Effects of spatial patterning within working pine forests on priority avian species in Mississippi

Within dynamic ecosystems, research into how land use changes and patterns affect species diversity has led to a suite of ecological hypotheses to assess species-landscape associations. The Habitat Amount Hypothesis suggests that it is the total amount of habitat, regardless of configuration, whereas the Multi-dimensional Hypothesis suggests it is the suite of local, landscape, and landform characteristics that have the greatest influence on species diversity within its local landscape. Our study aims to understand how landscape pattern influences species abundance, in the context of these two competing hypotheses on priority avian species in working forest landscapes of the southeastern United States. To examine these hypotheses, we conducted bird point counts and vegetation surveys in short-rotation loblolly pine (Pinus taeda) forests in east-central Mississippi during 2019–2020 and used abundance and richness models to assess avian species associations to amount vs. configuration of habitat in a 2 km2 landscape. We found that habitat amount alone did not exhibit consistent positive associations with species abundance for both early-successional and mature forest associated avian communities. Most target species exhibited positive associations with patch proximity, measured by Euclidean distance, and proximity-area index. However, measures of species richness showed no association with amount or proximity. Associations with landform features generally had positive influences on early-successional species than mature-pine priority species. The dynamic mosaic of forest stand ages may be sufficient in providing habitat needs such that measures of richness are not driven by amount and proximity at the 2 km2 scale in this working landscape. However, influences of proximity and landform on priority species abundance warrant further research to assess potential drivers of associations with stand proximity and effects of amount and proximity on measures of species diversity across scales. Given the growing demand for forest products, sustainable forestry guidelines that consider proximity of stands in similar age classifications could enhance landscape suitability for some target species.

Objectives Our study aims to understand how landscape pattern influences species abundance, in the context of these two competing hypotheses on priority avian species in working forest landscapes of the southeastern United States. Methods To examine these hypotheses, we conducted bird point counts and vegetation surveys in short-rotation loblolly pine (Pinus taeda) forests in east-central Mississippi during 2019-2020 and used abundance and richness models to assess avian species associations to amount vs. configuration of habitat in a 2 km 2 landscape. Results We found that habitat amount alone did not exhibit consistent positive associations with species abundance for both early-successional and mature forest associated avian communities. Most target species exhibited positive associations with patch proximity, measured by Euclidean distance, and proximity-area index. However, measures of species richness showed no association with amount or proximity. Associations with landform features generally had positive influences on early-successional species than maturepine priority species. Conclusions The dynamic mosaic of forest stand ages may be sufficient in providing habitat needs such that measures of richness are not driven by amount and proximity at the 2 km 2 scale in this working landscape. However, influences of proximity and landform on priority species abundance warrant further research to assess potential drivers of associations with stand proximity and effects of amount and

Introduction
Concepts relating patch size and isolation as direct determinants of species diversity became a leading premise in ecology since MacArthur and Wilson (1967) presented their theory of island biogeography. Building on their legacy, numerous theories and hypotheses have been suggested to characterize effects of ecosystem fragmentation on landscapes (Wiens 1976;Jones et al. 2009;Haddad et al. 2017), many focusing on patch-scale dynamics within forested systems. Fahrig (2013) proposed the Habitat Amount Hypothesis (HAH), which suggests the total amount of habitat (defined therein as the particular cover types used by a given species or species group) within its local landscape has the greatest influence on species diversity. Under the HAH, configuration of habitat patches in a local landscape is suggested to have less influence than total habitat amount. This implies that in many studies, fragmentation effects are blurred by sample area effects when comparing biodiversity in large continuous patches vs. biodiversity in small-fragmented patches (Fahrig 2013). Since it was first proposed, several studies showed support for the HAH (e.g., Melo et al. 2017;Rabelo et al. 2017;Seibold et al. 2017). However, its validity has been questioned by others (e.g., Evju and Sverdrup-Thygeson 2016;Haddad et al. 2017;Lindgren and Cousins 2017). Many critics of HAH suggest that only considering one attribute of landscapes (e.g., habitat amount) and only one aspect of species diversity (e.g., species richness) is too simplistic to fully determine effects of fragmentation on biodiversity (e.g., Haddad et al. 2017). However, Fahrig (2013) claimed that this simplification of fragmented ecological systems is needed in the context of pressing conservation challenges. Years prior another hypothesis that sought to explain how species may be affected by landscape scale factors was Mitchell et al. (2006) Multi-dimensional Hypothesis (MDH). In this hypothesis neither amount nor configuration have the greatest effect on species richness. Instead, a multi-dimensionality of landform, landscape, and stand level factors are posited to affect species richness, and the influence of these factors varies among species guilds. The relevance of this hypothesis may be of particular importance as it is tested within managed pine forest of the Southeastern U.S. region and provides tangible connections to forest management practices and sustainability certification standards.
