- Top of page
- Materials and Methods
- Supporting Information
Forest growth ranks amongst the most important processes that determine the carbon balance of terrestrial ecosystems. The magnitude and dynamics of the forest carbon sink strongly depend on carbon allocation to different storage pools (Litton et al., 2007) and their responses to key determinants, such as climate (Babst et al., 2013), land use change (Kaplan et al., 2012), tree age (Genet et al., 2010), forest disturbances (Kurz et al., 2008; Amiro et al., 2010), management practices (Kowalski et al., 2004; Fahey et al., 2009), nutrient and light competition (Wolf et al., 2011; Sardans & Penuelas, 2013) and intensive seed production (i.e. masting years in broadleaf species; Mund et al., 2010). These mechanisms form a complex set of drivers for carbon allocation, which is still relatively poorly understood at large scales (Brüggemann et al., 2011). In particular, the mechanisms linking photosynthesis and carbon storage with above- and below-ground tree growth and ecosystem respiration remain uncertain (Kuptz et al., 2011). In this regard, integrative studies are needed to better constrain the amount of CO2 captured by different forest carbon pools, and to determine how these pools interact and vary across different spatiotemporal scales.
Since pioneer measurements of turbulent fluxes over tall vegetation, eddy-covariance (EC) has been widely used as a standard method for the estimation of seasonal fluctuations in carbon exchange between forest ecosystems and the atmosphere (Baldocchi, 2003). In conjunction with forest inventories (Etzold et al., 2011), EC data have greatly improved the understanding of the terrestrial carbon budget and its climate sensitivity at local to global scales (Baldocchi, 2003; Reichstein et al., 2007). Flux towers, however, essentially provide integral measurements above the canopy and leave uncertainties concerning the magnitude and inter-annual variability of carbon allocation within the respective ecosystems. Furthermore, the fraction of CO2 entering different long-term storage pools is challenging to quantify (Litton et al., 2007; Luyssaert et al., 2007) and may not be constant in time (Campioli et al., 2011) or across ecosystems (Pan et al., 2011). A combination of biometric and EC-based forest productivity assessments could theoretically help to overcome these limitations, but local studies have been variably successful in linking tree growth and carbon stock changes (i.e. allocation of photosynthates to wood production) with flux tower measurements (Barford et al., 2001; Curtis et al., 2002; Gough et al., 2008; Gielen et al., 2013). Conclusions ranged from finding nearly no link to observing high coherence between biometric and EC data. For example, Rocha et al. (2006) found no significant relationship between tree-ring width (TRW) chronologies from selected mature black spruce (Picea mariana) trees and annual ecosystem carbon gain in central Canada. By contrast, Zweifel et al. (2010) reported remarkably close links between stem radius changes and gross primary productivity (GPP) at hourly to inter-annual time-scales in a Swiss subalpine Norway spruce (Picea abies) forest, driven by a combination of growth, stem water balance and frost-induced shrinkage. Ohtsuka et al. (2009) observed a significant relationship of net ecosystem productivity (NEP) with woody biomass increment, but not with foliage production, at a central Japanese flux site. Results from these and other studies hint at the complexity of the processes and ecosystems under question, especially as wood formation is additionally supported by stored carbohydrate reserves (Richardson et al., 2013). Ilvesniemi et al. (2009) found reasonable agreement between tree-ring and EC-based productivity estimates in a Scots pine (Pinus sylvestris) forest in southern Finland. Yet, decreasing coherence with increasing distance between sampled trees and the flux tower suggest rather local representation of these data. Granier et al. (2008) found high correlations between biometric and EC estimates at weekly to monthly time-scales in a young beech (Fagus sylvatica) stand in northern France. The disappearance of these links in the second half of the growing season, and thus at annual time-scales, suggests that the timing of wood formation plays a key role in the quantification of forest carbon assimilation. Given the current state of literature on biometric and EC comparisons, it is premature, if not impossible, to make generalizations with regard to the complementarity and compatibility of these two different ecosystem perspectives. Comparable investigations at multiple flux tower sites across different biomes are needed.
Above-ground woody biomass increment in trees can be calculated as the product of the volume increase and wood density. All existing biometric studies rely on stem diameter changes derived from tree rings (Rocha et al., 2006), dendrometer data (Zweifel et al., 2010) or repeated inventories of tree girth (Ohtsuka et al., 2007). Some studies (Wirth et al., 2004; Wutzler et al., 2008) additionally consider tree height to avoid a priori and potentially site-specific parametric relationships between diameter and stem volume. Wood density is generally assumed to be constant in biomass assessments, thereby neglecting its inter-annual to centennial variation owing to both climate and tree age (Bouriaud et al., 2004). The extent to which changes in wood density and its between-tree variability induce errors into local (Ilvesniemi et al., 2009) to national (Nepal et al., 2012) forest carbon inventories remains unclear. Investigations of TRW and maximum latewood density have shown that these parameters are most sensitive to environmental conditions during different times of the year (Briffa et al., 2002; Frank & Esper, 2005). We thus hypothesize that different climatic controls apply to radial growth and average ring density (XD), and therefore the variability of both is needed to accurately estimate the annual biomass increment in trees.
