Content uploaded by William J. Zielinski
Author content
All content in this area was uploaded by William J. Zielinski on Jan 14, 2015
Content may be subject to copyright.
1
Effects of Forest Restoration on
Mesocarnivores in the Northern Redwood
Region of Northwestern California
Final Report
28 November 2010
Keith M. Slauson1, William J. Zielinski1, and Thomas A. Kirk2
1USDA Forest Service, Pacific Southwest Research Station,
Redwood Sciences Laboratory, Arcata, California.
2Lassen National Forest, Susanville, CA 96130
2
ABSTRACT
Restoration efforts are underway in the logged second growth forests in the Redwood
National and State Parks complex to accelerate the return of old growth forest conditions that
have been lost. Mesocarnivores are ideal focal suite of species to evaluate the effects of forest
change because they include species such as the Humboldt marten, highly adapted to complex
coastal forest conditions of hunting and opportunistic generalists such as the gray fox that
typically respond positively to human-altered ecosystems. From 2009-2010 we studied how
several species of mesocarnivores and key habitat components for the Humboldt marten have
responded to old growth and second growth forest characteristics, forest roads, and restorative
thinning. Stands restoratively thinned 15-30 years ago have regenerated moderately dense shrub
layers composed of native, shade tolerate shrub species approaching the conditions in old growth
redwood stands and stands used by the Humboldt marten. Suitable resting structures for
martens, including large diameter trees, snags, and downed logs with cavities or platforms have
been significantly reduced in second growth stands compared to old growth stands and stands
used by martens. Regional occupancy modeling revealed that dense shrub cover, measured at a
spatial scale close to each species‟ home range size, most significantly increased the probability
of marten occurrence and decreased the probability of occurrence of the fisher and gray fox.
Road density only affected martens, where it reduced the probability of occurrence. Camera
survey results for carnivores on paired road and creek locations revealed that 80% of the
detections of generalist carnivores occurred on roads while 80% of detections of habitat
specialist carnivores occurred off-roads, along creeks. Restorative thinning and road removal
will affect mesocarnivores over the short term (1-3 decades) by restoring dense shrub cover and
reducing its fragmentation, factors that will likely benefit habitat specialists like the Humboldt
marten by reducing the distribution and abundance of larger-bodied generalist mesocarnivores
that can kill them. However, because mesocarnivores have large home ranges, restoration
actions will have to be strategically located to have the greatest impact. Large-diameter standing
and downed woody structures, critical for providing resting and denning locations for the
Humboldt marten are depauperate in second growth landscapes. Because the restoration of the
natural recruitment of these structures will likely take >200 years, alternatives for creating
suitable structures will be necessary to improve habitat conditions in the interim.
BACKGROUND
The goals of forest restoration efforts in the redwood region are to return natural conditions
to sites where they have been significantly altered. To date, forest restoration has focused on
restorative thinning to accelerate the recruitment of late-seral forest conditions and the closure or
removal of roads. It is unclear how wildlife species will respond to these changes.
Intermediately sized mammalian carnivores (mesocarnivores) are important indicators of
ecosystem integrity due to their wide-ranging habits, reliance on numerous prey populations, and
often highly specialized habitat requirements. Understanding their ecology in relation to changes
in the structure and function of the habitats they require will allow managers to predict their
responses to restoration.
3
Redwood forests are more than big trees to terrestrial mesocarnivores, they are highly
productive and structurally complex ecosystems. For terrestrial carnivores that chase down
smaller-bodied prey, old growth redwood forests offer a challenging obstacle course of dense
stems and foliage near the ground in which to capture a meal. The Humboldt marten (Martes
americana humboldtensis), the smallest of all marten subspecies, appears to have adapted to this
environment and its small size allows it to negotiate thick understory tangles (Hagmeier 1961).
Unfortunately, few areas of sufficient size and suitability -- dense spatially extensive shrub cover
underneath an old growth tree canopy -- remain in the redwood region to support marten
populations (Thornburg et al. 2000).
The Humboldt marten is endemic to the redwood region (Grinnell and Dixon 1926) and was
common in the early twentieth century (Grinnell et al. 1937) but was feared extinct towards the
end of the century due to the absence of verifiable detections for over a 50 year period (Zielinski
and Golightly 1996). Remarkably, in 1996, the first marten was detected from the only known
population remaining, currently occupying an area <5% of the historical range of the Humboldt
marten, adjacent to second growth forest in Redwood National and State Parks (RNSP; Slauson
2003). Surveys in 2002 found that martens still had not recolonized RNSP (including the Mill
Creek addition), despite the presence of a nearby population located <2 kilometers to the west of
the Rock Creek area and well within their dispersal range (Slauson and Zielinski 2003). The
marten is one of the most highly specialized species in the redwood region. Martens appear to
require numerous large and old woody structures (e.g., live trees snags with cavities, downed
logs with cavities) for resting refugia and large patches of dense, spatially extensive shrub cover
to exclude to exclude larger-bodied mesocarnivores (Slauson et al. 2007, Slauson and Zielinski
2009a). Restoration of marten populations to more of their former range represents one of the
best indicators of ecosystem recovery in the redwood region and should be one of the long-term
goals guiding restoration efforts.
The distribution of carnivores is largely determined by the interaction of a number of factors
including habitat structure, prey availability, mesocarnivore community interactions, and
physical factors. The marten is smaller than most mesocarnivores and thus is likely to be
excluded by larger-bodied species through interference competition and direct killing. Martens
in North America typically occupy forests in locations that receive frequent deep, soft snowfall
during the winter (e.g., high elevations of the Sierra Nevada and Cascade mountains). Martens
are well adapted to cope with deep, soft snow due to having one of the lowest foot loadings (low
mass per unit foot surface area) of all mesocarnivores (Krohn et al 2004). Where snowfall is not
so frequent or deep, dense fisher populations have been shown to limit the distribution of marten
populations (Krohn et al. 2004). Deep, soft snow is a physical factor, that limits the ability of
larger-bodied mesocarnivores (e.g., fisher and gray fox) with higher foot loading to occupy these
regions. Snowfall is rare in the redwood region, but dense shrub cover is not, and shrubs likely
represent another physical factor that discriminates against the larger-bodied mesocarnivores. In
the last century this dense, spatially extensive shrub layer has been highly fragmented by road
building and reduced by stand management methods developed to maximize wood production.
Site preparation after logging typically reduces shrub cover. The high densities of replanted
trees limit the amount of light reaching the ground and further reduce shrub cover. Species of
mesocarnivores that were not historically common in the redwood region (i.e., fisher, gray fox)
4
have now become quite common in these second growth habitats with highly altered shrub
structure (Grinnell et al. 1937, Klug 1996, Slauson and Zielinski 2004).