Understanding the extent of how landscape or landform features may affect species richness and abundance has been a major focus of ecological and biogeographical research, given its central importance for conservation planning and landscape management. Working pine (Pinus spp.) forests in the U.S. Southern Coastal Plain region (SCP) are generally comprised of a shifting mosaic of forest management practices promoting a patchwork of forest stand sizes, ages, and adjacencies (Miller 2009;Demarais et al. 2017). This patchwork allows assessing biodiversity outcomes in local landscapes with varying amount and configuration of forest patches. Many studies examined effects of landscape composition and configuration within mature and early-successional pine forest stands (Askins 2001;Berglind 2004;Demarais et al. 2017;Greene et al. 2019). However, few have examined effects of composition and configuration of forest patches providing early-successional conditions on biodiversity in the context of these ecological hypotheses, particularly for avian communities (Torrenta and Villard 2017;De Camargo et al. 2018). This is of high practical importance in working landscapes, as it determines whether conservation strategies should focus only on total habitat amount as part of the shifting forest mosaic or include its spatial configuration (Lindenmayer and Fischer 2007).
It is estimated that 89% of working pine forests in the southeastern United States are privately owned (Oswalt et al. 2019). The southeastern U.S. experienced steep declines in avian populations over the last 50 years, particularly in species adapted to disturbance-mediated vegetation types (e.g., grasslands and shrublands ;Askins 2001;Sauer et al. 2017). The dynamic mosaic created by a working pine system can provide conditions favorable for several avian species guilds, particularly early-successional vegetative structure created by forest management regimes which produce ephemeral vegetative conditions for both facultative grassland and scrub-successional species at different stages in a typical forest rotation (Miller et al. 2009;Evans et al. 2021). However, research into the landscape-level effects of pine forest management on avian communities is limited (e.g., Mitchell et al. 2001Mitchell et al. , 2006Tappe et al. 2004;Loehle et al. 2005).
To better understand how landscape pattern influences avian communities we examined priority bird species response to early-successional and mature pine habitat amount and patch configuration in two privately-owned working pine forest sites managed primarily for timber production in east-central Mississippi in the context of these two hypotheses. We predicted that while the amount of habitat components within a species' local landscape will be foundational, the configuration of forest stands will be influential to avian species richness and abundance due to variation in dispersal capabilities and other factors in these dynamic systems that drive patch configuration. Specifically, species with low vagility may be more sensitive to patch proximity. Additionally, those species that rely on certain habitat components may be influenced more by local level characteristics (i.e., vegetation structure characteristics within the patch). Furthermore, we predicted that differences in landform between the two study sites would have differing effects on bird communities.

Study area
Our study sites were located within two geographically distinct areas comprised predominantly of privately-owned working pine forests within the SCP in east-central Mississippi, USA. Our sites included hilly pine woodlands of Webster, Calhoun, and Chickasaw Counties (lat. 33.689069, long. -89.193069) and pine flatwoods of Kemper and Noxubee County (lat. 32.841059, long. -88.53836; Fig. 1). Study sites included mid-late rotation post-thinned loblolly pine (P. taeda) ≥ 15 years' post-establishment (hereinafter, mature) and early-successional (recently clearcut and young planted pine; 0-3 years' post-harvest/planting) forest stands owned and managed by Weyerhaeuser Company. Sites were characterized by mostly contiguous blocks of working pine forests and represented approximately 116 km 2 and 76 km 2 of loblolly pine stands in Kemper/Noxubee and Webster/Chickasaw/ Calhoun Counties, respectively. Landscapes consisted of 85% working loblolly pine forest stands of various ages; 12% mature pine-hardwood or hardwood stands, primarily along streams, and 3% non-forested areas (right of ways, food plots, and log decks), on average.

