Potentiel d’utilisation de la diversité fonctionnelle des arbres dans l’aménagement durable des forêts tempérées nordiques et boréales

Potentiel d’utilisation de la diversité fonctionnelle des arbres dans l’aménagement durable des forêts tempérées nordiques et boréales

Response diversity, functional redundancy and post-logging productivity in northern temperate and boreal forests

Résumé

Le développement d’indicateurs de résilience des écosystèmes est essentiel pour l’amélioration des stratégies d’aménagement durable des forêts. Des études récentes suggèrent que la redondance fonctionnelle (RF) et la diversité des réponses (DR), deux indicateurs de la diversité des arbres associés au fonctionnement de l’écosystème, pourraient être utilisées comme indicateurs de la résilience des communautés forestières face à diverses perturbations. L’objectif de ce chapitre était d’examiner la relation entre ces deux indicateurs et la productivité suivant une coupe totale dans les forêts du Québec. La productivité a été mesurée par la variabilité saisonnière du Enhanced Vegetation Index, une mesure courante de la productivité primaire de la biomasse végétale estimée à partir d’images satellites. De plus, j’ai testé l’hypothèse que les essences de feuillus et de conifères contribuent de façon distincte à la productivité après coupe en mesurant les deux indicateurs de diversité séparément pour chaque groupe d’essences. Le rôle de la richesse spécifique et des effets d’identité des espèces sur la réponse de la productivité après coupe ont aussi été examinés.
Cette analyse a démontré la nature complémentaire des indicateurs traditionnels de diversité et des indicateurs basés sur les traits fonctionnels lors de l’analyse de la réponse des communautés forestières à une perturbation sévère. J’ai trouvé une relation significative et positive entre la DR avant-coupe des feuillus et la productivité après coupe, ainsi qu’une relation significative et négative entre la DR des conifères et la productivité. Cette relation négative avec la productivité après coupe est probablement causée par la régénération plus lente des conifères et par la courte taille de l’intervalle de temps analysé. Les effets négatifs d’identité causés par l’épinette noire reflètent probablement la plus grande susceptibilité des peuplements dominés par cette espèce à des facteurs menant à une réduction de la productivité après coupe, comme la paludification et l’invasion par des éricacées. La diversité des réponses était un meilleur indicateur de la productivité après coupe que la redondance fonctionnelle. Cette étude est parmi les premières à trouver une relation significative négative entre la DR et le fonctionnement d’un écosystème, c’est-à-dire la DR de communautés des conifères et leur productivité après coupe.

Abstract

The development of efficient ecosystem resilience indicators has been identified as one of the key research priorities in the improvement of existing sustainable forest management frameworks. Two indicators of tree diversity associated with ecosystem functioning have recently received particular attention in the literature: functional redundancy (FR) and response diversity (RD). We tested the hypothesis that these indicators could be used to predict post-logging productivity in forests of Québec, Canada. We analysed the relationships between pre-logging FR and RD in temporary sample plots and post-logging productivity, measured as seasonal variation in Enhanced Vegetation Index obtained from MODIS satellite imagery. The effects of the deciduous and coniferous tree components in our pre-disturbance diversity assessments were isolated in order to examine the additional hypothesis that they have different impacts on post-disturbance productivity. The role of tree species richness and species identity effects were also examined.
Our analysis revealed the complementary nature of traditional biodiversity indicators and trait-based approaches in the study of biodiversity-ecosystem functioning relationships in dynamic ecosystems. We report a significant and positive relationship between pre-disturbance deciduous RD and post-disturbance productivity, as well as an unexpected significant negative effect of coniferous RD on productivity. This negative relationship with post-logging productivity likely results from slower coniferous regeneration speeds and from the relatively short temporal scale examined. Negative black spruce-mediated identity effects were likely associated with increased stand vulnerability to paludification and invasion by ericaceous shrubs that slow down forest regeneration. Response diversity outperformed functional redundancy as a measure of post-disturbance productivity most likely due to the stand-replacing nature of the disturbance considered. To the best of our knowledge, this is among the first studies to report a negative significant relationship between a component of RD and ecosystem functioning, namely coniferous RD and forest ecosystem productivity after a stand-replacing disturbance.