The fisher (Martes pennanti) was historically distributed throughout much of northwestern
California, but it was less common in the coastal redwood forests than it was in the interior
Douglas-fir forests (Grinnell et al. 1937). During the late nineteenth and early twentieth
centuries nearly all records (95%) for fishers in Humboldt and Del Norte counties were >20 km
from the coast. However, during the latter portion of the twentieth century the fisher has
expanded its range in the redwood region and now occupies many areas of second growth forest
in the northern redwood region (Slauson and Zielinski 2004). The gray fox has followed a
similar pattern, although historical data is not as complete as for the fisher. Nevertheless,
contemporary occurrence data suggest that fishers and gray foxes are either absent from, or
extremely rare, in the largest remnant patches of fog-influenced old growth forests in the
redwood region within Redwood National and State Parks (Slauson and Zielinski 2003), the
Smith River National Recreation Area, and Six Rivers National Forest (Slauson and Zielinski
2004). While the fisher has received deserved conservation attention, due to its more severe
decline throughout the Pacific states, its presence in most of the northern redwood region appears
to be without precedent and appears to represent an expansion into areas formerly occupied by
martens. If the goal of the restoration of second growth is to return old growth conditions
(including dense shrub layers) where they have been lost, these efforts will have beneficial
effects on marten and will probably discourage larger-bodied competitors like the fisher and gray
fox.
Using mesocarnivores like the marten, fisher, and gray fox as focal species to help predict,
design, and monitor restoration efforts will provide large-scale models of how wildlife responds
to such efforts. We quantitatively assess the current suitability of second growth areas of
Redwood National and State Parks for the Humboldt marten at three spatial scales: microhabitat,
stand, and home range. Secondly, we will investigate the effect of road density on the
occurrence of martens, fishers, and gray foxes in the redwood region. Finally, we will use the
information gained from these new efforts to evaluate how future restoration efforts can best be
planned to benefit restoration of the Humboldt marten to Redwood National and State Parks.
Project Objectives
1. Compare the availability of potential marten rest structures in stands occupied by martens to
their availability in second growth stands.
2. Determine the current habitat suitability, using previously developed methods (Slauson et al.
(2007), of all recently restoratively thinned stands.
3. Determine how road density affects the occurrence of martens, fishers, and gray foxes.
Compare the use of roads versus the interior of stands with dense shrub layers for several species
of mesocarnivores.
5
4. Use the information from this study to predict the future responses of fishers, martens, and
gray foxes to the current plans for restoration in Redwood National and State Parks and to
identify strategic areas for restoration to facilitate re-colonization for the Humboldt marten.
METHODS
We compared the habitat characteristics at areas currently occupied by martens to second
growth areas in Redwood National and State Parks currently unoccupied by martens. We
included a representative sample of restoratively thinned stands, including recent (n = 21) and
15-30 year old (n = 6; Veirs 1986, Keyes 2005) thinned stands, as well as a sample of adjacent
unthinned stands (n = 15). In each stand we sampled the density of microhabitat elements (e.g.,
large downed logs) and stand characteristics (e.g., seral stage, shrub cover). At the microhabitat
scale we compared the densities of potential rest structures in second growth stands and stands
known to be occupied by martens. At the larger, stand and home range scales, we applied
predictive habitat models, developed from the areas occupied by martens, to determine the
change in habitat suitability following restoration. Finally, to assess the effects of road density,
we measured mesocarnivore use of roads and also took a more regional approach, using our
systematic survey database of mesocarnivore detections for the northern redwood region to
evaluate the relative effects of road density and other key habitat characteristics on
mesocarnivore distributions.
Habitat Assessments
Microhabitat: Rest and Den Site Availability
We established reference conditions for the availability of potential resting and denning
structures by estimating their density in stands within known marten home ranges to the east of
Redwood National and State Parks. Rest structures included large live trees with defects, snags,
logs, and rock piles, the most common resting structures used by martens in the coastal
population (Slauson and Zielinski 2009a). Each log, snag, and live tree structure must have been
>60 cm in maximum diameter or diameter at breast height (DBH) and have a detectable feature
(e.g., chamber, cavity, platform) which a marten could use for resting or denning. Each
restoratively thinned and paired unthinned second growth stand was also sampled to compare
potential resting and denning structure density. We recognized that it was necessary to reconcile
the fact that places where martens occur may be different than the redwood second growth areas.
To do so we referenced previously collected data from Slauson and Zielinski (2003) to provide
additional reference conditions for the density of potential resting and denning structures in old
growth stands throughout Redwood National and State Parks.
Density of potential resting structures was determined using multiple variable-length belt
transects following the approach of Bate et al. (1999). Transect length was determined by the
dimensions of each stand and contained enough total length to cover >50% of each stand.
Transect width differed for trees and snags (wider) versus downed logs and rock piles (narrower)
due to differences in detectability, each width was determined in each stand based on the
distance the observer could reliably detect each structure type.
6
Shrub cover for each stand was measured along the same transects as potential resting
structures were sampled. At each 10m interval along the transect, shrub cover and species
dominance was visually estimated in a 1x1m quadrat. The quadrat was placed on alternating
sides (right or left) of the transect at successive intervals. Mean shrub cover for each stand was
estimated by taking the mean of all quadrat estimates. Shrub species rank dominance for each
stand was calculated by summing all the ranks (1-3) from each quadrat.
Stand Scale Habitat Suitability
To determine the current habitat suitability of restoratively thinned and unthinned stands we
applied a previously developed predictive model for the Humboldt marten (Slauson et al. 2007):
Prob. of Occupancy = 1/(1+exp[-4.16 + (Seral Stage) + (0.036* % Shrub Cover)
+ (2.41 * % Conifer)]
This stand model combines information on the seral stage, density of the shrub layer, and conifer
dominance of the overstory to produce a score, ranging from 0-1, representing the likelihood that
a marten would occupy a stand. Stand scale habitat suitability was determined for all
restoratively thinned stands and their adjacent unthinned stands. To provide a set of old growth
redwood stands to evaluate habitat suitability, we included all the old growth stands (n = 69)
throughout RNSP sampled by Slauson and Zielinski (2003), representing a random set of
redwood reference stands. Finally, we included all 26 marten occupied and 133 unoccupied
stands sampled earlier by Slauson et al. (2007).