Landscape sampling and landcover variables
Within our study sites, we used a multi-stage stratified random sampling design following recommendations for assessing habitat amount and proximity outlined by Fahrig (2013). We included loblolly pine stands classified to early-successional or post-thinned mature stands that met the criteria for total amount and proximity. We then assigned each stand within the sampling frame a centroid point from which radial landscapes were assessed. To best represent the landscape and species dispersal capabilities, we tested landscape characteristics at various scales (0.5, 1, 1.5, and 2 km) and chose a 2 km radial buffer which best quantified the surrounding landscape and coincided with our focal species with the largest home range, American Kestrel (Falco sparverius).
Once the land cover classification was completed within each 2 km radial buffer, we used FRAGSTATS (McGarigal et al. 2012) and the Landscapemetrics package (Hesselbarth et al. 2019) in R (R Development Core Team 2016) to calculate landscape metrics and separate each landscape into one of four strata, two-stage (e.g., high amount-high proximity; high amount-low proximity; low amount-high proximity; low amount-low proximity) as described by Fahrig (2013). Within each 2 km radial buffer, we calculated the total area (a proxy for habitat amount under the HAH) using total class area for both early-successional and mature forest, and the proximity of patches to its neighbors of the same class using Euclidean Nearest Neighbor (ENN). Additionally, we calculated a proximity index (PROX) which is the sum of patch area (m 2 ) divided by the nearest edge-to-edge distance squared (m 2 ) between patches to evaluate if it was a combination of both area and distance factors that influenced avian communities (McGarigal et al. 2012). We used value cut-offs to categorize stands into high to low amounts of early-successional or mature pine and high to low proximity to neighboring stands of similar age class. To build upon the HAH, we obtained landform information which included: mean elevation using 0.7 m Digital Elevation Model tiles obtained from Mississippi Automated Resource Information System (MARIS 2012). Using the slope and aspect tool in ArcGIS, we calculated mean slope and aspect within a 250 m buffer to determine influence of topographic variables on avian species at the point they were detected.

Avian and vegetation sampling
We conducted a single, 10-min variable radius point transect bird survey at each stand centroid during the breeding season (May-July) in 2019 and 2020. We performed surveys from sunrise to 10:00 am CST, after which bird activity significantly declined, and on days with no precipitation and wind speeds < 19 kph to avoid unequal detectability (Robbins 1981).
Observers recorded all unique visual and aural detections of singing males into 1-min intervals for a 10-min survey period. At that time, observers placed individuals detected into distance bands (0-25, 25-50, 50-100, 100-250, > 250 m) using the regular grid spacing of planted pine trees (1.5 × 6.1 m) as a reference.
We also sampled vegetation structure at each bird survey point to facilitate analysis of locallevel associations with bird occurrence. For both early-successional and mature pine strata, we estimated percent ground cover (vine, dead down wood, herbaceous, forb, woody, dead grass, live grass, litter, bare ground) at each bird survey point using a 1 m 2 Daubenmire frame for ground nesting species. We recorded ocular estimates of percent shrub/understory cover in 0-1 m height categories within a 3.6 m radius of the center point (12.96 m 2 ) of each stand to assess associations for shrub nesting species based on previous vegetation sampling protocols for open pine systems (Nordman et al. 2016). We also measured horizontal cover/ visual obstruction in 10 cm increments using a Robel pole to examine cover provided for ground foraging species (Robel 1970) and measured mean sapling height in each early-successional stand by walking a 10 m transect in each cardinal direction from the stand centroid, measuring heights (m) of all saplings within 1 m of the transect. We calculated mean overstory canopy height (m) in mature pine stands by averaging measurements of the three tallest trees within a 11.3 m radius plot (127.7 m 2 ).We estimated live tree species basal area (m 2 /ha) for pine and hardwood species using a 10-factor angle gauge and estimated percent overhead canopy cover by using a concave spherical crown densitometer (Lemmon 1956; Forestry Suppliers, Inc. Model-C). Because retained structures are important for cavity nesting species, we also recorded number of snags > 20 cm DBH within a 15.3 m radius around the stand center (Bull et al. 1990).

