, the vegetation-specific V_cmax values were inversed for each PFT. The Markov Chain-Monte Carlo (MCMC) procedure was used for the parameter inversion, and the Metropolis-Hastings (M-H) algorithm was used as the MCMC sampler [17]–[18] (see Xu et al. [19] and Yuan et al. [20] for detailed descriptions of the MCMC procedure). We conducted 10000 samples for each site and assigned the V_cmax value with the highest frequency as the optimal value of V_cmax. Finally, we ran the model using the inversed V_cmax for each PFT as the IBIS-Type simulation.

Download:

Table 1. The values of V_cmax set in the IBIS model and the initial range of MCMC inversion.

https://doi.org/10.1371/journal.pone.0110407.t001

2.2 EC-LUE model

A satellite-based model (i.e. EC-LUE) [5], [21]–[23] was used to compare the local and global GPP simulations with those of the IBIS model. The EC-LUE model is based on two assumptions: (1) ecosystem GPP has a direct relationship with the absorbed photosynthetically active radiation (APAR) via light use efficiency (LUE), where LUE is defined as the amount of carbon produced per unit of APAR; and (2) realized LUE may be reduced below its theoretical potential value by environmental stressors, such as low temperatures or water shortages. The EC-LUE model is driven by four variables: the normalized difference vegetation index (NDVI), photosynthetically active radiation (PAR), air temperature, and the ratio of sensible to latent heat flux (Bowen ratio).

2.3 Data

We used the eddy covariance (EC) data to validate the IBIS model. We used data obtained from the LaThuile dataset (http://www.fluxdata.org). The daily GPP values were estimated from the eddy covariance measurements using a community standard method [24]. Briefly, GPP was estimated from the equation:(6)where is daytime NEE. Daytime ecosystem respiration was estimated using daytime temperature and an equation describing the temperature dependence of respiration, which was formulated from nighttime NEE measurements. For further details on the algorithm, see Reichstein et al. (2005) [24]. The gap filling and quality control of this dataset are conducted according to standard criteria [25]–[26], and the uncertainty in annual GPP can be controlled to some extent below 100 g C m⁻² year⁻¹ [25]. This dataset is widely used in model studies. In the present study, we selected 62 EC sites for model validation (Table S1). The selected data covered six major terrestrial biomes: evergreen needleleaf forest (ENF), deciduous broadleaf forest (DBF), evergreen broadleaf forest (EBF), mixed forest (MF), grassland (GRA) and savanna (SAV). Additional information on the vegetation, climate and soil characteristics of each site was collected from the associated metadata of the LaThuile dataset. Daily average, maximum and minimum temperature, relative humidity, precipitation, cloud fraction, photosynthetically active radiation, latent heat and sensible heat were used to drive the IBIS, and GPP was used to evaluate its performance.

To decrease model uncertainty, we used the satellite-based leaf area index (LAI) from the Moderate Resolution Imaging Spectroradiometer (MODIS) as a model input. The Normalized Difference Vegetation Index (NDVI) data derived from MODIS were used to drive the EC-LUE model. The 8-day MODIS-NDVI data (MOD13) and the MODIS-LAI (MOD15) data with 1-km spatial resolution were used for model verification at the EC sites. Quality control (QC) flags, which signal cloud contamination in each pixel, were examined to filter out NDVI and LAI data of insufficient quality. We temporally filled missing or unreliable values for each 1-km MODIS pixel based on their corresponding quality assessment data fields, as proposed by Zhao et al. [27]. In addition, data on soil properties, including soil texture, organic carbon content and nitrogen content, were required to input soil information into the model, and were therefore collected from the sites where EC towers are established.

For global simulation, we used meteorological datasets from the Modern Era Retrospective Analysis for Research and Applications (MERRA) archive for 2000–2006 to drive the IBIS and EC-LUE models. MERRA is a NASA reanalysis dataset for the satellite era which uses a new version of the Goddard Earth Observing System Data Assimilation System Version 5 (GEO-5). We used climate conditions at 10 meters above the land surface and at a resolution of 0.5° latitude by 0.6° longitude. The Global Gridded Surfaces of Selected Soil Characteristics datasets were used to supply soil properties for the IBIS model; detailed information is available from the website (http://ww.isric.org). The global distribution of plant functional types (PFTs) was derived by overlapping the MODIS land-cover type product with the Köppen-Geiger climate classification map, with the land cover classifications aggregated into nine PFTs (Table 1). The IBIS model was unable to simulate carbon cycle processes in cropland or wetland; therefore, these two vegetation types were replaced with C₃ grassland in this study.

2.4 Statistical analysis

Three metrics were used to evaluate model performance:

(1) the coefficient of determination, R², which represents how much variation in the observations is explained by the model simulations;

(2) the root mean square error (RMSE), which represents the total difference between the simulated and measured values;

(3) the relative predictive error (RPE), which represents the ratio of error to observation. It is computed as(7)where and are the mean simulated and measured values, respectively.

One-way ANOVA was employed using SPSS software to test the significance of the differences in optimal V_cmax for each biome in the IBIS-Type scheme, and the paired-samples t-test was used to test the significance of the difference in statistical metrics between different models.

Results

3.1 Model validation at EC towers

The overall comparison of the estimated GPP with the EC measurements showed that the IBIS model performed well in capturing the variability in GPP. Across all study sites, the IBIS model explained approximately 60% of the variation in site-averaged GPP (Fig. 1). The coefficient of determination (R²) varied from 0.11 at the ES-LMa site to 0.94 at the CA-Man site, with a mean value of 0.71 across all EC sites. The mean R² values were 0.66, 0.55, 0.81, 0.77, 0.76 and 0.48 for deciduous broadleaf forest, evergreen broadleaf forest, evergreen needleleaf forest, mixed forest, grassland and savanna, respectively (Fig. 2). The root mean square error (RMSE) varied from 1.48 to 2.71 g C m⁻² day⁻¹ among the six vegetation types (Fig. 2). On average, the mean relative predictive error (RPE) of all sites was −9.68%, and the RPE at most sites was less than 20% (Table 2).

