Content uploaded by Holly Elizabeth Loraine Gamblin
Author content
All content in this area was uploaded by Holly Elizabeth Loraine Gamblin on Jul 29, 2024
Content may be subject to copyright.
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
1
Holly E. L. Gamblin1, Department of Wildlife, California State Polytechnic University
1
Humboldt, 1 Harpst Street, Arcata, California 95521, USA
2
Keith M. Slauson2, USDA Forest Service, Pacific Southwest Research Station, 1700 Bayview
3
Drive, Arcata, California 95521, USA
4
and
5
Micaela Szykman Gunther, Department of Wildlife, California State Polytechnic University
6
Humboldt, 1 Harpst Street, Arcata, California 95521, USA
7
8
Habitat Use and Distribution of a Recently Discovered Population of Humboldt Martens
9
10
Running footer: Humboldt Marten Habitat Use
11
12
2 tables, 3 figures
13
14
1Author to whom correspondence should be addressed. Email: gamblinh@myumanitoba.ca
15
1Current Address: University of Manitoba, 212 Biological Sciences Building, 50 Sifton Rd,
16
Winnipeg, Manitoba Canada R3T 2N2
17
2California Department of Parks and Recreation, North Coast Redwoods District, 3401 Fort
18
Avenue, Eureka, California, 95503, USA
19
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
2
Abstract
20
The Humboldt marten (Martes caurina humboldtensis) has declined from over 95 % of its
21
historic range and currently occurs in just four extant population areas (EPAs). Prior to their
22
listing under the Endangered Species Act, a conservation strategy was developed to identify key
23
conservation needs for this species. This assessment identified an area near the California–
24
Oregon (CA–OR) border as the second EPA in California, yet little was known about the overall
25
distribution or habitat used by this population. This prompted our investigation to provide the
26
first systematic survey of the CA–OR EPA and to assess habitat use under an occupancy
27
modeling framework. Between 2017–2018 we surveyed 51 survey units in and around the EPA
28
and detected martens at 20 (39.2 %). We found that occupancy was most influenced by the
29
spatial scale-specific amount of low-elevation late-seral old-growth forest habitat, riparian
30
habitat, and mid-seral forest habitat. Occupancy by marten was greatest in low-elevation (< 800
31
m) habitat and was positively associated with late-seral forest habitat at the 1,170-m home range
32
scale (Odds Ratio [OR] = 35.31, 95 % CI = 1.30–958.07), riparian habitat at the 1,170-m home
33
range scale (OR = 3.20, 95 % CI = 1.01–10.1), and increased amounts of mid-seral forest habitat
34
at the 50-m microhabitat scale (OR = 1.28, 95 % CI = 0.95–1.73). Our findings identified habitat
35
types important for explaining the distribution of this understudied population, addressing two of
36
the highest priority research needs identified in the Humboldt marten conservation strategy.
37
Key Points
38
• Our surveys detected Humboldt martens in areas beyond the previously mapped
39
California–Oregon extant population area, expanding the known distribution.
40
• Martens were detected at 20 of 51 (39 %) survey units in and around the California–
41
Oregon extant population area, suggesting a patchy distribution.
42
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
3
• Occupancy by marten was influenced by low-elevation late-seral forest and riparian
43
habitat (home range scale), as well as mid-seral forest habitat (microscale).
44
Keywords: coastal marten, late-seral forest, Martes caurina humboldtensis, mesocarnivore,
45
Pacific marten
46
Introduction
47
Over the last few hundred years, the global loss of biodiversity has occurred at an alarming rate
48
(Estes et al. 2011, Segan et al. 2016), and this trend is particularly profound for rare species
49
(Dirzo and Raven 2003). Rare species are inherently vulnerable to population declines due to
50
their limited distributions and low abundances (Drever et al. 2012). Furthermore, the difficulties
51
associated with studying elusive species can pose challenges in developing timely conservation
52
initiatives (Martin et al. 2022). Understanding habitat use of at-risk species is an important first
53
step in identifying key areas for management and recovery (Krausman 1999), yet lack of
54
sufficient data is a common challenge in modeling habitat use for rare species (Hamilton et al.
55
2015, Todman et al. 2023).
56
The Humboldt marten (Martes caurina humboldtensis), also known as the coastal marten,
57
is a subspecies of the Pacific marten (M. caurina) and is an example of a rare and elusive species
58
for which knowledge of key population dynamics is lacking (Martin et al. 2022). The Humboldt
59
marten is a medium-sized forest carnivore that historically occurred throughout the coastal
60
forests of northwestern California and Oregon and has declined from > 95 % of its historic range
61
(Slauson et al. 2018, Moriarty et al. 2021). Signs of decline began to appear in the early 1900s
62
due to the unregulated and excessive trapping for their fur (Grinnell et al 1937), while continued
63
declines and lack of recovery following cessation of trapping has been attributed to extensive
64
timber harvesting that followed throughout the latter 1900s (USFWS 2015, Slauson et al. 2018).
65
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
4
After 50 years without verifiable detections, the Humboldt marten was considered extirpated
66
throughout its California range (Zielinski and Golightly 1996). However, in 1996 the subspecies
67
was rediscovered in remote portions of its historical range in northwestern California (Zielinski
68
et al. 2001).
69
Contemporary surveys conducted throughout the historical range of the Humboldt marten
70
in California and Oregon have identified four extant population areas (EPAs): two disjunct EPAs
71
have been identified in Oregon along the central and southern coast range, and two disjunct
72
EPAs in California, one in the northern coast range and the other farther inland near the
73
California–Oregon border (CA–OR EPA; Slauson et al. 2018). Despite extensive survey efforts,
74
there is still uncertainty about the exact distributions, population sizes, and habitat use of the few
75
populations of Humboldt martens that remain (Moriarty et al. 2016, Slauson et al. 2019). In
76
2009, the northern coastal California EPA was estimated to contain fewer than 100 individuals
77
(Slauson et al. 2009), and in 2018 the population size of the central coastal Oregon EPA was
78
estimated at 71 individuals (95 % CI = 41–87; Linnell et al. 2018). Concerns over the persistence
79
of this subspecies, known from only a few small and geographically isolated populations,
80
prompted the listing of Humboldt martens as Endangered under the California Endangered
81
Species Act in 2018 (CDFW 2019) and Threatened under the federal Endangered Species Act in
82
2020 (83 FR 50574).
83
With Humboldt martens occupying < 5 % of their historic range, it is critical to
84
understand the habitat conditions important for supporting the few existing populations.
85
Humboldt martens are considered habitat specialists and like other carnivores have large home
86
ranges relative to their small body size (Lindstedt et al. 1986). Consistent with marten species
87
across much of their North American range, Humboldt martens are known to occur in
88
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
5
structurally complex, late-seral and old-growth forests (Andruskiw et al. 2008, Kirk and
89
Zielinski 2009, Thompson et al. 2012). This habitat type contains large trees and snags for
90
resting and denning, prey resources, tree canopy and shrub cover for protection from aerial
91
predators, and downed woody debris near the forest floor that helps to improve hunting success
92
(Andruskiw et al. 2008, Kirk and Zielinski 2009, Thompson et al. 2012).
93
Surveys of the northern coastal California EPA found that Humboldt martens were
94
primarily associated with late-seral forest habitats, but they have also been detected in two low
95
productivity forest habitat types: shore pine (Pinus contorta) dominated coastal forest habitat
96
found only on stabilized dunes (Linnell et al. 2018, Moriarty et al. 2019), and serpentine forest
97
habitat found only on ultramafic soils (Slauson et al. 2019). The central coastal Oregon EPA
98
persists entirely in young, coastal forest habitat (< 70 years old; Eriksson et al. 2019), and
99
detections in serpentine forest habitat have occurred in both the southern coastal Oregon EPA
100
and northern coastal California EPA (Moriarty et al. 2019, Slauson et al 2019). Collectively,
101
these two low productivity forest habitat types are endemic to their parent soil types and are
102
limited to < 8 % of the Humboldt marten’s historic range (Slauson et al. 2019). Martens can
103
persist in these two less productive habitat types, so long as key habitat types are available for
104
supporting resting, denning, and prey resources (Slauson et al. 2007, Moriarty et al. 2016,
105
Moriarty et al. 2021). However, martens do not occur in their structural analogs (i.e., forest
106
habitat with small diameter or young trees) in the productive forest habitats that comprise the
107
majority (> 90 %) of their historical range where they have been largely extirpated (Slauson et al.