Effect of Roads
Spatial Analysis
We used our existing survey database for the northern redwood region to investigate the
relationships of mesocarnivore distribution to road density and other home range scale metrics of
habitat structure and composition. The database used for this analysis contained 112 survey
locations from 3 survey efforts. The first two survey efforts were the 1996-1997 “systematic
surveys” (n = 78; Carroll et al. 1999) and 2002 Redwood National and State Park Surveys (n =
22; Slauson and Zielinski 2003) both which used the same survey protocol, a sample unit
consisting of 6 track plate stations spaced ~500m apart in a pentagonal array. The third survey
effort was the 2000-01 Humboldt marten population survey (n = 10; Slauson et al. 2007) which
used 2 track plate stations per sample unit, spaced 250m apart. All three survey efforts used the
same station-level protocol: once established each station was revisited every 2 days to remove
tracks and replace bait. All stations were baited with single chicken drumsticks and had an
olfactory lure (Gusto; Minnesota Trapline Products, Pennock, MN, USA).
Evaluation of the effect of road density on mesocarnivore distribution must include other
habitat characteristics known or hypothesized to be important to each species in order to best
understand the relative importance of roads. To create a suite of variables to model
mesocarnivore distributions, we developed 7 habitat covariates (2 related to roads, 5 related to
7
habitat composition), representing home-range scale characteristics of the areas within a 1 and
2.5-km radius of each sample unit (Table 1). We used 2 GIS coverages to derive the habitat
covariates. The first was a gradient nearest neighbor vegetation coverage (hereafter GNN layer),
created by the Landscape Ecology, Modeling, Mapping, and Analysis program
(http://www.fsl.orst.edu/lemma/splash.php) which uses a plot interpolation method to integrate
field plot, remotely sensed, and mapped environmental data to map current vegetation. This
vegetation layer is moderate to highly accurate (r2 = >0.60) for predicted plot values of the coarse
vegetation characteristics we selected. Three different structural habitat types were derived from
the GNN layer: old forest (OG) defined by the combination of giant and large tree size classes,
young forest (Y) defined by the combination of the shrub and pole developmental stages, and
shrub cover (S) defined by the presence of shrub cover >50%. The second coverage was a
transportation layer, used to assess road density, that included all types of forest roads (e.g.,
paved, gravel, unimproved) created by the Six Rivers National Forest. The linear amount of road
was converted to 30x30m pixel coverage, representing presence or absence of a road in each
pixel.
Each sample unit was buffered with a 1 and 2.5 km radius circle, representing areas (1 km =
314 ha) equivalent for a typical male marten home range in coastal California (1-km radius;
Slauson and Zielinski unpubl. data), equivalent to 2 home ranges for male gray foxes (1-km
radius; Fryxell 1982), and equivalent to 50% and 100% of an average male and female fisher
home range, respectively (2.5 km radius; Higley and Matthews 2009). Structured query
language (SQL) was used to make selections from the vegetation and road coverages to develop
predictors of mesocarnivore habitat. Selected habitat predictors were processed with a spatial
analysis program (FRAGSTATS v 3.3, University of Massachusetts Landscape Ecology
Program, http:/www.umass.edu/landeco/research/fragstats/ fragstats.html) to calculate landscape
metrics associated with each sample unit. These metrics included the percent of the circle
represented by roads (R) and each habitat type (e.g., OG, Y, S; Table 1) in the circle (PLAND),
and for OG only, the number of habitat patches (NP) and number of distinct core areas (NDCA;
Table 1).
Model Development, Selection, and Evaluation
We hypothesized that both positive and negative relationships existed between road and habitat
metrics and each species‟ distribution. We developed 8-12 a priori habitat models, based on
existing information and our hypotheses, describing the relationships between species occupancy
and home range scale metrics of habitat structure and composition (Burnham and Anderson
2002). We used resource selection functions (Manly et al 2002) to determine the habitat
characteristics most important for species occupancy.
We independently ranked each species‟ set of models using Akaike‟s Information Criterion for
small sample sizes AICc, (Burnham and Anderson 2002). We interpreted models by the
comparison of ΔAICc values, which provides a measure of fit of data to the model (Anderson et
al. 2000). To further interpret the relative importance of a model, given the a priori model set,
we calculated Akaike weights (wi) using ΔAICc values and created a 95% confidence set of
models by considering all models whose cumulative weights equaled 0.95 (Burnham and
Anderson 2002). To assess the relative importance of each variable in the selected models, we
8
assessed their frequency in the 95% confidence set of models and their effect sizes using odds
ratios.
Road Surveys Using Cameras
To determine the species that use roads, we established paired remote camera sample units.
One camera was located along a creek in the interior (>250 m from the edge/road) of a stand
with dense (>70% cover), spatially extensive native shrub layers and the second was located on
an adjacent, low human use, forest road. We selected creeks for comparison, because they
represent a naturally occurring linear feature. Each camera was run for a total of 70 days, the
first 28 days no olfactory lure was used, the second 28 days fisher, gray fox, and bobcat urine
was sprayed 2-3 m in front of each camera, and during the final 14 days 3 chicken drumsticks
and an commercial trapping lure (Gusto) was applied 2-3m in front of each camera. This
sequence provided an initial period to collect unbiased information on the frequency of use of
roads and creeks and then 2 levels of attractants to see if use patterns would change. Two types
of remote camera units were used, Cuddyback Excite 2.0 and Scoutguard 5.0 megapixel
cameras. Paired camera sample units were established in 3 different landscapes, defined by the
2.5 km radius surrounding the sample unit dominated by either old growth (>80%), old growth-
second growth mix (>30%, <70% each), or second growth (>80%). In addition, each landscape
surveyed had to have had a fisher or marten detected there within the last 10 years in order to be
able to interpret the results relative to the two mesocarnivore species of highest conservation
concern in the region.
An individual detection was recorded when a carnivore species was detected ≥1 time during
each 24-hour period. This approach minimized the influence of short-term responses to olfactory
lures. Detection rates were calculated by the following equation:
(Total Detections/Survey Duration)*10
Detection rates were compared for roads versus creeks for each species and species guilds
(habitat generalists versus specialists) using paired t-tests.
RESULTS
Habitat Assessments
Shrub Cover
We measured shrub cover along 1,060 m of transects in 6 stands occupied by martens, 900 m
in 9 second growth stands thinned 15-30 years ago and 2,100 m in 21 second growth stands
thinned in the last 2-7 years. Shrub cover was significantly greater in stands thinned recently (t =
3.16, df = 18, P =0.003) and 15-30 years ago (t = 3.88, df = 5, P =0.003) than in paired unthinned
stands (Figure 1, Table 2). However, mean shrub cover was still significantly lower in stands
thinned 15-30 years ago (t = 2.26, df = 6, P =0.03) and recently thinned stands (t = 5.97, df = 16,
P <0.001) compared to stands occupied by martens (Figure 1). Mean shrub cover in stands
9
thinned 30 years ago did not differ statistically from stands occupied by martens (t = -0.66, df =
3, P =0.066), but was still lower (Figure 1, Table 2) and overall more patchily distributed in
thinned stands (K. Slauson pers. obs.).