Download:

Figure 1. Comparison of predicted and measured GPP.

Comparison between gross primary production (GPP) estimated from Eddy Covariance (EC) measurements and GPP predicted from different model simulations: (a) IBIS, (b) EC-LUE and (c) IBIS-Type. The solid lines are the linear regression lines and the short dashed lines are the 1∶1 lines.

https://doi.org/10.1371/journal.pone.0110407.g001

Download:

Figure 2. Comparison among models for different PFTs.

Comparisons among IBIS, IBIS-Type and EC-LUE models for each plant functional type (PFT), where (a) and (b) are the results of R² and RMSE, respectively. Lowercase letters above the bars indicate significant differences among models.

https://doi.org/10.1371/journal.pone.0110407.g002

Download:

Table 2. The performance of the IBIS model and the EC-LUE model.

https://doi.org/10.1371/journal.pone.0110407.t002

Although the IBIS model explained most of the GPP variability at individual sites, large differences between the predicted and estimated GPP values from the EC measurements were apparent at some sites and for some vegetation types. The IBIS model underestimated GPP for the majority of PFTs and overestimated GPP in evergreen broadleaf forest. Specifically, over 40 of the 62 sites had negative RPE values, and the RPEs of 26 sites were below −20%. The largest underestimation occurred at two savanna sites, ZA-Kru and BW-Ma1, with PREs of −51.83% and −48.15%, respectively. The other 24 sites with RPEs below −20% were predominantly deciduous broadleaf forest (6 sites), evergreen needleleaf forest (7 sites) and grassland (8 sites). In addition, the model overestimated GPP at 6 sites with RPEs greater than 20%, four of which were evergreen broadleaf forest. Extreme overestimation occurred at two sites of evergreen broadleaf forest, FR-Pue and PT-Mi1, with RPEs of 59.41% and 72.49%, respectively.

3.2 Comparison of IBIS and EC-LUE

Compared to the satellite-based EC-LUE model, the IBIS model performed comparably at most sites, according to the R², RMSE and RPE values (Table 2). The IBIS mean R² values for evergreen broadleaf forest, evergreen needleleaf forest, mixed forest and grassland were similar to those of the EC-LUE model, which were 0.52, 0.79, 0.75, 0.83, respectively; no significant differences were found for most PFTs. Comparable results were also found for the RMSE, and no significant differences in RMSE were detected for any PFTs except for evergreen broadleaf forest.

However, the EC-LUE model had some advantages for some PFTs (Figure 2). For broadleaf forest, grassland and savanna, the R² of the EC-LUE was significantly higher than that of the IBIS, particularly for savanna, where the mean was 76% higher than that of the IBIS (0.74 for EC-LUE and 0.42 for IBIS). In addition, the EC-LUE had significantly low model error for evergreen broadleaf forest, with a mean RMSE 22% lower than that of the IBIS (2.12 g C m⁻² day⁻¹ for EC-LUE and 2.71 g C m⁻² day⁻¹ for IBIS).

3.3 IBIS Performance with inversed V_cmax

Model parameterization did not significantly improve the performance of the IBIS, as indicated by the R² and RMSE. The mean R² values of the IBIS-Type were 0.64, 0.51, 0.81, 0.77, 0.76, 0.41 for deciduous broadleaf forest, evergreen broadleaf forest, evergreen needleleaf forest, mixed forest, grassland and savanna, respectively. These values were very similar to those of the IBIS model; only the R² of the deciduous broadleaf forest differed significantly, being higher in the IBIS-Type model. The overall R² increased from 0.60 to 0.68 after revising V_cmax (Fig. 1). The RMSEs of the IBIS and the IBIS-Type models were also comparable. The mean RMSE of the IBIS-Type varied from 1.49 g C m⁻² day⁻¹ for evergreen needleleaf forest to 2.60 g C m⁻² day⁻¹ for deciduous broadleaf forest. Evergreen broadleaf forest was the only vegetation type yielding a significant difference in RMSE between the two models.

The IBIS employs a set of parameter values for a given PFT (Table 1). In the IBIS-Type scheme, the V_cmax value, which was set to as the mean value of V_cmax inversed from each site within a given PFT, was largely differentiated from the original parameter values. V_cmax of the IBIS-Type was 36.67%, 30.00%, 24.29%, 61.09% and 30.00% higher than that of the IBIS for deciduous broadleaf forest, evergreen needleleaf forest, mixed forest, grassland and savanna, respectively; it was 50% lower for evergreen broadleaf forest. In addition, the V_cmax value inversed at each site did not indicate significant differences among different PFTs (Fig. 3).

Download:

Figure 3. Comparison of V_cmax from the original scheme and the inversed V_cmax values.

Comparison of model V_cmax values with the inversed values of the IBIS-Type. Same letter on the top of the bar indicates no significant difference among biomes.

https://doi.org/10.1371/journal.pone.0110407.g003

3.4 Temporal and spatial patterns in global averaged GPP

The spatial pattern in average annual GPP estimated using the original IBIS, the IBIS-Type and the EC-LUE models from 2000 to 2006 were generally consistent (Fig. 4). The highest value was from the humid tropics (Amazonia, Central Africa and Southeast Asia), with an annual GPP over 2000 g C m⁻² year⁻¹. Temperate regions had intermediate levels of GPP, and the lowest GPP was found in both cold and arid regions.

Download:

Figure 4. The global pattern of GPP.