108
2018). Dense shrub cover, typically dominated by ericaceous species, is the most consistent
109
habitat feature within the three distinct habitat types used by Humboldt martens (Slauson et al.
110
2018, Moriarty et al. 2019).
111
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
6
The occurrence of martens in these three distinct habitat types demonstrate the variation
112
in habitat use between the EPAs (Slauson et al. 2007, Eriksson et al. 2019, Moriarty et al. 2021).
113
This variation highlights the importance of using localized data to model habitat use that may be
114
particular to each remnant population. With only a handful of verified detections near the
115
California–Oregon EPA, little is known about the habitat types that are most important for
116
Humboldt martens in this population (Slauson et al. 2018). The first verified detection of a
117
marten in the CA–OR EPA occurred in 2011, with subsequent surveys between 2012–2014
118
detecting martens at five additional locations (Slauson et al. 2018). No formal assessments of the
119
distribution or habitat associations of martens in this EPA have been conducted to date, and these
120
assessments have been identified as high-priority information needs in the Humboldt marten
121
conservation strategy (Slauson et al. 2018).
122
Our primary objective was to conduct the first systematic survey of the CA–OR EPA and
123
provide a formal assessment of the habitat use and distribution of Humboldt martens in the least
124
studied population. This population-level assessment provides an important clarification of the
125
habitat types that are used by marten in the CA–OR EPA. Understanding habitat requirements
126
for species of conservation concern is essential for developing effective management and
127
conservation actions. Our study addresses one of the most important information needs identified
128
in the Humboldt marten conservation strategy (Slauson et al. 2018).
129
Methods
130
Study Area
131
The CA–OR EPA is located primarily on federal lands managed by the Six Rivers and
132
Siskiyou National Forests in northwestern California, just south of the Oregon border (-123° 42’
133
58” W, 41° 53’ 41” N, Figure 1). The study area encompassed approximately 406 km2 and
134
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
7
ranged from 27 to 48 km inland from the Pacific Ocean. The climate was characterized by warm,
135
dry summers and cool, wet winters (3–30 °C, Jimerson 1989), with annual averages for
136
precipitation of 237 cm and snowfall of 6 cm.
137
The study area was composed mainly of two habitat types known to be used by
138
Humboldt martens: serpentine forest habitats found on low-productivity ultramafic soils (17.0%)
139
and productive forest habitats found on high-productivity soil types (83.0%; Soil Survey Staff
140
2022). The productive forest habitats were dominated by Douglas fir (Pseudotsuga menziesii),
141
incense-cedar (Calocedrus decurrens), Port Orford-cedar (Chamaecyparis lawsoniana), red fir
142
(Abies magnifica), and white fir (A. grandis) plant associations (USFS 2018, CDFW 2021).
143
Hardwoods, such as tanoak (Notholithocarpus densiflora), Pacific madrone (Arbutus menziesii),
144
and canyon live oak (Quercus chrysolepsis) were also subdominant in the tree overstory.
145
Ericaceous shrubs, such as evergreen huckleberry (Vaccinium ovatum) and salal (Gaultheria
146
shallon), dominated the shrub layers of the productive forest habitats. Serpentine forest habitats
147
were dominated by Jeffrey pine (Pinus jeffreyi), knobcone pine (P. attenuata), and Douglas fir
148
plant associations. The dominant shrub species in serpentine habitats were huckleberry oak (Q.
149
vacciniifolia), manzanita (Arctostaphylos spp.), bush tanoak (N. d. echinoides), and California
150
red huckleberry (V. parvifolium).
151
The study area was characterized by a mixture of forest seral stages (LEMMA 2017). The
152
tree size class attribute data characterized seral stages based on quadratic mean diameter (QMD)
153
and canopy cover, with early-seral stages represented by size class 0–3, mid-seral stages by size
154
class 4, and late-seral stages by size class 5–6. Overall, early-seral stages (59.3%, size class 0–3)
155
included 6.6 % classified as unvegetated or the shrub/seedling stage (size class 0–1, QMD 0–2.4
156
cm and canopy cover < 10.0 %), 28.8 % in the sapling/pole stage (size class 2, QMD 2.5–24.9
157
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
8
cm and canopy cover 10.0–24.9 %), and 23.9 % in the small tree stage (size class 3, QMD 25.0–
158
37.4 cm and canopy cover 25.0–37.4 %). Mid-seral forest habitat in the medium tree stage (size
159
class 4, QMD 37.5–49.9 cm and canopy cover 37.5–49.9 %) composed 17.3 % of the study area,
160
and late-seral forest habitat in the large and giant tree stages (size class 5–6, QMD ≥ 50.0 cm and
161
canopy cover ≥ 50 %) composed 23.5 % of the study area (LEMMA 2017).
162
Detection Surveys
163
We used the Humboldt marten population monitoring protocol to survey for martens
164
(Slauson and Moriarty 2014). This survey protocol is based on a 2-km systematic grid that
165
covers the entire historical range. The 2-km distance between grid points is larger than the
166
average radius of home ranges for male martens elsewhere in California (Moriarty et al. 2021),
167
likely ensuring spatial independence from detecting the same individual at adjacent survey units.
168
The survey period occurred during the latter half of the denning period (May–mid-August;
169
Delheimer et al. 2021) to increase the likelihood of detecting resident adults rather than
170
dispersing juveniles (Slauson and Moriarty 2014, Zielinski et al. 2015). At each central grid
171
point, we established a two-station survey unit: one placed on the central grid point (station A)
172
and the second placed 500 m away in a random direction (station B). In 2017, one remote camera
173
station and one track plate station were deployed within each survey unit. We randomly assigned
174
either a track plate or remote camera to station A, and station B was assigned the alternative
175
detection device. The Humboldt marten surveying protocol recommends the use of both remote
176
cameras and track plates as both device types yield similar detection probabilities for martens
177
(Gompper et al. 2006, Slauson and Moriarty 2014). However, we used remote cameras at all
178
stations in 2018 due to the difficulties of deploying track plates in our study area and the
179
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
9
similarities in detection events observed between device types within the survey units deployed
180
in 2017.
181
At stations with remote cameras, we used passive infrared-triggered cameras (Command
182
Ops Pro; Browning Trail Cameras, Morgan, Utah) programmed to take 8-shot photo bursts once
183
triggered. Cameras were placed in metal security boxes to prevent damage from black bears
184
(Ursus americanus) and mounted to trees using lag bolts and straps. Bait was mounted < 0.6 m
185
from the ground on a tree < 10 m away from the camera. Track plate stations consisting of an
186
open-ended Coroplast cubby were placed alongside a stable structure (i.e., tree, stump, rocks)
187
with sooted metal plates inside and set with sticky contact paper near the far end. Surrounding
188
debris was placed along the sides and top to minimize movement, and bait was placed inside
189
near the far end. Each station included two chicken drumsticks on the camera bait tree or in the
190
back of the track plate and a sponge soaked in commercial trapping lure (Gusto; Minnesota
191
Trapline Products, Pennock, MN) to attract martens (Baldwin and Bender 2008, Moriarty et al.
192
2018). The trapping lure was hung approximately 2-m above the ground in the tree or shrub
193
nearest to the camera or track plate station. Once established, each station was deployed for a
194
minimum of 21 days and revisited approximately every 3–5 days to replace bait, refresh lure, and
195
retrieve photographs on SD cards from camera stations or tracks on contact paper from track
196
plate stations. All survey methods were approved by the Humboldt State University Institutional
197
Animal Care and Use Committee (protocol 16/17.W.05-A).
