Site Selection

I randomly selected 45 sites from a subset of oak-dominated woodlands from a GIS database created by the Washington Natural Heritage Program. The Heritage Program defines oak-dominated stands as having greater than 25% crown canopy cover of Oregon white oak in the upper canopy layers and less than 25% crown cover of conifers with a minimum stand size threshold of approximately one acre. To reduce variation in within-stand characteristics, the subset consisted of woodlands between one and 10 ha. The stand size range was chosen to capture the majority of oak woodlands in the database while minimizing variation in woodland size. Of all the oak-dominated woodlands in the study area, 90% were smaller than 10 ha. Oak woodlands less than one ha were also excluded to avoid including woodlands with only a few oak trees. Ultimately, 62% of all the oak-dominated woodlands in the study area fell within the size range selected. All sites were located on public lands and sites were discarded from the sample if site visits revealed that they were not accessible or had understory dominated by manicured lawns or pavement. As a result, all sites had understory vegetation that would permit natural forest regeneration. A total of 30 sites met my criteria and were included in the study. Elevation of the study area ranges from 10 to 200 m a.s.l. and all woodlands were located on gravelly or fine sandy loam soils in the Everett-Spanaway-Nisqually complex (Pierce County soils data, JBLM soils data, and Thurston County soils data).

Landscape Measurement Methods

I defined a “landscape” as the area within a 0.5 km of each oak woodland site. I chose this distance because field research has indicated that it captures the approximate perceptual range of gray (Sciurus carolinensis) and fox (Sciurus niger) squirrels (Zollner 2000). Landscape areas did not overlap for any of the selected oak woodlands. I calculated the percent urban cover for each oak woodland landscape using ArcGIS 9.3 and a 30-m land-cover data layer derived from Landsat TM and ETM data for the Puget Sound region (Alberti et al. 2006). Urban cells were defined as those with > 50% impervious surface. Urban development was related to oak measures as both a continuous and categorical variable. The categorical variable classified each oak woodland as either urban (> 40% urban cover within 0.5 km) or non-urban (< 40% urban cover). The 40% threshold was chosen because none of the woodlands surveyed had between 25–40% urban cover creating a clear break between urban and non-urban sites. I calculated the area of each woodland in ArcGIS based on the boundaries delineated by the Heritage Program. I also used a dummy variable to indicate whether the woodland was located on or off JBLM.

Vegetation Survey Methods

I collected oak stand vegetation data within 10-m radius plots nested within each of the 30 oak woodland sites. Plot center locations were randomly generated using Hawths Tools in ArcGIS 9.3 and located in the field using a Garmin GPS. Each site had two to ten plots, depending on the woodland size. Within each plot, I collected diameter at breast height (DBH) for all oak trees with a DBH greater than 2.5 cm. DBH was measured for all other trees larger than 10 cm DBH. All oak saplings were counted and binned into two size classes. Class 1 (young) saplings were <2.5 cm DBH and 0.5–2 m tall. Class 2 (old) saplings were > 2 m tall and < 10 cm DBH. I counted oak seedlings (< 0.5 m tall) in four 1-m radius subplots. The center of each subplot was located 5 m in each cardinal direction (N, S, E, and W) of the 10-m radius plot center.

Canopy and understory composition were visually estimated for the 10-m plot and each of the 1-m seedling subplots. Both canopy and understory were analyzed separately as continuous (Table 1) and categorical variables. To create canopy categories, I binned the 1-m subplots into one of three canopy composition categories: 1) some oak canopy present (oak canopy > 0%), 2) some forest canopy but no oak canopy present (oak = 0% and non-oak forest > 0%), and 3) no forest canopy cover (open = 100%). These categories allowed me to distinguish between subplots located under or at the edge of oak canopy, under or at the edge of non-oak forest, and under no forest canopy (open). I also binned both subplots and plots into three canopy density categories: sparse canopy (≤ 30% forest canopy), mixed canopy (> 30% and < 60% forest canopy) and dense canopy (> 60% forest canopy). In these categories, oak and non-oak forest percentages are summed together.