Statistical analysis
We estimated abundance for priority bird species and richness for birds detected in early-successional and mature pine stands separately to assess effects of habitat amount and configuration in each age group. Fig. 1 Extent of our survey points and proportion of deciduous, evergreen, and mixed forest landcover within our five study counties, Kemper and Noxubee in the south and Calhoun, Chickasaw, and Webster in the north. Zoomed in 2 km radial buffers show satellite imagery of the local landscape in which landscape metrics were calculated ◂ Of the 12-priority grassland and scrub-shrub species identified in the East Gulf Coastal Plain (Greene et al. 2021), 6 were detected on our sites and had enough detections to be adequately modeled (Table 1). Of the 13-priority open pine and mixed pine-hardwood bird species identified in the East Gulf Coastal Plain (Greene et al. 2021), 7 were detected on our sites and had enough detections to be adequately modeled. However, we considered the Red-headed Woodpecker (Melanerpes erythrocephalus) an early-successional species since we had more detections in early-successional than mature stands.

Abundance model
We assessed species-specific associations with local and landscape characteristics for priority bird species by estimating abundance of birds using the Bayesian hierarchical extension of a combined distance-sampling and time-removal model for point-count data that estimates abundance and density of birds. This method allows for estimation of detection dependent on birds' availability and perceptibility (Amundson et al. 2014). This model was developed to accommodate single-visit, point-count data replicated at points in an area of interest for imperfectly detected species. We accounted for elements of detectability by incorporating measures of distance-based perceptibility (pd) and time removal-based availability (pa). The probability of availability followed Farnsworth et al. (2002), and probability of perceptibility, was the probability of detection within each distance class is a half-normal distance function with a scale parameter representing the rate of decay as a function of distance to each point. To estimate abundance, population size was modeled for each point N k as a Poisson distribution with mean expected value , N k ~ Poisson( k ) (Royle et al. 2004). All landscape, landform, and local level covariates were tested using a Pearson's correlation test and we dropped competing covariates from the model if significantly correlated (< 0.4). We created two models-one for each age class. Within each earlysuccessional and mature model, covariates included continuous measures of landscape (TotalArea, ENN, PROX), landform (slope, elevation, and aspect), and local-level vegetation. Although included covariates were the same for both age classes, values differed between the two models as size and configuration values varied with landscape pattern between the two stand types. Local-level covariates were also unique to each age class as vegetation structure and composition characteristics were sampled differently for the two age classes. Covariates included in the abundance model were species-specific and included in models using prior and presumed knowledge on species-habitat requirements (Tirpak et al. 2009). We analyzed each year separately due to the dynamic state of the working pine system. Thus, for each species in each year, the model was characterized as, where i is the abundance for species i, and β 1…x k are the coefficients for effects of proximity, amount, slope, elevation, and local-level covariates on points k. We first normalized point-level and landscape-level data using a log transformation and back-transformed the landscape and vegetation covariates before reporting them.

Richness model
We also assessed avian species richness within earlysuccessional and mature pine avian communities using a hierarchical fixed-effects community occupancy model within a Bayesian framework to examine landscape covariate effects on species richness. The communities included early-successional and mature pine and were defined by the species detected at either early-successional or mature pine survey points. The model provides estimates of species richness for each individual and the overall community responses to the environmental variables based on these estimates (Kery and Royle, 2016). All covariates in the richness model were the same as in the abundance model except for local level covariates, in which we used the top three vegetation characteristics that had the greatest effect on each community. This included pine seedling height, shrub cover 0-1 m in height, and herbaceous ground cover for the earlysuccessional community and pine basal area, percent canopy cover, and 0-1 m shrub cover for the mature pine community.
For both the abundance and richness model we ran three chains of 100,000 iterations for each species model in JAGS (Plummer 2003) called from within R using package jagsUI (Kellner 2015). We included a 10,000-iteration adaptation phase, a 1000-5000-iteration burn-in (depending on numbers of detections) and thinned every 5 iterations to reduce autocorrelation within chains. We assessed model convergence using the Gelman-Rubin potential scale reduction parameter ( R ), where R = 1 at convergence (Gelman and Rubin 1992). We accepted coefficient estimates with R values less than 1.1. Additionally, we examined trace and density plots of the posteriors to ensure proper mixing of chains. For the availability and detectability components of the models, we generated Bayesian P values from the posterior predictive distributions to assess goodness-of-fit (Gelman et al. 1996), where a P-value near 0.5 indicates a fitting model. We then extracted the mean probability of availability, mean probability of detectability, and density of birds per hectare based on posterior probabilities and Bayesian credible intervals by year from the model.

