Identify the Response of Forest Songbirds to Local Scale Forest Features and Complexity

Open Access
Author:
Barenblitt, Abigail Cohen
Graduate Program:
Wildlife and Fisheries Science
Degree:
Master of Science
Document Type:
Master Thesis
Date of Defense:
December 13, 2016
Committee Members:
  • Margaret Brittingham-Brant, Thesis Advisor
  • David Andrew Miller, Committee Member
  • Matthew Marshall, Committee Member
Keywords:
  • Forest bird communities
  • Forest bird species richness
  • Occupancy Models
  • Local Forest Features
  • Forest Complexity
Abstract:
I conducted a study to examine how local scale forest features and local forest complexity influence forest bird occupancy. In the first chapter, I utilized multi-species occupancy models to determine if I could identify local scale forest features associated with increased occupancy of groups of co-occurring species of forest birds. Over the 2015 and 2016 breeding seasons, I conducted 310 avian point counts within 77 sites across forested regions within the Allegheny Plateau and Ridge and Valley ecoregions in Pennsylvania. I returned to each of these point count locations to collect vegetation data after the breeding season in 2015 and 2016. For my analysis, I first used a general model that tested the same set of six covariates across six groups of co-occurring species: species associated with the canopy, midstory, understory and ground strata as well as conifers and dead wood. This model included the abundance of canopy, midstory, understory, and ground cover, as well as proportion of conifers in the canopy and abundance of snags. I then created group specific models for each of the six groups. For the canopy group, I created a model that included three covariates: canopy cover, the abundance of large trees, and abundance of understory cover. For the midstory group, I created a model that included three covariates: the abundance of midstory cover, abundance of small trees, and standard deviation of DBH. For the understory group, I created a model that included two covariates: the abundance of tall ericaceous shrubs and the abundance of ground cover. For the ground group, I created a model that included three covariates: the abundance of ground cover, abundance of hayscented fern, and litter depth. For the conifer group, I created a model that included two covariates: the proportion of hemlock and proportion of other conifers in the canopy. Finally, for the dead wood group, I created a model that included two covariates: the abundance of snags and abundance of woody debris. I found that different local scale forest features were associated with the increased or decreased occupancy of each group. From the results of each occupancy model, I was able to determine a posterior mean effect of each covariate. A positive posterior mean effect demonstrates an increase in predicted occupancy given a certain covariate while a negative posterior mean effect demonstrates a decrease in predicted occupancy. Each posterior mean effect size included 95% credible intervals, which can be used to evaluate strength of association. Intervals that do not overlap 0 or slightly overlap 0 indicate more significant results than credible intervals that largely overlap 0. On average, the predicted occupancy of understory associates was 0.29. Of the covariates I measured, the abundance of tall ericaceous shrubs had the strongest correlation with the predicted occupancy of understory associates and had a positive mean posterior effect, or β, of 0.59 with the 95% credible interval slightly overlapping 0 (95% CrI=-0.04 to 1.2). The predicted occupancy of five out of six understory associates, Common Yellowthroat (Geothlypis trichas), Hooded Warbler (Setophaga citrina), Chestnut-sided Warbler (Vireolanius melitophrys), Black-throated Blue Warbler (Setophaga caerulescens), Canada Warbler (Cardellina canadensis), and Eastern Towhee (Pipilo erythrophthalmus), increased with increasing abundance of tall ericaceous shrubs. However, the predicted occupancy of Indigo Bunting (Passerina cyanea) decreased (β=-0.67, 95% CrI=-1.37 to 0.00). Overall, the posterior mean effect of tall ericaceous shrubs was more positive than the posterior mean effect of the other covariate included in the understory associate model, abundance of ground cover (β=0.24, 95% CrI=-0.08 to 0.57). Our results indicate that understory associates utilize areas with increased abundance of tall ericaceous shrubs, such as mountain laurel. These shrubs occur on acidic soils that are considered poor quality and can suppress tree regeneration. Therefore, these shrubs are often considered detrimental to forest quality. However tall ericaceous are currently a dominant understory feature within Pennsylvania. The positive relationship we identified with this study indicate that despite management concerns, tall ericaceous shrubs may provide habitat for certain bird species. Therefore, the shrubs may provide an important structural feature for understory associates even in areas with poor quality soils. The predicted occupancy of midstory associates was 0.20 on average, slightly lower than that of understory associates. Of the covariates I measured, the standard deviation of DBH had the strongest correlation with the predicted occupancy of midstory associates and had a positive mean posterior effect, or β, of 0.16 with the 95% credible interval slightly overlapping 0 (95% CrI = -0.09 to 0.43). The posterior mean effect was positive for all five species included in this group and was highest for Acadian Flycatcher (Empidonax virescens) (β=0.23, 95% CrI=-0.07 to 0.54). The posterior mean effects for the other