The rank dominance of shrub species was similar between marten occupied, old growth
redwood stands, and restoratively thinned stands (Table 3). Control stands reveal the severe
decline and loss of the dominant shrub species found in marten occupied stands (e.g., salal
[Gaultheria shallon], rhododendron [Rhododendron macrophyllum], evergreen huckleberry
[Vaccinium ovatum]) and adjacent thinned stands (Table 3).
Fruiting and flowering by dominant shrub species (e.g., evergreen huckleberry, salal) was
rarely observed in thinned stands, with only 38% (5/13) of stands supporting flowering and
fruiting shrubs. Within these stands, flowering and fruiting was limited to individual shrubs with
the most solar exposure (least overhead canopy cover). No flowering or fruiting was observed in
any control stands, with many plants showing signs of moderate to severe decline in vigor and
little new shoot growth. In contrast, fruiting and flowering was observed in 100% (6/6) of stands
known to be occupied by martens.
Rest and Den Site Availability
Thinned stands contained significantly fewer potential resting and denning structures than
stands known to be occupied by martens (Table 4). Only stands thinned 15-30 years ago had
large logs at a similar densities to marten occupied stands, likely reflective of past practices of
leaving cut trees if they had significant heart rot. In other cases several logs had been piled
together, creating chambers potentially suitable for marten use.
Stand Scale Habitat Suitability
Application of the stand scale resource selection probability function revealed that the
changes to date from restorative thinning have increased stand suitability values for martens
beyond unthinned stands (Figure 2). However, structural conditions have not yet returned to
those within the range of stands used by martens (Figure 2). The trajectory of restoratively
thinned stands suggests that suitable shrub cover will establish within 10-20 years (Figure 1).
Although restoration of shrub cover will result in increases in stand habitat suitability values,
they will likely remain below the suitability range for martens until old growth tree
characteristics return to the stands.
Effect of Roads
Mesocarnivore Occupancy Modeling Results
Marten--Five models were included in the 95% confidence set, 2 of which best fit the data
(Models 1-2, Table 5). All models in the 95% confidence set contained the 3 variables with the
highest importance weights (Table 5): shrub cover, road density, and old forest. All five models
within the 95% confidence set included the shrub cover variable (Table 5).
10
In all top models, the odds of marten occurrence increased most significantly with dense
shrub cover and decreased with road density. Using the estimates from Model 2, a 10% increase
in the percent of dense shrub cover within a 1-km radius was associated with a 98% increase in
marten occurrence (odds = 1.98, 95% CI = 1.12 to 3.50), after accounting for road density. For
road density, a 0.5% increase in the amount of roads in a 1-km radius resulted in a 19% decrease
in marten occurrence (odds = 0.81, 95% CI = 0.67 to 0.97), after accounting for dense shrub
cover.
Fisher--The 95% confidence set contained only 3 models, 1 of which was well supported as
the top model (Model 1, Table 5). In all the models in the 95% confidence set, the probability of
fisher occurrence decreased with dense shrub cover, old forest, and young forest. Using the
estimates from the top model (Model1, Table 5), a 10% increase in the percent of dense shrub
cover within a 2.5-km radius was associated with a 61% decrease in fisher occurrence (odds =
0.39, 95% CI = 0.20 to 0.77), after accounting for old and young forest. For old forest, a 10%
increase in the amount of old forest in a 2.5-km radius resulted in a 50% decrease in fisher
occurrence (odds = 0.50, 95% CI = 0.30 to 0.90), after accounting for dense shrub cover and
young forest. For young forest, a 10% increase in the amount of young forest in a 2.5-km radius
resulted in a 15% decrease in fisher occurrence (odds = 0.85, 95% CI = 0.49 to 1.46), after
accounting for dense shrub cover and young forest.
Gray Fox--The 95% confidence set contained 8 models, 4 of which best fit the data (Table 5).
All 4 models included shrub cover. The probability of gray fox occupancy decreased with
increased shrub cover. Using the estimates in the top model (Model 1, Table 5), a 10% increase
in the percent of dense shrub cover within a 1-km radius was associated with a 50% decrease in
gray fox occurrence (odds = 0.50, 95% CI = 0.25 to 1.0), after accounting for the number of old
forest patches. For old forest patches, an increase of 10 in the number of old forest patches in a
1-km radius resulted in a 20% increase in gray fox occurrence (odds = 1.20, 95% CI = 0.97 to
1.48), after accounting for dense shrub cover.
We evaluated the scale sensitivity of the strength of the effect of the shrub cover variable for
each species by comparing the univariate p-values for each spatial scale (Figure 3). P-values
were minimized when dense shrub cover was measured closest to the home range size of each
species. This indicates how the amount of dense, spatially extensive shrub cover most directly
influences the process of home range establishment by individuals of each species.
Camera Surveys
From 01 May to 16 December 2009, 20 remote camera sample units were surveyed in
California State Parks (n = 8), Redwood National Park (n = 3), and Six Rivers National Forest (n
= 9; Figure 4). Seven carnivore species were detected; 3 habitat generalists and 4 habitat
specialists (Table 6). The survey protocol performed well for most species, with near perfect
probability of detection for all but 2 species, mountain lion (Felis concolor) and mink (Mustela
vison, Table 6).
Three species showed positive responses to both attractants: gray fox, fisher, and marten
(Table 7). Detection rates for these species increased when either the 3 urines or Gusto+chicken
11
were added. Marten and fisher both had the strongest positive response to Gusto+chicken (Table
7). Addition of urine did have a positive response for bobcats, but the presence of
Gusto+chicken did not alter their detection rate.
Habitat generalist carnivores were detected on roads significantly more than forest interiors,
while habitat specialists used forest interiors significantly more than roads (Figure 5, Table 5).
Although fishers and martens showed the strongest increase in detections when attractants were
added, the presence of Gusto+chicken did not increase their detections on roads, with only 2
fisher and 0 marten detections occurring on roads when Gusto+chicken used.
During our camera surveys, we detected at least one marten in two locations in Prairie Creek
Redwoods State Park (Figure 6). These detections are >9 km west of the nearest known marten
detections and are the first in recent times to occur south of the Klamath River. Surveys in 2010
have confirmed the continued presence of a marten in the same area (K. Slauson unpubl. data).
This appears to be a recent recolonization as surveys conducted in 2002, <200m from the
detection locations, did not detect martens.
DISCUSSION
Potential Marten Rest and Den Site Availability
Potential rest and dens structures have been severely reduced in logged stands. These
findings are consistent with the comparison of logged second growth and old growth stands in
Redwood National and State Parks (Slauson and Zielinski 2003). However, logged stands in the
Mill and Rock creek watersheds are more depauperate of potential rest structures than older
second growth stands (Veirs, Whiskey 40) sampled in this study and by Slauson and Zielinski
(2003) throughout Redwood National Park. This finding is undoubtedly linked to the reduction
in the practice of leaving felled trees with extensive heart rot.