198
Occupancy Modeling Approach
199
We used occupancy modeling to account for imperfect detection and to model the
200
influences of habitat characteristics on the probability of occupancy by marten using our
201
detection/non-detection data (MacKenzie et al. 2002). To create detection histories for each
202
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
10
survey unit we first defined our survey occasion and then identified whether a marten was (1) or
203
was not (0) detected during each occasion. Survey occasions were defined by each of the 3- to 5-
204
day station check intervals, for a total of 5 survey occasions for each station. Detections were
205
combined for both track plates and cameras to create a single detection history for each survey
206
unit. Since there were no instances of a track plate detecting a marten when the associated
207
camera did not, the resulting detection histories for each survey unit remained unchanged when
208
both detection methods were used. A survey unit was considered occupied if a marten was
209
detected at either station using either method on at least one survey occasion.
210
We used a hierarchical modeling approach to develop and evaluate our candidate
211
occupancy models by first modeling the detection process (p), and then using the top detection
212
probability model in all occupancy models (Ψ). We used an information-theoretic approach to
213
develop a candidate model set (Burnham and Anderson 2002) by first developing a set of a priori
214
models representing alternative hypotheses of the most influential variables on the detection
215
process and marten occurrence. Alternative a priori hypotheses were developed using variables
216
known to influence habitat use in the three other Humboldt marten EPAs (Slauson et al. 2007,
217
2019, Moriarty et al. 2019, 2021), expert opinion, and hypotheses developed while conducting
218
fieldwork in the study area (Supplementary Material 1).
219
Candidate Variable Selection
220
Twenty-three variables (3 detection, 20 occupancy) were considered for inclusion when
221
developing candidate models (Supplementary Material 1). To evaluate the influence of survey-
222
specific variables on detection probability, we included the variables survey month (June or
223
July–early August) to account for temporal variation and total survey duration (number of days),
224
and station check interval length (number of days) to account for any effects of differences in
225
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
11
overall survey duration. To account for potential heterogeneity in detection probability over the
226
survey occasions we considered both constant detection probability (p.) and occasion-specific
227
(i.e., time-varying) detection probability (pt). For occasion-specific detection probability models,
228
we incorporated the variable check interval length (check) to capture the realized differences in
229
the number of days between when stations at each survey unit were checked.
230
We calculated a number of physical and biological variables to represent the habitat
231
characteristics of the survey units (Supplemental Material 1). We used topographic and
232
environmental variables from USGS, TIGER, and PRISM, including elevation, slope, road
233
density, stream density, and precipitation (Supplementary Material 1). We used forest structure
234
and composition variables from the Gradient Nearest Neighbor (GNN; LEMMA 2017)
235
vegetation coverage: tree size classes (small, medium, and large), canopy cover,
236
dominant/codominant conifer QMD, snag density, regionalized old-growth structure index
237
(OGSI), late-seral old-growth forest (LSOG), mean forest ages, hard masting trees, coarse woody
238
debris, and pine basal area (Supplementary Material 1). We generated shrub cover using data
239
published for available understory shrub species in the study area (Prevéy et al. 2022). We used
240
the USDA Gridded Soil Survey Geographic Database (Soil Survey Staff 2022) and groups
241
associated with gabbro and serpentinite soil types to identify serpentine habitat. All geographic
242
information system (GIS) calculations were conducted in ArcMap 10.3 (ESRI 2015).
243
We evaluated each variable for inclusion in the candidate model set. Variables were
244
excluded if there was incomplete GIS coverage in our study area, there was redundancy with
245
other variables, or if they were inapplicable to our dataset. This included slope, precipitation,
246
small tree size classes, road density, snag density, coarse woody debris, forest age, serpentine,
247
and pine basal area. Using this approach, we retained 14 (3 detection, 11 occupancy) variables
248
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
12
(Supplementary Material 3). We evaluated correlations between the variables retained using the
249
‘corrplot’ package in RStudio (RStudio Team 2022). If a variable pair was highly correlated
250
(correlation coefficient |r| ≥ 0.6), those variables were not included in the same model. We used
251
the ‘car’ package in RStudio to test for collinearity among covariates within a single model by
252
evaluating variance inflation factor (VIF) values (Zuur et al. 2010). Covariates with VIF ≥ 2
253
were removed from the model.
254
We evaluated the inclusion of sample units dominated by serpentine habitat prior to
255
developing candidate models. We conducted an exploratory principal components analysis to
256
compare survey units located in low productivity serpentine habitat (n = 9) to those located in
257
high productivity forest habitat (n = 42) (Supplementary Material 2 Table 1). We found that 19
258
of the 20 candidate variables were significantly different between these unique habitat types
259
(Supplementary Material 2 Table 2). There was a small number of serpentine-dominated survey
260
units, thus we excluded these units from the occupancy analysis (see Supplementary Material 2).
261
We reported the means and standard errors for the variables for survey units composed primarily
262
of serpentine versus productive forest habitats separately and combined (Supplementary Material
263
2 Table 3).
264
Spatial Scale Optimization of Habitat Variables
265
Martens are known to exhibit habitat selection at multiple spatial scales (Slauson et al.
266
2007, Kirk and Zielinski 2009, Thompson et al. 2012). We used bi-variate spatial scale
267
optimization to identify the optimal spatial scale for each variable, which is a technique used to
268
capture scale-dependent effects of habitat selection for martens (Shirk et al 2014, Tweedy et al.
269
2019, Martin et al. 2021, Moriarty et al. 2021). We created 6 spatial scales represented by buffers
270
around the central grid point for each survey unit with radii of 50, 270, 500, 750, 1,170, and
271
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
13
3,000 m. The smallest spatial scale (50-m) represented fine-scale microhabitat types measured at
272
the station level. The 270-m and 500-m scales represented within-home range (core area) scales
273
(Tweedy et al. 2019, Slauson et al. 2019). The 750-m and 1,170-m scales represented the average
274
female and male home range size, respectively (Moriarty et al. 2021). Our broadest spatial scale
275
(3,000 m) incorporated landscape-level effects that may influence where martens position their
276
home ranges within the surrounding area (Slauson et al. 2019). All occupancy models included
277
only each variable’s optimal spatial scale (Supplementary Material 1).
278
Candidate Models
279
We developed 11 candidate models for detection probability and 26 candidate models for
280
occupancy to evaluate both additive and interactive effects of variables on the probability of
281
occupancy (Supplementary Material 3). Due to the small sample size, we limited the total
282
number of variables included in any occupancy model to ≤ 3 variables to reduce the risk of
283
overfitting (Burnham and Anderson 2002) and maintain a ratio of ≥ 10 observations per
284
estimated parameter. Models were fit using Program MARK (White 2001) and evaluated using
285
Akaike’s Information Criterion adjusted for small sample size (AICc). Models with ΔAICc < 2
286
units were considered to have substantial support (Burnham and Anderson 2002).
287
To interpret the relationship between each variable and marten occurrence or detection,
288
we calculated odds ratios for variables present in models with substantial support. Odds ratios
289
were calculated by exponentiating the beta coefficients to estimate the influence of a one-unit
290
shift on the odds of occurrence or detection. For variables where a one-unit shift was not
291
biologically meaningful (i.e., 1 m elevation), we adjusted the odds ratio to reflect a scale
292
appropriate to the range of the data by multiplying the beta coefficient by a more meaningful
293
value (i.e., 100-m change in elevation) and exponentiating the adjusted beta coefficient. To
294
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
14
evaluate the relative strength of each variable in the model set, we also calculated adjusted
295
variable importance weights by taking the sum of AICc weights for models containing the
296
variable and adjusting it relative to the number of models the variable appears in (Burham and
297
Anderson 2002). We created boxplots to visually examine the univariate relationship between
298
the scale-optimized variables at detected and non-detected productive forest habitat survey units
299
(Supplementary Material 4).