Introduction

Natural and anthropogenic transformations of forest ecosystems threaten their capacity to sustain the provision of numerous ecological services (MacDicken et al., 2015, Thom and Seidl 2016). In the face of such uncertainty, the development of efficient resilience indicators capable of predicting ecosystem response to disturbances has been identified as one of the key research priorities in the improvement of existing sustainable forest management frameworks (Mori et al., 2016). Specifically, developing indicators capable of predicting primary productivity following natural and anthropogenic disturbances could be a very useful step towards that goal. While primary productivity is not the only ecosystem service provided by forests, it is considered to be a main concern of forest management (Mori et al., 2016), since it is one of the key supporting ecosystem processes other services depend on (Millennium Ecosystem Assessment 2005). For instance, primary productivity regulates the strength of cascading effects of large herbivores on community function and structure (Pringle et al., 2007), and directly influences carbon dynamics (Hulvey et al., 2013). Preventing the decline of forest productivity following harvesting is therefore of particular interest to forest managers (Bose et al., 2014).
Tree diversity has been shown to play a fundamental role in temperate and boreal forest productivity (Liang et al., 2016). In these types of forest, species richness positively influences biomass production across all vegetation layers (Zhang et al., 2017). This relationship is stronger for overstory tree species richness, which is also positively associated with ecosystem multifunctionality and understory plant species richness (Gamfeldt et al., 2013). Primary productivity is also significantly affected by the traits of the dominant species due to species identity effects (Grossman et al., 2017). While compositional and functional tree diversity have been shown to contribute to forest productivity (Zhang et al., 2012), these two components of biodiversity can be independent of one another and have distinct impacts on ecosystem functioning. For instance, in an experimental study where functional diversity was allowed to vary independently from tree species richness, Tobner et al., (2016) showed that stands with identical species richness and greater functional diversity can be significantly more productive. Such discrepancies can be further magnified when the impact of disturbances is considered: following the flooding of a grassland experimental study, species richness reduced plant community resistance to flood while functional traits drove post-disturbance increase in biomass (Fischer et al., 2016).
Functional diversity is particularly suited to studying biodiversity-ecosystem functioning relationships under disturbances because they define mechanistic links between biodiversity and function, including responses to disturbances (Cadotte et al., 2011). However, due to the large number of indicators available (Laliberté & Legendre 2010), indicator selection is highly dependent on the objective of the study. Two particularly promising functional diversity-based indicators of ecosystem resilience to disturbances have been highlighted by recent reviews on this subject: functional redundancy (FR) and response diversity (RD; e.g. Mori et al., 2016, 2013). The concept of FR is based on the insurance hypothesis put forward by Yachi and Loreau (1999): ecosystem functioning should be less affected by the absence or extinction of a species that can be replaced by another one that contributes similarly to ecosystem functioning, than by one for which no analogue is readily available. RD represents the different capacity that species within a given functional effect group (i.e. species that perform similar functions) have to respond to disturbances (Mori et al., 2013). In theory, greater inter-specific variation in responses to environmental fluctuations within each functional effect group should prevent disturbances from eliminating the majority of a given functional group, thus ensuring the preservation of the corresponding ecosystem functions (Elmqvist et al., 2003).
To date, empirical evidence demonstrating the ability of RD and, to a lesser degree, FR, to be adequate indicators of resilience to disturbances is still scarce, especially for forest ecosystems (Mori et al., 2016). The few real-world studies that do exist tend to be limited in scope and scale (Mori et al.,2013), or tend to indirectly test the linkages between functional traits and ecosystem functioning (e.g. Laliberté et al., 2010). Additionally, studies of the importance of RD in other taxa are not always conclusive (e.g. Cariveau et al., 2013). Although experimental studies on the relationships between FR, RD and ecological resilience are more common, they usually focus on grassland ecosystems (e.g. Pillar et al., 2013) and their ability to infer real-world biodiversity-ecosystem functioning relationships is restricted. Among other issues, such studies typically exclude disturbances that substantially influence biodiversity-ecosystem functioning relationships (Brose and Hillebrand 2016).
We present a large-scale empirical study of northern temperate and boreal forest plots in Québec (eastern Canada) aiming to improve our understanding of the real-world relationships between pre-disturbance functional diversity indicators (functional redundancy, response diversity) and post-disturbance productivity trajectories within a management-relevant context. We treat deciduous and coniferous species as two functional effect groups because they contribute differently to ecosystem productivity, they are easily identified by forest managers, and changes in their composition should greatly influence ecosystem processes and services (see Functional effect groups section in the methodology). We examined forest plots recovering from logging to test the hypothesis that pre-disturbance coniferous and deciduous functional redundancy and response diversity are good indicators of post-disturbance productivity. We calculated these functional diversity indexes per functional effect group in order to test the additional hypothesis that these groups have different impacts on post-disturbance productivity. We also controlled for tree species richness and species identity effects in order to assess whether trait-based approaches improve our understanding of biodiversity-ecosystem functioning when these traditional approaches are considered. In order to test this hypothesis, we built 10-year time series of EVI (enhanced vegetation index), a remotely sensed primary productivity indicator (Huete et al., 2002), for forest plots of Québec that were clear-cut immediately before the beginning of the time series. We further discussed underlying mechanisms driving these relationships and explored the management implications of our results.