The global pattern of annual vegetation gross primary production (GPP) from 2000 to 2006. (a) estimated GPP using the IBIS model, (b) estimated GPP using the EC-LUE model, (c) estimated GPP using the IBIS-Type model, (d) the difference between the IBIS and EC-LUE models and (e) the spatial distribution of plant functional types (PFTs).

https://doi.org/10.1371/journal.pone.0110407.g004

The magnitude of GPP estimated by the IBIS and EC-LUE models were comparable, reaching 107.50±1.37 Pg C year⁻¹ and 109.39±1.48 Pg C year⁻¹ (mean value ± standard deviation) globally, respectively (Fig. 5). Two-model comparisons revealed consistent GPP estimations for the various PFTs (Table 3) with the exception of savanna, for which the IBIS model greatly underestimated GPP. The GPP estimate of the IBIS-Type scheme was much higher than those of the other two models, with a global value of 123.97±1.76 Pg C year⁻¹ (Fig. 5). Larger GPP estimations using the IBIS-Type simulation were found for most biomes (Table 3).

Download:

Figure 5. Interannual variability in GPP derived from different models.

Interannual variability in global mean gross primary production (GPP) derived from the IBIS, IBIS-Type and EC-LUE models.

https://doi.org/10.1371/journal.pone.0110407.g005

Download:

Table 3. The magnitude of gross primary production (GPP) in each plant functional type (PFT).

https://doi.org/10.1371/journal.pone.0110407.t003

Discussion

Process-based ecosystem models are one of the most important components of earth system models used to predict future climate change [1]. The IPCC AR5 requested that all earth system models integrate the global carbon cycle module [7]. Previous studies have shown that the uncertainty in carbon cycle models can produce 40% differences in the predicted temperature by 2100 [28]. The GPP is the total photosynthetic uptake or carbon assimilation by plants, and it is a key component of terrestrial carbon balance. Any errors in the GPP simulations will propagate through the model, introducing errors into the simulated biomass and net ecosystem fluxes. If the simulated GPP is too low or too high, predicted leaf area index, wood biomass, crop yield, and soil biomass may also be too low or too high [29].

In this study, we examined the performance of the IBIS, which has been widely used to evaluate the regional and global terrestrial ecosystem carbon balance and has been integrated into earth system models (e.g., Brazilian Earth System) [30]. Our results indicate that the IBIS model is a good candidate for simulating GPP at regional-to-global scales, and its performance was comparable to that of the satellite-based EC-LUE model based on EC site validation and comparison. The magnitude of global GPP estimated by the IBIS is also consistent with results of previous studies. Our estimate of annual global mean GPP was 107.50±1.37 Pg C year⁻¹(mean value ± standard deviation). Beer et al. estimated global GPP as 123.8 Pg C year⁻¹ [31]. Two satellite-based light-use efficiency models revealed similar estimates of global GPP: 109.29 Pg C year⁻¹ by the MODIS algorithm [27] and 111 Pg C year⁻¹ by the EC-LUE [21]. Interestingly, we found that the IBIS was consistent with the EC-LUE model across different PFTs (Table 3), despite the large differences between the two approaches (i.e., satellite-based vs. process-based).

However, the IBIS did not perform well for two PFTs: evergreen broadleaf forest and savanna. Among the four evergreen broadleaf forest sites, two (PT-Mi1 and FR-Pue) have subtropical Mediterranean climate with dry summers and wet winters [32]–[33], and the AU-Tum site also suffers drought in summer [34]. At these sites, rainfall is the key driver of water and carbon fluxes [35]. Leuning analyzed the CO₂ and H₂O fluxes at the AU-Tum site [34], and found that carbon uptake was more strongly constrained by water stress than by temperature, and strongly affected by soil water availability. Moreover, savannas are characterized by climate with distinct wet and dry seasons, and this climate forcing causes savanna to form open, heterogeneous woodland canopies with grass understories [36].

These particular ecosystem properties result in greater complexity of modeling fluxes [37]. Process-based models need to simulate variation in soil moisture and plant phenology. However, previous studies have identified significant biases when simulating soil moisture [38]. We examined the performance of the IBIS on soil water at the savanna sites and also found obvious differences between the simulated values and the EC measurements (Fig. 6). Moreover, the IBIS integrated temperature-dominated phenology algorithms developed by Botta et al. [39]. However, field studies suggest that for many drought-deciduous species, the first large precipitation event at the start of the rainy season initiates rapid leaf flush [40]–[41], and leaf senescence is closely related to soil water availability in the dry season [41]–[42]. This relationship may explain why the IBIS model did not effectively capture the variance in GPP at savanna sites.

Download:

Figure 6. Comparison of estimated and observed soil water.

Daily variation in the estimated soil water fraction of the IBIS model (i.e., fraction of soil pore space containing liquid water) and in the observed soil water content at EC sites. The solid lines represent simulation data, and the dotted lines represent the observed data.

https://doi.org/10.1371/journal.pone.0110407.g006

Parameterization is another large source of uncertainty in process-based models. V_cmax is the key parameter of the photosynthesis process [11], and a study by Bonan et al. [43] suggests that uncertainty in this parameter could account for 30 Pg C year⁻¹ variations in model estimation of global GPP. In the present study, the differences in setting V_cmax values between the IBIS and IBIS-Type schemes caused an increase in global GPP of over 16.5 Pg C year⁻¹, which also indicated that V_cmax is a salient parameter for simulating GPP. Unfortunately, the determination of V_cmax in current models contains large uncertainty. Rogers surveyed V_cmax in current state-of-the-art models [44] and found that V_cmax varied within a wide range of −46 to +77% of the PFT mean. Thus, the determination of V_cmax and the reduction of uncertainty in this parameter are important issues for model development.

Many parameters in process-based models are established by PFTs, which are based on the assumption that the same type of vegetation responds similarly to the environment. However, a current study found that model parameters were more variable than previously assumed within the given PFTs [45] and that categorization of vegetation into less than eight PFTs may result in artificial multiple steady-states in a model of the Earth's climate-vegetation system depending on the number of PFTs used [46]. In the present study, the variation analysis showed that V_cmax did not significantly differ among PFTs. The predetermined parameterization scheme that sets the V_cmax constant values for each PFT in the IBIS model may cause systematic error.

