globalchange  > 气候变化事实与影响
DOI: doi:10.1038/nclimate2621
论文题名:
Using ecosystem experiments to improve vegetation models
作者: Belinda E. Medlyn
刊名: Nature Climate Change
ISSN: 1758-906X
EISSN: 1758-7026
出版年: 2015-05-21
卷: Volume:5, 页码:Pages:528;534 (2015)
语种: 英语
英文关键词: Biogeochemistry ; Ecological modelling
英文摘要:

Ecosystem responses to rising CO2 concentrations are a major source of uncertainty in climate change projections. Data from ecosystem-scale Free-Air CO2 Enrichment (FACE) experiments provide a unique opportunity to reduce this uncertainty. The recent FACE Model–Data Synthesis project aimed to use the information gathered in two forest FACE experiments to assess and improve land ecosystem models. A new 'assumption-centred' model intercomparison approach was used, in which participating models were evaluated against experimental data based on the ways in which they represent key ecological processes. By identifying and evaluating the main assumptions causing differences among models, the assumption-centred approach produced a clear roadmap for reducing model uncertainty. Here, we explain this approach and summarize the resulting research agenda. We encourage the application of this approach in other model intercomparison projects to fundamentally improve predictive understanding of the Earth system.

The response of the terrestrial biosphere to increasing atmospheric CO2 concentration (Ca) is a major uncertainty in models projecting future climate change, because of the critical feedback between terrestrial ecosystem carbon (C) cycling and the atmosphere1, 2, 3. Current Earth system models disagree strongly on the size of this feedback2. This disagreement results from differences in the projected increase of plant production owing to CO2 fertilization, including its interactions with terrestrial nitrogen (N)4, 5, 6 and phosphorus7 cycles, as well as differences in the turnover times of C in terrestrial ecosystems8. These differences among models imply that our predictive understanding of the effects of Ca on terrestrial C storage remains very low.

FACE experiments aim to investigate how terrestrial ecosystems respond to elevated atmospheric CO2 concentration (eCa). In general, FACE experiments are ecosystem-scale, long-term, multi-investigator experiments that provide an extraordinarily rich source of data on plant and soil processes mediating ecosystem-level responses to eCa (refs 9,10). However, the full richness of these experiments has rarely been exploited to constrain model uncertainty, with model outputs typically only being compared against the response of net primary production (for example, refs 11,12,13).

The FACE Model–Data Synthesis (FACE-MDS) project14 aimed to benefit from the wide range of complementary data sets available from these experiments to better constrain eCa responses in models. To do so, we followed an 'assumption-centred' approach, in which we studied the underlying assumptions that models use to represent key ecosystem processes, and evaluated which assumptions best represented the experimental data. We used 11 process-based models, including four stand-scale ecosystem models (DAYCENT15, ED216, GDAY17 and TECO18), five land surface models (CABLE19, CLM420, EALCO21, ISAM22 and O-CN23) and two dynamic vegetation models (LPJ-GUESS24 and SDGVM25). These models were compared with data from two temperate FACE experiments on established forest plantations: the Duke FACE experiment on Pinus taeda26 and the Oak Ridge National Laboratory (ORNL) FACE experiment on Liquidambar styraciflua27 (Fig. 1). These two experiments have the advantage of being stand-level, ecosystem experiments in established forests that are readily comparable with ecosystem-scale models. Both experiments ran for more than ten years, during which time data were collected on many aspects of ecosystem function.

Figure 1: Aerial views of FACE experiments.
Aerial views of FACE experiments.

a, ORNL FACE experiment; b, Duke FACE experiment.

© CURTIS BOLES, ORNL © WILL OWENS, DUKE UNIVERSITY

Model intercomparisons against data often use a 'benchmarking' approach38 (Fig. 2a), in which models are compared against a suite of observed system characteristics, and then ranked according to how well they replicated the observed patterns. While this approach can identify a 'best' model for a given data set (or suite of data sets), it rarely leads to model improvement because it does not identify the reasons for good or bad model performance. It also overlooks the problems of equifinality and parameter tuning, which can lead to models performing well for the wrong reasons39.