Resting structures are a critical component of habitat for martens and are used daily, to
provide thermal benefits and security from predation while resting. Arboreal resting structures
constituted 54% of those used by martens in coastal northwestern California in the summer and
fall (Slauson and Zielinski 2009a). Elsewhere in their range, martens shift to using more ground-
based resting structures covered by snow during the winter due to their increased thermal
benefits (Schumacher 1999). However, this seasonal shift is not expected in the coastal forests
martens inhabit, due to the lack of significant snow pack. The minimum ages of live and dead
woody structures used by martens in coastal northwestern California was 176 and 254 years,
respectively (Slauson and Zielinski 2009a). Given the time required to naturally regenerate
suitable rest structures, alternative approaches will be necessary to provide the adequate types
and number of resting structures during the next 1-2 centuries.
One alternative approach could involve developing artificial structures for resting. Slash
piles, left after logging often have chambers, and have been shown to be used by martens. Use
of slash piles is typical during the summer, likely due to their poor thermal benefits during winter
(Raphael and Jones 1994). However, the creation of inner chambers to provide protection from
12
rain and increased thermal cover may increase their use in winter. British biologists have
designed a „marten den box‟ (Vincent Wildlife Trust 2010) used by pine martens (Martes
martes) in young forests. These boxes could be used in slash piles as well as elevated on live
trees. The interstitial spaces in large rock piles have also been used by martens (Slauson and
Zielinski 2009a) and could be an additional longer-lived alternative to slash piles. Existing large
logs and trees could be enhanced through cavity creation, which has been successful in the
restoration of limited resources for an endangered forest bird, the red-cockaded woodpecker
(Picoides borealis). Finally, large logs with natural or created cavities could also be moved into
sites where they are especially depauperate.
Response of Shrub Cover to Thinning
Shrub cover positively responded to thinning of the overstory. Stands thinned 2-7 and 15-30
years ago had a 3 and 5-10 times increase in mean shrub cover compared to their respective
unthinned control stands. If left unthinned, many control stands will lose most individual shrubs
and head on a developmental trajectory very different from surrounding reference conditions.
Only the oldest thinned stands (Veirs, Whiskey) showed regeneration of the shrub layer to the
extent present in marten occupied stands, suggesting that 20-30 years may be required after
thinning to regenerate suitably dense and spatially extensive shrub cover.
Dominant shrubs that produce flowers and fruits support additional foraging opportunities for
martens. Martens frequently consume fruit, which appeared in >80% of 420 scats samples
primarily from June until November (Slauson unpubl. data). Indirectly, flowering and fruiting
shrubs support numerous prey populations (e.g., bald-faced hornets [Dilovespula maculata]) and
attract migratory species (e.g., band-tailed pigeons [Columbia fasciata], varied thrushes [Ixoreus
naevius]) that collectively increase the foraging value of stands where flowering and fruiting
occurs.
Most shrubs showed responses to thinning with increased shoot growth, but the presence of
flowering and fruiting was nearly absent from recently thinned and very sparse in old thinned
stands. In comparison, shrub layers in all marten occupied stands showed fruiting and flowering.
While recently thinned stands may simply require more time for shrubs to have the resources to
support flowering and fruiting, the old thinned stands have had 1-3 decades. Fruit production is
related directly to the availability of soil nutrients and the amount of photosynthate available,
thus canopy closure directly affects this latter resource and can limit fruit production (Lee 1988).
This suggests that stands thinned 20-30 years ago have either reduced soil nutrients or that
overstory canopies have become/remained restrictive to flower and fruit production. Given that
some shrubs in recently thinned stands had responded by fruiting, the reduction in soil nutrients
does not appear to be the cause. We suspect that the single layer tree canopies in regenerating
stands versus multi-layered of old growth, marten occupied stands may result in markedly
different solar radiation exposure when overall canopy cover is similar. This has been shown to
be the case for how other soft-masting species, which decrease fruit production in even-aged
versus uneven-aged silvicultural systems as canopy cover increases (Perry et al. 1999).
Restoration of fully productive shrub layers may require increased solar radiation to the shrub
layer either through time as the canopy becomes multi-layered or through additional thinning.
13
Effects of Roads and Dense Shrub Cover on Mesocarnivore Distribution
The pattern of road use by habitat generalists versus specialists was striking in that habitat
generalists dominated the use of roads and habitat specialists dominated the riparian areas in the
forest interior. This supports the hypothesis that roads benefit species least adapted for capturing
prey in the dense shrub understory of coastal forests. We cannot determine whether species are
simply adjusting their movements to use roads or whether roads are facilitating species use of
coastal forest landscapes. However, roads favor the generalists (i.e., bobcats, gray foxes) which
also prey on martens and fishers (the forest specialists). Habitat specialists that must cross roads
to move between habitat patches are more likely to encounter predators there, farther away from
escape cover.
The most important factor determining the distribution of all 3 mesocarnivores was the
presence of dense shrub cover. This provides strong support for the importance of the
maintenance and restoration of this structural layer for marten conservation and restoration.
Furthermore, it highlights the importance of a physical factor structuring the mesocarnivore
community and mediating carnivore interactions.
At the home range scale, road density was an important factor contributing to limiting marten
distribution. Although road density is correlated with other factors known to negatively affect
martens (e.g., percent of the landscape logged, Slauson et al. 2007), our findings for high road
use by potential marten predators suggests that the potentially negative effects of roads should be
further investigated.
Reponses of Mesocarnivores to Current and Future Restoration
Overall our findings illustrate the multiscale responses of meoscarnivores to changes in forest
structure and composition from logging in the northern redwood region. Second growth
landscapes are used by larger-bodied mesocarnivores that are not typically detected in
contiguous patches of old growth coastal forest and show a negative association with one of the
main structural components of these forests, dense shrub cover. Furthermore, forest roads are
frequently used by generalist mesocarnivores and may be facilitating their increased distribution
in coastal forests. Both the loss of old growth forest habitat, old forest elements (large live and
dead trees), and the increased abundance of larger-bodied mesocarnivores have likely
contributed to the decline and lack of recovery of the Humboldt marten. However, restoration
actions, road removal and thinning, are likely to benefit martens by returning larger portions of
the landscape to conditions favoring their size and foraging adaptations and reducing larger-
bodied generalists (e.g., gray foxes, bobcats) that can kill them.
The challenges for restoration are great in the Redwood National and State Park complex.