Understory types included: shrub, dense invasive, grass, and open/herb. Some plots and subplots included areas of lawn, gravel road, tree trunks and brush piles. These are areas where regeneration cannot take place. Consequently, any such plots or subplots were removed from the analysis. Invasive species found during the survey include Scotch broom (Cytisus scoparius (L.) Link), blackberry (Rubus armeniacus Focke [Rubus discolor]), ivy (Hedera hibernica (G. Kirchn.) Bean), and reed canary grass (Phalaris arundinacea L.). Except for Scotch broom, these species were individually present in very few plots. Therefore, Scotch broom was lumped with other shrubs due to their similar growth form, while the other invasive species, which create dense cover conditions, were lumped into their own “dense invasive” category. Scotch broom was also analyzed separately as its own category.

To create understory categories, I binned each plot and subplot into their dominant understory types. An understory type was considered dominant if it covered more than 60% of the plot or subplot area. For this analysis, Scotch broom, shrub and dense invasive were all considered “shrub.” This resulted in four understory categories: shrub, grass, open/herb, and mixed (when no one understory type was >60%). At the plot level, only 5 plots were categorized as open/herb so these were removed from that analysis.

Statistical Analysis

I applied general linear mixed models to analyze relationships between seedling and sapling abundance and subplot, plot, and landscape characteristics. Mixed models account for the nested data structure of subplots within plots within oak woodlands. Dependent variables include: the quadratic mean adult tree DBH, total oak basal area per plot, seedling counts in subplots, and class 1 and 2 sapling counts in plots. Independent variables included woodland area, landscape urban cover, and plot and subplot understory and canopy estimates (Table 1).

Seedling counts for each subplot were tested against understory and canopy measures at the subplot level and landscape measures at the woodland level. In these models, both plot and woodland were included as random effects. Sapling and adult measures were tested against understory and canopy measures at the plot level and landscape measures at the woodland level. In these models, woodland was included as a random effect. Models using count data (e.g., seedling and sapling counts) were fit using the Laplace approximation and Poisson distribution (lme4 package, R 2.11.1 software). Models using continuous data (e.g., oak mean DBH) were fit by maximizing the restricted log-likelihood and the Gaussian distribution (nlme package, R 2.11.1 software). Significance for all models was set at α ≤ 0.05.

TABLE 1.

Direction and significance of linear mixed-effects models for seedling, young (class 1) sapling, and old (class 2) sapling abundance. Significance for all models was set at α ≤ 0.05. P-values of 0.1 or less are reported, n.s. indicates a p-value > 0.1. Sign indicates the direction of the relationship.

Effects of Canopy Cover

The data support my first hypothesis that seedlings are more abundant under forest canopy cover. Seedlings were more abundant with increasing amounts of oak canopy cover, and were less abundant with increasing amounts of open (non-forested) canopy (Table 1). Using canopy composition as a categorical variable showed that oak seedlings were least abundant under non-forest canopy (Figure 2). Comparing seedling abundance under closed, mixed and open canopy conditions again showed that abundance is lowest in open conditions (Figure 3). Oak seedling abundance appeared to peak under mixed canopy closure conditions, but this was not significantly different from closed canopy. Similarly, old (class 2) saplings were positively correlated with percent oak canopy, negatively correlated with percent open canopy (Table 1), and least abundant in plots with open forest canopy (Figure 4). Young (class 1) saplings showed no relationship to canopy cover (Table 1, Figure 5).

Figure 2.

Mean number of seedlings (+1 sd) by canopy type. Oak = some oak canopy present (oak canopy > 0%), non-oak forest canopy =some forest present, but no oak canopy present (oak = 0% and conifer + deciduous + shrub > 0%), and open = no forest canopy cover (open = 100%). Categories labeled with the same letter are not significantly different (P > 0.05).

Figure 3.