300
Model Fit
301
Individual model fit was evaluated in program PRESENCE (MacKenzie and Hines 2006)
302
using a parametric bootstrap goodness of fit test with 10,000 simulations. The goodness of fit test
303
was used to generate an estimate of overdispersion, ĉ, to evaluate whether the top model
304
adequately fit the data. The general approach for this method is to run the test on the global
305
model. However, when the number of parameters in the global model is too large this results in
306
reduced precision in the estimate of ĉ, which can make it difficult to detect lack-of-fit. We used
307
the most parsimonious model to assess model fit, as that method is recommended when the
308
global model has a large number of parameters (MacKenzie and Bailey 2004). The goodness of
309
fit test generated an overdispersion estimate (ĉ) of 0.67 for the most parsimonious model, which
310
is generally considered to reflect underdispersion (Cooch and White 2001). When ĉ < 1 it is
311
recommended to set ĉ = 1 and proceed with model interpretation, and so we followed this
312
guideline before interpreting parameter estimates (Cooch and White 2001).
313
Results
314
Occupancy Surveys
315
During June–August in 2017 and 2018 we surveyed 51 survey units (21 in 2017, 30 in
316
2018). Survey durations differed somewhat from the protocol, averaging 20 days (range = 14–28
317
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
15
days) and 5 survey occasions (range = 4–7 occasions). Stations with fewer than the
318
recommended 21 days of survey effort occurred due to a nearby wildfire that required the
319
removal of stations for safety concerns, or due to camera malfunctions. Survey durations were
320
extended beyond 21 days at some stations to increase the chances of capturing hair samples for a
321
complementary study.
322
Overall, martens were detected at 20 of 51 survey units (39.2 % naïve occupancy; Figure
323
1). Martens were detected at a total of 24/102 stations across all two-station survey units, with
324
only four survey units (20 %) detecting martens at both stations and 16 survey units (80 %)
325
detecting martens at only one station. At stations where martens were detected, detections
326
occurred on an average of 2 survey occasions (range = 1–6 survey occasions). Mean latency to
327
the first detection was 6 days (range = 1–13 days). Martens were detected at four of the nine
328
survey units that were dominated by serpentine habitat (44.4 % naïve occupancy). Martens were
329
detected at 16 of 42 survey units dominated by productive forest habitat (38.1 % naïve
330
occupancy). Limited road access and hazardous terrain limited our ability to survey substantial
331
portions of the eastern part of the CA–OR EPA; therefore, approximately half of the survey units
332
occurred within the CA–OR EPA boundary and the rest were immediately adjacent on the
333
western edge of the boundary (Figure 1).
334
Occupancy Analysis
335
Of the 11 models for estimating detection probability, only one model showed substantial
336
support (ΔAICc < 2; Supplementary Material 3). The top model for detection probability
337
included survey month and total survey duration (Table 1), indicating these two variables
338
accounted for sources of heterogeneity realized in the detection process. This model was used as
339
the base detection probability model for all occupancy models.
340
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
16
The odds of detecting a marten during surveys conducted in July–early August were 281
341
% greater than in surveys conducted in June (OR = 3.81, 95% CI = 1.31–11.10), after accounting
342
for the effects of survey duration. The estimated detection probability for each survey occasion
343
was 0.23 in June (95 % CI = 0.12–0.38) and 0.53 in July–early August (95 % CI = 0.34–0.71).
344
For each additional survey day added to the mean survey duration of 20 days, the odds of
345
detection increased by 14 % (OR = 1.14, 95 % CI = 1.02–1.28), after accounting for the effects
346
of the month when the surveys were conducted (Table 1).
347
Of the 26 models evaluated for estimating the probability of occupancy by marten, three
348
models showed substantial support (ΔAICc < 2; Table 1). The top-ranked model included the
349
variables elevation (Elev) and mid-seral forest habitat (SC_Med). The second most competitive
350
model included the variables riparian habitat (Stream) and late-seral forest habitat (LSOG), and
351
the third most competitive model included an interaction between late-seral forest habitat and
352
elevation (Table 1).
353
The amount of late-seral forest habitat and elevation had the greatest importance weights
354
relative to occupancy of a survey unit by marten, followed by the amounts of mid-seral forest
355
and riparian habitat, respectively (Table 2). The mean amount of late-seral forest habitat
356
measured at the 1,170-m spatial scale was greater at survey units where martens were detected
357
(mean = 46.0 % [197.6 ha], SE = 1.8 %, range = 35.0–58.8 % [150.7–252.8 ha]) compared to
358
units where they were not detected (mean = 35.8 % [154.1 ha], SE = 2.5%, range = 16.3–66.0 %
359
[70.2–283.9 ha]; Table 2, Figure 2b). Using the beta estimates from the second-ranked model
360
(Table 1), for every 5 % (21.5 ha) increase in the amount of late-seral forest habitat at the 1,170-
361
m scale, the odds of marten occurrence was 35.3 times greater (OR = 35.3, 95 % CI = 1.3–958.0;
362
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
17
Figure 3d). Martens were not detected in high productivity survey units composed of < 35% (150
363
ha) late-seral forest habitat at the optimal spatial scale (1,170-m).
364
Martens were detected at survey units located at lower elevations (mean = 582-m, SE =
365
36.9-m, range = 362–858-m; survey units with no detection: mean = 964-m, SE = 67.3-m, range
366
458–1,655-m; Table 2). Using the beta coefficients from the best-supported model (Model 1,
367
Table 1), a 100-m increase in elevation was associated with a 67.1 % decrease in odds of
368
occurrence (OR = 0.33, 95 % CI = 0.13–0.81, Figure 3a). The influence of elevation and the
369
amount of late-seral forest habitat on occupancy by marten appeared to be interactive as one of
370
the highly competitive models included their interaction term (Model 3, Table 1). Most marten
371
detections occurred in survey units with greater amounts of late-seral forest habitat located at the
372
lowest elevations (Figure 2b). There was a 69.4 % decrease in odds of occurrence of marten for
373
every 100-m increase in elevation (OR = 0.301, 95 % CI = 0.207–0.404, Figure 3e) when using
374
the beta coefficients from the interactive model (Model 3, Table 1) and modeling the interacting
375
variable at its mean value. Similarly, using the beta coefficients from the interactive model, for
376
every 5 % (21.5 ha) increase in the amount of late-seral forest habitat at the 1,170-m scale, the
377
odds of marten occurrence were 198 % greater (OR = 2.98, 95 % CI = 2.88–3.08, Figure 3f).
378
The mean amount of mid-seral forest habitat measured at the 50-m spatial scale was
379
greater at survey units where martens were detected (mean = 17.0 % [0.14 ha], SE = 5.7 %,
380
range = 0.0–87.5% [0.0–0.69 ha]) compared to survey units where they were not detected (mean
381
= 13.2 % [0.10 ha], SE = 3.4%, range = 0.0–62.5 % [0.0–0.49 ha]; Table 2, Figure 2a). Using the
382
beta coefficients from the best-supported model (Model 1, Table 1), a 5 % (0.04 ha) increase in
383
mid-seral forest habitat at the 50-m spatial scale was associated with a 28.4 % increase in odds of
384
occurrence (OR = 1.28, 95 % CI = 0.95–1.73; Figure 3b).
385
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
18
Riparian habitat at the 1,170-m spatial scale was more abundant at survey units where
386
martens were detected (mean = 1.55 km/km2, SE = 0.09 km/km2, range = 0.75–1.96 km/km2)
387
compared to survey units where they were not detected (mean = 1.17 km/km2, SE = 0.09
388
km/km2, range = 0.16–2.06 km/km2; Table 2, Figure 2c). Using the beta coefficients from the
389
second best-supported model (Model 2, Table 1), every 100 m/km2 increase in the amount of
390
riparian habitat resulted in the odds of marten occurrence increasing by 220 % (OR = 3.20, 95 %
391
CI = 1.01–10.1, Figure 3c). No martens were detected in high productivity survey units
392
composed of < 0.75 km/km2 riparian habitat at the optimal spatial scale (1,170-m).