Collectively, the complex includes >24,000 hectacres of logged second growth forest and
includes >500 km of logging roads. To date, road removal has dominated most of the restoration
14
actions, with >300 and >50 km of roads removed by the national and state parks, respectively.
Thinning has just begun on a large scale, with ~4% and ~6% of second growth stands treated by
2010 in the National and State Parks, respectively.
To be of the greatest benefit, restoration actions should select combinations of roads for
removal and stands for restorative thinning that will: (1) buffer existing old growth patches and
(2) connect adjacent areas of old growth. For example, Redwood National Park‟s restoration
plan for the second growth and road removal in the Lost Man Creek watershed achieves the first
of these criteria by explicitly targeting restoration of stands and removal of roads in proximity to
the remaining old growth there. Although we recognize there are multiple objectives for road
removal and restorative thinning, we have yet to see the considerations of large-scale
connectivity guide the selection of restoration areas in any redwood national or state park units
(RNP 2008, CSP 2010).
How roads are removed will likely have alternative effects on mesocarnivores. Roads that are
completely removed using landform restoration will reduce access by generalist mesocarnivores
but road decommissioning (blocking vehicle passage but leaving the road) will offer no change
in access for use by mesocarnivores until the road bed is revegetated. The cost of landform
restoration will likely limit the annual amount of road removal using this technique, which
emphasizes the need to strategically select sections of roads that can buffer existing old growth
patches and help restore connectivity between old growth patches.
One of the most critical elements lacking in the second growth landscapes are large woody
structures with suitable resting and denning locations for martens and other wildlife. Because
natural regeneration of these structures may take >200 years, alternatives will be required to
provide adequate numbers of suitable structures until natural processes of recruitment are
restored. Much like the need to return large wood to creeks to restore critical habitat for
salmonids, large woody structures or surrogate structures (e.g., marten den boxes) need to be
strategically returned to the landscape to accelerate the restoration of one of the most threatened
mammals in the redwood region. Given that the only remnant Humboldt marten population is
estimated to number <100 individuals (Slauson et al. 2009b) and individuals have been detected
<1 km from the Mill Creek acquisition, short term measures that can increase habitat suitability
may be critical to ensuring persistence and recovery of this population.
The scale at which restoration is needed and is being undertaken in the northern redwood
region is both daunting and globally unprecedented. While the research we have described
herein contributes to the beginning of our understanding of mesocarnivore response to
restoration, we will need to continue to add to this knowledge base to refine management and
restoration actions.
ACKNOWLEDGEMENTS
We would like to thank Save-the-Redwoods League for funding. We thank Jay Harris and
Amber Transou of the Northcoast Redwoods District of California State Parks and Kristin
Schmidt of Redwood National Park for logistical support while working on the respective Park
15
units. We would like to thank Jan Werren and Ric Schlexer, U.S. Forest Service, Pacific
Southwest Research Station, for GIS and logistical support.
LITERATURE CITED
Anderson, D. R., K. P. Burnham, W. L. Thompson. 2000. Null hypothesis testing: problems,
prevalence, and an alternative. Journal of Wildlife Management 64: 912-923.
Bate, L. J., E. O. Garton, and M. J. Wisdom. 1999. Estimating snag and large tree densities and
distributions on a landscape for wildlife management. Gen. Tech. Rep. PNW-GTR-425.
Portland, OR: U.S. Department of Agriculture, Forest Service, Pacific Northwest
Research Station. 76 p.
Burnham, K. P., and D. R. Anderson. 2002. Model selection and multi-model inference: A
practical information-theoretic approach. 2nd Edition. Springer-Verlag, New York.
California State Parks. 2010. Del Norte Coast Redwood State Park. Preliminary General Plan
Amendment / Final Environmental Impact Report : Mill Creek Acquisition. 384 p.
Carroll, C., W. J. Zielinski, and R. F. Noss. 1999. Using presence-absence data to build and test
spatial habitat models for the fisher in the Klamath Region, U.S.A. Conservation Biology
13(6): 1344-1359.
Fryxell, E. 1982. Urocyon cinereoargenteus. Mammalian Species. 189: 1-8.
Grinnell, J., and J. S. Dixon. 1926. Two new races of the pine marten from the Pacific Coast of
North America. Zoology 21(16):411-417.
Grinnell, J., J. S. Dixon, and J.M. Linsdale. 1937. Fur-bearing mammals of California. Vol. 1.
University of California Press, Berkeley, CA. 375 pp.
Hagmeier E (1961) Variation and relationships in North American marten. Can Field Nat
75:122–137
Higley, M. and S. Matthews. 2009. Fisher habitat use and population monitoring on the Hoopa
Valley Reservation, California. Hoopa Tribal Forestry, Final Report. 82 p.
Keyes, C. R. 2005. Ecological basis for old growth redwood forest restoration: 25 year
assessment of redwood ecosystem response to restorative thinning. A report to Save-the-
Redwoods League. Humboldt State University, Arcata, CA.
Klug, R. R. 1997. Occurrence of Pacific fisher (Martes pennanti pacifica) in the Redwood Zone
of northern California and the habitat attributes associated with their detections. M.S.
thesis. Humboldt State University, Arcata, CA. 50 p.
16
Krohn, W. B., C. Hoving, D. Harrison, D. Phillips, and H. Frost. 2004. Martes foot-loading
and snowfall patterns in eastern North America: Implications to broad-scale
distributions and interactions of mesocarnivores. Pages 115-131 in: D. Harrison, A.
Fuller, and G. Proulx (eds). Martens and fishers (Martes) in human altered
landscapes: an international perspective. Springer Science+Buisness Media, Inc, New
York, NY, USA.
Lee, T. D. 1988. Patterns of fruit and seed production. Plant Reproductive Ecology. Patter-ns
and Strategies J. Lovett Doust & L. Lovett Doust (eds), pp. 179-201. Oxford University
Press, New York, NY.
Manly, B. F. J., L. L. McDonald, and D. L. Thomas. 2002. Resource selection by
animals: statistical design and analysis for field studies. Second edition. Kluwer
Academic Publishers, New York, New York, USA.
Perry, R.W., Thill, R.E., Peitz, D.G., Tappe, P.A., 1999. Effects of different silvicultural systems
on initial soft mast production. Wildlife Society Bulletin 27: 915–923.
Raphael, M. G. and L. C. Jones. 1997. Characteristics of resting and denning sites of American
martens in Central Oregon and Western Washington. In: Proulx, G., H. N. Bryant, and P.
M. Woodard (eds). Martes: taxonomy, ecology, techniques, and management.
Provincial Museum of Alberta, Edmington, Alberta, Canada. Pages 146-165.
Schumacher, T. V. 1999. A multi-scale analysis of habitat selection at den and resting sites of
American martens in southeast Alaska. M.S. Thesis, University of Wyoming, Laramie.