Mean number of seedlings (+1 sd) by canopy closure type: sparse canopy (≤ 30% forest canopy), mixed canopy (> 30% and ≤ 60% forest canopy) and dense canopy (> 60% forest canopy). Categories labeled with the same letter are not significantly different (P > 0.05).

Effects of Understory Type

The most prevalent understory types were shrub (including Scotch broom) and grass. Scotch broom alone was a somewhat common understory component (73 subplots or 13%). Other invasive species were only present in a few subplots: blackberry (10 subplots), ivy (16 subplots), and reed canary grass (3 subplots). Scotch broom was not significantly related to seedling abundance but was positively correlated with young (class 1) and old (class 2) saplings abundance (Table 1). The dense invasive category was only present in 27 subplots, but was negatively correlated with old (class 2) saplings. However, it is questionable whether this sample is sufficient to identify valid relationships.

Figure 4.

Mean number of old (class 2) saplings (+1 sd) by canopy closure type: sparse canopy (≤ 30% forest canopy), mixed canopy (> 30% and ≤ 60% forest canopy) and dense canopy (> 60% forest canopy). Categories labeled with the same letter are not significantly different (P > 0.05).

Figure 5.

Mean number of young (class 1) saplings (+1 sd) by canopy closure type: sparse canopy (≤ 30% forest canopy), mixed canopy (> 30% and ≤ 60% forest canopy) and dense canopy (> 60% forest canopy). Categories labeled with the same letter are not significantly different (P > 0.05).

The data do not support my second hypothesis as seedling abundance differed between understory types. Seedling abundance was strongly positively correlated with grass understory (Table 1), and seedlings were most abundant in grass-dominated and mixed subplots (Figure 6). In contrast, old (class 2) saplings were negatively correlated with grass. Abundance of both young (class 1) and old (class 2) saplings increased with increasing percent shrub cover (Table 1). In addition, both classes were most abundant in plots dominated by shrub cover (Figure 7 and 8). These results do not support my third hypothesis. While saplings did respond differently to understory conditions compared to seedlings, they showed a tighter relationship to disperser species habitat preferences than seedlings, as rodents are known to prefer shrub understory. Seedlings and saplings were all negatively correlated with open/herb understory, although this relationship was only significant for seedlings and young (class 1) saplings (Table 1).

Figure 6.

Mean number of seedlings (+1 sd) dominant understory category. Categories labeled with the same letter are not significantly different (P > 0.05).

Figure 7.

Mean number of young (class 1) saplings (+1 sd) by dominant understory category. Categories labeled with the same letter are not significantly different (P > 0.05).

Figure 8.

Mean number of old (class 2) saplings (+1 sd) by dominant understory category. Categories labeled with the same letter are not significantly different (P > 0.05).

Effects of Urban Development

Urban development, as either a continuous or categorical variable, was not significantly related to any of the dependent variables measured, except young saplings. There was a marginally significant, negative relationship between young (class 1) sapling abundance and percent urban land cover (Table 1, P = 0.06). Classifying each woodland into either urban (> 40% urban cover) or non-urban (< 40% urban cover) demonstrated that young (class 1) oak saplings were significantly less abundant in oak woodlands surrounded by more than 40% urban cover (Figure 9).

Figure 9.

Comparison of young (class 1) sapling abundance in urban and non-urban oak woodlands. Categories labeled with different letters are significantly different (P = 0.05).

Effects of Canopy Cover

Recruitment patterns within oak stands appear to change overtime. Oak seedlings were most abundant under oak canopy (Table 1, Figure 2), but older (class 2) saplings were less abundant with increasing adult oak basal area and DBH (Table 1). One interpretation of these findings is that although the majority of acorns germinate beneath adult trees, survival of seedlings is lower in these areas. Brudvig and Asbjornsen (2005) found the highest abundance of Quercus alba seedlings within existing oak stands, but noted low increases in height and basal area in these areas compared to seedlings located in canopy gaps. A lack of regeneration within existing oak stands and apparent suppression of conspecifics by adult trees has been documented in other oak systems as well (Callaway and Davies 1998). Density-dependent mortality is common in plant species and seed dispersal away from parent trees is seen as an important adaptation to escape this mortality (Janzen 1970).