393
Discussion
394
This study provides the first systematic survey of the CA–OR EPA and addresses two of
395
the key information needs identified in the Humboldt marten conservation strategy: 1) to
396
determine the distribution of martens in the CA–OR EPA, and 2) to identify habitat types that
397
most influence the distribution of marten in this area. Martens were detected both in and adjacent
398
to the previously mapped EPA boundary, suggesting the population was distributed more
399
broadly than initially predicted and reported in the Humboldt marten conservation strategy
400
(Slauson et al. 2019). We suspect that the distribution of this population may exist most
401
significantly to the south, east, and southwest of the area we surveyed, based on the presence of
402
similar habitat conditions to where most martens were detected during our efforts. Overall,
403
occupancy of habitat by marten was most influenced by productive forest habitats located at
404
lower elevations, with greater amounts of late-seral forest and riparian habitat at the home range
405
scale (1,170-m) and greater amounts of mid-seral forest habitat at the microscale (50-m).
406
The amount of late-seral forest habitat at the home range scale and elevation collectively
407
had the greatest influence on the occupancy of productive forest by Humboldt marten. The
408
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
19
importance of late-seral forest for this population was consistent with habitat selection by
409
martens in the larger California population of Humboldt martens (Slauson et al. 2007) and
410
elsewhere for Pacific martens (Buskirk and Ruggiero 1994, Kirk and Zielinski 2009, Delheimer
411
et al. 2019). Humboldt martens have been found to occur at all elevations present within their
412
historical range, from sea level to approximately 1,500-m (Slauson et al. 2018), yet martens in
413
the CA–OR EPA primarily occupied low-elevation areas. However, the CA–OR EPA is located
414
further from the coast than most of the northern coastal California EPA and the two Oregon
415
EPAs, and it occurs in a more xeric climate than the other EPAs. The CA–OR EPA is one of the
416
most inland locations where Humboldt martens have been found within their historic range, and
417
these low-elevation (< 800-m) forest habitats may provide mesic microclimatic conditions that
418
support more productive habitat for this more inland EPA.
419
The amount of mid-seral forest habitat and riparian habitat were present in the top two
420
occupancy models, suggesting that occupancy of lower elevation sites by marten may be
421
influenced by more productive habitat. Similar to the two Oregon EPAs (Eriksson et al. 2019,
422
Moriarty et al. 2021), we found that Humboldt martens in the CA–OR EPA used areas associated
423
with greater amounts of mid-seral forest habitat. However, the influence of mid-seral forest was
424
only significant at the microscale (50-m) which represented < 1 % of a typical marten home
425
range. With such a small amount of habitat represented by the 50-m scale, this association may
426
reflect micro-habitat use rather than the influence of mid-seral forest on home range occupancy
427
in the CA–OR EPA.
428
We used stream density as an indicator of the amount of riparian habitat, as riparian
429
zones are known to support increased vegetation productivity and truffle production leading to
430
higher densities of prey (Doyle 1990, Waters et al. 2001). Riparian areas are known to be
431
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
20
important foraging areas for martens (Zielinski 2014), and these areas provide mesic
432
microenvironments for thermoregulation that can be especially important during the warmest
433
periods of the year. Riparian habitat has also been shown to be positively associated with
434
Humboldt marten occurrence at the core area scale (500-m radius) in broader habitat modeling
435
efforts (Slauson et al. 2019), although its influence was much less than the amount of late-seral
436
forest habitat in widespread productive forest habitats and the amount of serpentine habitat in the
437
limited distribution of low productivity habitats. The importance of riparian habitat may increase
438
with distance from the coast or other dominant orographic features, such as major river valleys,
439
as key habitat elements for Humboldt martens (e.g., dense, spatially extensive ericaceous shrub
440
cover) are influenced by factors such as moisture and summer fog, which are less prevalent
441
further inland.
442
Martens select resources at multiple spatial scales and therefore habitat models
443
accounting for this scale-dependency can provide stronger relationships between resources and
444
animal occurrences than single-scale models (Shirk et al. 2012). We tested a range of spatial
445
scales (n = 6, 50–3,000-m) that were applied in other analyses of habitat use by Humboldt
446
marten (Slauson et al. 2019, Moriarty et al. 2021). However, the use of the smaller scales (50–
447
270-m) departed from those theorized or demonstrated to influence home range scale habitat
448
selection. Thompson et al.’s (2012) review of scale-specific habitat use by martens across North
449
America found that habitat selection was strongest at the landscape scale, suggesting a robust
450
connection between home range composition and individual fitness. Two of the most influential
451
habitat variables in our analyses, late-seral forest and riparian habitat, were consistent with this
452
home-range scale pattern of importance for key resources, while elevation and mid-seral forest
453
habitat showed scale-specific optimization at the smallest microhabitat scale (50-m).
454
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
21
While elevation was statistically optimized at the 50-m scale, it was only marginally
455
more significant than larger spatial scales. Moreover, nearly all topographic variables had the
456
strongest statistical differences at the smallest spatial scales, raising further questions about the
457
biological relevance of these increasing statistical differences for smaller spatial scales. Finally,
458
the interaction between elevation and late-seral forest habitat suggested that lower elevation late-
459
seral forest at the home range scale was most influencing site occupancy by marten rather than
460
the elevation of a small portion (< 1 %; 50-m scale) of the home range.
461
The significance of mid-seral forest habitat at the 50-m scale may represent patterns of
462
within-home range use, but because the scale represents < 1 % of a marten home range its
463
biological relevance for home range selection and composition is questionable. While martens,
464
like most animals, select resources at multiple spatial scales, they do not exhibit selection at all
465
spatial scales at the same time (Mayor et al. 2009). Selection of resources to incorporate into a
466
home range to provide for an animal’s year-round resource needs may happen once in an
467
individual’s life, while selection of specific habitat types at the microscale may happen on a daily
468
or hourly basis while they are foraging (Rettie and Messier 2000, Mayor et al. 2009). Therefore,
469
it is critical to identify and constrain the selection of spatial scales for evaluation in multi-scale
470
habitat modeling to those that the dataset is capable of addressing. In our study, we compared the
471
portions of the study area occupied by martens to those not occupied by martens, essentially
472
comparing where marten home ranges occurred versus where they did not. The spatial scales
473
most relevant for modeling resource influence on home range occupancy should therefore be
474
constrained to those representing significant portions of the study area (e.g., core areas, the entire
475
home range, or the larger landscape area encompassing the home range). Although recent
476
examples of modeling with spatial scale optimization for Humboldt martens include all 6 spatial
477
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
22
scales (Slauson et al. 2019, Moriarty et al. 2021) and we sought to follow these methods, it may
478
have been more appropriate to exclude the use of the smaller spatial scales (50–270-m) as these
479
did not match the scales of habitat selection we were explicitly modeling. We recommend that
480
the spatial scales used in multi-scale habitat analyses carefully evaluate scales of habitat
481
selection that the study design and dataset can address and select only spatial scales for
482
consideration that are relevant to the specific research objectives.
483
Although the majority of Humboldt marten detections in the CA–OR EPA occurred in
484
high productivity low-elevation forest habitats, four marten detections also occurred in low-
485
productivity serpentine forest habitats. This confirms that the two distinct habitat types present in
486
the CA–OR EPA that are known to be used by Humboldt martens elsewhere are also used by
487
martens in this population. However, despite the large amount of serpentine habitat present in the
488
broader region around the CA–OR EPA, previous research suggests the use of serpentine forest
489
habitat may depend on its spatial juxtaposition to areas with large patches of late-seral productive
490
forest (Slauson et al. 2018). The significant structural and compositional differences in the tree
491
characteristics, primarily age and size classes/seral stages, between high-productivity and low-
492
productivity forest habitat used by Humboldt martens have prompted researchers to assess
493
characteristics for these distinct habitat types separately (Slauson et al. 2007). Our exploratory
494
analysis of the differences in characteristics of the locations where martens were detected in each
495
of these habitat types confirmed the stark differences between these habitat types
496
(Supplementary Material 2 Table 2). Our limited sample size for survey units dominated by
497
serpentine habitat (n = 9) precluded our inclusion of these unique areas in this analysis.