We attempted to test the impact of V_cmax on model performance. However, variations in V_cmax cannot explain the overall uncertainty of the IBIS model. Validation of the IBIS on three flux sites demonstrated that parameterization and formulations of phenology also limit the model's ability to capture seasonal fluctuations in carbon and water exchange [38]. Particularly in regions with summer drought, phenology is primarily controlled by water supply rather than temperature [34]. This relationship hinders model simulation because biases in phenology and the dynamics of the leaf area index can affect the simulation of evapotranspiration. Such errors can pass to simulations of soil water content and other variables associated with the water and carbon cycles [38]. Therefore, revisions of not only the parameterization of V_cmax but also other parameters and formulations are needed for model improvement.

Summary

Process-based models are important tools for carbon cycle research, but current models incorporate substantial uncertainty. This study examines the performance of the IBIS model at global EC sites. Our results showed that the IBIS model explained 60% of the variation in GPP at all EC sites and performed comparably to the EC-LUE model, which explained 80% of the variation in observed GPP. At the global scale, the magnitudes of GPP estimated by the IBIS and EC-LUE models were comparable, being 107.50±1.37 Pg C year⁻¹ and 109.39±1.48 Pg C year⁻¹, respectively. The parameter V_cmax is a key parameter in the photosynthesis model. In the IBIS model, V_cmax was set as a constant for each PFT. The inversed V_cmax value was largely differentiated from the original setting, and no significant differences were detected among PFTs.

Supporting Information

Table S1.

Information of EC sites used in this study.

https://doi.org/10.1371/journal.pone.0110407.s001

(DOCX)

Acknowledgments

This work used eddy covariance data acquired by the FLUXNET community; in particular, by the following networks: AmeriFlux (U.S. Department of Energy, Biological and Environmental Research, Terrestrial Carbon Program (DE-FG02-04ER63917 and DE-FG02-04ER63911), AfriFlux, AsiaFlux, CarboAfrica, CarboEuropeIP, CarboItaly, CarboMont, ChinaFlux, Fluxnet-Canada (supported by CFCAS, NSERC, BIOCAP, Environment Canada, and NRCan), GreenGrass, KoFlux, LBA, NECC, OzFlux, TCOS-Siberia, USCCC. We acknowledge the financial support of the eddy covariance data harmonization provided by CarboEuropeIP, FAO-GTOS-TCO, iLEAPS, Max Planck Institute for Biogeochemistry, National Science Foundation, University of Tuscia, Université Laval, Environment Canada and the US Department of Energy. We also acknowledge the database development and technical support from Berkeley Water Center, Lawrence Berkeley National Laboratory, Microsoft Research Science, Oak Ridge National Laboratory, University of California–Berkeley and the University of Virginia.

Author Contributions

Conceived and designed the experiments: WY. Performed the experiments: DL. Analyzed the data: DL. Wrote the paper: DL WC JX WD GZ YC HZ WY.

