Although ‘shade tolerance’ has featured prominently in the vocabulary of foresters and ecologists for a century, we have yet to agree on a standardized method for quantifying this elusive property. The ‘whole-plant compensation point’, interpolated from stem growth measurements across a wide range of light environments, has been proposed as a simple, robust measure of species shade tolerance. Others have argued that shade tolerance is primarily a function of differential ability to survive periods of slow growth (‘suppression’), implying that measurements of survival are vital.
We measured growth of juveniles (500–1000 mm tall) of five evergreen trees over 12 months in a cool-temperate rain forest in New Zealand, to determine whether whole-plant compensation points predicted species differences in occupancy of understorey light environments, which were quantified using hemispherical photography.
The five species encompassed 3·5-fold variation in whole-plant compensation points. Compensation points of most species fell within the first quartile of the distribution of light environments occupied by juveniles; they were also correlated with low-light mortality rates of juveniles, estimated from permanent plot data archived in the National Vegetation Survey Databank. Compensation points were also significantly positively correlated with height growth rates in high light, confirming the presence of the growth vs. shade tolerance trade-off detected in many other forest tree assemblages.
Results show that, in temperate evergreen forests, the whole-plant compensation point distinguishes reliably between species of differing shade tolerance. Excepting situations involving parameterization of demographic models, shade tolerance can therefore be assessed without survival measurements. However, estimating whole-plant compensation points may prove more difficult in deciduous forests, where seasonal variation in understorey light transmission poses additional challenges.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
The concept of shade tolerance was probably first introduced to ecology in the early 20th century by Warming (1909). Species differences in the ability to survive and grow in the shade have been a frequent focus of foresters interested in predicting and manipulating regeneration of timber species, and have also fascinated ecologists seeking to understand succession and mechanisms of species coexistence (Valladares & Niinemets 2008). A negative correlation between juvenile shade tolerance and maximum growth rate of tree species is regarded as one of the main drivers of secondary succession (Pacala et al. 1994), and likely contributes to tree species coexistence in old-growth forests (Denslow 1980). North American and European foresters and ecologists have devised shade tolerance scales or ratings based on semi-quantitative assessments of factors thought to be associated with survival and growth of juvenile trees and shrubs under dense canopies (e.g. Spurr & Barnes 1980; Burns & Honkala 1990; Ellenberg 1991). The most common factors used in assessing tolerance are the relative abundance of juveniles beneath closed canopies, the ability to respond to release from suppression and tree crown depth and density (Lorimer 1983).
It is only in the last few decades that more quantitative approaches to shade tolerance have been facilitated by technology permitting integrated measurements of understorey light environments. Although the application of hemispherical photography for this purpose dates back to Anderson's (1964) seminal work, its use did not become widespread until computerized processing of images (e.g. Chazdon & Field 1987; Frazer, Canham & Lertzman 1999) and digital cameras enabled enormous gains in speed and convenience. Hemispherical photography has since been used to relate the distribution, growth and survival of juvenile plants to estimates of light availability in diverse vegetation types (e.g. King 1994; Kobe et al. 1995; Kobe 1999; Baltzer & Thomas 2007). The LAI-2000 Canopy Analyzer (Li-Cor, Inc., Lincoln, NE, USA) gave researchers another option for quantifying the light environments of understorey plants (e.g. Lusk & Reich 2000; Claveau et al. 2002), although an important limitation is its inability to measure the direct beam radiation considered especially important for light-demanding species (e.g. Lusk, Kooyman & Sendall 2011).
Despite progress resulting from technological advances, debate has continued over the nature of shade tolerance and how best to measure it (e.g. Baltzer & Thomas 2007; Poorter & Kitajima 2007). Givnish (1988) developed a model of the whole-plant compensation point – the light level required for zero net carbon gain at whole-plant level – as a measure of shade tolerance. This approach originally ignored carbon costs resulting from turnover of leaves and fine roots. However, a study of sapling growth in a Panamanian rain forest (King 1994) provided support for the general principle of this approach: species regarded as light demanding (e.g. Cecropia insignis) required more light to achieve positive growth than those regarded as shade-tolerant (e.g. Tachagalia versicolor). Baltzer & Thomas (2007) reported threefold interspecific variation in whole-plant compensation points of saplings in a tropical rain forest in Borneo, associated with species crossovers in the relationship of growth with light. Furthermore, whole-plant compensation points correlated well with leaf-level traits likely to influence low-light net carbon gain and growth, especially leaf life span and dark respiration rates. A less encouraging result was reported by Martin, Stedman & Thomas (2011) who found that juvenile growth of seven Caribbean rain forest trees was only weakly related to light availability, with most relationships not significant. An experimental study comparing seedling growth of 15 tropical forest trees in six light treatments reported that whole-plant compensation points discriminated broadly between fast-growing pioneers and primary forest species, but did not permit more subtle differentiation among species (Poorter 1999).
