Project topics and materials
Get Complete Project Material File(s) Now! »
Responses of species presence to woody substrate type and mean substrate diameter
The effect of mean substrate diameter on species presence was modeled both as a fixed linear effect and a quadratic effect (Table 6). The quadratic response type had better DIC values for the majority of species (13/20). When compared with models without mean diameter (Table 3), DIC values preferred models with mean diameter in only six species (N. complanata, F. tamarisci, U.bruchii, I. myosuroides, D. scoparium, and O. lyellii); these models all had mean diameter modeled as a quadratic response. The response of species presence probability to substrate type in models containing both substrate type and mean substrate diameter variables were very similar in effect magnitude as compared to models testing only substrate type (not shown). The numeric value of the effects of the multivariate model were slightly weaker, as expected since these models also had mean substrate diameter to explain a portion of the variation in species presence probability. Other differences between substrate effects when models include the mean diameter variable include the weakening of the effect magnitude of branches in P. filiforme and B. capillare, and the weakening of the effect magnitude of snags in T. tamariscinum. In one species, U. bruchii, including the mean diameter variable caused the positive effect magnitudes of branches and snags to increase from (+) and (0) to (++).
All effects of mean substrate diameter, once substrate type was accounted for, on species presence were negligible or inconclusive (Table 7). The one exception is P. filiforme whose presence was found to increase slightly (+) when substrate diameter is increased from initially small values.
Discussion
Model comparisons for species richness
When modeled as either fixed linear or random linear, the woody substrate type variable’s response type of best fit differed between total species richness and forest species richness. For total species richness, substrate type was best modeled as a random linear variable because substrate types had different effects depending on the forest. This could be an indication of the beta species diversity between forests if these differences between forests are due to different bryophyte communities in each forest responding to substrate types in different ways. Conversely, if bryophyte communities in each forest are very similar, the differences in woody substrate type effects may come from environmental differences between forests that in turn alter the capacity of substrates to sustain of bryophyte species. This has been shown to be the case for vascular plants (Zilliox and Gosselin 2014). Forest species richness’s preference for substrate type as a fixed linear variable suggests that the different forests’ environmental characteristics did not greatly impact woody substrate type effects. Forest species are a smaller group, of more particular ecological characteristics, so it could be expected that forest species communities respond similarly from one forest to another, even if these communities have high beta species diversity between forests.
Substrate effects on species richness
For the effect of each substrate type on species richness, we see that logs have a positive effect on forest species diversity, but a negligible effect on total species richness. This could denote a difference in the ecology of forest species compared to all epiphytic bryophytes. We may expect this difference in forest species diversity if it were particularly rich in saproxylic species that associate with dead wood, but of the 69 species comprising of the forest species subgroup (as classified by Schmidt et al. 2013 for Germany), only 23 species are particularly associated to dead wood (as classified by Hill et al. 2007 for Britain Isles). Therefore it is not clear why there is a particular preference for logs in forest species richness.
This could be due to the fact that the databases of ecological attributes we used are not completely accurate for French forests and the bryophytic associations therein.
Meanwhile, branches had a consistent negative effect on species richness compared to the strongly positive effects of living trees and snags. The difference between branch, snag, and log effects brings to bear the importance of substrate size on species richness. A question for forest management practices is how much and what kind of dead wood to leave behind when removing trees to mitigate biodiversity loss. These findings suggest that branches are not an optimal woody substrate to leave behind, but that larger standing or fallen dead wood offer a stronger biodiversity benefit.
Model comparisons for species presence:
Although the 20 species tested for species presence did not all prefer models with support type as a fixed linear variable, this response was used for all species to streamline comparisons. Furthermore, three of the four species preferring the null model, where substrate type was not taken into account – I. myosuroides , D. montanum ,and O. lyellii – were the three least frequent species of the chosen set. These species were simply never observed in certain forests; I. myosuroides, for example, was entirely absent in Aubrive, Ballons Comtois, and Citeaux – three of the six forests surveyed. In such cases, the forest variable would be very indicative of species presence probability, which may explain why substrate type did not further explain the variation in these species according to the DIC values.
It is worth mentioning that while DIC offers insight on which model structure best fits the data, it is also important to consider response magnitude – even more so for conservation and management practices. Decisions that affect the system are done in situ without complete certainty or information (McCarthy and Burgman 1995). The role of the forest manager is one of compromise and taking risks; therefore, a statistically significant parameter that has a weak response may be of less interest from a management perspective than a less statistically certain parameter with a large response.
Substrate effects on species richness and species presence
As expected from the results on species richness, almost all individual species tested responded negatively to branches. This could an indication of a preference for substrates of larger size, because branches have the smaller surface areas than other substrates and may be able to sustain fewer species. In this case, we would expect an effect of diameter in the models that include substrate diameter as a predictive variable, but this turns out to not be the case (see next section).
Three species responded positively to branches: F. tamarisci, U.bruchii, and O. lyellii. However, each of these species also associated positively to living trees. It is therefore possible that seeing these species on fallen branches is a result of them colonizing branches in living tree canopies, before the branch fell.
A question we asked was whether congenerics would fall into the same prospective specialist group. The findings had mixed results in this regard. Both U. crispa and U.bruchii appeared as living tree specialists, whereas only one species of the other genera with two species represented (Isothecium, Frullania, and Dicranum) appeared as a specialist. Of these three pairs of species, the rarer species was the specialist in only one case, with I. myosuroides appearing as a branch avoider. The living wood specialist F. dilatata and the dead wood specialist D. scoparium, on the other hand, were more frequent than their more generalist congenerics. Studies on species presence according to wood decomposition reveal similar specialist groups for the species mentioned (Odor and van Hees 2004, Sabovljevic et al. 2010).