References

1. IPCC (2007) Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change [Solomon S, Qin D, Manning M, Chen Z, Marquis M, et al. (eds).]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA.
2. Raupach MR, Canadell JG, Quere CL (2008) Anthropogenic and biophysical contributions to increasing atmospheric CO₂ growth rate and airborne fraction. Biogeosciences 5(6): 1601–1613.
- View Article
- Google Scholar
3. Rojstaczer S, Sterling SM, Moore NJ (2001) Human appropriation of photosynthesis products. Science 294(5551): 2549–2552.
- View Article
- Google Scholar
4. Imhoff ML, Bounoua L, Ricketts T, Loucks C, Harriss R, et al. (2004) Global patterns in human consumption of net primary production. Nature 429(6994): 870–873.
- View Article
- Google Scholar
5. Yuan WP, Liu SG, Zhou GS, Zhou GY, Tieszen LL, et al. (2007) Deriving a light use efficiency model from eddy covariance flux data for predicting daily gross primary production across biomes. Agricultural and Forest Meteorology 143(3–4): 189–207.
- View Article
- Google Scholar
6. Zhang FM, Chen JM, Chen JQ, Gough CM, Martin TA, et al. (2012) Evaluating spatial and temporal patterns of MODIS GPP over the conterminous US against flux measurements and a process model. Remote Sensing of Environment 124: 717–729.
- View Article
- Google Scholar
7. Friedlingstein P, Cox P, Betts R, Bopp L, von Bloh W, et al. (2006) Climate-carbon cycle feedback analysis: Results from the C4MIP model intercomparison. Journal of Climate 19(14): 3337–3353.
- View Article
- Google Scholar
8. Schwalm CR, Williams CA, Schaefer K, Anderson R, Arain MA, et al. (2010) A model-data intercomparison of CO₂ exchange across North America: Results from the North American Carbon Program site synthesis. Journal of Geophysical Research: Biogeosciences 153(G3): G00H05.
- View Article
- Google Scholar
9. Knorr W, Heimann M (2001) Uncertainties in global terrestrial biosphere modeling 1. A comprehensive sensitivity analysis with a new photosynthesis and energy balance scheme. Global Biogeochemical Cycles 15(1): 207–225.
- View Article
- Google Scholar
10. Wolf A, Akshalov K, Saliendra N, Johnson DA, Laca EA (2006) Inverse estimation of V_cmax, leaf area index, and the Ball-Berry parameter from carbon and energy fluxes. Journal of Geophysical Research: Atmosphere 111(D8) DOI:10.1029/2005JD005927.
11. Friend AD (2010) Terrestrial plant production and climate change. Journal of Experimental Botany 61(5): 1293–1309.
- View Article
- Google Scholar
12. Friend AD, Arneth A, Kiang NY, Lomas M, Ogee J, et al. (2007) FLUXNET and modeling the global carbon cycle. Global Change Biology 13(3): 610–633.
- View Article
- Google Scholar
13. Baldocchi D, Falge E, Gu LH, Olson R, Hollinger D, et al. (2001) 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 82(11): 2415–2434.
- View Article
- Google Scholar
14. Foley JA, Prentice IC, Ramankutty N, Levis S, Pollard D, et al. (1996) An integrated biosphere model of land surface processes, terrestrial carbon balance, and vegetation dynamics. Global Biogeochemical Cycles 10(4): 603–628.
- View Article
- Google Scholar
15. Farquhar GD, Caemmerer SV, Berry JA (1980) A biochemical-model of photosynthetic CO₂ assimilation in leave of C-3 species. Planta 149(1): 79–90.
- View Article
- Google Scholar
16. Collatz GJ, Ball JT, Grivet C, Berry JA (1991) Physiological and environmental-regulation of stomata conductance, photosynthesis and transpiration - a model that includes a laminar boundary-layer. Agricultural and Forest Meteorology 54(2–4): 107–136.
- View Article
- Google Scholar
17. Metropolis N, Rosenbluth AW, Rosenbluth MN, Teller AH, Teller E (1953) Equation of state calculations by fast computing machimes. Journal of Chemical Physics 21(6): 1087–1092.
- View Article
- Google Scholar
18. Hastings WK (1970) Monte-Carlo sampling methods using Markov Chains and their applications. Biometrika 57(1): 97.
- View Article
- Google Scholar
19. Xu T, White L, Hui DF, Luo YQ (2006) Probabilistic inversion of a terrestrial ecosystem model: Analysis of uncertainty in parameter estimation and model prediction, Global Biogeochemical Cycles 20(2) DOI:10.1029/2005GB002468.
20. Yuan WP, Liang SL, Liu SG, Weng ES, Luo YQ, et al. (2012) Improving model parameter estimation using coupling relationships between vegetation production and ecosystem respiration. Ecological Modelling 240: 29–40.
- View Article
- Google Scholar
21. Yuan WP, Liu SG, Yu GR, Bonnefond JM, Chen JQ, et al. (2010) Global estimates of evapotranspiration and gross primary production based on MODIS and global meteorology data. Remote Sensing of Environment 114(7): 1416–1431.
- View Article
- Google Scholar
22. Yuan W, Luo Y, Li X, Liu S, Yu G, et al. (2011) Redefinition and global estimation of basal ecosystem respiration rate. Global Biogeochemical Cycles 25(4) DOI:10.1029/2011GB004150.
23. Li X, Liang S, Yu G, Yuan W, Cheng X, et al. (2013) Estimation of gross primary production over the terrestrial ecosystems in China. Ecological Modelling 261–262(0): 80–92.
- View Article
- Google Scholar
24. Reichstein M, Falge E, Baldocchi D, Papale D, Aubinet M, et al. (2005) On the separation of net 20 ecosystem exchange into assimilation and ecosystem respiration: review and improved algorithm. Global Change Biology 11: 1424–1439.
- View Article
- Google Scholar
25. Papale D, Reichstein M, Aubinet M, Canfora E, Bernhofer C, et al. (2006) Towards a standardized processing of Net Ecosystem Exchange measured with eddy covariance technique: algorithms and uncertainty estimation. Biogeosciences 3: 571–583.
- View Article
- Google Scholar
26. Moffat AM, Papale D, Reichstein M, Hollinger DY, Richardson AD, et al. (2007) Comprehensive comparison of gap-filling techniques for eddy covariance net carbon fluxes. Agricultural and Forest Meteorology 147(3–4): 209–232.
- View Article
- Google Scholar
27. Zhao M, Heinsch FA, Nemani RR, Runing SW (2005) Improvements of the MODIS terrestrial gross and net primary production global data set. Remote Sensing of Environment 95(2): 164–176.
- View Article
- Google Scholar
28. Huntingford C, Lowe JA, Booth BBB, Jones CD, Harris GR, et al. (2009) Contributions of carbon cycle uncertainty to future climate projection spread. Tellus B 61(2): 355–360.
- View Article
- Google Scholar
29. Schaefer K, Collatz JG, Tans P, Denning SA, Baker I, et al. (2008) Combined Simple Biosphere/Carnegie-Ames-Stanford Approach terrestrial carbon cycle model. Journal of Geophysical Research-Biogeosciences 113(G3) DOI:10.1029/2007JG000603.
30. Nobre P, Siqueira LSP, de Almeida RAF, Malaguitti M, Giarolla E, et al. (2013) Climate simulation and change in the Brazilian Climate Model. Journal of Climate 26(17): 6716–6732.
- View Article
- Google Scholar
31. Beer C, Reichstein M, Tomelleri E, Ciais P, Jung M, et al. (2010) Terrestrial gross carbon dioxide uptake: Global distribution and covariation with climate. Science 329(5993): 834–838.
- View Article
- Google Scholar
32. Reichstein M, Tenhunen J, Roupsard O, Ourcival JM, Rambal S, et al. (2003) Inverse modeling of seasonal drought effects on canopy CO₂/H₂O exchange in three Mediterranean ecosystems. Journal of Geophysical Research Atmospheres 108(D23): 4726.
- View Article
- Google Scholar
33. David TS, Ferreira MI, Cohen S, Pereira JS, David JS (2004) Constraints on transpiration from an evergreen oak tree in southern Portugal. Agricultural and Forest Meteorology 122(3–4): 193–205.
- View Article
- Google Scholar
34. Leuning R, Cleugh HA, Zegelin SJ, Hughes D (2005) Carbon and water fluxes over a temperate Eucalyptus forest and a tropical wet/dry savanna in Australia: measurements and comparison with MODIS remote sensing estimates. Agricultural and Forest Meteorology 129(3–4): 151–173.
- View Article
- Google Scholar
35. Yuan W, Luo Y, Richardson AD, Oren R, Luyssaert S, et al. (2009) Latitudinal patterns of magnitude and interannual variability in net ecosystem exchange regulated by biological and environmental variables. Global Change Biology 15(12): 2905–2920.
- View Article
- Google Scholar
36. Eamus D, Prior L (2001) Ecophysiology of trees of seasonally dry tropics: Comparisons among phenologies. Advances in Ecological Research 32: 113–197.
- View Article
- Google Scholar
37. Baldocchi D, Xu L, Kiang N (2004) How plant functional-type, weather, seasonal drought, and soil physical properties alter water energy fluxes for an oak-grass savanna and an annual grassland. Agricultural and Forest Meteorology 123(1–2): 13–39.
- View Article
- Google Scholar
38. Kucharik CJ, Barford CC, Maayar ME, Wofsy SC, Monson RK, et al. (2006) A multiyear evaluation of a Dynamic Global Vegetation Model at three AmeriFlux forest sites: Vegetation structure, phenology, soil temperature, and CO₂ and H₂O vapor exchange. Ecological Modelling 196(1–2): 1–31.
- View Article
- Google Scholar
39. Botta A, Viovy N, Ciais P, Friedlingstein P, Monfray P (2000) A global prognostic scheme of leaf onset using satellite data. Global Change Biology 6(7): 709–725.
- View Article
- Google Scholar
40. Monasterio M, Sarmiento G (1976) Phenological strategies of plant species in the tropical savanna and the semi-deciduous forest of the Venezuelan Llanos. Journal of Biogeography 3(4): 325–355.
- View Article
- Google Scholar
41. Borchert R (1994) Soil and stem water storage determine phenology and distribution of tropical dry forest trees. Ecology 75(5): 1437–1449.
- View Article
- Google Scholar
42. Childes SL (1988) Phenology of nine common woody species in semi-arid, deciduous Kalahari Sand vegetation. Vegetatio 79(3): 151–163.
- View Article
- Google Scholar
43. Bonan GB, Lawrence PJ, Oleson KW, Levis S, Jung M, Reichstein M, et al. (2011) Improving canopy processes in the Community Land Model version 4 (CLM4) using global flux fields empirically inferred from FLUXNET data. Journal of Geophysical Research-Biogeosciences 116(G2) DOI:10.1029/2010JG001593.
44. Rogers A (2013) The use and misuse of V_cmax in Earth System Models. Photosynthesis Research 1–15.
45. Groenendijk M, Dolman AJ, van der Molen MK, Leuning R, Arneth A, et al. (2011) Assessing parameter variability in a photosynthesis model within and between plant functional types using global Fluxnet eddy covariance data. Agricultural and Forest Meteorology 151(1): 22–38.
- View Article
- Google Scholar
46. Kleidon A, Fraedrich K, Low C (2007) Multiple steady-states in the terrestrial atmosphere-biosphere system: a result of a discrete vegetation classification. Biogeosciences 4(5): 707–714.
- View Article
- Google Scholar
47. Kucharik CJ, Foley JA, Delire C, Fisher VA, Coe MT, et al. (2000) Testing the performance of a Dynamic Global Ecosystem Model: Water balance, carbon balance and vegetation structure. Global Biogeochemical Cycles 14(3): 795–825.
- View Article
- Google Scholar