An alternative view is that interspecific variation in the ability to survive periods of slow growth is a core component of shade tolerance, due to differences in carbon allocation patterns. Shade-tolerant species may allocate more photosynthate to storage than light demanders, enabling them to survive periods of low carbon gain and slow growth (Kobe 1997; Poorter & Kitajima 2007). If light demanders, in contrast, rapidly commit most photosynthate to new growth, slow growth would reflect a precarious carbon balance, and insufficient reserves to cope with any additional stresses, or to recover from severe damage by herbivores or falling debris (Poorter et al. 2010). This view is embodied by the SORTIE approach, which involves modelling seedling growth of each species as a function of light availability, and then developing species-specific models of survival as a function of growth (Pacala, Canham & Silander 1993). The SORTIE approach has been widely applied to the dynamics of temperate and boreal forests (e.g. Kobe et al. 1995; Coates et al. 2003; Astrup, Coates & Hall 2008; Kunstler, Coomes & Canham 2009). The fundamental tenet of this approach is that survival data are required to determine the relative shade tolerance of different species.
Here, we test the ability of the whole-plant compensation point to predict the minimum light levels occupied by juveniles of five cool-temperate rain forest trees. If light demanders require more light to achieve positive net growth than their shade-tolerant associates, compensation points should correspond well with the minimum light environments occupied by juveniles (Fig. 1a). If compensation points are a reliable indicator of species light requirements, we would also expect them to be positively correlated with growth in the open, as a trade-off between maximum growth rate and shade tolerance has been reported in many humid forest assemblages (Hubbell & Foster 1992; Kobe et al. 1995; Lin et al. 2002; Sterck, Poorter & Schieving 2006; Seiwa 2007); crossovers should therefore be observed in the relationship of growth with light (Sack & Grubb 2001; Baltzer & Thomas 2007). If, on the other hand, shade tolerance is primarily a function of species differences in the ability to survive slow growth, juveniles of light-demanding species should be absent from some light environments above their predicted compensation points (Fig. 1b), where net carbon gain, although apparently adequate for positive net growth, is insufficient for long-term survival.
Materials and methods
Study Site and Species
The study was carried out in cool-temperate rain forest in Tongariro National Park, in the central North Island of New Zealand. The study site was located on the south bank of the Mangaturuturu stream near Pokaka (39°18′22″S, 175°23′25″E, elevation 800 m a.s.l.), on the ring-plain built up by a succession of Quaternary lahars from Mt. Ruapehu (Lecointre, Neall & Palmer 1998), an andesitic stratovolcano whose crater lies about 15 km to the east. Although Ruapehu frequently produces small ash eruptions and lahars, the last lahar episode to have left significant deposits at the site has been dated at 8470 BP (Lecointre, Neall & Palmer 1998). Drainage is poor (Atkinson 1981) and although the terrain is near-flat, there is marked pit-and-mound microtopography generated by treefalls. Mean annual temperature is about 9·5 °C (NIWA 2005b), and mean annual precipitation is estimated at >2000 mm including occasional snowfalls in winter (NIWA 2005a).
The tree assemblage is a mixture of Nothofagus, podocarps and broad-leaved hardwoods (Atkinson 1981). In view of the reputation of Nothofagus species as the most light demanding canopy-forming element in New Zealand forests (Wardle 1984), this assemblage was considered likely to encompass a wide range of shade tolerance. The main canopy species is Nothofagus solandri var. cliffortioides up to c. 20 m tall; Weinmannia racemosa also reaches the main canopy as well as forming a sparse subcanopy in places. The podocarps Prumnopitys ferruginea and Podocarpus hallii are also significant canopy components, with Dacrydium cupressinum occurring mainly as scattered emergents. These five principal overstorey species were chosen as the study species; henceforth, they are referred to by their generic names. The dense understorey is dominated by Coprosma spp. Juveniles of all three podocarps and Weinmannia are common throughout, but regeneration of Nothofagus is mainly restricted to gaps and forest margins. Nomenclature follows Allan (1961), except where superseded by de Laubenfels (1969).