covariates included in the midstory associate model, midstory cover and small tree abundance, were both negative and had credible intervals that overlapped 0 (β=-0.15, 95% CrI =-0.46 to 0.14; β=-0.15, 95% CrI =-0.50 to 0.23). Consequently, our results demonstrate a positive correlation between the standard deviation of DBH and the predicted occupancy of midstory associates. Management techniques such as uneven-aged management may prove useful in creating important structural features, such as a variety of tree sizes. Monitoring features like standard deviation of DBH following management may allow managers to determine how successful certain prescriptions are for creating quality habitat. Additionally, identifying areas with high variation of tree sizes may allow managers to identify areas that have higher quality habitat for midstory associates. Canopy associates were the most abundant group of species with a predicted occupancy of 0.42. Of the covariates I measured, the abundance of understory cover had the strongest correlation with the predicted occupancy of canopy associates and had a negative mean posterior effect, or β, of -0.24 (95% CrI=-0.49 to -0.01). The posterior mean effect of understory cover was negative for all six species included in this group and was most negative for Rose-breasted Grosbeak (Pheucticus ludovicianus) (β=-0.33, 95% CrI =-0.67 to -0.02). While the abundance of understory itself likely does not have a direct negative impact on canopy associates, more abundant of understory cover likely indicates larger canopy openings and reduced availability of canopy habitat. The posterior mean effects of the other covariates included in the canopy associate model, canopy cover and abundance of large trees, were considered negligible because they did not demonstrate as strong of a negative effect as understory cover and had credible intervals that largely overlapped 0 (β=-0.01, 95% CrI =-0.31 to 0.31; β=-0.08, 95% CrI =-0.38 to 0.23). Ground associates were the second-most abundant group with a predicted occupancy of ground associates was 0.35. However, the local scale forest features I included in my ground associate model had minimal effect on birds within this group. The posterior mean effects of the covariates included in the ground associate model, abundance of ground vegetation cover (β=-0.08, 95% CrI =-0.54 to 0.36), litter depth (β=-0.05, 95% CrI =-0.30 to 0.20), and the abundance of hayscented fern (β=0.00, 95% CrI =-0.25 to 0.25), were all considered negligible with small effect sizes and credible intervals that largely overlapped 0. Therefore, it is difficult to draw conclusions about the direct impact of local forest features on the occupancy of ground associates. Conifer associates were our rarest group with a predicted occupancy of conifer associates was 0.16. Of the covariates I measured, the proportion of hemlock in the canopy had the strongest correlation with the predicted occupancy of conifer associates and had a positive mean posterior effect, or β, of 0.34 (95% CrI=0.09 to 0.59). The posterior mean effect of hemlock was positive for 10 out of the 11 species we included in this model and was highest for Blue-headed Vireo (Vireo solitaries) (β=0.62, 95% CrI =0.29 to 0.97). The posterior mean effect of hemlock was slightly negative for Yellow-rumped Warbler (Setophaga coronata), but considered negligible as the effect size was small and the credible intervals largely overlapped 0 (β=-0.02, 95% CrI =-0.66 to 0.54). The other covariate included in the conifer associate model, proportion of other conifers in the canopy, had a positive posterior mean effect (β=0.13, 95% CrI =-0.10 to 0.36), however the credible intervals overlapped 0 and this posterior mean effect was not as strong as the effect of hemlock in the canopy. The posterior mean effect of other conifers in the canopy was highest for Blue-headed Vireo (β=0.42, 95% CrI =0.10 to 0.77). The positive correlation between the abundance of hemlock in the canopy and the predicted occupancy of conifer associates indicates another forest feature that managers should monitor. The abundance of hemlock in the canopy may decrease as pests like hemlock woolly adelgid damage hemlock stands, while the abundance of woody debris may increase with hemlock die-off. My results indicating a positive effect of hemlock on conifer associates provide further evidence that hemlock woolly adelgid will likely negatively affect conifer associates as forest composition changes. In addition to protecting extant stand of hemlock, managers should examine other forest features that may provide similar habitat for conifer associates. While the positive correlation we identified between the predicted occupancy of conifer associates and other conifers was not as strong as the relationship with hemlock, these findings indicate that certain species within this group may be able to utilize other coniferous species as hemlocks decline. Hemlock is considered valuable to wildlife because it retains its lower branches, creating complex vertical structures from close to the base of each tree to the crown. Therefore, it may also be worthwhile for future research to explore how conifer associates utilize other structural features within the forest that could be managed as hemlock continues to decline. On average, the predicted occupancy of dead wood associates was 0.17. Of the covariates I measured, the abundance of woody debris had the strongest correlation with the predicted occupancy of dead wood associates and had a positive mean posterior effect size, or β, of 0.18, although credible intervals slightly overlapped 0 (95% CrI=-0.02 to 0.39). The