498
However, these data will be valuable when combined with larger samples for areas dominated by
499
low productivity serpentine habitats.
500
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
23
This study represents the first stage of determining the spatial extent of martens in this
501
population and provides a timely assessment of habitat use in this area. We provide evidence that
502
martens in the CA–OR EPA primarily occupy productive forest habitats located at low
503
elevations and composed of large amounts of late-seral forest, mid-seral forest, and riparian
504
habitat. In addition, some martens in the CA–OR EPA also occupy low-productivity forest
505
composed of serpentine habitat. The CA–OR EPA has been affected by multiple recent wildfires
506
since the completion of our surveys (USFS 2020), providing an opportunity to assess the short-
507
term influence of mixed-severity wildfires on this population. Nearly all of the EPA burned
508
between 2018–2023. Our surveys provide a pre-fire baseline of occupancy of habitat by marten
509
in the CA–OR EPA that can be used to compare the distribution and post-fire habitat use, and to
510
evaluate the effects of fire-severity on post-fire occupancy patterns. Managers can help maintain
511
and promote the expansion of Humboldt martens in and around the CA–OR EPA by using our
512
results to prioritize the maintenance and restoration of habitat management areas that are
513
composed of: 1) large patches of low-elevation (< 858-m) late-seral forest habitat (> 197.6 ha
514
within 1,170-m radius areas), 2) large amounts of riparian habitat (>1.55 km/km2 within 1,170-m
515
radius areas), and 3) adjacent areas of low-productivity serpentine habitat.
516
Acknowledgments
517
We thank all those who contributed and supported us through this project. We thank T. Bean, B.
518
Devlin, D. Barton, S. Hart, A. Benn, B. Carniello, K. Wright, and the many volunteers who
519
contributed their time to the project. We also thank the U.S. Fish and Wildlife Service, the U.S.
520
Forest Service, and the Humboldt State University Sponsored Programs Foundation for their
521
financial support.
522
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
24
Conflict of Interest
523
The authors declare that the research was conducted in the absence of any commercial or
524
financial relationships that could be construed as a potential conflict of interest.
525
Data Availability Statement
526
The datasets generated during the study are available from the corresponding author upon
527
reasonable request.
528
Supplementary Materials
529
Supplementary materials are hosted online by BioOne.
530
Author Contributions
531
HELG: Conceptualization, data collection, writing – original draft, visualization, validation,
532
formal analysis. KMS: Conceptualization, data collection, writing – review and editing,
533
visualization, validation, formal analysis. MSG: Conceptualization, writing – review and editing,
534
validation, supervision, funding acquisition.
535
References Cited
536
Andruskiw M, Fryxell JM, Thompson ID, Baker JA. 2008. Habitat-mediated variation in
537
predation risk by the American marten. Ecology 89:2273-2280. DOI: 10.1890/07-1428.1
538
Baldwin RA, Bender LC. 2008. Distribution, occupancy, and habitat correlates of American
539
martens (Martes americana) in Rocky Mountain National Park, Colorado. Journal of
540
Mammalogy 89:419-427. DOI: 10.1644/07-MAMM-A-053R1.1
541
Burnham KP, Anderson DR. 2002. Model selection and multi-model inference: A practical
542
information-theoretic approach. Second edition. Springer-Verlag, New York, USA.
543
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
25
Buskirk SW, Ruggiero LF. 1994. The American marten. In LF Ruggiero, KB Aubry, SW
544
Buskirk, LJ Lyon, and WJ Zielinski (editors), American marten, fisher, lynx, and
545
wolverine in the western United States. USDA Forest Service GTR-RM-254, Rocky
546
Mountain Forest and Range Experiment Station, Fort Collins, Colorado, USA. Pp. 7-37.
547
California Department of Fish and Wildlife, Biogeographic Data Branch [CDFW]. 2021.
548
California Wildlife Habitat Relationship System, Version 10.1.29. Sacramento, CA.
549
https://wildlife.ca.gov/Data/CWHR (accessed 6 December 2023)
550
California Department of Fish and Wildlife [CDFW]. 2019. Title 14: OAL Matter Number: 2019
551
− 0201 − 02 Endangered Status for Humboldt marten, tri-colored blackbird, fisher
552
southern Sierra ESU. Sacramento: CDFW18 March 2019 Amendment.
553
Cooch E, White G. 2001. Program Mark: A gentle introduction. Colorado State University.
554
<http://www.phidot.org/software/mark/docs/book/> Accessed 07 Jun 2024.
555
Delheimer MS, Moriarty KM, Linnell MA, Woodruff BV. 2019. If a tree falls in a forest:
556
Implications of forest structure persistence for the Pacific marten (Martes caurina).
557
Ecosphere 10:e02819. DOI: 10.1002/ecs2.2819
558
Delheimer MS, Moriarty KM, Slauson KM, Roddy AM, Early DA, Hamm KA. 2021.
559
Comparative reproductive ecology of two subspecies of Pacific marten (Martes caurina)
560
in California. Northwest Science 94:271-285. DOI: 10.3955/046.094.0305
561
Delheimer MS, Slauson KM, Szykman Gunther M, Zielinski WJ. 2018. Use of artificial cavities
562
by Humboldt marten. Wildlife Society Bulletin 42:510-517. DOI: 10.1002/wsb.900
563
Dirzo R, Raven PH. 2003. Global state of biodiversity and loss. Annual Review of Environment
564
and Resources 28:137-167. DOI: 10.1146/annurev.energy.28.050302.105532
565
Doyle A. 1990. Use of riparian and upland habitats by small mammals. Journal of Mammalogy
566
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
26
71:14-23. DOI: 10.2307/1381312
567
Drever RC, Drever MC, Sleep DJH. 2012. Understanding rarity: A review of recent conceptual
568
advances and implications for conservation of rare species. The Forestry Chronicle
569
88:165-175. DOI: 10.5558/tfc2012-033
570
Eriksson CE, Moriarty KM, Linnell MA, Levi T. 2019. Biotic factors influencing the unexpected
571
distribution of a Humboldt marten (Martes caurina humboldtensis) population in a young
572
coastal forest. PLOS ONE 14:e0214653. DOI: 10.1371/journal.pone.0214653
573
ESRI. 2015. ArcGIS Desktop: Release 10.3 Environmental Systems Research Institute.
574
Redlands, California, USA.
575
Estes JA, Terborgh J, Brashares JS, Power ME, Berger J, Bond WJ, Carpenter SR, Essington TE,
576
Holt RD, Jackson JBC, Marquis RJ, Oksanen L, Oksanen T, Paine RT, Pikitch EK,
577
Ripple WJ, Sandin SA, Scheffer M, Schoener TW, Shurin JB, Sinclair ARE, Soulé ME,
578
Virtanen R, Wardle DA. 2011. Trophic downgrading of planet Earth. Science 333:301-
579
306. DOI: 10.1126/science.1205106
580
Gompper ME, Kays RW, Ray JC, LaPoint SD, Bogan DA, Cryan JR. 2006. A comparison of
581
noninvasive techniques to survey carnivore communities in northeastern North America.