Measuring Juvenile Tree Growth
Stem growth rates of juveniles 500–1000 mm tall were measured over a 12-month period. Shade tolerance differences between species become more accentuated in larger size classes (Lusk et al. 2008), whereas smaller juveniles are more tractable because of their greater abundance and ease of measurement; our choice of size class strikes a balance between these two conflicting priorities. At least 40 individuals of each species were chosen haphazardly across a wide range of light environments as possible, sampling along forest margins as well as in the extensive old-growth stands present at the site.
Change in length and basal diameter of the longest stem was used to estimate the relative growth rate of each juvenile tree (Baltzer & Thomas 2007). These stem dimensions have been shown to correlate well with actual measurements of whole-plant relative growth rate (Kohyama & Hotta 1990). In order to facilitate accurate remeasurement, white acrylic paint was used to mark the point at which the initial diameter measurements (to the nearest 0·1 mm) were made using digital callipers. This mark was also used as the endpoint of stem length measurements. Relative growth rates were thus estimated as:
where li and lf are the initial and figure stem lengths, and di and df are the initial and final basal diameters.
Relating Growth to Light Availability
Hemispherical photos were taken immediately above each juvenile tree at both the start and conclusion of the measurement period, in order to determine relationships of growth with light availability. Photos were taken with a Nikon Coolpix 4500 digital camera and an EC-08 fisheye adaptor. Photos were oriented to magnetic north and levelled with a bubble level. The captured images were processed using Gap Light Analyzer (Frazer, Canham & Lertzman 1999), using cloud cover frequency estimates from MODIS satellite photos to cloudiness index (Kt), beam fraction and spectral fraction by month (Iqbal 1983). This enabled estimates of mean daily photon flux reaching each plant over the 12-month observation period, averaging figures obtained from the initial and final photos.
Several different functions were considered for modelling the response of relative growth rate to light availability. In the past, Michaelis–Menten and simple regression against log(Light) have been found to approximate plant growth responses to light (Sack & Grubb 2001; Kitajima & Bolker 2003). We explored both of these options.
Kobe (1999, equation 10) writes the Michaelis–Menten model as:
where L is the available light, P1 is the limiting growth rate as L becomes very large and P2 is the slope of G(L) at zero light. This is always positive for at all light levels, but we needed a form capable of giving negative values at light levels below the compensation point. Such a form is
where the additional parameter C is the compensation point and the parameters P1 and P2 retain their meanings. Another form is given by:
where γ is the logarithm of the compensation point. This last form has the advantage that a scientifically meaningless negative estimate for the compensation point can never be obtained when this form of the model is fitted to data.
A competing model is a simple regression G(L) = a + b · log (L). The compensation point C then must satisfy 0 = a + b · log (C), so that we may write this model in the form:
The parameters C and γ are now nonlinear, but this form makes comparison with the Michaelis–Menten easier.
The Akaike Information Criterion (AIC) is commonly used for model selection balancing fit against complexity (Burnham & Anderson 2002). The simple log regression model gave a lower AIC value than the Michaelis–Menten model (−247·7 and −245·2, respectively; delta AIC = 2·5); thus, the former was selected for use. The R script used to carry out this analysis is included in the Supporting Information (Appendix S1).
Comparing Compensation Points
We used a likelihood ratio test to test the hypothesis that the compensation points of the five species were equal. This test compares two models: one that allows the five compensation points to differ, and one in which they are constrained to be equal. All other aspects of the two models are the same. The test statistic is twice the difference in log-likelihood between the two models. Under the null hypothesis, the test statistic has a large-sample distribution, that is, Chi-square with degrees of freedom equal to the difference in the number of parameters in the two models (Pawitan 2006, p. 266). In our case, there are four degrees of freedom. The R script used to carry out this test is included in the Supporting Information (Appendix S2).