Slauson, K. M. 2003. Habitat Selection by American Martens (Martes americana) in Coastal
Northwestern California. M. S. Thesis, Oregon State University, Corvallis, USA. 120 p.
Slauson, K. M. and W. J. Zielinski. 2003. Distribution and habitat associations of the
Humboldt marten and Pacific Fisher in Redwood National and State Parks.
Final Report. USDAFS, Pacific Southwest Research Station, Arcata, CA.
29 p.
Slauson, K. M. and W. J. Zielinski. 2004. Conservation status of American martens and fishers
in the Klamath-Siskiyou bioregion. In: K. Merganther, J. Williams, and E. Jules, eds.
Proceedings of the 2nd Conference on Klamath-Siskiyou Ecology, Cave Junction, OR,
USA. 29-31 May, 2003. Siskiyou Field Institute, Cave Junction, Oregon.
Slauson, K. M., W. J. Zielinski, and J. P. Hayes. 2007. Habitat selection by American
martens in coastal California. Journal of Wildlife Management 71: 458-468.
Slauson, K. M. and W. J. Zielinski. 2009a. Characteristics of summer/fall resting
structures used by American martens in coastal northwestern California.
Northwest Science 83: 35-45.
17
Slauson, K. M. and W. J. Zielinski. 2009b. Status and estimated size of the only remnant
population of the Humboldt subspecies of the American marten (Martes americana
humboldtensis) in northwestern California. USDAFS, Pacific Southwest Research
Station, Arcata, CA. 28 p.
Thornburg, D. A., R. F. Noss, D. P. Angelides, C. M. Olson, F. Euphrat, and H. W.
Welsh. 2000. Managing redwoods. In: R. F. Noss (ed.). The Redwood Forest:
History, ecology, and conservation of the Coast Redwoods. Island Press, Covelo, CA.
339 pp.
U. S. Department of Interior. 2008. Redwood National Park South Fork Lost Man Creek
Second Growth Forest Restoration Environmental Assessment. Redwood National
Park. 107 p.
Veirs, S.D. 1986. Redwood second-growth forest stand rehabilitation study, Redwood
National Park: Evaluation of 1978-79 thinning experiments. Redwood National Park,
Arcata, CA.
Vincent Wildlife Trust. 2010. Prospects for pine martens: constructing, erecting, and
monitoring pine marten den boxes. Accessed 30 October, 2010.
http://www.pinemarten.info/downloads/Constructing,%20erecting%20and%20monitorin
g%20den%20boxes.pdf
Zielinski, W.J., and R.T. Golightly. 1996. The status of marten in redwoods: is
the Humboldt marten extinct? Pages 115-119 in: LeBlanc, John, ed.,
Conference on Coast Redwood Forest Ecology and Management, 1996
June 18-20, Humboldt State University, Arcata, CA.
18
Table 1. Definitions and abbreviations for variables measured at 1-km and 2.5-km radius circles
around each sample unit used to model mesocarnivore distributions from surveys conducted
from 1996-2002 in coastal northwestern California, USA.
______________________________________________________________
Variable Abbreviation
______________________________________________________________
%Landscape Old Foresta OG_Pland
Number of Habitat Patches of Old Foresta OG_NP
Number of Distinct Core Areas of Old Foresta OG_NDCA
% Landscape with Dense Shrub Covera S_Pland
% Landscape of Young Forest Habitata Y_Pland
% Landscape with Roadsb R_Pland
% Landscape Young Forest and Roadsa YR_Pland
_______________________________________________________________
a Measured from the gradient nearest neighbor (GNN) vegetation coverage produced by the
Landscape Ecology, Modeling, Mapping, and Analysis program
(http://www.fsl.orst.edu/lemma/splash.php).
b Measured from the transportation layer provided by the Six Rivers National Forest.
19
Table 2. Mean shrub cover measured in forest stands from 2002-2010 in coastal northwestern
California, USA. Stands were sampled in 3 groups, Reference stands that included old growth
redwood stands and old growth stands occupied by martens, and stands restoratively thinned
either 15-30 years ago (Old Thin)or 207 years ago (New Thin). The 2 thinned stand groups also
included measurements of paired control stands, where thinning was not conducted.
.
Group Treatment (n) Mean Shrub 95% Confidence
Cover (SD) Interval
____________________________________________________________________________
Reference
Marten Occupied (6) 71.7% (10) 63 – 80%
Marten Occupied (26)* 74.0% (20) 66 -- 82%
Old Growth Rwd (69)** 86.3% (15) 83 -- 90%
Old Thin
Total (6) 46.8% (25) 22 – 72%
Veirs (3) 66.5% (12)
Whiskey 40 (3) 33.0% (19)
Control (3) 6.2% (4) 2 – 10%
New Thin
Total (13) 31.9% (19) 20 – 44%
Control (8) 11.8% (10) 4 – 44%
______________________________________________________________________________
*From Slauson et al. (2007).
** From Slauson and Zielinski (2003).
20
Table 3. Rank-order shrub species dominance for 6 most dominant shrub, shrub form tree, and fern species determined from transect
sampling from 2002-2010 in coastal northwestern California, USA. For each species the sum of the ranks and the number of stands it
was present, in parenthesis, are presented. The 3 most dominant shrub species in each stand type are highlighted.
____________________________________________________________________________________________________________
Thinned 15-30 Thinned 2-7 Marten Occupied Old Growth
Plant Species Thin (n=5) Control (n=3) Thin (n=13) Control (n=7) Stands Slauson Rwd (n=90)*
(n=6) 2003 (n =26)**
___________________________________________________________________________________________________________
Evergreen huckleberry 82 (4) 21 (3) 86 (9) 53 (4) 0 (0) 35 (14) 143 (NA)
Salal 47 (5) 6 (2) 55 (6) 3 (1) 97 (2) 27 (10) 69 (NA)
Rhododendron 21 (4) 0 (0) 22 (5) 10 (2) 76 (2) 19 (9) 85 (NA)
Tanoak 6 (1) 0 (0) 19 (3) 2 (1) 55 (4) 14 (6) NA
Huckleberry oak 0 (0) 0 (0) 0 (0) 0 (0) 61 (2) 17 (6) NA
Sword Fern 17 (2) 3 (1) 51 (7) 24 (4) 5 (1) NA 189 (NA)
____________________________________________________________________________________________________________
* Results from Slauson and Zielinski (2003).
**Results from (Slauson 2003) are for rank-order estimates from single plots in each stand.
21
Table 4. Mean densities (#/ha) of potential resting and denning structures in recently thinned,
not recently thinned, and old growth stands in coastal northwestern California. Each structure
must have been >60cm in maximum diameter and have a detectable feature (e.g., chamber,
cavity, platform) that a marten could use for resting or denning.
