Results and Discussion

Forest Remnants Moderate Temperatures

We observed lower temperatures in remnant interiors than adjacent disturbances, and both remnant types buffered temperatures to similar extent (Fig 11, Table 2). We can also see strong edge effects around 0 m, where temperatures decrease rapidly  moving into the forest. 

When comparing remnants to reference forest, we found that while remnants had higher soil temperatures compared to reference forest at the edge, at 15m into the forest those differences disappeared (Fig 12, Table 2). The overall difference between reference and remnant sites may be due to the 100m plot in the rference, which tended to be cool (Fig 12). Combined with higher edge temperatures in remnants, the overall tempeartures in reference sites may be cooler, according to the model. Temperatures at 45m inside the forest were similar to temperatures well into the forest interior. These results corroborate the observations of other studies, which have found relatively shallow penetration of edge effects in boreal forests (Jansson 2009 ). 

Fig 11. Effect of edge distance on relative temperatures in harvest and fire remnants and adjacent disturbance. Temperature represents difference from reference temperature. Edge distance refers to distance from the nearest forest edge, with negative values in the disturbed area adjacent to the remnant and positive values within the remnant. 

Fig 12. Effect of edge distance on temperature in remnants and reference forest. Temperature represents the daily average from June-August/ Edge distance referes to distance from the nearest forest edge.

Table 1. Effects of edge distance, disturbance type (fire vs. harvest) and their interaction on air and soil temperatures.

Table 2. Effects of edge distance, forest type (remnant vs. reference) and their interaction on air and soil temperatures.

Effects of Forest Structure on Temperature in Fire vs. Harvest Remnants

Forest structure changed with edge distance, but varied little between the two remnant types (Fig 13). We can see that deadwood volume (DWVol), graminoid cover (Gram) and forb cover (Forb) are associated with the disturbance plots while moss cover (Moss) and canopy cover (CanopyClose) are associated with interior plots. Increasing temperatures are associated with disturbance plots. 

Fig 13. Indirect gradient analysis showing correlations between forest structure elements (gray vectors) and temperature response variables (black) in fire and harvest remnants and the adjacent disturbance. Ellipses represent 80% of the observations found in each combination of plot location and distubance type.

Air Temperatures

Table 3. Results of mixed linear models investigating effects of forest structure on relative air temperatures.

Minimum air temperatures, which typically occur at night, depend on the insulating capacity of the forest to hold heat obtained during the day and the extent to which the ground radiates heat back as air temperatures cool  (Weil and Brady, 2017). The best fit model showed a significant correlation of minimum air temperature with litter cover (Table 3). High litter cover is correlated with a lack of other vegetation and thin organic layers (Fig 13), which facilitates heat transfer from ground to air during the night (Weil and Brady, 2017).  This also suggests that in our study, closed canopy forests did not retain heat any more heat during the night. 

Maximum air temperatures usually occur during midday, when solar radiation is high. Tall vegetation intercepts incoming radiation which limits heating of the air below. We found significant correlations between maximum air temperature, canopy cover and cover of tall shrubs and mosses (Table 3). Moss cover may be a response to temperature but could also absorb solar radiation that does reach the forest floor which may otherwise be reflected back into the forest.

Soil Temperatures

Table 4. Results of mixed linear models investigating effects of forest structure on relative soil temperatures. 

Daily maximum and minimum soil temperatures were significantly correlated with canopy closure and litter cover (Table 4). Increasing canopy cover reduces the amount of solar radiation reaching the soil during the day. High litter reflects a lack of low vegetation, which allows radiation passed though the canopy to heat the soil directly. 

Effects of Forest Structure on Temperature in Remnant vs. Reference Forest

Forest structure varied slightly with edge distance, but varied little between the two forest types (Fig 14). Increasing deadwood volume (DWVol), forb cover (Forb) and short shrub cover (SS) are associated more closely with edge plots while moss cover (Moss) and canopy cover (CanopyClose) are associated with interior plots.

Fig 14. Indirect gradient analysis showing correlations between forest structure elements (gray vectors) and temperature response variables (black) in remnant and reference forest. Ellipses represent 80% of the observations found in each combination of plot location and forest type.

Air Temperatures

Table 5. Results of mixed linear models investigating effects of forest structure on relative air temperatures.

Models for maximum and minimum air temperature in reference and remnant forest indicate similar mechanisms to what was seen in the remnants and disturbance plots.  Canopy cover remained the most significant influence on maximum air temperatures (Table 5),  leading to higher temperatures in edge plots than interior plots  (Fig 14).

Soil Temperatures

Table 6. Results of mixed linear models investigating effects of forest structure on relative soil temperatures. 

In the reference-remnant comparison, since canopy cover and litter was more consistent across the sample, moss was most significant factor influencing soil temperatures (Table 6). High moss cover has the opposite effect of litter, preventing solar radiation which reaches the forest floor from heating the soil, and insulating the soil from high air temperatures. 