[ref1] 1. IPCC (2007) Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change [Solomon S, Qin D, Manning M, Chen Z, Marquis M, et al. (eds).]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA.

[ref2] 2. Raupach MR, Canadell JG, Quere CL (2008) Anthropogenic and biophysical contributions to increasing atmospheric CO₂ growth rate and airborne fraction. Biogeosciences 5(6): 1601–1613.
View Article
Google Scholar

[3] View Article

[4] Google Scholar

[ref3] 3. Rojstaczer S, Sterling SM, Moore NJ (2001) Human appropriation of photosynthesis products. Science 294(5551): 2549–2552.
View Article
Google Scholar

[6] View Article

[7] Google Scholar

[ref4] 4. Imhoff ML, Bounoua L, Ricketts T, Loucks C, Harriss R, et al. (2004) Global patterns in human consumption of net primary production. Nature 429(6994): 870–873.
View Article
Google Scholar

[9] View Article

[10] Google Scholar

[ref5] 5. Yuan WP, Liu SG, Zhou GS, Zhou GY, Tieszen LL, et al. (2007) Deriving a light use efficiency model from eddy covariance flux data for predicting daily gross primary production across biomes. Agricultural and Forest Meteorology 143(3–4): 189–207.
View Article
Google Scholar

[12] View Article

[13] Google Scholar

[ref6] 6. Zhang FM, Chen JM, Chen JQ, Gough CM, Martin TA, et al. (2012) Evaluating spatial and temporal patterns of MODIS GPP over the conterminous US against flux measurements and a process model. Remote Sensing of Environment 124: 717–729.
View Article
Google Scholar

[15] View Article

[16] Google Scholar

[ref7] 7. Friedlingstein P, Cox P, Betts R, Bopp L, von Bloh W, et al. (2006) Climate-carbon cycle feedback analysis: Results from the C4MIP model intercomparison. Journal of Climate 19(14): 3337–3353.
View Article
Google Scholar

[18] View Article

[19] Google Scholar

[ref8] 8. Schwalm CR, Williams CA, Schaefer K, Anderson R, Arain MA, et al. (2010) A model-data intercomparison of CO₂ exchange across North America: Results from the North American Carbon Program site synthesis. Journal of Geophysical Research: Biogeosciences 153(G3): G00H05.
View Article
Google Scholar

[21] View Article

[22] Google Scholar

[ref9] 9. Knorr W, Heimann M (2001) Uncertainties in global terrestrial biosphere modeling 1. A comprehensive sensitivity analysis with a new photosynthesis and energy balance scheme. Global Biogeochemical Cycles 15(1): 207–225.
View Article
Google Scholar

[24] View Article

[25] Google Scholar

[ref10] 10. Wolf A, Akshalov K, Saliendra N, Johnson DA, Laca EA (2006) Inverse estimation of V_cmax, leaf area index, and the Ball-Berry parameter from carbon and energy fluxes. Journal of Geophysical Research: Atmosphere 111(D8) DOI:10.1029/2005JD005927.