582
Wildlife Society Bulletin 34:1142-1151. DOI: 10.2193/0091-7648
583
Hamilton SH, Pollino CA, Jakeman AJ. 2015. Habitat suitability modelling of rare species using
584
Bayesian networks: Model evaluation under limited data. Ecological Modelling 299:64-
585
78. DOI: 10.1016/j.ecolmodel.2014.12.004
586
Harrison S, Safford H, Grace J, Viers J, Davies K. 2006. Regional and local species richness in
587
an insular environment: Serpentine plants in California. Ecological Monographs 76:41-
588
56. DOI: 10.1890/05-0910
589
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
27
Jimerson TM. 1989. Snag densities in old-growth stands on the Gasquet ranger district Six
590
Rivers National Forest, California. Unpublished report PSW-196 on file at USDA Forest
591
Service, Pacific Southwest Forest and Range Experiment Station.
592
Kirk TA, Zielinski WJ. 2009. Developing and testing a landscape habitat suitability model for
593
the American marten (Martes americana) in the Cascades mountains of California.
594
Landscape Ecology 24:759-773. DOI: 10.1007/s10980-009-9349-5
595
Krausman PR. 1999. Some basic principles of habitat use. Grazing Behavior of Livestock and
596
Wildlife 70:85-90.
597
LEMMA. 2017. GNN species and structure maps. Landscape ecology, modeling, mapping, and
598
analysis group. Oregon State University, Corvallis.
599
http://lemma.forestry.oregonstate.edu/data/structure-maps (accessed 6 December 2023)
600
Lindstedt SL, Miller BJ, Buskirk SW. 1986. Home range, time, and body size in
601
mammals. Ecology 67:413-418. DOI: 10.2307/1938584
602
Linnell MA, Moriarty K, Green DS, Levi T. 2018. Density and population viability of coastal
603
marten: A rare and geographically isolated small carnivore. PeerJ 6:e4530. DOI:
604
10.7717/peerj.4530
605
MacKenzie DL, Bailey LL. 2004. Assessing the fit of site-occupancy models. Journal of
606
Agriculture, Biological, and Environmental Statistics 9:300-318. DOI:
607
10.1198/108571104X3361
608
MacKenzie DL, Hines JE. 2006. Program PRESENCE. Version 2.0. https://www.mbr-
609
pwrc.usgs.gov/software/presence.html (accessed 1 January 2019)
610
MacKenzie DL, Nichols J, Lachman G, Droege S. 2002. Estimating site occupancy rates when
611
detection probabilities are less than one. Ecology 83:2248-2255. DOI: 10.1890/0012-
612
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
28
9658
613
Martin ME, Delheimer MS, Moriarty KM, Early DA, Hamm KA, Pauli JN, Mcdonald TL,
614
Manley PN. 2022. Conservation of rare and cryptic species: Challenges of uncertainty
615
and opportunities for progress. Conservation Science and Practice 4:e12809. DOI:
616
10.1111/csp2.12809
617
Martin ME, Moriarty KM, Pauli JN. 2021. Landscape seasonality influences the resource
618
selection of a snow-adapted forest carnivore, the Pacific marten. Landscape Ecology
619
36:1055-1069. DOI: 10.1007/s10980-021-01215-9
620
Mayor SJ, Schneider DC, Schaefer JA, Mahoney SP. 2009. Habitat selection at multiple scales.
621
Écoscience 16:238-247. DOI: 10.2980/16-2-3238
622
Moriarty KM, Bailey JD, Smythe SE, Verschuyl J. 2016. Distribution of Pacific marten in
623
coastal Oregon. Northwestern Naturalist 97:71-81. DOI: 10.1898/NWN16-01.1
624
Moriarty KM, Linnell MA, Thornton JE, Watts III GW. 2018. Seeking efficiency with carnivore
625
survey methods: A case study with elusive martens. Wildlife Society Bulletin 42:403-
626
413. DOI: 10.1002/wsb.896
627
Moriarty KM, Thompson J, Delheimer M, Barry BR, Linnell M, Levi T, Hamm K, Early D,
628
Gamblin H, Szykman Gunther M, Ellison J, Prevéy JS, Hartman J, and Davis R. 2021.
629
Predicted distribution of a rare and understudied forest carnivore: Humboldt marten
630
(Martes caurina humboldtensis). PeerJ 9:e11670. DOI: 10.7717/peerj.11670
631
Moriarty KM, Verschuyl J, Kroll AJ, Davis R, Chapman J, Hollen B. 2019. Describing
632
vegetation characteristics used by two rare forest-dwelling species: Will established
633
reserves provide for coastal marten in Oregon? PLOS ONE 14:e0210865. DOI
634
10.1371/journal.pone.0210865
635
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
29
National Research Council [NRC]. 2002. Structure and functioning of riparian areas across the
636
United States. In Crossgrove RD (editor), Riparian areas: Functions and strategies for
637
management. National Academy, Washington, D.C., USA. Pp. 49-144.
638
Prevéy JS, Parker LE, Brodie LC, Harrington CA. 2022. Habitat suitability models for
639
understory shrubs of western North America: U.S. Geological Survey data release. DOI:
640
10.5066/P9Z2TVGA
641
PRISM Climate Group. 2004. PRISM climate data. Oregon State University, Corvallis.
642
http://prism.oregonstate.edu/normals/ (accessed 6 December 2023)
643
Rettie WJ, Messier F. 2000. Hierarchical habitat selection by woodland caribou: Its relationship
644
to limiting factors. Ecography 23:466-478. DOI: 10.1111/j.1600-0587.2000.tb00303.x
645
RStudio Team. 2022. RStudio: Integrated development environment for R. RStudio, PBC,
646
Boston, Massachusetts, USA.
647
Segan DB, Murray KA, Watson JEM. 2016. A global assessment of current and future
648
biodiversity vulnerability to habitat loss–climate change interactions. Global Ecology and
649
Conservation 5:12-21. DOI: 10.1016/j.gecco.2015.11.002
650
Shirk AJ, Raphael MG, Cushman SA. 2014. Spatiotemporal variation in resource selection:
651
Insights from the American marten (Martes americana). Ecological Applications
652
24:1434-1444. DOI: 10.1890/13-1510.1
653
Slauson KM. 2017. Linking landscape pattern to population processes for a carnivorous
654
mammal. PhD dissertation, University of Montana, Missoula, Montana.
655
Slauson KM, Baldwin JA, Zielinski WJ, Kirk TA. 2009. Status and estimated size of the only
656
remnant population of the Humboldt subspecies of the American marten (Martes
657
americana humboldtensis) in northwestern California: Final report. Unpublished report
658
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
30
on file at U.S. Department of Agriculture, Forest Service, Pacific Southwest Research
659
Station, Arcata, California.
660
Slauson KM, Howard B, White AM, Maxwell C, Holland T. 2022. Evaluating the effects of
661
alternative landscape management scenarios on three old-forest-associated predators over
662
100 years in the fire-prone forests of the Sierra Nevada, USA. Ecology and Society
663
27:28. DOI: 10.5751/ES-13362-270328
664
Slauson KM, Moriarty KM. 2014. Determining the distributions, sizes, and trends of Humboldt
665
marten populations: Survey design and field survey protocol. Unpublished report on file
666
at Pacific Southwest Research Station, Arcata, California.
667
Slauson KM, Schmidt GA, Zielinski WJ, Detrich PJ, Callas RL, Thrailkill J, Devlin-Craig B,
668
Early DA, Hamm KA, Schmidt KN, Transou A, West CJ. 2018. A conservation
669
assessment and strategy for the Humboldt marten (Martes caurina humboldtensis) in
670
California and Oregon. Unpublished report PSW-GTR-260 on file at U.S. Department of
671
Agriculture, Forest Service, Pacific Southwest Research Station, Arcata, California.
672
Slauson KM, Zielinski WJ. 2009. Characteristics of summer and fall diurnal resting habitat used
673
by American martens in coastal northwestern California. Northwest Science 83:35-45.