Methodology

Study area

We quantified pre-disturbance tree functional diversity metrics from temporary sampling plots selected from the Quebec provincial forest inventory programs conducted between 1992 and 2009 (MFFP 2016). Among these plots, a subset was selected according to four criteria. First, plots had to have been clear cut between 2000 and 2006 after they had been measured, so that an uninterrupted 10-year post-disturbance time series of MODIS data was available. Hence, the 10-year time series started between 2000 and 2006 and ended between 2011 and 2015. Second, in order to test the hypothesis that functional effect groups were important for determining post-disturbance productivity, plots needed to include at least one species from each group (deciduous and coniferous species). Both groups had to be present because Rao quadratic diversity, the measure used to calculate response diversity, can only be calculated when all functional groups are present. Otherwise, missing values are generated (Laliberté & Legendre 2010). Third, we discarded plots sampled over 10 years before they were clear cut. For all other plots, we considered unlikely that forest communities underwent key changes within 10 years if no stand-replacing disturbances occurred. Finally, only temporary plots located in MODIS (Moderate-resolution imaging spectroradiometer) 250 m pixels where over 80% of the area had undergone the same disturbance were kept. If the majority of a neighboring MODIS pixel had been clear cut but the pixel in which the temporary plot was positioned had not, the plot was associated with the neighboring MODIS pixel instead.
A total of 796 plots were selected according to these criteria (Figure 2.1). These plots were spread across a latitudinal gradient that encompasses multiple bioclimatic domains: sugar maple-bitternut hickory (Carya cordiformis), sugar maple-yellow birch (Betula alleghaniensis), balsam fir (Abies balsamea)-yellow birch, balsam fir-white birch (Betula papyrifera) and black spruce (Picea mariana)-feathermoss (Robitaille and Saucier 1998). Each plot consisted of a circular area of 400 m2.
Within this area, all trees whose diameter at breast height (DBH) was greater than 9 cm were recorded and their DBH was measured (MFFP 2016). A smaller 40 m2 circular plot in which all saplings (DBH≤ 9 cm) were identified and counted by DBH class was located at the center of each 400 m2 plot. Sampling effort varied between main vegetation zones (deciduous, coniferous and mixed) and followed a random stratified design (MFFP 2016).