mean posterior effect of woody debris was highest for Black-and-white Warbler (β=0.13, 95% CrI =0.02 to 0.54). The mean posterior effect of the other covariate included in the dead wood associate model, the abundance of snags, was slightly positive (β=0.09, 95% CrI =-0.18 to 0.38), but had credible intervals that overlapped 0 and was not as high as the mean posterior effect of woody debris. The abundance of woody debris was a stronger predictor of increased occupancy of dead wood associates than snags. Our results suggest that the abundance of woody debris is a better indicator of the availability of dead wood features than the abundance of snags. Within the forest there are many trees that are dying from disease, yet are not considered snags. These trees may create the same nesting and foraging materials as snags. Collectively, woody debris may be the best indicator of habitat features used by dead wood associates given these condition. Dead wood associates may also benefit in the short term as dying hemlock creates more dead wood material. However, these species may decline in the long term as diseased trees are replaced and no longer contributes to the amount of coarse woody debris within a forest. Tracking these local scale features may allow managers to determine the current suitability of forest habit for forest bird communities and make more effective management decisions to conserve quality habitat as these processes continue to shape forested habitat. These results indicate that local scale forest features may be useful in predicting increased occupancy of groups of co-occurring species. Of the covariates we included in our models, the abundance of tall ericaceous shrubs, hemlock in the canopy, and woody debris, as well as the standard deviation of DBH appeared to positively correlate with the predicted occupancy of understory, conifer, dead wood, and midstory associates respectively. The increased abundance of these forest features may therefore be indicators of higher quality habitat for these groups of habitat associates. These features may be monitored to determine the current quality of forested habitat for certain groups of interest, as well as the efficacy of management prescription. Each of these features are already monitored through surveys such as the Continuous Forest Inventory. Additionally, managers may benefit from utilizing forest bird monitoring programs and point count data to identify areas where these features may already be present by examining which forested areas are occupied by the groups of co-occuring species. Moving beyond a focus on different groups within the forest bird community, in the second chapter I used multi-species occupancy models to examine how local forest complexity affects overall species richness of forest specialists. I used the same avian point count and vegetation data from the first chapter to examine if forest bird species richness increased as local compositional and structural complexity increase. I created a model that included two measures of structural complexity as well as four measures of compositional complexity: the standard deviation of DBH, the volume of vegetation below the canopy, tree species richness, shrub species richness, the number of ground cover types, and the number of forest cover types within 500 m. I also examined the nestedness of our points to examine whether species depauperate (i.e., not species rich) sites were subsets of species rich sites. Our results indicate that structural complexity had a positive effect on overall species richness. The standard deviation of DBH had the strongest positive correlation with bird species richness compared to other measurements of structural complexity and across 47 species had an average mean posterior effect size of 0.09 (95% CrI=0 to 0.18). The standard deviation of DBH had a positive correlation with the occupancy of 44 of the species we examined. The standard deviation of DBH had the strongest positive correlation with the occupancy of the midstory (β=0.16, 95% CrI = -0.10, 0.41), conifer (β=0.18, 95% CrI = -0.01, 0.38), and dead wood (β=0.15, 95% CrI = -0.07, 0.36) associated groups, although credible intervals overlapped 0. These results indicate that uneven-aged management or other management strategies that can create more variation of tree DBH may be a useful tool for maintaining forest bird species diversity. Managers monitoring the success of management prescriptions may also benefit from tracking the standard deviation of DBH as a metric. Prescriptions that results in a higher standard deviation of DBH will likely be more effective for creating stands that support higher bird species richness. While the volume of vegetation below the canopy did not have as strong of a positive relationship with bird species richness as standard deviation of DBH, this features also hada positive correlation with bird species richness and had an average mean posterior effect of 0.06, although credible intervals overlapped 0 (95% CrI= -0.06 to 0.19). This feature had a positive correlation with the occupancy of about half of the species we examined (n=23) and had the strongest positive correlation with the occupancy of understory associates (β=0.35, 95% CrI = 0.00, 0.65). Certain measures of compositional complexity also had a positive correlation with bird species richness. The number of cover types within a 500 m buffer around each point had the strongest positive effect on species richness compared to other measurement of compositional complexity. The number of cover types within a 500 m buffer had an average mean posterior effect size of 0.10, although credible