Exploring How Functional Traits Modulate Species Response to Topography in Baxian Mountain, North China


 The associations between functional traits and species response to environments have aroused more and more ecologists’ interest and can provide insights into understanding and explaining how plants respond to the environment. Here, we applied a hierarchical generalized linear model to quantifying the role of functional traits in plants response to topography. Functional traits data, including specific leaf area, maximum height, seed mass and stem wood density, together with elevation, aspect and slope were used in the model. In our results, species response to elevation and aspect were modulated by maximum height and seed mass. Shorter-statured tree species had a more positive response than taller ones to an increase in elevation. Compared to light-seeded trees, heavy-seeded trees responded more positively to more southerly aspects where the soil was drier. In this study, the roles of maximum height and seed mass in determining species distribution along elevation and aspect gradients were highlighted respectively where plants are confronted with low-temperature and soil moisture deficit conditions. This work contributes to the understanding of how traits may be associated with species responses along mesoscale environmental gradients.

functional traits in plants response to topography. Functional traits data, including 21 specific leaf area, maximum height, seed mass and stem wood density, together 22 with elevation, aspect and slope were used in the model. In our results, species 23 response to elevation and aspect were modulated by maximum height and seed 24 mass. Shorter-statured tree species had a more positive response than taller ones 25 to an increase in elevation. Compared to light-seeded trees, heavy-seeded trees 26 responded more positively to more southerly aspects where the soil was drier. In 27 this study, the roles of maximum height and seed mass in determining species 28 distribution along elevation and aspect gradients were highlighted respectively 29 where plants are confronted with low-temperature and soil moisture deficit 30 conditions. This work contributes to the understanding of how traits may be 31 associated with species responses along mesoscale environmental gradients.

36
Functional traits are associated with environmental conditions and can provide 37 insights into understanding and explaining how plants respond to environments. 38 A trait-environment association is a consistent and general pattern linking a 39 biological attribute and an environmental gradient without considering taxonomic 40 identity (Díaz, Cabido, & Casanoves, 1999). Trait-environment associations may mean that only species with particular traits have the opportunity to become 42 abundant under certain environmental conditions. For instance, high-SLA 43 (specific leaf area) species that have fast growth rates and take up nutrients 44 quickly have an advantage in resource-rich environments (Ordoñez et al., 2009; 45 Westoby, Falster, Moles, Vesk, & Wright, 2002). In contrast, low-SLA species, 46 which have long-lived leaves and a low resource turnover rate, are more tolerant 47 of resource-poor conditions (Ordoñez et al., 2009;Westoby et al., 2002).