Functional effect groups

We adopted a hierarchical effect-response functional trait framework to analyse the relationships between tree functional redundancy, response diversity and ecosystem productivity (e.g. Laliberté et al., 2010). We first employed an unsupervised hierarchical clustering algorithm to identify functional effect groups using 6 effect traits: (i) average maximum height; (ii) leaf phenology (whether species lose all foliage seasonally or not); (iii) nutrient uptake strategy (presence of arbuscular mycorrhiza, ectomycorrhizal, or both); (iv) nitrogen content per leaf mass unit; (v) wood density; and (vi) leaf mass per area (R’s stats package hclust algorithm; R Core Team 2016). These traits are publicly available (e.g., Paquette et al., 2015) and have been suggested as effect traits associated with tree growth, photosynthetic rate and productivity (Cornelissen et al., 2003). The clustering algorithm followed Ward’s minimum variance method and was applied to a Gower dissimilarity matrix (caret R package, version 6.0-62; Kuhn, 2015). A visual inspection of the resulting dendrogram (Appendix 2.A) revealed two main functional effect groups: coniferous and deciduous species.
Considering deciduous and coniferous species as having distinct effects on forest productivity and, more generally, on function, makes sense for a number of reasons. First of all, species within these two groups support distinct animals (e.g., Drapeau et al., 2000) and can be conceived of as ‘umbrella’ species that reflect functional diversity in the understory layers (Fourrier et al., 2015). Secondly, coniferous species have greater leaf mass per area, which is associated with longer leaf lifespan, increased leaf defences and reduced decomposition, growth and maximum photosynthetic rates (Appendix 2.B; Cornelissen et al., 2003). Thirdly, wood density, which is associated with carbon storage capacity, relative growth rate and stem defences (Cornelissen et al., 2003), tends to be greater in deciduous species (Appendix 2.B). Finally, coniferous species have a narrower range of average maximum height (Appendix 2.B), a functional trait that has been linked to competitive vigour, stress response strategies and aboveground biomass (Cornelissen et al., 2003).

Functional redundancy and response diversity

We measured each deciduous and coniferous pre-disturbance functional redundancy (FR) and response diversity (RD) on each identified plot following the framework and code provided by Ricotta et al., (2016). FR was estimated with the six previously mentioned effect traits used to establish the functional effect groups: average maximum height, leaf phenology, nutrient uptake strategy, nitrogen content per leaf mass unit, wood density and leaf mass per area. The following seven functional response traits were used to estimate RD: average maximum height, growth rate, wood density, vegetative reproduction capacity, seed mass, shade tolerance and capacity to establish seed banks. These response traits are directly associated with tree regeneration speed and strategies: these influence tree species ability to colonize sites after disturbance and are key aspects of post-disturbance recovery (Cornelissen et al., 2003). Deciduous and coniferous response diversity were calculated for each plot as the Rao quadratic diversity, according to the following equation (2.1): =∑ ∑ , where RD is the Rao quadratic diversity, pi is the relative abundance of species i, pj is the relative abundance of species j, and δij is the pairwise functional dissimilarity between species i and j. If there is only a single species in the plot, species j equals species i. This functional dispersion-based indicator estimates the average distance between two randomly selected individuals within the functional trait space (Botta-Dukat 2005). Functional redundancy was calculated as 1 minus the ratio of Rao quadratic diversity and the Simpson index according to the following equation (2.2): =1− , where FR is functional redundancy, Q is Rao quadratic diversity, and D is the Simpson index. Thus, this indicator compares the observed functional diversity with that of the most functionally distinct community possible that shares an identical abundance distribution (Ricotta et al., 2016). Hence, plots dominated by a single species have a FR of 1 and a RD of 0. Gower dissimilarity matrices were chosen because they can handle both missing values and mixed variable types (continuous, ordinal and categorical). These two metrics were not re-measured after clear-cutting took place (post-disturbance period). Species response trait values were weighted by basal area relative abundance. Most trait values were collected by a previous study (Paquette et al., 2015). Missing values and additional variables were collated from other online data sources (Appendix 2.B).