Figure 2: Comparison of the benchmarking and model–data synthesis approaches to model intercomparison.
Comparison of the benchmarking and model-data synthesis approaches to model intercomparison.

a, Model benchmarking/validation; b, Model–experiment synthesis. The assumption-centred approach translates model evaluation into hypothesis testing, allowing a two-way flow of information between modelling and experiment to improve our predictive understanding of the system. Figure adapted with permission from ref. 14, AGU. Tree diagram © nikitinaolga/iStock/Thinkstock.

For the following processes (shown in green in Fig. 3), it was possible to distinguish model assumptions that best captured the experimental responses, leading to clear recommendations for the types of models to use.

Stomatal conductance. There is a relationship between stomatal and photosynthetic responses to Ca, with the strongest reductions in stomatal conductance occurring when photosynthetic increases are smallest and vice versa41. Models thus commonly represent stomatal responses to Ca as a function of assimilation. The most widely used stomatal conductance models assume that the ratio of assimilation to stomatal conductance is proportional to Ca, but some variants of these models assume a different relationship (see Box 1). Data from the two forest FACE experiments support proportionality, as do a wide range of other experiments42, 43, suggesting that the most widely used models are appropriate29.

Allocation. Allocation describes the distribution of NPP among the different plant organs (leaves, wood and fine roots). Of the allocation assumptions considered, we found that 'functional relationship' allocation models, in which allocation is calculated to give allometric relationships among plant tissues that vary with resource availability, were best able to capture the general features of the observations. Some models assumed that the fractions of NPP allocated to each tissue were constant, but allocation in the FACE experiments responded dynamically to eCa, with significantly greater root allocation at ORNL44 and slightly greater wood allocation at Duke26, 30, so the 'constant coefficient' models did not perform well. Similarly, models that used allocation coefficients that were unconstrained by relationships among plant biomass components gave results that were inconsistent with data. We thus recommend allocation models that include dynamic allometric constraints30. Such models may include empirical, optimization or competitive approaches45.

N limitation. The ecosystem models differed in how quickly N availability declined owing to eCa such that it limited plant production. In two models, N limitation was assumed to effectively preclude any stimulation of productivity even at the start of the experiments. This assumption is not supported by either experiment, as site-average productivity was strongly stimulated in the first years at both sites. Limitation of the eCa effect by N availability occurred at the ORNL site as a gradual process in subsequent years27. These results clearly indicate that models need to allow for a degree of flexibility in the coupling of the C and N cycles28.

There were a number of processes (shown in red in Fig. 3) for which it was found that no model correctly captured the behaviour seen in the experiments. These cases indicate areas where new theory is needed.

Leaf mass per area. Leaf mass per area (LMA) is important in determining LAI, a key ecosystem property. The experimental data showed an increase in LMA in eCa, particularly at the single-species, more homogeneous ORNL FACE, which meant that LAI did not respond as strongly to eCa as did foliage biomass. An increase in LMA is a common observation in eCa experiments (for example, ref. 9) but was not captured by any of the models; in fact, most of the models treat LMA as a constant. Further theory is needed to predict such changes in LMA and avoid over-prediction of eCa effects on LAI31.

Flexibility of plant stoichiometry. Increasing tissue C/N ratios is one mechanism by which plants can maintain high productivity under nutrient limitation. The experimental data showed a consistent decline of the mass-based foliar N concentration with eCa. A subset of the ecosystem models under investigation included this acclimation process, which qualitatively increased the agreement with observations. However, all of these models overestimated the extent of stoichiometric acclimation, suggesting that the current models lack an appropriate representation of the fundamental trade-offs governing foliar N allocation28. Theories on foliar N demand are in development (for example, ref. 46) and may help to determine foliar N demand beyond simple stoichiometric coefficients.

Priming of soil N release. The models underestimated the observation-based net transfer of N from soil organic matter to vegetation associated with eCa, and thus suggested stronger than observed N limitation at the Duke site, where this net N transfer was substantial. This model failure is very probably owing to a missing representation of the increase in soil organic matter turnover with increased plant rhizodeposition47, 48. Such an increase was observed at both sites48, 49. However, the magnitude of this effect differed strongly between the sites, alleviating N limitation in the Duke evergreen, needle-leaved site, but not the ORNL deciduous, broad-leaved forest. New theory is needed for the models to incorporate this effect (see refs 50, 51). In addition, slow accrual rates and large standard errors in observations of soil matter content made it difficult to quantify the extent of the model failure, suggesting that improving the accuracy of soil organic matter records is pivotal28.