Juvenile Distributions in Relation to Light Availability
Distributions of juveniles were described by recording presence or absence of each species in randomly placed plots, and estimating light availability above plots by hemispherical photography. A series of transects were run through old-growth stands and along forest margins, in order to maximize the range of light environments sampled. At random intervals of between 5 and 10 m, presence or absence of 500–1000 mm tall juveniles of each species was recorded in an 800 mm diameter circular plot. A hemispherical photo was taken immediately above the tallest juvenile present on the plot, within the studied height range of 500–1000 mm. A total of 414 plots were sampled.
Juvenile Mortality Rates
Juvenile mortality rates are widely used as a measure of species light requirements. Accordingly, we examined relationships of compensation points with mortality rates of shaded juveniles reported by Russo et al. (2010), derived from permanent plot data archived in the National Vegetation Survey Databank (Wiser, Bellingham & Burrows 2001). In contrast to our data, these mortality data were collated from hundreds of plots distributed throughout New Zealand. Russo et al. (2010) used neighbourhood basal areas to identify juvenile trees of each species likely to be growing in environments with low light availability; the chosen criterion for selection of juvenile trees was the 85th percentile of neighbourhood basal areas of overtopping trees. As the variance of compensation point estimates varied widely across species, we used weighted least squares regression (Neter, Wasserman & Kutner 1996) to examine the relationship of low-light mortality with compensation points. Compensation points were weighted by the inverse of their variances.
Testing for the Growth Vs. Shade Tolerance Trade-Off
We computed the major axis of the relationship between compensation points and growth in high light (Warton et al. 2006). Very few juveniles of Dacrydium or Prumnopitys were found growing at >20% light (5·13 mol m−2 day−1), so growth at this level of illumination was used as our measure of high-light performance. As height growth rates of Dacrydium were not significantly correlated with light availability (Supporting Information, Fig. S2), the mean height growth rate of this species was used instead.
Relative growth rates of all species were significantly correlated with light availability, with mean daily photon flux explaining between 15 and 40% of variation in growth (Fig. 2). Growth of most species was more closely related to total mean daily photon flux than to its direct or diffuse components (data not shown). Whole-plant compensation points estimated from these relationships differed significantly among species (P = 0·020), estimates ranging from 0·47 moles m−2 day−1 in Dacrydium to 1·66 in Nothofagus.
Compensation points of four out of five species fell within the first quartile of the range of light environments occupied by juveniles of the same species (Fig. 3). The compensation point of the remaining species (Weinmannia) lay within the second quartile. Compensation points of the five species were also correlated with low-light mortality rates obtained from nation-wide permanent plot data (Fig. 4).
Predicted height growth rates at 20% of full sunlight ranged from c. 64 mm year−1 in Dacrydium to 119 mm year−1 in Nothofagus. Recall that height growth rates of Dacrydium were not significantly correlated with light availability, so the value given for this species is actually the mean height growth rate. These height growth rates were strongly positively correlated with species whole-plant compensation points (Fig. 5), consistent with the observed rank changes in species relative growth rates along the gradient of light availability (Fig. 2).
Whole-plant compensation points were good predictors of the minimum light environments occupied by juveniles of most species. Compensation points differed widely among species and in most cases occurred towards the lower end of the range of light environments occupied by juveniles (Fig. 4). For the most part, the rank order of compensation points also corresponded well with previous reports of the relative shade tolerance of our five study species. Nothofagus spp. have been regarded as the most light demanding of New Zealand's major canopy trees, with other lines of evidence suggesting that collectively they are less shade-tolerant than both Weinmannia (Wardle 1984) and tall podocarps such as Dacrydium and Prumnopitys (Lusk & Smith 1998). Compensation points of the three podocarps did not differ significantly among each other (P > 0·05), consistent with previous evidence that they have very similar light requirements (Kunstler, Coomes & Canham 2009; Lusk, Duncan & Bellingham 2009). Although the data presented here suggest Weinmannia is somewhat less shade-tolerant than the three podocarps, there is less agreement over this point in the literature (e.g. Lusk & Ogden 1992; Kunstler, Coomes & Canham 2009; Lusk, Duncan & Bellingham 2009), and in our study, the estimated compensation point of Weinmannia did not differ significantly from that of any of the podocarps (P > 0·05).