[ref11] 11. Friend AD (2010) Terrestrial plant production and climate change. Journal of Experimental Botany 61(5): 1293–1309.
View Article
Google Scholar

[28] View Article

[29] Google Scholar

[ref12] 12. Friend AD, Arneth A, Kiang NY, Lomas M, Ogee J, et al. (2007) FLUXNET and modeling the global carbon cycle. Global Change Biology 13(3): 610–633.
View Article
Google Scholar

[31] View Article

[32] Google Scholar

[ref13] 13. Baldocchi D, Falge E, Gu LH, Olson R, Hollinger D, et al. (2001) 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 82(11): 2415–2434.
View Article
Google Scholar

[34] View Article

[35] Google Scholar

[ref14] 14. Foley JA, Prentice IC, Ramankutty N, Levis S, Pollard D, et al. (1996) An integrated biosphere model of land surface processes, terrestrial carbon balance, and vegetation dynamics. Global Biogeochemical Cycles 10(4): 603–628.
View Article
Google Scholar

[37] View Article

[38] Google Scholar

[ref15] 15. Farquhar GD, Caemmerer SV, Berry JA (1980) A biochemical-model of photosynthetic CO₂ assimilation in leave of C-3 species. Planta 149(1): 79–90.
View Article
Google Scholar

[40] View Article

[41] Google Scholar

[ref16] 16. Collatz GJ, Ball JT, Grivet C, Berry JA (1991) Physiological and environmental-regulation of stomata conductance, photosynthesis and transpiration - a model that includes a laminar boundary-layer. Agricultural and Forest Meteorology 54(2–4): 107–136.
View Article
Google Scholar

[43] View Article

[44] Google Scholar

[ref17] 17. Metropolis N, Rosenbluth AW, Rosenbluth MN, Teller AH, Teller E (1953) Equation of state calculations by fast computing machimes. Journal of Chemical Physics 21(6): 1087–1092.
View Article
Google Scholar

[46] View Article

[47] Google Scholar

[ref18] 18. Hastings WK (1970) Monte-Carlo sampling methods using Markov Chains and their applications. Biometrika 57(1): 97.
View Article
Google Scholar

[49] View Article

[50] Google Scholar

[ref19] 19. Xu T, White L, Hui DF, Luo YQ (2006) Probabilistic inversion of a terrestrial ecosystem model: Analysis of uncertainty in parameter estimation and model prediction, Global Biogeochemical Cycles 20(2) DOI:10.1029/2005GB002468.

[ref20] 20. Yuan WP, Liang SL, Liu SG, Weng ES, Luo YQ, et al. (2012) Improving model parameter estimation using coupling relationships between vegetation production and ecosystem respiration. Ecological Modelling 240: 29–40.
View Article
Google Scholar

[53] View Article

[54] Google Scholar

[ref21] 21. Yuan WP, Liu SG, Yu GR, Bonnefond JM, Chen JQ, et al. (2010) Global estimates of evapotranspiration and gross primary production based on MODIS and global meteorology data. Remote Sensing of Environment 114(7): 1416–1431.
View Article
Google Scholar

[56] View Article

[57] Google Scholar

[ref22] 22. Yuan W, Luo Y, Li X, Liu S, Yu G, et al. (2011) Redefinition and global estimation of basal ecosystem respiration rate. Global Biogeochemical Cycles 25(4) DOI:10.1029/2011GB004150.

[ref23] 23. Li X, Liang S, Yu G, Yuan W, Cheng X, et al. (2013) Estimation of gross primary production over the terrestrial ecosystems in China. Ecological Modelling 261–262(0): 80–92.
View Article
Google Scholar

[60] View Article

[61] Google Scholar

[ref24] 24. Reichstein M, Falge E, Baldocchi D, Papale D, Aubinet M, et al. (2005) On the separation of net 20 ecosystem exchange into assimilation and ecosystem respiration: review and improved algorithm. Global Change Biology 11: 1424–1439.
View Article
Google Scholar

[63] View Article

[64] Google Scholar

[ref25] 25. Papale D, Reichstein M, Aubinet M, Canfora E, Bernhofer C, et al. (2006) Towards a standardized processing of Net Ecosystem Exchange measured with eddy covariance technique: algorithms and uncertainty estimation. Biogeosciences 3: 571–583.
View Article
Google Scholar

[66] View Article

[67] Google Scholar

[ref26] 26. Moffat AM, Papale D, Reichstein M, Hollinger DY, Richardson AD, et al. (2007) Comprehensive comparison of gap-filling techniques for eddy covariance net carbon fluxes. Agricultural and Forest Meteorology 147(3–4): 209–232.
View Article
Google Scholar

[69] View Article

[70] Google Scholar

[ref27] 27. Zhao M, Heinsch FA, Nemani RR, Runing SW (2005) Improvements of the MODIS terrestrial gross and net primary production global data set. Remote Sensing of Environment 95(2): 164–176.
View Article
Google Scholar

[72] View Article

[73] Google Scholar

[ref28] 28. Huntingford C, Lowe JA, Booth BBB, Jones CD, Harris GR, et al. (2009) Contributions of carbon cycle uncertainty to future climate projection spread. Tellus B 61(2): 355–360.
View Article
Google Scholar

[75] View Article

[76] Google Scholar

[ref29] 29. Schaefer K, Collatz JG, Tans P, Denning SA, Baker I, et al. (2008) Combined Simple Biosphere/Carnegie-Ames-Stanford Approach terrestrial carbon cycle model. Journal of Geophysical Research-Biogeosciences 113(G3) DOI:10.1029/2007JG000603.