Ecosystem productivity metric

Post-disturbance 10-year productivity time series were built using 16-day MODIS EVI (enhanced vegetation index) data. EVI is a productivity indicator based on the surface reflectance of solar radiation that has clear links to primary productivity (Pettorelli et al., 2005). The MODISTools R package (Tuck et al., 2014) was used to download each temporary plot’s 250 m pixel EVI and pixel reliability data from the MOD13Q1 MODIS product (Didan et al., 2015). The original 16-day time series data were smoothed using TIMESAT (Jönsson and Eklundh 2004). EVI data contribution to the smoothing functions was weighted using the complementary reliability layer and outliers were removed by multiplying the weights from a seasonal trend decomposition with the original weights. Asymmetric Gaussian functions were then fit to the data. This type of function was chosen because it has been found to be among the top two performing smoother functions for this kind of dataset (Hird and McDermid 2009) and appeared to perform slightly better than the double logistic function for our dataset. The seasonal variation of EVI (maximum EVI – minimum EVI; svEVI), a productivity measure that has been shown to be significantly correlated with gross primary productivity in North America (Sims et al., 2006), was then calculated. In order to remove any badly smoothed data from the dataset, svEVI points above the 0.999 and below the 0.001 percentiles of the distribution were removed.

Statistical analysis

Autoregressive linear regression models of svEVI (nlme R package; Table 2.1; Pinheiro et al., 2015) were built with the following explanatory variables: (i) yearly climate variables; (ii) pre-disturbance stand characteristics; (iii) year of logging; (iv) site post-disturbance land cover class at the year 2013 (deciduous, coniferous or shrubland); (v) pre-disturbance tree species richness; (vi) pre-disturbance relative abundance of the most abundant tree species of each functional group (black spruce and white birch); (vii) pre-disturbance coniferous and deciduous FR and RD; (viii) number of years since disturbance as a numeric variable; and (ix) a binary categorical variable (set at 1 for 2 to 5 years after disturbance and 0 for the remaining years) that was introduced to allow the models to more realistically consider the relationships between FR, RD and time since disturbance observed in the data. The first year after disturbance was not included in this categorical variable because the noise introduced by the time lag between the time since disturbance and the first growing season (from a few weeks up to several months) had a greater impact on the first measurement of productivity.
Plot-level yearly climate variables (average annual temperature, annual precipitation, growing degree days over 5ºC, potential evapotranspiration, water balance and growing season length) were calculated using BioSIM (version 10; Régnière et al., 2014), a software tool that uses geographical coordinates, elevation, slope and aspect to interpolate climate data. Pre-disturbance stand characteristics (age class, height class, density class and cover type) were directly measured in the sampling plots (MFFP 2016). The top 3 principal components of climate and pre-disturbance stand characteristics were then extracted through Principal Component Analysis (PCA). For the PCAs, a categorical variable (cover type) was transformed into three binary variables, ordinal variables were first transformed into numeric variables according to the middle point of each class, numeric variables were log-transformed, and data were centered by subtracting the mean and scaled by dividing the predictor values by the standard deviation (caret R package, version 6.0-62; Kuhn, 2015).
Considering that site productivity is greatly influenced by different post-logging regeneration trajectories, post-disturbance land cover class was estimated using the land cover type 3 of the MODIS MCD12Q1 product (Friedl et al., 2010). This MODIS product provides an estimate of the land cover class at a spatial resolution of 500m, which is larger than the one svEVI was estimated at, but is still likely to reflect the type of regeneration of the disturbed area. Although this MODIS product was not available for the whole time series (it ended in the year 2013 and was not available for the whole time series up to that year), its value by the year 2013 was included in the regression models as an indicator of the type of vegetation the disturbed area was likely regenerating into.
Since the identity of the tree species present can have considerable impacts on ecosystem processes (Hooper and Vitousek 1997) and it has been suggested that primary productivity can be substantially influenced by the traits of the dominant species (Grime 1998), the possible influence of species identity effects on post-disturbance productivity was also assessed. In order to do this, within-group pre-disturbance basal area relative abundance of the most abundant species within each functional group (deciduous: white birch; coniferous: black spruce) was added to the candidate model set.
A total of 10 autoregressive linear regression models with svEVI as the response variable were built. Models shared all previously mentioned variables except FR, RD and species relative abundances. These variables were added to distinct models as combinations of two groups of variables: (i) functional diversity metrics (FR and RD, FR only and RD only); and (ii) species relative abundance (none, black spruce and white birch). A null model with tree species richness but without FR, RD nor species relative abundance was also included in the candidate model set. As we were comparing models with a similar random effect structure and different fixed effect structures, models were fit with log-likelihood maximization (Burnham & Anderson 2002). All input variables were standardized to a mean of 0 and a standard deviation of 0.5 in order to set all effect sizes on comparable scales and facilitate their interpretation (Grueber et al., 2011). The temporal correlation structure of our dataset was accounted for by specifying unique plot ID as a random effect with an autocorrelation structure of order 1. The best models were selected according to the AICc change (second-order Akaike Information Criterion; Burnham & Anderson 2002). Model variances were homogenous, the model residuals were normal and no multicollinearity was detected among explanatory variables (all variance inflation factors were lower than 2; Marquardt, 1970).