Prospective specialist species found in these analyses are not final definitions of ecological functional groups. To fully understand the unique ecology of epiphytic bryophyte species in French ecosystems, we would need to do more pointed studies on these particular species and determine their preferred habitats. The broad overview of the GNB project and my analyses thereof allows us to pinpoint species of interest that may be specialists.
Mean diameter effects on species richness
Curiously, mean substrate diameter did not show a positive non-negligible effect in species presence when increased by a notable 23.8617cm, despite our expectations to the contrary (see also Király et al. 2013, Heilmann-Clausen et al. 2005). The only species to show a non-negligible response to diameter was P. filiforme. A preference for larger diameters may be expected for species for which branches had a strong negative effect on species presence because branches are the smallest substrate type. However, P. filiforme was one of the few species to not have a negative response to branches.
The inherent repartitioning of substrate diameters among the different substrate types may in fact be the reason why we don’t see an effect in the diameter variable. The maximum mean diameter of branches is 27cm, while for the other substrate types we may see mean diameter measurements from 30cm to 158cm. Therefore, it is possible that the negative effects seen in branches are an inadvertent combination of both substrate effect and small diameter. If this is the case, it would explain why the diameter variable explains so little additional variance after substrate type had been taken into account. To address this, it may be worth testing the effect of mean substrate diameter alone, without the effect of substrate type. Another possibility is to add an interaction variable between diameter and substrate type. This way, we may be able to determine whether – and if so, how – diameter effects change depending on the substrate type. In this case, for example, we may expect branches to still have a generally negative effect on species presence, but for large branches to have a more positive effect than small branches.
An additional problem encountered had to do with the lower ranges of diameter measurements between substrate types. According to the GNB protocol (Annex 1), logs and snags were defined as substrates measuring at least 30cm in diameter, but our data shows the logs and snags with diameters inferior to this threshold, indicating erroneously labeled branches. This error will be corrected for in future analyses. That said, this accidental blurring of substrate type effects did not prevent us from seeing a clear difference between branches and other substrate types.
Prospective analyses: next steps
As of the current writing of this report, there is still over a month in my internship during which I will continue working on this project. During this time, I will continue to explore different model structures and predictive variables to model species richness and species presence. The different response types tested so far have been fixed linear, random linear, and quadratic. Another response type for quantitative variables is the threshold response, wherein the effect of the predictive variable on the response variable stays constant up until a certain value, after which the effect jumps up or down. For example, this type of model would predict that species richness would not change with mean substrate diameter until the diameter was of a certain minimal, or threshold, value. A last possible response type for quantitative variables is the sigmoid response. Here, the effect on the response variable encounters first a lag phase as the predictive variable increase, then rises sharply at an inflection point, then finally reaches a plateau. This can be seen as a smoother, less sudden presentation of the threshold response. This more gradual response may be more biologically accurate and may provide insight on the dynamics of species richness and presence indicators. Conversely, threshold responses are of interest for management purposes, because their delineation of when an effect increases offers forest managers clear goals.
Along with different response types, there are different possible predictive variables at the substrate level to test. Although logs and snags are necessarily dead wood, an indicator of interest is the level of decomposition of the woody substrate. Different species are linked to certain decomposition stages (Odor and Van Hees 2004, Sabovljevic et al. 2010, Andersson and Hytteborn 1991). Dead woody substrates are therefore subject to species turnover as substrates decompose ( Botting and DeLong 2009), and individual species presence as well as species assemblies can be studied as a function of progressive decomposition to determine saproxylic specialists.
The decomposition of the tree bark is of specific interest for studying bryophyte ecology. As nonvascular plants without a root system, bryophytes are sensitive to desiccation (Barkman 1958, Söderström 1987), and the capacity of bark to retain water is beneficial to epiphytic bryophytes (Hauck et al. 2000). Therefore, as bark degrades and becomes less present on further decomposed substrates, we can expect to see changes in species composition from species associated with living trees covered in bark to more saproxylic specialists colonizing bare wood (Jansova & Soldan 2006, but see also Botting and DeLong 2009)
We used diameter as the measure of substrate size in our current models, but it is possible to use substrate surface area or volume instead. Since epiphytic bryophytes colonize the habitable surface of a substrate, it is possible that surface area is the most telling indicator. This consideration may give a more nuanced understanding on the effects of branches, for example: if surface area is an important indicator of species presence or richness, then both diameter and branch length, which can vary, must be taken into account.
Another variable to consider is the tree species of the woody substrate. As well as having different bark texture and chemistry (Barkman 1958, Hauck and Javkhlan 2009), the different evolutionary histories of tree species allows for unique associations between epiphytic bryophytes and host trees, which can be explored for both living trees and dead woody substrates.
Table of contents :
Introduction
GNB
Materials and Methods
Study area:
Sampling design
Data analyses
Model selection
Magnitude of variable response
Results
Responses of species presence and species richness to woody substrate type
Responses of species presence to woody substrate type and mean substrate diameter
Discussion
Prospective analyses: next steps
Mean diameter effects on species richness
Substrate effects on species richness and species presence
Model comparisons for species presence
Substrate effects on species richness
Model comparisons for species richness
Bibliography
Annex 1 : GNB protocol of bryophyte surveys
Annex 2 : Occurrence of each epiphytic bryophyte species inventoried in the GNB survey