Tree-ring analyses permit fully compatible measurements of TRW and XD from the same samples, and are thus a valuable archive of annually resolved variability in stem biomass (Bouriaud et al., 2004) and growth over inter-annual (Babst et al., 2012a) to centennial (Esper et al., 2002) time-scales. Large TRW networks have been compiled to address forest growth variability on continental scales (Gedalof & Berg, 2010; Babst et al., 2013). These datasets, however, are not suitable to infer stand biomass changes. As outlined in Babst et al. (2012b), there is little control concerning the number of trees, their dimensions and social status, the research area or even which trees from a stand were sampled. This information is required to transform tree-ring parameters into biomass increments using allometric biomass functions (Zianis et al., 2005; Tabacchi et al., 2011). Furthermore, sample collection is usually oriented towards individual project goals and may influence or even severely bias the quantification of growth variability (Melvin, 2004).
Aiming to reconcile the quantification of carbon cycling from biometric and EC techniques, we measured radial tree growth and wood density at five long-term EC forest sites. The resulting records were used to calculate the annual above-ground woody biomass increment (i.e. stems and branches) and associated carbon uptake. We corrected tree-level biomass increments for inter-annual variations in wood density before upscaling to stand-level carbon uptake in above-ground woody tissues. Inter-annual growth variability was verified with neighboring tree-ring chronologies (Babst et al., 2013) to test the spatial representativeness of individual sites. Subsequent comparisons with the sum of monthly to seasonal CO2 flux measurements (i.e. NEP) and derivatives thereof (GPP; terrestrial ecosystem respiration, TER) were performed to assess: (1) the seasons in which EC and tree-ring data correspond best in different parts of Europe; and (2) the fraction of eddy fluxes that is associated with changes in above-ground woody carbon stocks. Our efforts contribute to reducing the uncertainties in estimates of the carbon allocation processes in forested ecosystems.
- Top of page
- Materials and Methods
- Supporting Information
This study quantified carbon sequestration in tree stems and branches at five European flux tower sites using a combination of extensive tree-ring and biometric measurements, and linked these measurements with long-term EC datasets aiming to better constrain carbon allocation to above-ground woody tissues. In line with results from earlier studies (Curtis et al., 2002; Peichl et al., 2010), we observed large variation in the NEP fraction explained by tAWI from year to year. The variability tended to be larger at sites with lower mean tAWI to NEP ratios. By contrast, the fraction of GPP used for wood production is small and the tAWI to GPP ratios observed at our study sites (Fig. 7) are within the range reported in a recent global investigation (Vicca et al., 2012). Seasonal relationships between biomass increment and GPP were weaker than with NEP, which indicates that most of NEP goes to wood production, not to other biomass pools or soils (Vicca et al., 2012). However, the relationship between NEP and tAWI was strongest at our sites if at least one growing season month in which NEP is mostly GPP driven was considered. This finding, together with the modest TER fraction and variability represented by tAWI, points to a strong influence of GPP on tAWI. The common sensitivity of assimilation and radial growth to the water balance (Beer et al., 2007) may serve as an explanation here. NEP, however, is driven by a complex set of processes, including more temperature-sensitive mechanisms (i.e. TER; Reichstein et al., 2005), which are not so closely coupled to radial tree growth. The observed relationships represent a step towards resolving forest carbon dynamics using in situ measurements, and may also help to improve vegetation model parameterizations and hence decrease uncertainties in model projections of terrestrial carbon cycling (Keenan et al., 2012).
The TRW and XD datasets produced herein provide insights into a wide span of ecological, competitive, climatic and management regimes (see references in Table 1). In addition, the XD measurements permitted us to quantify and mitigate biases introduced by the assumption of constant wood density in most allometric models (Zianis et al., 2005; Wutzler et al., 2008). This XD correction assumes similar inter-annual wood density variability in stem and branch wood, which was reported for Scots pine (Mäkinen, 1999), but lacks testing in many tree species and forest types. Relationships between TRW and XD differed between study sites, suggesting that biometric studies exclusively based on changes in tree diameter may provide a locally biased picture of the sensitivity and resilience of woody biomass production to climate variation and extremes. The effect of XD variability, however, is smaller than that of tAWI variability, and the sign of their correlation may reflect dependence on site properties (e.g. sparse vs dense stands in BE-Bra and FI-Hyy, respectively, both Scots pine), as well as on species-specific characteristics. These relationships will need to be tested with additional sites, species, forest types and climate zones to uncover more general patterns and to quantify their impact on estimations of forest growth, for example, based on dendrometer bands (Zweifel et al., 2010).