Beyond the oak canopy, seedlings were significantly more abundant under or near forest canopy than in non-forested areas (Figure 2). Regan (2001) also found that Q. garryana seedlings on JBLM were most abundant under oak, then conifer canopy and least abundant in open areas. Three factors may lead to a lack of seedlings in non-forested areas: 1) reduced dispersal to these areas, T) poor environmental conditions for seedling emergence in these areas, and 3) heavy shoot herbivory in these areas. The first explanation is supported by several studies reporting that jays and rodents prefer to cache acorns under forest cover (Fuchs et al. 2000b, Gomez 2003, Pons and Pausas 2007, Gomez et al. 2008). The second explanation is not supported by a study by Fuchs et al. (2000a), who demonstrated that experimental acorn plantings survived and produced seedlings in most habitat types regardless of percent cover overstory vegetation. Work by Callaway (1992) supports the third explanation, which found very low survival of Q. douglasii seedlings in open grassland habitats compared to shrub habitats, primarily due to shoot herbivory. Based on these other studies, it is evident that additional experimental studies are needed to determine the relative importance of dispersal, environmental conditions, and herbivory as drivers of seedling abundance in this system.

Effects of Understory Type

The finding that seedlings were more abundant in grass- and mixed- understory types, rather than shrubs (Figure 6), may be related to seed predation behavior of dispersers and other animals. Several studies have found that rodent seed predation is heavier in areas with dense shrub cover (Smit et al. 2001, Kollmann and Buschor 2002, Muñoz and Bonal 2007, Perez-Ramos and Marañón 2008), and is generally reduced in open habitats (Hulme 1998). Fuchs and others (2000b) found that acorn predation on Q. garryana tended to be low in herb and high in shrub understory. Callaway (1992) found that acorn predation was the primary cause of mortality for Q. douglasii in shrublands, but shoot herbivory was more common in open grasslands.

In contrast to seedlings, the high abundance of both classes of saplings in areas with higher shrub cover suggests that shrub cover may favor regeneration (Figures 7 and 8). Similar results were found in a meta-analysis of shrubs as nurse plants in Spain, where Gomez-Aparicio et al. (2004) found that shrubs had a positive effect on woody plant seedling growth and survival for 18 tree species and that Quercus spp. showed the greatest positive response. Other studies have documented that shrubs facilitate oak recruitment in Quercus agrifolia (Callaway and Davis 1998) and Q. douglasii (Callaway 1992) in California and Q. humilis in France (Rousset and Lepart 1999).

Shrubs potentially facilitate oak regeneration through two primary mechanisms: 1) protecting seedlings from herbivores and 2) improving microclimatic conditions by providing shade. Oak leaves are particularly nutritious browse for deer and may experience heavy grazing pressure from herbivores (Larsen and Morgan 1998). Rousset and Lepart (1999) found that shrubs effectively protected oak seedlings from sheep herbivory, thereby increasing survival and growth. Callaway (1992) found that experimentally excluding herbivores improved Q. douglasii survival somewhat, but that excluding herbivores and providing artificial shade was necessary to increase seedling survival to levels comparable to those in shrub habitat. In contrast, Fuchs and others (2000a) found that while desiccation is a significant source of Q. garryana seedling mortality, there was no relationship between shade cover and desiccation mortality for experimentally planted seedlings.

My findings suggest that while acorns in grasslands may experience lower seed predation and therefore have a higher chance of germinating, seedlings in shrub habitat have a higher chance of surviving to the sapling stage. Gomez-Aparicio and others (2004) argue that shrub cover, which is commonly removed as a forest competitor, may actually enhance natural forest regeneration in the Mediterranean. While more experimental work is needed, my results suggest that shrubs may be important for Q. garryana regeneration as well. In this study, Scotch broom, an invasive species and the subject of significant removal efforts in this region, was included as a “shrub.” Both young and old saplings were more abundant in plots with Scotch broom, suggesting that it may provide benefits similar to other native shrubs in the region. However, managers will need to weigh the benefits Scotch broom potentially provides to oak regeneration against the costs it incurs to other native species in the oak ecosystem.