Vegetation Community Responds to Microclimate

Microclimate and Plant Communities Respond to Edge Effects

Results of perMANOVA show the influence of microclimate and forest structure on plant community composition (Table 7). Maximum air temperature was the most important predictor variable, accounting for about 7% of the variation observed in plant communities. Minimum soil temperature was second, accounting for 5.4% followed by maximum soil temperature which accounted for 3.4% of the variation (Table 7). Minimum air temperature had no significant effect. 

Of the forest structure, variables, edge distance (EdgeDist), spruce basal area (SpruceBA), deadwood volume (DWVol) and pine basal area (PineBA) were significant, each accounting for about 2% of the observed variation (Table 7). Disturbance type (DistType), organic layer depth (OrgLayer), canopy closure (CanopyClose) were non-significant, but only just. When analyzing our full spectrum of vegetation community data (without microclimate) the increased replication may allow us to determine the effects (if any) of those elements.  

Table 7. Results of perMANOVA on Hellinger-transformed plant community data using the Bray-Curtis distance. 

Results of a divisive cluster analysis show that the three clusters vary with microclimate with cluster 3 associated with high temperatures and cluster two with lower temperatures. Cluster two is also associated with increasing edge distance (forest interiors), deep organic layers and greater spruce basal area. Clusters 2 and 3 are significantly different from one another, with cluster 1 being intermediate between the two (Fig 15, Table 7). 

Fig 15. Results of an MDS of Hellinger-transformed plant community data using the Bray-Curtis distance. Ellipses represent 95% confidence intervals of clusters determined via divisive cluster analysis (kmeans). 

Fig 16: Pie charts showing the average composition of cover classes and species in each cluster, as well as the distribution of plot locations (disturbance, edge, interior) and forest types (disturbance, island, reference) in each.  See table 8 for key to cover classes.

Clusters are influenced by the relative proportion of moss and litter, which were the two most abundant cover classes in our plots. Labrador Tea, Dwarf Bilberry, Velvetleaf Blueberry and  Bunchberry are also present in all three clusters. Clusters 1 and 3 had higher cover of Fireweed and Hairy Wildrye, which are both well-adapted to disturbance (Fig 16).

Clusters also appear to be influenced by plot location, with cluster 3 having the highest proportion of disturbance plots, and cluster 2 containing the highest proportion of interior plots while cluster 1 is more strongly associated with edge plots (Fig 16).

Clusters one and three each contain an nearly equal portion of reference and island plots, while cluster 2 is comprised entirely of reference and island plots. These results show that plant community composition can be similar in all forest types, and it is the proportion of disturbance sites that drives the differences in the clusters, though according to a perMANOVA, forest type does explain a small but significant portion of the variation in plant communities (Table 9). This may be due to the fact that our sample is skewed toward harvest sites, where the boundaries of cutblocks and remnants are based on regulations and value of timber. Working with our full data set in a future analysis, we expect to be able to better determine difference in plant community compositions in remnant and reference forest differs based on disturbance type.

Table 9: Results of perMANOVA comparing Hellinger-transformed vegetation community data in remnants and reference forest.

Table 8: Key to the cover classes referenced in Fig 16. 

Conclusion

Our analysis shows how differences in forest structure, particularly canopy cover and ground cover, influence soil and air temperatures, which in turn are significantly correlated with the composition of vegetation communities. 

We found no differences in microclimate buffering capacity between fire skips and harvests remnants. Remnants, however, tended to have higher temperatures at edges when compared to reference forest, which may explain the small difference in plant community composition between the types revealed by perMANOVA. This difference could also be the result of management decisions related to timber harvesting. 

Our study only considered microclimate during the hottest months of the year, but it is possible winter microclimates also influence the plant community composition. In any comparison of fire and harvest sites, we must acknowledge that while fire is a somewhat random process, timber harvests are carefully planned to maximize value within regulatory constraints and other management objectives. This results in harvest unites whose boundaries may already correspond with variations in topography and forest communities. Our analysis of forest structure on microclimate was constrained by our study design, which was not initially intended to answer this question. Other studies show that edge aspect is a significant driver of microclimate, but we did not have sufficient replication on all aspects to detect this effect. Additionally, due to the outlier plant communities of our EC site, we did not have a lot of statistical power to investigate interactions of microclimate and disturbance type on plant community.

We found evidence to support our hypothesis that forest remnants maintain similar microclimates and plant communities as larger forest tracts. This supports the notion that these remnants could be important refugia for interior-adapted species and potential sources for seed and animal dispersal as the adjacent disturbance recovers. With climate change bringing increasing temperatures and more frequent drought, forest remnants could serve vital functions on post-disturbance landscape. Forest remnants should continue to be studied with respect to other taxa and ecosystems and given greater consideration in a range of forest management applications.