Results
We sampled 248 points across the two study sites, including 93 early-successional and 155 mature pine stands. We detected 58 species in early-successional stands in 2019 and 62 species in 2020, with an average number of priority species across points of 4.42 (SE = 0.18) and 5.01 (SE = 0.19), respectively. In the mature pine stands, we detected 63 and 61 species in 2019 and 2020, respectively, with an average number of priority species across all points of 2.56 (SE = 0.21) and 3.85 (SE = 0.18), respectively. The most notable changes in detections across years were observed in Kentucky Warbler (Geothlypis formosa), whose detections increased from 20 to 50 and Wormeating Warbler (Helmitheros vermivorum), which increased from 2 to 26 detections. Markov chains in models for each species in each year converged ( R < 1.1), and Bayesian P values for availability (P a ), and detectability (p d ) indicated model fit for all species (Table 1). Additionally, Pearson's correlation test showed that no landscape or landform variables were highly correlated.
Our abundance model coefficients of total area (a proxy for habitat amount under the HAH) were positive in 6 of 13 and 7 of 13 species in 2019 and 2020, respectively, and the directionality of the relationship changed in 4 of 13 species across years (Table 2). A greater number of early-successional associated species had positive associations with total area than mature associated species across both years, but mature species had more consistent positive associations with total area between both years. The proximity of stands was an important factor for predicted  avian abundance. Most species exhibited negative associations with increased distance to their nearest early-successional or mature stand neighbor in both 2019 and 2020 (Table 2). In fact, only three and four species showed positive associations with Euclidean nearest neighbor distance in 2019 and 2020, respectively. The importance of proximity to nearby habitat is also mirrored in species associations with proximity index. Except for Field Sparrow (Spizella pusilla) and Kentucky Warbler, all species showed positive associations with proximity index in 2019 (Table 2). Field Sparrow and Kentucky Warbler also showed a negative relationship with total area and Euclidean nearest neighbor, which may indicate that local-level factors have a greater influence on these species. Associations with total area, proximity index, and Euclidean nearest-neighbor distance were distributed differently for each species with some exhibiting linear relationships and others exhibiting quadratic or exponential relationships in some cases (Figs. 2 and 3). We found community and species-level differences in associations among landform variables and the predicted abundance of the early-successional and mature pine priority species. In general, priority species associated with early-successional conditions had more detections and greater densities in the hilly pine sites whereas mature pine priority species exhibited more detections and greater densities in the pine flatwoods site (Supplemental Table 1). This is consistent with early-successional priority species having greater positive associations with measures of slope, elevation, and aspect (Table 3). This also varied by year, with much greater negative associations with slope across most species in 2020 than in 2019. Aspect shows the greatest and most consistent positive associations with predicted species abundance, which may coincide with vegetation structural characteristics.
When we combined our species into communities and analyzed landscape, landform, and local level covariates on species richness, variables had little effect on richness within both early-successional and mature pine communities (Fig. 4). For the early-successional community, the greatest positive effect to species richness resulted from proximity index and aspect, similar to our single species abundance model. Landscape variables for total area, and Euclidean nearest neighbor seemed to have the greatest negative effects on species richness, but effect sizes were small for all variables (Table 4). For the mature pine community, the greatest positive effect on species richness came from landscape covariates for total area and proximity index of mature pine stands within the landscape. This again mirrors our abundance model as proximity and total size of patches was more important for mature pine species, but effect sizes were more subdued at the community level than at the population level. Landform covariates for elevation, slope, and aspect seemed to have the greatest negative effects on species richness in mature pine stands, though again effect sizes were small. All species associations with local vegetation structural characteristics can be found in Supplemental Table 2.

Discussion
Our predictions that abundance of avian species was driven not only by total amount of a target habitat type, but by proximity of habitat patches within a species' local landscape was supported by our analyses.