Results

Functional redundancy and response diversity

The model including coniferous and deciduous pre-disturbance response diversity and black spruce relative abundance was the top model within the main candidate model set (Table 2.1; AICc weight
= 0.97). Our analysis also shows that functional diversity can complement species richness and species identity effects in explaining post-disturbance productivity: the top model was more parsimonious than models that did not include either functional diversity or species richness, and models that considered the tree community as a whole, instead of splitting it into two functional effect groups (Appendix 2.C). The fixed component of the top model explained approximately 48% of the observed variation and the whole model explained approximately 71% (Table 2.1). All models were substantially more parsimonious than the null model that did not consider FR, RD nor any species identity effects (Table 2.1).
Our analyses revealed a significant negative relationship between coniferous RD and post-disturbance productivity (Table 2.2). Two dominant species mixtures likely mediate this relationship. First, black spruce-balsam fir dominated stands are common throughout our study area (Figure 2.2h). These plots tend to have relatively low levels of pre-disturbance coniferous RD and high levels of post-disturbance productivity (top parabola in Figures 2.3a & b). Second, balsam fir-white spruce stands are also relatively common and have high levels of pre-disturbance coniferous RD and low levels of post-disturbance productivity (bottom parabola in Figure 2.3a).
We found a significant positive relationship between pre-disturbance deciduous RD and post-disturbance productivity (Table 2.2). This result supports the initial hypothesis that increased RD should lead to increased post-disturbance productivity. While significant, the p-value of this relationship is relatively close to the commonly accepted significance threshold of 0.05. The p-value is only marginally significant probably because of the large number of plots whose deciduous component is dominated by a single deciduous species, namely white birch (Figures 2.2a & e): these plots have low values of deciduous RD and post-disturbance productivity (Figure 2.3c). Deciduous and coniferous functional redundancy were not present in the top model of our candidate model set (Table 2.1).