The correspondence with low-light mortality rates provides further evidence that the whole-plant compensation is a reliable measure of species shade tolerance (Fig. 4). Species rank order differed slightly between these two variables: although Dacrydium had the lowest estimated compensation point of the five species, its low-light mortality rate exceeded that of Prumnopitys. However, as pointed out above, the compensation points of these two species were statistically indistinguishable, and in view of the large standard errors associated with the estimated compensation points of some of our species (especially Dacrydium), we place more confidence in the species rank order apparent in the mortality data. This also corresponds better with the general view among New Zealand ecologists and foresters that Prumnopitys ferruginea is, if anything, slightly more shade-tolerant than Dacrydium cupressinum (Wardle 1991; Lusk & Ogden 1992; Ogden & Stewart 1995).
The anomalously high compensation point of Weinmannia is attributable to browsing by introduced mammals. On face value, about 30% of juveniles of this species were found at light levels below that required for positive net growth (Fig. 4). However, Weinmannia was the only study species to suffer conspicuous browsing by vertebrates during the study period. Spoor of deer (Cervus elephus scoticus) and feral pigs (Sus scrofa) was abundant throughout one half of our study site (which straddled a highway and a railroad), and Weinmannia foliage is known to be a preferred food of deer (Nugent, Fraser & Sweetapple 2001). Although we rejected heavily browsed plants, individuals suffering only minor damage were included in order to avoid a further reduction in the sample size of this species. The loss of resources suffered by these browsed plants likely had the effect of increasing the light level required to achieve positive stem growth. Because the New Zealand flora evolved in the absence of browsing mammals, many native plant species are highly palatable to vertebrates introduced in the 19th century (Veblen & Stewart 1982; Nugent, Fraser & Sweetapple 2001). In the absence of significant browsing by vertebrates, we would expect the whole-plant compensation point of Weinmannia to lie somewhere in the first quartile of the distribution of light environments occupied by juveniles, as seen in the other four species (Fig. 3).
Measurement error and spatial variation in drainage may partly explain the relatively weak relationships of some species' growth with light availability. Although all relationships were statistically significant, light explained as little as 15% of variation in the relative growth rates of some species (Fig. 2) – lower than most coefficients of determination reported by Baltzer & Thomas (2007), though stronger than most relationships reported by Martin, Stedman & Thomas (2011). Measurement error is undoubtedly responsible for some of this scatter, as understorey saplings with complex growth histories often have crooked and irregular-shaped stems that are difficult to measure accurately. In addition, the ubiquitous pit-and-mound microtopography at the site means that individual saplings growing only a few metres apart could experience very different drainage conditions, those individuals rooted on mounds likely experiencing conditions more conducive to growth than those rooted in swales subject to intermittent waterlogging. As well as restricting root development, high water tables are known to influence rates of nitrogen mineralization in some ecosystems (van Bodegom et al. 2006). On infertile sites such as the one we worked at, trenching and nutrient amendments have shown that growth of understorey seedlings and saplings is sometimes colimited by nutrient availability (Platt et al. 2004; Carswell et al. 2012), giving further credence to the notion that spatial variation in edaphic factors contributed to the relatively weak relationships of some species' growth with light availability. Another possible source of error in our data set is size-related variation in relative growth rates (Evans 1972; Philipson et al. 2012) and light requirements (Lusk et al. 2008), although we note that our study was restricted to a narrower range of plant heights than previous work on whole-plant compensation points (Baltzer & Thomas 2007; Martin, Stedman & Thomas 2011).
It may be more difficult to calculate whole-plant compensation points in forests with a significant deciduous canopy component. In North American deciduous forests, budbreak of juveniles often precedes that of overstorey trees (Gill, Amthor & Bormann 1998; Augspurger & Bartlett 2003), and there is evidence that understorey saplings gain an important fraction of their annual carbon income prior to canopy closure (Augspurger 2008). Accurately estimating the mean daily photosynthetic photon flux received by understorey plants is therefore a more daunting task in forests with a significant deciduous component, probably requiring multiple hemispherical photographs at different times of the year. However, we note that other approaches to shade tolerance face essentially similar problems posed by the complexity of understorey light environments in deciduous forests.