Predicted abundance decreased for almost all species when proximity decreased (i.e., patch distance increased). Additionally, proximity index, which considers both patch size and proximity, exhibited the greatest number of positive effects among priority species. Since proximity index includes both metrics for area and proximity, we used ENN and total area to discern between these differences and found that proximity has a greater influence on species abundance. Therefore, the configuration of patches is important for these species, but size should not be disregarded, especially in those species which may be area-sensitive (e.g., Kentucky Warbler and Field Sparrow). However, our observed associations were species-specific and varied between years. Thus, broad generalizations regarding avian response to landscape characteristics should be avoided. This is reinforced by results of our community model, where we found landscape characteristics had little effect on species richness in early-successional and mature forest communities. The incorporation of all species' responses to landscape pattern averages out the effect so that we see no effect at the community level and between communities. We tested the effects of landscape pattern in the context of both abundance and richness and found species specific responses to size and isolation of patches, which little effect on species richness for our two communities. Under this premise, we cannot confirm Fahrig's (2013) hypothesis that it is the total area of habitat alone in a species local landscape that has the largest influence on a species richness or abundance as we have shown the importance of configuration for priority bird abundance. Although we do agree with Fahrig's notion that we need to shift away from thinking of a patch as an entity that delimits bounds of a species' movements and instead focus on how the local landscape influences each species differently within and across a gradient of different matrix types and quality (Fahrig 2013). The difficulty in testing how landscape scale factors affect biodiversity or species richness within a community is that effects are blurred from the incorporation of all species' responses. This is shown in our single species abundance model where species have varying responses to landscape patterns which may temporally fluctuate. We must be cognizant that differences in degrees of stand size and configuration appear to benefit certain species but are potentially costly for others. Further complexity is added by the dynamic shifting mosaic of working pine systems.
Although we evaluated associations within working pine forests with proximity to nearby similar vegetation conditions to test the HAH, microsite conditions and proximity to different vegetation types may have equal or greater effects on predicted abundance for some species. In this case, elements brought forward by the MDH in terms of a suite of Table 3 Beta coefficients means from the species abundance model of slope, elevation, aspect, and associated standard errors in parenthesis for 13 priority early-successional and mature pine associated bird species within the 2019 and 2020 breeding season in working pine forests in east-central Mississippi * Early-successional species local, landscape, and landform (which may also influence local characteristics) conditions that influence the abundance of species may have stronger support for certain species. We also assessed associations with landform as identified in the MDH (Mitchell et al. 2006) and found differences both between and within the two communities. Given that the differences in topography varied latitudinal among our sites in east-central Mississippi, it is difficult to discern whether associations with landform were associated with topography or species distributional ranges during the breeding season. Landform likely plays a much stronger role outside of flatwoods and hilly pine woodland systems of Mississippi, where topography will have greater influence on the structure and function of vegetative communities. However, understanding processes driving distributional patterns across a range of ecosystems is likely to enable better predictions of species' abundances in response to patterns in landform.
Perhaps the complexity of the MDH is better suited to explain these complex systems as species also do not only respond to landscape or landform, but the multiple vegetation structural characteristics that may be influenced by them. This is supported by results for Northern Bobwhite a species widely documented to require multiple cover types during its full life history (Guthery 1997(Guthery , 1999. Our assessment suggested mixed associations with early-successional vegetation amount, positive associations with proximity index (which incorporates area and patch proximity), and negative associations with distance to nearest patch, suggesting there is some importance of proximity of both early-successional and mature pine stands for Northern Bobwhite, a ground-dwelling species numerous known for limited dispersal distances (Fies et al. 2002). However, this species also exhibited consistent negative associations with dense vegetation (as measured by visual obstruction) and positive associations with bare ground cover, suggesting a mix of landscape configuration and local vegetation structural characteristics were influential (Guthery 1999;White et al. 2005).
