The comparison of continuous stem radius changes (DR) of individual trees and the EC-based NEP of the corresponding forest ecosystem revealed an intriguingly close relationship between the two measures DR and NEP on three different time scales (Table 1). This close relationship was not expected a priori as one measure (DR) is an averaged physiological response of individual mature trees, whereas the other (NEP) integrates over the entire ecosystem represented in the EC footprint (Kljun et al., 2004). This finding is even more surprising when considering that DR is the product of growth and water-related processes in the stem and that both processes contribute to the close relationship to NEP with changing weights depending on integration time. There is no simple physiological approach available that could fully explain this relationship in all of its detail (Table 1). However, there is good evidence that at least part of this relationship between NEP and DR can be functionally understood. In the following, we discuss to what degree physiological causes and effects can build a link between the two measures.
Integration time-dependent information content of DR
Stem radius changes are determined by water- and growth-related processes (Daudet et al., 2005; Steppe et al., 2006). The analytical distinction between the two fractions of DR is possible with either statistical (Zweifel et al., 2005; Deslauriers et al., 2007) or process-based (Zweifel et al., 2007; Steppe et al., 2008) modelling approaches. However, such a distinction is never absolutely precise and depends, to a certain degree, on assumptions about wood and phloem properties (Zweifel et al., 2006). In general, the longer the time period for data aggregation, the stronger the importance of growth-related processes in DR (Zweifel et al., 2006; Steppe et al., 2008). Conversely, the shorter the time period, the stronger the influence of water-related processes (Zweifel et al., 2001). This is particularly true for our slow-growing spruce trees with a maximum growth rate of c. 10–20 μm d−1 (data not shown), whereas water-related changes in DR are a magnitude larger at 100–300 μm d−1 (Zweifel & Häsler, 2001). Understanding how these two integration time-dependent processes alter DR is essential for the physiological interpretation of relationships between DR and NEP.
Close relationship between DR and NEP
As NEP represents the budget of CO2 fluxes in and out of the ecosystem, it was reasonable to expect that only long-term, growth-related (and thus carbon-related) fractions of DR would show a close relationship with NEP. Such a relationship on an annual time scale has been reported previously in a study of tree ring widths in combination with NEP (Rocha et al., 2006). The interdependence of the two measures in a black spruce stand in central Manitoba, Canada, was less strong (R2 = 0.73) than in our study (adj. R2 = 0.85), but still convincing. Nevertheless, the statistically highly significant regression found between annual NEPyr and DRyr was expected to disappear with shorter integration times as a result of the increasing weight of water-related changes of DR. However, this was not the case (Table 1). Our analyses have clearly shown that the relationship remained close from annual to half-hourly time scales, yet with a switch of the algebraic sign for the regression from positive to negative. This switch can be attributed to the changing weights of growth-related (positive regression) and water-related (negative regression) contributions to DR from long (annual) to short (half-hourly) integration times.
Approaching causalities on different time scales
On annual time scales, wood growth – known as the main carbon sink of a forest ecosystem (Barford et al., 2001; Rocha et al., 2006) – is a main driver of DRyr and NEPyr, and therefore responsible for close relationships between the two measures (Fig. 5). However, the correlation between the two measures was closer than that reported in a tree ring study by Rocha et al. (2006), and was also closer than that found in own preliminary tree ring data (data not shown). As DR includes not only woody growth but also growth (and death) of phloem cells, we hypothesize that annual phloem size changes may have substantially contributed to the patterns found. However, at the current state of knowledge, this remains speculative because there are few data available on inter-annual phloem size changes (Gricar & Cufar, 2008; Gricar et al., 2009). It is interesting that neither GPPyr nor TERyr showed a similar close relationship to DRyr at this time scale (Table 1). This supports recent findings that wood growth is not controlled directly by GPP (Rocha et al., 2006; Stoy et al., 2009), but involves complex processes, such as carbohydrate storage and allocation (Hoch et al., 2003; Carbone et al., 2007). Wood growth is thus affected by environmental conditions that occur weeks, months or years ago (Gough et al., 2008). Overall, it became evident that wood and phloem growth, independent of water-related fluctuations of DR, represented NEP up to an unexpectedly high degree on the annual scale.