92
As explained above, we generated 10 species occurrence datasets to fit the 93 topographic model, according to the Moran's I results, there are no spatial 94 autocorrelations in the residuals of these 10 models (p>0.05). Here we picked the 95 first one to present and discuss the results, and the averaged coefficient results 96 among these ten can be found Supplementary Fig. S1  mongolica were the most prevalent species (Fig 2). Aspect had greater influence 105 on species occurrence than the other two topographic factors, and its effect was also more consistent across species (Table 1). The results from 10 sample datasets 107 also indicated that ( Supplementary Fig. S1 online). In contrast, the effect of 108 elevation on occurrences differed more widely from one species to another (Table   109 1). 110 Species responses to elevation had the highest observed level of trait modulation, 111 following by aspect and slope (Fig 3). Maximum height interacting with elevation 112 and seed mass interacting with aspect stood out from our topographic model 113 (  Supplementary Fig. S1 online). 117 Traits explained a larger part of the variation in species responses to elevation 118 than those to aspect and slope (Fig 3). According to the association's coefficient shorter-statured tree species were more common on high-altitude sites than the 123 low-altitude, while for taller-statured tree species, we could more easily find them 124 on the lower-altitude sites. 125 In addition, the association between seed mass and aspect also had a large and 126 significant coefficient (0.29, SE=0.10), so seed mass modulates the response to 127 aspect more than that to elevation and aspect. Furthermore, the association between seed mass and aspect (Fig 4, the last row, the second column) indicated 129 that trees with heavy seed had a more positive response to aspect than those with 130 an average seed mass, and most trees with small seed responded to aspect 131 negatively. Thus, species with heavy seeds were more likely to be more common 132 on south-facing sites compared to north-facing slopes, while species with small 133 seeds were more likely to have the opposite response to aspect.  Maximum height modulating species response to elevation 141 Our results showed that shorter-statured tree species had more positive responses 142 to higher-altitude elevations than taller ones. Maximum height represented 143 several ecological strategies (Westoby, 2002). First of all, taller species have a 144 greater chance of getting light before their neighbors do (Westoby, 2002). 145 Sunshine to high mountains is less likely to be blocked by their surroundings so 146 that all plants there usually are exposed to adequate light. By contrast, plants at 147 lower altitude lose such topologic advantage and the weakness in height can 148 critically affect the chance of shorter-statured tree species reach the light, while taller plants are more competitive here and distribute more. Second, growing 150 taller than its usual can be attained at the cost of the plant stem diameter growth, 151 and result in less mechanically and physiologically support to the crown (King,   152 1981; Mäkelä, 1986). However, the living condition at high mountains is usually 153 not such friendly to those thin stem plants. They are more likely to be broken by 154 strong winds or lighting strike, but those shorter and sturdy individuals are more  Table S2 online, Supplementary Fig. S3 online). 175 Seed mass modulating species response to aspect 176 According to our results, heavier-seeded trees responded more positively to more 177 southerly aspects than trees with lighter seeds. It can be explained by "seed mass In our study, this trend was indicated by the negative coefficient of the interaction between seed mass and soil moisture in the microclimatic data fitted model, 194 although not very certain (Supplementary Table S2   Species occurrence data collection 246 We sampled three, one-hectare plot sets along topographic gradients, including 247 100, 10 m x 10 m plots (Fig 1) in each set. Moreover, we broadly located 69, 10 248 m x 10 m plots outside those three sets along topographic gradients (Fig 1). In 249 order to avoid the many plots from the three one-hectare plot sets inducing 250 significant spatial autocorrelation, we resampled from those sets by putting a 3 *  Functional trait data collection 260 We followed the Leaf-Height-Seed (LHS) scheme (Westoby, 1998) been incorporated into our model as "trait-environment" interactions rather than 318 fixed effect terms. It means functional traits influence species occurrence through modulating their response to environments rather than influence their occurrence 320 directly. 321 We used blme (Chung, Rabe-Hesketh, Dorie, Gelman, & Liu, 2013) package to 322 fit our model in a Bayesian setting, which allowed us to specify a particular form 323 of weak prior to getting an approximate Bayesian maximum posterior estimation. 324 The prior distribution for the species covariance of random effects was an inverse here is equal to AUPRC of a random classifier (Saito & Rehmsmeier, 2015), to 337 show how many times the model's prediction is better than a random classifier. 338 We fitted the model with four traits (Fig 2) of 31 species and three topographic 339 variables from 150 plots. Based on the 10 datasets from grid sampling process, 340 we built up 10 topographic models, and averaged the coefficients for each fixed effect terms for visualization ( Supplementary Fig. S1 online).    Figure 1 Baxian Mountain National Nature Reserve and plot sites. Three black square boxes in the second map show the location of three plot sets, and each of them has one hundred 10 m x 10 m plots. Those red dots in 100 plots set are plots we sampled and picked. Effect of environmental variables on the occurrence of 31 species given their traits. Species names were shortened following Supplementary Table S1 online.

Figure 3
The explanatory power of traits for the response of species occurrences to topographic variables. First, we calculated the variance of species responses to environmental variables. The model's xed effect coe cients caught all the variance of species response to the environment. Then, from the model we built, we extracted the residual standard deviation by eliminating the modulation from traits. The difference among these two standard deviations was the variation explained by traits, which was shown as the light grey area.

Figure 4
The relationships between environmental responses and species traits at species (points) and mean (line) levels. These partial dependence plots show the estimated response given the trait, with all other traits and environmental variables held at their means. The environmental variables were centered, so species with values above zero have positive responses to environmental variables. Species names were shortened following Supplementary Table S1 online. For a better illustration, the values of seed mass were log transformed.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.