Species identity effects

Since functional effect groups were often dominated by a single species, species identity effects were widespread in our study area. The pre-disturbance deciduous effect group of approximately 38% of all plots (n = 299) was dominated by a single species (white birch). This resulted in a large number of plots with maximal pre-disturbance deciduous functional redundancy (FR; Figure 2.2a) and minimal pre-disturbance deciduous response diversity (RD; Figure 2.2c). White birch is a species of particular interest, since it dominated the deciduous component in 25.6% of the plots in our dataset (n = 204; Figure 2.2e).
The coniferous functional effect group was less dominated by any single species, but significant species identity effects were still present. Hence, fewer plots had maximal levels of pre-disturbance coniferous FR (Fig. 2b) and minimal levels of pre-disturbance coniferous RD (Fig. 2d). Nevertheless, black spruce and balsam fir were particularly dominant within the coniferous functional effect group: over three quarters of the coniferous basal area was occupied by black spruce and balsam fir in 69.6% of our plots (n = 554; Fig. 2h).
Black spruce identity effects appear to play an important role in post-disturbance productivity: black spruce relative abundance was a significant variable negatively correlated with post-disturbance productivity (Table 2.2). Black spruce relative abundance also appears to be negatively associated with coniferous RD, as plots where black spruce was less dominant tended to have greater levels of pre-disturbance coniferous RD (Figure 2.3b). Direct white birch identity effects on post-disturbance productivity do not appear to be significant: models containing this variable were considerably less parsimonious than the top model (Table 2.1).

Discussion

The complementarity between traditional species-oriented biodiversity indicators, such as species richness, and functional diversity indicators reported in this study highlights the contribution of trait-based approaches to the study of biodiversity-ecosystem functioning relationships. While the complementarity between these facets of biodiversity and tree productivity has been reported elsewhere (e.g. Paquette & Messier 2011), few studies have explicitly addressed these relationships with respect to disturbances. For example, a simulation study of a temperate central European forest found that the positive effects of tree species richness on net primary productivity resilience to simulated natural disturbances was likely mediated by changes in functional diversity (Pedro et al., 2015). However, direct effects of changes in functional diversity were not assessed. As functional diversity indicators can be independent of species richness (Ricotta et al., 2016), it is important to take into account both contributions when studying biodiversity-ecosystem relationships.
The absence of pre-disturbance functional redundancy (FR) from the top model is possibly due to the severity of the analysed disturbance. According to Yachi & Loreau (1999), FR should be an efficient indicator of ecosystem resilience to disturbances because the impact of the loss of a given species on ecosystem functioning should be reduced when there are multiple species performing similar functions. Indeed, this hypothesis appears to hold when partial disturbances are considered (e.g. Pillar et al., 2013). However, as our results suggest, this is unlikely to be the case when severe stand-replacing disturbances occur: as these types of disturbances tend to extirpate all species from a given area, a larger pool of species performing similar functions is unlikely to improve post-disturbance performance of those ecosystem functions.
In the face of such severe disturbances, response diversity (RD) is more likely to influence post-disturbance ecosystem functioning. Instead of measuring how similarly species perform a given ecosystem function, this indicator quantifies how species within the same functional group respond to various types of disturbances (Elmqvist et al., 2003). As disturbance response is trait-dependent (Mori et al., 2013), an appropriate trait selection can allow researchers to successfully quantify species responses to stand-replacing disturbances. The relevance of this variable is supported by the significant negative and positive relationships reported between coniferous and deciduous RD and post-disturbance productivity, respectively. The observed differences in effect direction likely result from the different regeneration speeds of these two functional groups and the relatively short temporal scale examined (up to 10 years after clear cutting took place). Considering that, following severe disturbances, deciduous species tend to occupy the upper canopy cover faster than the main shade-tolerant coniferous species within our study area (Pothier and Auger 2009), we are more likely to detect the effect of multiple deciduous regeneration strategies on post-disturbance productivity within such a relatively short period of time. As their coniferous counterparts are slower to recover, we were probably only able to detect the effect of a limited number of regeneration strategies on post-disturbance productivity. A bias in the relationships between the seasonal variation of EVI and gross primary productivity across vegetation composition types could have also potentially influenced the observed results: while the EVI-gross primary productivity relationship is strong in coniferous forests, it is generally stronger in deciduous forests (Huete et al., 2010). These results might also have been affected by the response traits used: even though all chosen response traits are associated with tree regeneration speeds and strategies (Cornelissen et al., 2003), their importance might vary according to forest type.
Black spruce-mediated identity effects influenced the magnitude of the effect of coniferous RD and FR on post-disturbance productivity. The observed negative black spruce-productivity relationship is supported by the mass-ratio hypothesis, which proposes that some ecosystem functions are mainly dictated by the dominant species (Grime 1998). This hypothesis has previously been proposed as an alternative mechanism to explain productivity in grassland ecosystems (Sasaki and Lauenroth 2011) and has been promoted as an important underlying process driving productivity in forests (Mori et al., 2016, Grossman et al., 2017). Black spruce-dominated stands are at a considerably greater risk of decreased productivity due to paludification and invasion by ericaceous shrubs (Thiffault et al., 2013). For instance, mixed trembling aspen-black spruce stands are less vulnerable to paludification than pure black spruce stands as a result of increased nutrient cycling (Légaré et al., 2005), while ericaceous shrubs produce slowly-decomposing litter that sequesters soil nitrogen (Joanisse et al., 2009), compete with black spruce for soil resources and excrete harmful allelochemicals (Yamasaki et al., 2002). Therefore, the observed negative relationship between black spruce abundance and post-disturbance productivity likely reflects this increased risk of paludification and invasion by ericaceous shrubs. A study comparing black spruce, trembling aspen and jack pine-dominated stands also found that black spruce-dominated stands were the least productive in terms of annual aboveground net primary productivity (Reich et al., 2001).
While our results suggest that common species play a more critical role in ecosystem functioning, the number and nature of the services in question needs to be taken into account before strong conclusions are drawn. In the literature, it is unclear whether common or rare species are more important in dictating biodiversity-ecosystem functioning relationships. Common species appear to be more influential when productivity is the ecosystem function under consideration (e.g. Vile et al., 2006), but some regulating and recreational services appear to be more dependent on rare species (e.g. Zavaleta, 2004). These relationships are further complicated when multifunctionality, one of the main objectives of sustainable forest management (Gustafsson et al., 2012), is directly acknowledged. Even though primary productivity directly influences other ecosystem functions (e.g. carbon storage; Hulvey et al., 2013), multifunctionality was not directly considered in our analyses. Further studies are needed to disentangle these complex relationships, as the few published scientific articles on this issue have found evidence on the importance of both rare (Soliveres et al., 2016) and dominant species (Lohbeck et al., 2016) in driving ecosystem multifunctionality.