In several cases (shown in orange in Fig. 3), the reasons for discrepancies among models were not specific to eCa, but related to model representation of plant ecophysiological function in general. In these cases, broader data sets and data syntheses are needed to constrain the models.

The relative importance of electron transport and Rubisco limitations to photosynthesis. Most models employ the Farquhar–von Caemmerer model of photosynthesis52, in which photosynthesis is determined by the most limiting of two processes, electron transport and Rubisco activity. As Rubisco-limited photosynthesis responds more strongly to changes in Ca than electron-transport-limited photosynthesis, models in which the Rubisco limitation predominates predict larger direct responses of canopy photosynthesis to eCa28. Direct empirical tests of theoretical predictions for how the ratio of the two limitations varies on the leaf scale (for example, ref. 53) and on the canopy scale (for example, ref. 54) could help identify the best way to parameterize these processes and thereby reduce intermodel differences.

Sensitivity of transpiration to stomatal conductance. An important cause of differences in the predicted eCa effect on WUE among models was that the sensitivity of canopy transpiration to stomatal conductance varied dramatically among the models. Although most models predicted that the stomatal conductance would be reduced significantly with eCa (see 'Stomatal conductance' section), the resultant change in transpiration varied from close to proportional to the change in stomatal conductance, to almost none. Given that there has been much previous work on the strength of coupling of transpiration to canopy conductance (for example, refs 55, 56, 57), this discrepancy seems remarkable, and should be resolvable from existing data on canopy coupling29.

Interception. Models disagreed on what fraction of rainfall was intercepted (and evaporated) by the canopy, and how canopy gas exchange was affected when the canopy is wet. Both components noticeably affected the overall water-budget response to eCa, as eCa affected the foliar projected cover. A model-oriented review of data on wet canopy function would help to reduce uncertainty around the representation of this component of the water balance29. This issue is particularly important for moist canopies with high leaf area index, such as tropical rainforests.

Drought. Models disagreed on whether low soil-moisture availability affected the ratio of stomatal conductance to photosynthesis, or the biochemistry of photosynthesis, or both29. This assumption strongly affects the WUE response to eCa under low soil-moisture availability. Evidence emerging from other studies of drought impacts on gas exchange indicates that both processes are impacted by drought58, 59, suggesting that models should include both effects.

Turnover. The effect of eCa on biomass accumulation is strongly affected by the rate at which plant organs, particularly wood, turn over. Most models represented this process with a relatively simple parameterization, and the parameter values chosen differed strongly among models, indicating large uncertainty about this process30. However, the timescale of the FACE experiments (10 years) is clearly too short to constrain this long-term process. Data to constrain this aspect of the models need to come from observational studies rather than manipulative experiments (for example, refs 60, 61).

Ecosystem N losses. The models disagreed strongly on the magnitude of the eCa effect on ecosystem losses of mineral N through leaching. On the ten-year time scale of this experiment, this disagreement had only a small effect on plant N uptake, because changes in soil organic N turnover had a stronger effect28. However, as changes in ecosystem N losses accumulate over time, for longer-t

URL: http://www.nature.com/nclimate/journal/v5/n6/full/nclimate2621.html
Citation statistics:
资源类型: 期刊论文
标识符: http://119.78.100.158/handle/2HF3EXSE/4734
Appears in Collections:气候变化事实与影响
科学计划与规划
气候变化与战略

Files in This Item:
File Name/ File Size Content Type Version Access License
nclimate2621.pdf(1041KB)期刊论文作者接受稿开放获取View Download

Recommended Citation:
Belinda E. Medlyn. Using ecosystem experiments to improve vegetation models[J]. Nature Climate Change,2015-05-21,Volume:5:Pages:528;534 (2015).
Service
Recommend this item
Sava as my favorate item
Show this item's statistics
Export Endnote File
Google Scholar
Similar articles in Google Scholar
[Belinda E. Medlyn]'s Articles
百度学术
Similar articles in Baidu Scholar
[Belinda E. Medlyn]'s Articles
CSDL cross search
Similar articles in CSDL Cross Search
[Belinda E. Medlyn]‘s Articles
Related Copyright Policies
Null
收藏/分享
文件名: nclimate2621.pdf
格式: Adobe PDF
此文件暂不支持浏览
所有评论 (0)
暂无评论
 

Items in IR are protected by copyright, with all rights reserved, unless otherwise indicated.