[ref30] 30. Nobre P, Siqueira LSP, de Almeida RAF, Malaguitti M, Giarolla E, et al. (2013) Climate simulation and change in the Brazilian Climate Model. Journal of Climate 26(17): 6716–6732.
View Article
Google Scholar

[79] View Article

[80] Google Scholar

[ref31] 31. Beer C, Reichstein M, Tomelleri E, Ciais P, Jung M, et al. (2010) Terrestrial gross carbon dioxide uptake: Global distribution and covariation with climate. Science 329(5993): 834–838.
View Article
Google Scholar

[82] View Article

[83] Google Scholar

[ref32] 32. Reichstein M, Tenhunen J, Roupsard O, Ourcival JM, Rambal S, et al. (2003) Inverse modeling of seasonal drought effects on canopy CO₂/H₂O exchange in three Mediterranean ecosystems. Journal of Geophysical Research Atmospheres 108(D23): 4726.
View Article
Google Scholar

[85] View Article

[86] Google Scholar

[ref33] 33. David TS, Ferreira MI, Cohen S, Pereira JS, David JS (2004) Constraints on transpiration from an evergreen oak tree in southern Portugal. Agricultural and Forest Meteorology 122(3–4): 193–205.
View Article
Google Scholar

[88] View Article

[89] Google Scholar

[ref34] 34. Leuning R, Cleugh HA, Zegelin SJ, Hughes D (2005) Carbon and water fluxes over a temperate Eucalyptus forest and a tropical wet/dry savanna in Australia: measurements and comparison with MODIS remote sensing estimates. Agricultural and Forest Meteorology 129(3–4): 151–173.
View Article
Google Scholar

[91] View Article

[92] Google Scholar

[ref35] 35. Yuan W, Luo Y, Richardson AD, Oren R, Luyssaert S, et al. (2009) Latitudinal patterns of magnitude and interannual variability in net ecosystem exchange regulated by biological and environmental variables. Global Change Biology 15(12): 2905–2920.
View Article
Google Scholar

[94] View Article

[95] Google Scholar

[ref36] 36. Eamus D, Prior L (2001) Ecophysiology of trees of seasonally dry tropics: Comparisons among phenologies. Advances in Ecological Research 32: 113–197.
View Article
Google Scholar

[97] View Article

[98] Google Scholar

[ref37] 37. Baldocchi D, Xu L, Kiang N (2004) How plant functional-type, weather, seasonal drought, and soil physical properties alter water energy fluxes for an oak-grass savanna and an annual grassland. Agricultural and Forest Meteorology 123(1–2): 13–39.
View Article
Google Scholar

[100] View Article

[101] Google Scholar

[ref38] 38. Kucharik CJ, Barford CC, Maayar ME, Wofsy SC, Monson RK, et al. (2006) A multiyear evaluation of a Dynamic Global Vegetation Model at three AmeriFlux forest sites: Vegetation structure, phenology, soil temperature, and CO₂ and H₂O vapor exchange. Ecological Modelling 196(1–2): 1–31.
View Article
Google Scholar

[103] View Article

[104] Google Scholar

[ref39] 39. Botta A, Viovy N, Ciais P, Friedlingstein P, Monfray P (2000) A global prognostic scheme of leaf onset using satellite data. Global Change Biology 6(7): 709–725.
View Article
Google Scholar

[106] View Article

[107] Google Scholar

[ref40] 40. Monasterio M, Sarmiento G (1976) Phenological strategies of plant species in the tropical savanna and the semi-deciduous forest of the Venezuelan Llanos. Journal of Biogeography 3(4): 325–355.
View Article
Google Scholar

[109] View Article

[110] Google Scholar

[ref41] 41. Borchert R (1994) Soil and stem water storage determine phenology and distribution of tropical dry forest trees. Ecology 75(5): 1437–1449.
View Article
Google Scholar

[112] View Article

[113] Google Scholar

[ref42] 42. Childes SL (1988) Phenology of nine common woody species in semi-arid, deciduous Kalahari Sand vegetation. Vegetatio 79(3): 151–163.
View Article
Google Scholar

[115] View Article

[116] Google Scholar

[ref43] 43. Bonan GB, Lawrence PJ, Oleson KW, Levis S, Jung M, Reichstein M, et al. (2011) Improving canopy processes in the Community Land Model version 4 (CLM4) using global flux fields empirically inferred from FLUXNET data. Journal of Geophysical Research-Biogeosciences 116(G2) DOI:10.1029/2010JG001593.

[ref44] 44. Rogers A (2013) The use and misuse of V_cmax in Earth System Models. Photosynthesis Research 1–15.

[ref45] 45. Groenendijk M, Dolman AJ, van der Molen MK, Leuning R, Arneth A, et al. (2011) Assessing parameter variability in a photosynthesis model within and between plant functional types using global Fluxnet eddy covariance data. Agricultural and Forest Meteorology 151(1): 22–38.
View Article
Google Scholar

[120] View Article

[121] Google Scholar

[ref46] 46. Kleidon A, Fraedrich K, Low C (2007) Multiple steady-states in the terrestrial atmosphere-biosphere system: a result of a discrete vegetation classification. Biogeosciences 4(5): 707–714.
View Article
Google Scholar

[123] View Article

[124] Google Scholar

[ref47] 47. Kucharik CJ, Foley JA, Delire C, Fisher VA, Coe MT, et al. (2000) Testing the performance of a Dynamic Global Ecosystem Model: Water balance, carbon balance and vegetation structure. Global Biogeochemical Cycles 14(3): 795–825.
View Article
Google Scholar

[126] View Article

[127] Google Scholar

Figures

Abstract

Introduction

Data and Methods

2.1 IBIS model and parameter inversion

2.2 EC-LUE model

2.3 Data

2.4 Statistical analysis

Results

3.1 Model validation at EC towers

3.2 Comparison of IBIS and EC-LUE

3.3 IBIS Performance with inversed Vcmax

3.4 Temporal and spatial patterns in global averaged GPP

Discussion

Summary

Supporting Information

Table S1.

Acknowledgments

Author Contributions

References

3.3 IBIS Performance with inversed V_cmax