The direct comparison between tAWI and EC was challenged by five principal factors which we will address. (1) tAWI is mainly related to biomass increment, whereas EC data are an integral measurement of physiological processes related to carbon assimilation, allocation and use (Baldocchi, 2003). Earlier studies have shown that stem growth alone is an insufficient proxy for total biomass production (Mund et al., 2010). Consequently, the relationship between tAWI and NEP is expected to be weakened by carbon allocation dynamics to other storage pools (Brüggemann et al., 2011), by potential changes in growth dynamics along the stem (Bouriaud et al., 2005) and by the heterotrophic respiration component of NEP – factors which all respond to different environmental drivers. These dynamics further involve temporally lagged processes, with a considerable fraction of the annual stem growth affected by the previous growing season (Skomarkova et al., 2006), thus leading to a strong control of stored carbon on spring fluxes (Kuptz et al., 2011). (2) Uncertainty may be introduced by the spatial and seasonal (in-)compatibility of tree-ring and EC datasets. As the footprint area varies in time, size and location as a result of weather conditions (Chen et al., 2009), the sampling plots may be more or less representative of the EC fluxes. Without fully sampling all trees within the footprint area, some errors caused by spatial variations in species composition, stem distribution and competition will always exist in sampling approaches. However, multiple sampling plots at individual sites provided similar tAWI estimates (Babst et al., 2012b), suggesting that the forest structure is rather homogeneous and is captured reasonably well by our approach. (3) The applicability of the biometric functions for our specific sites could not be considered in this study because functions are often not published with sufficient detail. Our estimates of residual error propagation are thus not function specific and the uncertainty ranges around the mean tAWI may in reality be different from those shown in Fig. 4. Validation against individual functions indicated either an overestimation (Wutzler et al., 2008) or an underestimation (Zianis, 2008) of the residual errors. (4) Management practices may exhibit strong impacts on both EC fluxes (Etzold et al., 2011) and the growth performance of remaining trees (Skomarkova et al., 2006). Consequently, NEP and tAWI may be partly decoupled before the last thinning events, as the flux measurements include signals from trees not present in the sampling year and thinning also alters the growth physiology (e.g. may reduce competition and accentuate climatic growth limitations). As evidenced in Fig. 6, relationships may thus be considerably stronger in the most recent period. (5) Importantly, the short time periods (9–13 yr) assessed through EC measurements result in few degrees of freedom to obtain statistically significant correlations (Table S1). Although our comparisons between biometric and EC-based carbon sink estimates provide comprehensive and mechanistically straightforward seasonal links, the inferred relationships should be revisited at these sites as EC records become longer and tested at independent locations.
Despite the above challenges, we found reasonably strong seasonal agreement between biometric measurements and NEP at most study sites. Comparisons of these results with existing studies are challenging because of inconsistent approaches for the assessment of radial tree growth and because of the different bioclimatic zones investigated (Curtis et al., 2002; Rocha et al., 2006; Gough et al., 2008; Granier et al., 2008; Ilvesniemi et al., 2009; Ohtsuka et al., 2009; Zweifel et al., 2010; Gielen et al., 2013). Our findings indicate a strong relationship among study sites between tAWI and NEP during the January to June/July season, where most of the radial growth is realized and soil water reserves are high. This result is consistent with earlier studies reporting robust links between radial growth and spring to early summer EC measurements (Granier et al., 2008). In addition, we observed reasonably high positive correlations between XD and late summer (i.e. August–September) NEP at three of the five sites. This seasonal partitioning in carbon allocation: (1) points towards a specific timing of wood production and carbon assimilation as photosynthesis continues after radial growth has stopped; (2) is mechanistically straightforward in terms of cell formation and maturation processes (Moser et al., 2010); and (3) suggests that carbon sequestered after June/July is mostly used for cell wall thickening processes (i.e. XD increase; Lupi et al., 2012) and/or stored in above- and below-ground nonstructural carbohydrate reserves (Granier et al., 2008; Wolf et al., 2011; Richardson et al., 2013). The latter fraction can be rather large and is known to support next year's spring growth (Skomarkova et al., 2006) or used up to several years later (Richardson et al., 2013). In the present study, relationships between tAWI and XD with the previous year's fluxes were not robust (Fig. S6), and the explicit quantification of carbon storage will require longer term assessments of all forest carbon pools and their differing turnover rates. Interestingly, the tAWI to NEP ratio was higher at the more productive sites, which also showed the strongest inter-annual variability in wood formation (Fig. 4). The forest carbon sink at these productive sites is thus probably most susceptible to extreme climatic events (Reichstein et al., 2007, 2013).
Our synthesis of above-ground biomass increment from the most important European climate zones and tree species helps us to better understand seasonal carbon allocation processes, and demonstrates that variability in tree-ring and monthly to seasonal EC measurements are largely compatible and complementary. Yet, carbon allocation to above-ground woody tissues may be altered by climate warming with different impacts expected in boreal and temperate regions (Lindner et al., 2010). For instance, warmer temperatures may enhance root and foliage growth in response to an earlier start of the growing season (Kalliokoski et al., 2012; Lapenis et al., 2013). In addition, the considerable inter-annual variations in above-ground biomass increment emphasize the relevance of, for example, extreme climatic events for the terrestrial carbon balance and the need for extensive in situ studies of climate–growth interactions. Our assessments provide a framework to link future biometric and EC measurements that will contribute to a better quantification of long-term changes in terrestrial carbon uptake and will reduce uncertainties for carbon cycle–climate feedbacks.