On a monthly scale, a mixture of growth- and water-related processes determined DRm with distinctly different ratios in summer and winter. The combination of these two processes was therefore also responsible for close relationships between DRm, on the one hand, and NEPm, GPPm and TERm, on the other (Table 1). Only on this time scale did all three EC-based measures show significant correlations with DR. The reason for this is not completely clear, but might be explained by split contributions of growth and tree water relations to DRm. During the summer, DRm increased up to 750 μm month−1 (Fig. 4), about two-thirds of which was attributable to radial growth and one-third to water-related fluctuations (Zweifel & Häsler, 2001). Positive NEPm in summer is thus linked to good wood growth conditions and water relations that do not induce large DRm-reducing tree water deficits (Zweifel et al., 2005). Such conditions lead to increased NEPm, GPPm and TERm, but in a ratio that favours assimilation over respiration, as shown by the positive NEPm values (Fig. 4).
In winter, DRm decreased with a rate of 400 μm month−1 (averages, Fig. 4) and even with > 1000 μm month−1 for individual trees (Zweifel & Häsler, 2000, 2001). There was no growth at this time of the season and such shrinkage is mainly attributable to freezing processes in the stem which lead to very rapid dehydration of elastic bark tissue with a consequent decrease in DR (Zweifel & Häsler, 2000). This winter shrinkage is thus a water-related process, although it is mainly induced by freezing temperatures (Ameglio et al., 2001; Mayr et al., 2007). The negative NEPm in winter, when TERm exceeds GPPm, as reported previously by, for example, Monson et al. (2006) or Lipson et al. (2009), was not expected a priori to be so closely linked to negative DRm, as lower temperatures should also lead to decreasing respiration (Schwalm et al., 2010). However, this was not the case, as shown by our results (Fig. 3). Furthermore, our finding at monthly scales produces an apparent discrepancy with winter findings at shorter time scales, as discussed below. As a synthesis of our results at monthly resolution, we conclude that growth- and water-related processes of trees in summer and temperature-induced, water-related processes in winter contribute to the close relationships observed. A clear mechanism, however, explaining these correlative findings, with physiological causes and effects, awaits further research.
On half-hourly time scales, the relationship between DRhh and NEPhh is highly water-related because of the influence of changing water contents on mainly bark and thus DR (Steppe et al., 2006; Zweifel et al., 2007). In contrast with the longer integration times investigated, the linear regression was found to be negative (Fig. 3), which is not surprising when considering the dominant physiological processes at this temporal scale. The sunnier the conditions, the more trees transpire and, consequently, the more depleted is tree internal water storage, for example in the bark (Steppe et al., 2006). The same conditions also lead to higher assimilation rates by photosynthesis and, consequently, an increased ecosystem CO2 uptake in parallel with shrinking stem radii. This relationship can be altered by drought stress-related stomatal closure (Pena-Rojas et al., 2004; Buckley, 2005). Although midday stomatal closure on sunny days has been reported for trees at our site (Zweifel et al., 2002), it apparently did not reduce NEPhh to an extent that would have forced the discussed relationship to disappear. Overall, the longer and more pronounced the shrinking of stems over a day, the larger the assimilation and therefore NEPhh (Figs 3, 6). Thus, any causal link between DRhh and NEPhh must relate to biotic and physical conditions (Stoy et al., 2009) that induce large diurnal stem radius fluctuations which do not represent tree net assimilation only, but also the productivity of the entire forest ecosystem. Such a general relationship between DRhh and NEPhh has been found for days with temperatures above the freezing point, and is further supported by the close correlation between DR and GPP at half-hourly scales (Table 1).
Exceptions and anomalies
There are also exceptions from this general relationship between DR and NEP. Such anomalies indicate that, although the two measures have similar drivers, they may be linked only indirectly to each other. One such exception was present during May, the month with the highest NEPm on average, but with little change in DRm (Fig. 4). Wood growth had not yet started, but stem rehydration of winter shrinkage, mostly occurring in April at this site, had nearly finished. This succession of stem rehydration followed by initial wood growth has also been found in other studies (Larcher, 2003; Monson et al., 2005; Zweifel et al., 2006), and shows the importance of balanced tree water relations for the initiation of wood growth (Lockhart, 1965; Steppe et al., 2006, 2008; Turcotte et al., 2009), in addition to the well-known limitations by temperature (Rossi et al., 2008). The relationships of DRm to GPPm and TERm were much less affected by the month of May (data not shown). However, increased GPPm, together with slightly decreased TERm, led to over-proportionally large NEPm in relation to DRm. This delay in RWG after NEPm explains, at least partially, the somewhat disturbed relationship between DRm and NEPm in May (Fig. 3, Table 1), and is a strong indication for two, at least partially decoupled, mechanisms determining DRm and NEPm based on the same climatic drivers.