We found evidence of a trade-off between shade tolerance and high-light growth, within a single juvenile size class (Fig. 5). This pattern has now been shown in diverse humid forests in tropical and temperate regions (Hubbell & Foster 1992; Kobe et al. 1995; Lin et al. 2002; Sterck, Poorter & Schieving 2006; Seiwa 2007), suggesting the existence of a fundamental constraint on adaptation of woody plants to light environments, which likely contributes significantly to species coexistence (Kobe 1999; Montgomery & Chazdon 2002). Greater complexity has been suggested by another New Zealand study reporting that this trade-off was evident only when shade tolerance of seedlings was compared with height growth of larger plants (‘saplings’) (Kunstler, Coomes & Canham 2009). That result may reflect time lags between the costs and benefits associated with certain traits. For example, although long leaf life spans likely help juvenile evergreens tolerate shade by reducing the annual costs of crown maintenance (e.g. King 1994; Lusk 2002), this advantage will only make itself felt over time frames >1 year, whereas the high initial construction cost of long-lived leaves has an immediate negative impact on plant carbon balance. A previous study in another temperate rain forest reported that, although rank order of species shade tolerance changed little with plant size, interspecific differences were less marked in seedlings than in larger size classes (Lusk et al. 2008), a finding that informed our choice of size class.
A more structured sampling regime might give more accurate and more stable estimates of compensation points. A haphazard sampling resulted in our growth measurements being obtained from (roughly) log-normal distributions of light environments (Fig. 2), meaning that the estimates of compensation points were strongly influenced by the growth of the relatively few plants found in both tails of the distribution. Stratified random sampling might yield growth measurements from something closer to a rectangular distribution of light environments. However, the feasibility of this approach will depend on the spatial configuration and disturbance history of forest stands; an ideal situation for a stratified sampling regime would be an old-growth stand with long forest margins encroaching upon open areas. Transects run parallel to the forest margin would thus provide ready access to juvenile trees growing in a wide range of light environments.
Our results are consistent with the proposal that shade tolerance is primarily of function of species differences in long-term net carbon gain in low light (King 1994; Baltzer & Thomas 2007). Our measurements integrate the consequences of species differences in leaf-level net carbon gain as well as biomass turnover and losses due to insect herbivory, pathogens and stem breakage. The wide range of whole-plant compensation points observed, associated with conspicuous crossovers in species growth rates (Fig. 3), suggest that shade tolerance can be assessed by measurements of growth alone, without the need to measure survival. Although it is quite possible that growth-survival relationships differ significantly across our five species, the occurrence of most species compensation points within the first quartile of the distribution of light environments occupied by juveniles (Fig. 4) suggests that none of our study species can tolerate sustained periods of nil or negative net growth (Baltzer & Thomas 2007).
We conclude that the whole-plant compensation point is a very promising option for measuring shade tolerance in comparative studies. Estimates of juvenile survival are an essential component of demographic approaches to the study of shade tolerance (Pacala, Canham & Silander 1993); although of clear utility for parameterizing models of forest dynamics, this approach requires researchers to either monitor very large numbers of plants or else harvest many live plants as well as ‘recently dead’ individuals (e.g. Kobe et al. 1995; Kunstler, Coomes & Canham 2009). A nondestructive alternative for describing shade tolerance differences (rather than parameterizing demographic models) is to quantify the distribution of juveniles in relation to light availability, and choose some distributional parameter as an indicator of shade tolerance (e.g. Figueroa & Lusk 2001; Lusk et al. 2008). However, the choice of which distributional parameter to use (e.g. 5th percentile, 10th percentile, mean) involves some subjectivity, which might hinder efforts to develop a standardized criterion facilitating comparisons across studies. In contrast, the unequivocal nature and clear biological significance of the whole-plant compensation point make it an attractive option, despite its time-consuming nature. Organ-level traits offer some potential for the eventual development of easily measured proxies of the whole-plant compensation point (Baltzer & Thomas 2007), though further work is required to determine the extent to which such proxies can be applied across sites.
We thank the Australian Research Council for funding through ARC Discovery Grants for support through Discovery Projects 0878209 and 1094606, Jennifer Baltzer and Lawren Sack for their constructive feedback and Peter Clarke for assuming leadership of the project in 2011 and for commenting helpfully on the draft manuscript.