Table des matières

Introduction générale
Le nouveau paradigme forestier
Les forêts du Québec
Les perturbations naturelles
Chablis
Épidémies de tordeuses des bourgeons de l’épinette
Feu
Les services écosystémiques
La biodiversité et les services écosystémiques
Le potentiel de la diversité fonctionnelle
Objectifs de la thèse
Chapitre 1: Stand height and cover type complement forest age structure as a biodiversity indicator in boreal and northern temperate forest management
Résumé
Abstract
Introduction
Methodology
Study area
Beta diversity analysis
Alpha diversity measures
Alpha diversity statistical analysis
Results
Beta diversity analysis
Alpha diversity analysis
Discussion
Conclusions
Acknowledgements
References
Tables and figures
Supplementary material
Chapitre 2: Response diversity, functional redundancy and post-logging productivity in northern temperate and boreal forests
Résumé
Abstract
Introduction
Methodology
Study area
Functional effect groups
Functional redundancy and response diversity
Ecosystem productivity metric
Statistical analysis
Results
Functional redundancy and response diversity
Species identity effects
Discussion
Conclusions
Acknowledgments
References
Tables and figures
Supplementary material
Chapitre 3: Mitigation of water loss and xylem resistance to cavitation influence the response of stand mortality to severe drought, but not productivity, in northern temperate and boreal forests
Résumé
Abstract
Introduction
Methodology
Study area
Identification of drought conditions
Productivity and mortality
Functional traits
Structural equation modelling
Results
Permanent sample plot dataset
Structural equation models
Discussion
Conclusion
Acknowledgments
References
Tables and figures
Supplementary material
Conclusion générale
Implication des résultats
Limites
Perspectives
Bibliographie

Télécharger le rapport complet