intervals slightly overlapped 0 (95% CrI=-0.01 to 0.21). The number of cover types had the strongest positive correlation with the occupancy of the understory (β= 0.14, 95% CrI = -0.11), ground (β=0.16, 95% CrI = -0.19, 0.49), and conifer (β=0.20, 95% CrI = -0.05, 0.45) groups and had a positive correlation with the occupancy of 45 species. This positive correlation between the number of forest cover types within a 500 m buffer and bird species richness indicates that spatial heterogeneity is also important for maintaining species diversity. The results of our other measurements of compositional complexity provide more evidence supporting the importance of spatial heterogeneity as well. Measurements of local compositional complexity, including tree species and shrub species richness and number of ground cover types did not appear to have a strong positive effect on species richness. Across 47 bird species, tree species richness had an average mean posterior effect of 0.04 with credible intervals largely overlapping 0 (95% CrI=-0.11 to 0.20), and shrub species richness had an average mean posterior effect of -0.04 with credible intervals also overlapping 0 (95% CrI=-0.19 to 0.10). Tree species richness had the strongest positive correlation with the occupancy of midstory-associated species (β=0.39, 95% CrI = -0.13, 0.94) and the strongest negative correlation with the occupancy ground-associated species (β=-0.12, 95% CrI = -0.43, 0.17), although credible intervals overlap 0. Shrub species richness had the strongest positive correlation with the occupancy of the midstory and understory groups (β=0.16, 95% CrI = -0.13, 0.46; β=0.23, 95% CrI = -0.06, 0.50) and the strongest negative correlation with the occupancy of the canopy, conifer, and dead wood groups (β=-0.26, 95% CrI = -0.67, 0.21; β=-0.18, 95% CrI = -0.46, 0.11; β=-0.17, 95% CrI = -0.55, 0.19), although credible intervals overlap 0. The number of ground cover types had an average mean posterior effect of -0.05 with credible intervals slightly overlapping 0 (95% CrI=-0.15 to 0.05), however this covariate appeared to have a negligible effect on the occupancy of all groups of co-occurring species. For each of these compositional covariates, the credible intervals largely overlapped 0 and did not appear to affect overall species richness positively or negatively. These results indicate that changes in compositional complexity may lead to changes in forest bird community composition, but not overall richness. The varying effect of measures of compositional complexity on the predicted occupancy of the six groups of interest, as well as our results from the first chapter linking species groups to local forest features, indicate that microhabitats within the forest are important to specific groups of co-occuring species. Therefore, spatial heterogeneity would increase the availability of a variety of these microhabitat features across a landscape and allow managers to maintain biodiversity across forests. Our nestedness results also indicate that managing certain for features that increase species richness, such as the standard deviation of DBH and spatial heterogeneity, stands to benefit the forest bird community as a whole. Of the points we surveyed, species rich and species depauperate points were nested based on the expected nesting from a random distribution (p<0.001). 48.23% of our points demonstrated an overlap pattern consistent with nestedness, whereas 25.08% of cells in a random matrix demonstrated overlap consistent with a nested pattern. Therefore, the overlap of species within our points deviated from what would be expected in a random, non-nested matrix (p<0.001). These results indicate that managing for species richness of forest specialists meets the goal of keeping common species common, while also creating habitat for rare species of interest. This demonstrates that managing for forest bird species richness may be a useful strategy for managing habitat for the forest bird community as a whole. Therefore, using management tools that create higher variability of tree sizes and also encourage spatial heterogeneity across a forested landscape would be useful for balancing the needs of both rare and common forest bird species on interest. The results of this project indicate that a number of forest features may be useful for monitoring the current quality of forested habitat, determining the efficacy of management techniques, and creating areas of higher diversity. Each of the forest features that demonstrated a positive correlation with the predicted occupancy of groups of co-occurring species are already monitored through surveys such as the Continuous Forest Inventory and are highly accessible to forest managers. The results of Chapter 1 demonstrating that measures of ericaceous shrubs, hemlock abundance, standard deviation of DBH, and woody debris abundance predict increased occupancy for understory, conifer, midstory, and dead wood associates respectively indicate the importance of microhabitat features for certain groups of co-occurring species. Chapter 2 demonstrated that the standard deviation of DBH and the number of cover types within 500 m positively correlated with predicted bird species richness. This demonstrates that local structural complexity as well as spatial heterogeneity may be useful to indicate area of forest that can support higher biodiversity. Across a forested landscape, managers may be able to maintain robust and diverse forest bird communities by maintaining a variety of microhabitats and forest cover types across a landscape as well as managing for higher variation of tree sizes within forest stands.