Landscape and landform processes have effects on numerous other processes besides avian richness and abundance. Species interactions such as predation and conspecific attraction are thought to be closely tied with patch size (Fletcher 2009;Butcher et al. Fig. 4 Estimated avian species richness in relation to total area, Euclidean nearest neighbor (ENN), and proximity index for early-successional avian communities on top and mature pine avian communities on bottom within working pine forests in northern Mississippi, conditional on species detected at least once during the 2019-2020 surveys. Symbols denote point estimates with 95% credible intervals (CRI's) from fixed effect community model. Gray line under the blue line is a spline smooth with weights equal to the reciprocal of the squared posterior standard deviations. Blue line is the linear regression line estimated in the meta-analysis that accounts for both estimation error (posterior standard deviations) and residual variation around the regression line. The dashed blue lines give the 95% CRI of the prediction 2010). Additionally, the movement or restriction of species between sub-populations due to barriers or landscape pattern has been linked in determining species population dynamics (Koh et al. 2010;Kennedy et al. 2011), genetic variability (Keyghobadi 2007), and territoriality (Kie et al. 2010). Furthermore, the ability for a species to disperse from one landcover to another is becoming increasingly important considering expected range shifts with climate change (McLaughlin et al. 2002). The importance of landscape pattern cannot be overlooked when it comes to its effect on numerous ecological processes or how landform shapes these patterns and processes (Swanson et al. 1988). Our results show the important role of patch proximity in dynamic, working pine systems with rapid phases of succession. The importance of patch proximity extends beyond its influence on avian abundance; it also affects species movement and persistence, pollination, and trophic dynamics (Haddad et al. 2015).
In our assessment, species abundance varied in response to landscape, landform, and local-level covariates and exhibited some temporal variation which may be why richness within each community showed little effects to these variables. Further breaking down these communities into smaller guilds may reveal the factors that influence avian richness. Though the HAH provides a step forward from the thinking of island biogeography it may be too simplistic in its view that a single component (i.e., habitat amount) is the main determinant of species richness. It ignores the numerous population level processes that are affected by patch size, isolation, topography, and the suite of local level features that are explained by the MDH. Although in such dynamic systems, one must be cautious when extrapolating from the findings of short-term studies to longer temporal scales, especially when prescribing forestry practices to achieve specific conservation goals. Weather, forest management activities, or other factors may have driven variation in landscape and vegetation associations observed across the two years of our study. Additionally, the scale at which landscape factors influence avian species should not be overlooked as the heterogeneity in the landscape can have positive, negative, or neutral effects at different scales (Addicott et al. 1987;Smith et al. 2011;Frishkoff et al. 2019). Thus, it is important to understand a species' area requirement and how conditions within a species' local and surrounding landscape can impact its persistence.
We presented a small piece of how landscape-scale factors such as size, configuration, and topography of forest patches influence avian abundance and richness in working pine forests and focused on regionally prioritized species, when possible. Efforts to integrate management of timber and conservation of upland birds must be species specific in its scope of providing habitat requirements and be cognizant of the distribution of these requirements on the landscape. We examined these assocations during breeding seasons within working pine forest. However, questions remain on associations during the non-breeding season, or across a multitude of systems and taxa in which much of these methodologies can be adapted. The fact that total area alone did not have the greatest levels of positive effects to abundances of species in our analysis suggests that habitat configuration may be an important factor, especially at low or Table 4 Beta coefficient estimates from the species richness model for landscape variables of total area, Euclidean nearest neighbor (ENN), and proximity index (PROX) and landform covariates elevation, slope, and aspect with associated standard errors (SE) for early-successional and mature pine communities for the combined 2019 and 2020 breeding season in working pine forests of east-central Mississippi  (Radford et al. 2005;Martensen et al. 2012;Ochoa-Quintero et al. 2015;Richmond et al. 2015). Thus, when managing for a species, one must consider proximity of habitat patches when implementing management practices while keeping in mind those species that may exhibit area-sensitivities which may be negatively impacted. When managing for a community, we also recommend breaking species down into guilds as lumping a community together blurs the response to management activities Although no single hypothesis or theory can be used to explain all the nuances of a species' habitat selection, there need not be a divide between ecological theory and management. From these results, we have demonstrated how we can use landscape ecology hypotheses and theory as a backbone for forest management and conservation planning. We have also shown that forests managed for timber can contribute to species conservation. Knowing how landscape patterns influence species richness and abundance may enable managers to plan and optimize forest management strategies to enhance conservation for our at-risk species.