Effects of Urban Development

Young sapling abundance was significantly lower in urban areas (Figure 9) possibly due to concomitant differences in urban herbivore populations. Deer often reach high densities in urbanized regions due to reduced predation and abundant nutritious browse (Etter et al. 2002, Bender et al. 2004). High deer densities can cause significant damage to vegetation through overgrazing (Stromayer and Warren 1997, Horsley et al. 2003, Côtè et al. 2004). In addition to deer, eastern cottontail rabbits (Sylvilagus floridanus) spread rapidly throughout western Washington after their introduction in 1927 (Dalquest 1941). Although little is known about urban populations of S. floridanus in the study area, this species has adapted well to fragmented landscapes in Massachusetts (Smith and Litvaitus 2000) and is another potentially unchecked herbivore.

Urban areas have very different climatic, hydrologic and chemical characteristics compared to rural areas (McDonnell et al. 1997), so it is possible that these microclimatic conditions adversely affect seedling survival. Urban soils are often drier due to both increased temperatures in urban landscapes and increased water runoff due to impervious surfaces (Pickett and Cadenasso 2009). These changes in soil moisture may reduce seedling survival and subsequent transition to the sapling stage, given that desiccation is one of the most common causes of oak mortality (Fuchs et al. 2000a, Johnson et al. 2002). It is possible that a combination of local environmental changes and altered herbivore pressure result in the reduced sapling abundance in urban areas observed in this study. Lastly, while the areas included in this study all had natural understory and unmowed grass, it is possible that managers periodically mow tall grasses in urban sites, which would significantly reduce sapling survival.

The lack of a relationship between older saplings and landscape urban cover may be due to the accumulated effects of urban development over time. These effects may not have been significant when the older saplings were developing, but have become more intense in recent years. Alternatively, urban environments may be more favorable, or less detrimental, to sapling survival in later life history stages. For example, residential and commercial development may increase edge habitat, which could benefit oak saplings by providing greater access to sunlight. More research into sapling growth and survival in urban landscapes is needed to understand these patterns.

Human and Management Impacts

All of the sites included in this study have been heavily used and managed by people for decades, if not centuries. The urban sites located off of JBLM are all in public parks, where they experience different levels of recreational use as well as vegetation management. Conversations with park managers indicated that the understory of the areas included in this study was not actively managed. Forestry officials on JBLM do actively manage for oak habitat, including periodic burning, selective thinning, and mowing Scotch broom. In addition, areas of JBLM experience significant disturbance from military training activities. None of the study sites were burned or mowed immediately prior to conducting the vegetation surveys. However, training activities were ongoing during the survey. These activities clearly have a potentially strong influence on oak stand structure and cannot be discounted.

Regan (2001) found that burning significantly reduces the number of oak saplings between 0.1 and 10 cm DBH and has some beneficial effects on seedling growth. In this study, I found no difference in seedling abundance on or off JBLM (Table 1). This either suggests that burning is not applied widely enough on JBLM to raise the average seedling abundance, or that conditions off JBLM somehow compensate for a lack of burning. Alternatively, my sample may not be large enough to capture real differences, because seedling abundance is highly variable. Regan (2001) notes that the relationship between conifer canopy cover and seedling abundance is possibly obscured by prescribed burning, which also influences the behavior of seed predators. Here, I found the same positive relationship between canopy cover and seedling abundance (Figure 2) regardless of the application of burn management.

Similarly, if burning on JBLM reduces sapling abundance, then I would expect to see a higher abundance of saplings in the unburned sites off JBLM. Instead, young saplings are significantly less abundant in unburned, urban sites (Figure 9), and older sapling abundance is not significantly different (Table 1). Finally, controlling for location on and off JBLM does not change the observed relationships between canopy cover and understory for any oak age class.