In contrast with that discussed for monthly values in wintertime, half-hourly DRhh and NEPhh values were found to be uncorrelated on days with freezing conditions (Fig. 3d). This anomalous short-term behaviour can be interpreted as a stem dehydration effect, which decouples frost-induced stem radius fluctuations from respiration processes and thus NEPhh. This should not be surprising per se as DR is expected to be decoupled from transpiration (and photosynthesis) under such cold conditions, and a direct coupling between DR and transpiration (and thus NEP) is only expected during the physiologically active period (Zweifel & Häsler, 2000). However, despite this short-term decoupling of DRhh from NEPhh, it is surprising that dehydrated stems at cold temperatures tend to lead to higher winter ecosystem respiration on a monthly scale. In this case, increased ecosystem respiration might not be functionally linked to stem physiological processes.
The ecological relevance of DR measurements for NEP
In summary, DR is, with some exceptions, closely correlated with NEP at all time scales investigated. This has not been explained mechanistically in full yet and was not expected beforehand. However, it raises the question of why NEP measured by EC over this ecosystem seems to be driven mostly by tree metabolism. Other ecosystem components, namely heterotrophic (soil) respiration and understorey vegetation carbon turnover, could have added substantial contributions to NEP (Carbone et al., 2007; Paterson et al., 2009; Subke et al., 2009). Hence, our results suggest that these contributions are proportional to changes in stem radius, either contributing very little to NEP or, more likely, being in phase with the mature trees measured. Similar close relationships between wood growth and NEP have been found in a boreal forest in central Manitoba, Canada (Rocha et al., 2006). On the one hand, this may be a coincidence and, on the other, might be a characteristic of subalpine (Davos) and boreal (Manitoba) regions, clearly differing from lowland deciduous forests where spring ephemerals can be a relevant short-term carbon sink which is not synchronous with the dominant tree species (Knohl et al., 2003). In particular, freezing conditions in winter and spring seem to play an important role for annual DRyr and NEPyr, as indicated by the high predictive power of the compensation day of the cumulative WRES (Table 1). Overall, the course of DR appeared to be a fingerprint of the ecosystem’s NEP at all temporal scales investigated, describing the short-time metabolism, the tree water relations and, most remarkably, the annual ecosystem productivity.
NEP of the subalpine coniferous forest Seehornwald Davos in the Swiss Alps is intriguingly highly predictable from DR at various integration times. This strong correlation between an integrative measure of NEP (which represents the whole ecosystem) and DR, a measure from individual trees (which reflects a subordinate component of the ecosystem), suggests that tree water relations and stem growth are representative for the productivity of this forest ecosystem. Both fractions of DR, tree water relations and stem growth, have considerable explanatory power for NEP. Furthermore, there is reasonable indication that phloem growth, so far an unquantifiable fraction of stem growth, could be a reason for the higher explanatory power of DR for NEP in comparison with the pure wood-related tree ring widths. The relationships between DR and NEP, however, are a function of the time scale under consideration. In general, we observed a shift from a water-related dominance on short-term DR changes towards a more growth-related dominance at seasonal to annual time scales.
Furthermore, tree physiological responses to winter and spring conditions play a decisive role in the seasonal cycles of DR and NEP for this subalpine forest. However, exceptions from the typically close relationship between DR and NEP also indicate that NEP is physiologically not directly, but indirectly, linked to DR via its climatic and biotic drivers. What are the biotic drivers and through what mechanisms they are linking DR and NEP are topics of future investigations. In particular, it remains to be tested at other localities, with different forest types and climates, whether such close correlations between DR and NEP are specific to cold climates (subalpine, boreal) or whether they are more abundant.