Structure Type
#/ha (SE)
__________________________________________________________
Stand Type (n) Logs Snags Live Trees
____________________________________________________________________________
Thinned <9 Years Ago (17) 0.98 (0.48)* 0.07 (0.07)* 0.00*
Thinned 15-30 Years Ago (9) 6.02 (1.82) 0.18 (0.19)* 0.00*
Marten Occupied (7) 5.14 (1.72) 2.10 (0.54) 0.74 (1.13)
_____________________________________________________________________________
*Significantly lower t-test results for comparisons of means to marten occupied stands at
P <0.05.
22
Table 5. 95% confidence set of models predicting the probability of each species occurrence
using regional survey data from coastal northwestern California and southwestern Oregon, USA.
Each model is ranked according to ΔAICc value. OG = old growth, S = Shrub cover >50%, R =
roads, Y = young forest, Pland= percent of the landscape, LPI = large patch index, NP = number
of patches, Fisher = presence of fisher at the sample unit.
Model Parameters Model Ranking
___________________________________________________________________
Species Model Variables ΔAICc wia Kb
______________#____________________________________________________
Marten 1 OG_Pland*S_Pland R_Pland 0.00 0.33 5
2 S_Pland R_Pland 0.59 0.32 3
3 OG_Pland*S_Pland 1.95 0.13 4
4 OG_LPI*S_Pland 2.86 0.08 4
5 S_Pland 3.27 0.07 2
Fisher 1 S_Pland OG_Pland Y_Pland 0.00 0.80 4
2 S_Pland OG_Pland 3.85 0.11 3
3 S_Pland *OG_Pland 5.46 0.05 5
Gray Fox 1 S_Pland OG_NP 0.00 0.26 5
2 S_Pland 0.92 0.16 4
3 S_Pland Y_Pland 1.32 0.13 5
4 S_Pland Fisher 1.52 0.12 8
5 S_Pland Y_Pland R_Pland 2.23 0.09 4
6 S_Pland OG_Pland 2.92 0.06 5
7 S_Pland R_Pland 3.32 0.05 8
8 S_Pland*OG_Pland 4.32 0.03 8
____________________________________________________________________
a wi = Akaike weight, corrected for small sample sizes.
b K = number of parameters in each model.
23
Table 6. Carnivore species detected at remote camera sample units surveyed from May-December 2009 in coastal northwestern
California. Detection probability represents the probability that a species would be detected using the survey protocol, if in fact it was
present.
.
Number of Observed Detections
___________________________________________
Species Detection Probability Road (% total det.) Forest Interior (% of total det.)
_________________________________________________________________________________________________
Habitat Generalists
Gray Fox 100% 11 (69%) 5 (31%)
Bobcat 95% 5 (100%) 0 (0%)
Mountain lion 37% 7 (88%) 1 (12%)
Totals: Habitat Generalists 23 (79%) 6 (21%)
Habitat Specialists
American marten 99% 2 (25%) 6 (75%)
Fisher 85% 3 (30%) 7 (70%)
Raccoon 94% 1 (14%) 6 (86%)
Mink 54% 0 (0%) 3 (100%)
Totals: Habitat Specialists 6 (21%) 22 (79%)
__________________________________________________________________________________________________
24
Table 7. Detection rates, calculated as # detections/10 survey days, for carnivore species at
remote camera sample units from May-December 2009 in coastal northwestern California.
Bolded detection rates are significantly different paired t-test results for roads versus forest
interior (Interior) for each species at P ≤ 0.05.
Survey Treatment
_____________________________________________
Species N No Attractants Urine Gusto+Chicken
__________________________________________________________________________
Gray Fox 10 Roads 0.88 (0.91) 0.45 (0.55) 1.36 (1.26)
Interior 0.12 (0.29) 0.84 (0.18) 0.14 (0.45)
Bobcat 5 Roads 0.23 (0.15) 0.77 (0.89) 0.29 (0.64)
Interior 0.0 (0.0) 0.05 (0.11) 0.0 (0.0)
Mountain lion 6 Roads 0.21 (0.15) 0.11 (0.20) 0.08 (0.22)
Interior 0.0 (0.0) 0.0 (0.0) 0.08 (0.22)
Habitat Generalists 21 Roads 0.46 (0.66) 0.39 (0.59) 0.65 (1.00)
Interior 0.04 (0.19) 0.09 (0.20) 0.08(0.01)
American marten 7 Roads 0.05 (0.12) 0.0 (0.0) 0.0 (0.0)
Interior 0.05 (0.13) 0.73 (0.62) 0.86 (0.65)
Fisher 9 Roads 0.0 (0.0) 0.0 (0.0) 0.23 (0.36)
Interior 0.11 (0.16) 0.0 (0.0) 1.29 (1.94)
Raccoon 7 Roads 0.04 (0.11) 0.0 (0.0) 0.0 (0.0)
Interior 0.33 (0.5) 0.33 (0.28) 0.75 (0.19)
Habitat Specialists 23 Roads 0.03 (0.09) 0.0 (0.0) 0.09 (0.25)
Interior 0.16 (0.31) 0.33 (0.47) 0.79 (1.33)
__________________________________________________________________________
25
Figure 1. Box and whisker plots of percent mean shrub cover in forest stands in coastal
northwestern California. Heavy black lines indicates the median values, boxes indicates upper
and lower quartiles, lines represent maximum and minimum values, and open circles outlier
values.
26
Figure 2. Box and whisker plots for the probability of stand scale occupancy for American
martens in coastal northwestern California. Heavy black lines indicates the median values,
boxes indicates upper and lower quartiles, lines represent maximum and minimum values, and
open circles outlier values.
27
Figure 3. Scale-sensitivity for key habitat metrics, demonstrating the higher significance of the
occupancy effect (positive for marten, negative to fisher and gray fox) when shrub cover is
measured at spatial scales closer to each species‟ average home range size than at scales larger
(marten and gray fox) or smaller (fisher). Y-axis equals the p-value for logistic regression
models using the shrub cover variable measuerd at 1-km and 2.5 km radii around sample units.
0
0.05
0.1
0.15
0.2
0.25
1-km
2.5-km
p-value
Radius Around Sample Unit
Fisher
Marten
Gray Fox
28
Figure 4. Distribution of remote camera sample units surveyed in coastal northwestern
California in 2009.
29
Figure 5. Percent of mesocarnivore station detections on paired road versus forest interior
locations from May-December 2009 in coastal northwestern California.
Figure 6. American marten detected by a remote camera in Prarie Creek Redwoods State Park,
July 2009.
n =23
n = 6
n =6
n = 22
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
Habitat Generalists
Habitat Specialists
Percent of Station Detections
Road
Forest Interior