674
DOI: 10.3955/046.083.0104
675
Slauson KM, Zielinski WJ, Hayes JP. 2007. Habitat selection by American martens in coastal
676
California. Journal of Wildlife Management 71:458-468. DOI: 10.2193/2004-332
677
Slauson KM, Zielinski WJ, Kirk TA, LaPlante DW. 2019. A landscape habitat suitability model
678
for the Humboldt marten (Martes caurina humboldtensis) in coastal California and
679
coastal Oregon. Northwest Science 93:30-51. DOI: 10.3955/046.093.0104
680
Soil Survey Staff. 2022. Gridded Soil Survey Geographic (gSSURGO) Database for California.
681
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
31
United States Department of Agriculture, Natural Resources Conservation Service.
682
https://gdg.sc.egov.usda.gov/ (accessed 6 December 2023)
683
Thompson ID, Fryxell J, Harrison DJ. 2012. Improved insights into use of habitat by American
684
martens. In Aubry KB, Zielinski WJ, Raphael MG, Proulx G, Buskirk SW (editors),
685
Biology and Conservation of Martens, Sables, and Fishers: A New Synthesis. Cornell
686
University Press, Ithaca, USA. Pp. 209-230.
687
Todman LC, Bush A, Hood ASC. 2023. ‘Small Data’ for big insights in ecology. Trends in
688
Ecology & Evolution 38:615-622. DOI: 10.1016/j.tree.2023.01.015
689
Tweedy PJ, Moriarty KM, Bailey JD, Epps CW. 2019. Using fine scale resolution vegetation
690
data from LiDAR and ground-based sampling to predict Pacific marten resting habitat at
691
multiple spatial scales. Forest Ecology and Management 452:117556. DOI:
692
10.1016/j.foreco.2019.117556
693
Waters JR, Zabel CJ, McKelvey KS, Welsh Jr. HH. 2001. Vegetation patterns and abundances of
694
amphibians and small mammals along small streams in a northwestern California
695
watershed. Northwest Science 75:37-52.
696
United States Census Bureau [USCB]. 2022. MAF/TIGER Line Geodatabases.
697
https://www.census.gov/geographies/mapping-files/time-series/geo/tiger-geodatabase-
698
file.html (accessed 6 December 2023)
699
United States Forest Service [USFS]. 2018. Existing Vegetation Mid Region 5 North Coast East.
700
https://data.fs.usda.gov/geodata/edw/datasets.php (accessed 6 December 2023)
701
United States Forest Service [USFS]. 2020. Slater Fire Rapid Assessment Team Report.
702
Unpublished report on file at Wild Rivers and Siskiyou Mountains Ranger Districts,
703
Jacksonville, Oregon.
704
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
32
Unites States Forest Service [USFS]. 2022. Overview for Geospatial Data.
705
https://www.fs.usda.gov/detail/r5/landmanagement/gis (accessed 6 December 2023)
706
United States Fish and Wildlife Service [USFWS]. 2015. Coastal Oregon and northern Coastal
707
California populations of the Pacific marten (Martes caurina). Unpublished report on file
708
at U.S. Fish and Wildlife Service, Region 8, Arcata, California.
709
White GC. 2001. Program MARK. Version 6.2.
710
http://www.phidot.org/software/mark/downloads/ (accessed 1 January 2019)
711
Zielinski WJ. 2014. The forest carnivores: Fisher and marten. In Long JW, Quinn-Davidson L,
712
Skinner CN (editors), Science Synthesis to Support Socioecological Resilience in the
713
Sierra Nevada and Southern Cascade Range. Unpublished report PSW-GTR-247 on file
714
at USDA Forest Service, Pacific Southwest Research Station, Albany, California. Pp.
715
395-435.
716
Zielinski WJ, Golightly RT. 1996. The status of marten in redwoods: Is the Humboldt marten
717
extinct? In LeBlanc J (editor), Conference on Coast Redwood Forest Ecology and
718
Management. Humboldt State University, Arcata, California. Pp. 115–119.
719
Zielinski WJ, Moriarty KM, Baldwin J, Kirk TA, Slauson KM, Rustigian-Romsos HL, Spencer
720
WD. 2015. Effects of season on occupancy and implications for habitat modeling: The
721
Pacific marten (Martes caurina). Wildlife Biology 21:56-67. DOI: 10.2981/wlb.00077
722
Zielinski WJ, Slauson KM, Carroll CR, Kent CJ, Kudrna DG. 2001. Status of American martens
723
in coastal forests of the Pacific states. Journal of Mammalogy 82:478-490. DOI:
724
10.1644/1545-1542
725
Zuur AF, Ieno EN, Elphick CS. 2010. A protocol for data exploration to avoid common
726
statistical problems. Methods in Ecology and Evolution 1:3-14. DOI: 10.1111/j.2041-
727
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
33
210X.2009.00001.x
728
729
Submitted 12 March 2024
730
Accepted 26 July 2024
731
732
733
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
34
Figures
734
735
Figure 1. Study area and locations of survey units sampled in and around the California–Oregon
736
Extant Population Area (CA–OR EPA) in northern California, USA, 2017–2018, depicting
737
survey units with Humboldt marten detections (n = 20, closed circles) and non-detections (n =
738
31, open circles).
739
740
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
35
741
Figure 2. The habitat values associated with Humboldt marten detections (n = 16, closed
742
diamonds) and non-detections (n = 26, open circles) in northern California, USA, 2017–2018, for
743
the scale-optimized habitat variables present in the top three occupancy models: (a) elevation at
744
the 50-m scale (Elevation_50) and mid-seral forest habitat at the 50 m scale (Size Class
745
Medium_50), (b) elevation at the 50-m scale (Elevation_50) and late-seral forest habitat at the
746
1,170 m scale (LSOG_1170), and (c) riparian habitat at the 1,170-m scale (Stream_1170) and
747
late-seral forest habitat at the 1,170-m scale (LSOG_1170).
748
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
36
749
Figure 3. Probability of occupancy (Ψ) by Humboldt marten in northern California, USA, 2017–
750
2018, along with associated 95 % confidence intervals for habitat variables in the top three
751
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
37
occupancy models (AICc < 2) while holding the other variables present within the model at their
752
average values. The top model depicts Ψ as a function of (a) elevation (Elev) and (b) the amount
753
of size class medium trees (SC_Med) present at the 50-m scale. The second best-supported
754
model depicts Ψ as a function of (c) riparian habitat (Stream) and (d) the amount of late-seral
755
old-growth (LSOG) habitat present at the 1,170 m scale. The third best-supported model depicts
756
Ψ as a function of (e) elevation (Elev) at the 50 m scale and (f) the amount of late-seral old-
757
growth (LSOG) present at the 1,170 m scale.
758
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a recently discovered population of Humboldt
martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest Science. Copy-editing may lead to
differences between this version and the final published version.
38
Tables
759
Table 1. Beta estimates and odds ratios (OR) for the top detection probability (p) and occupancy (Ψ) models for Humboldt martens
760
monitored in northern California, USA, 2017–2018, along with associated standard error (SE) and 95 % lower (LCI) and upper
761
confidence intervals (UCI). The optimal spatial scale (m) for each occupancy variable is included in the parameter name.
762
Model
Rank
Model Name
Parameter
Beta
SE
95 %
LCI
95 %
UCI
OR
95 %
LCIOR
95 %
UCIOR
1
p (month + dur)
p_intercept
-3.96
1.26
-6.43
-1.48
0.02
0.002
0.23
month
1.34
0.54
0.27
2.41
3.81
1.31
11.10
dur
0.13
0.06
0.02
0.25
1.14
1.02
1.28
1
Ψ (Elev_50 + SC_Med_50)
Ψ_intercept
6.89
2.96
1.09
12.68
981
2.99
3.23e5
SC_Med_50
4.99
3.04
-0.96
10.95
1.28*
0.95
1.73
Elev_50
-0.01
0.005
-0.02
-0.002
0.33*
0.13
0.81
2
Ψ (Stream_1170 + LSOG_1170)
Ψ_intercept
-43.61
20.65
-84.08
-3.14
>0.001
>0.001
0.04
Stream_1170
11.64
5.87
0.14
23.14
3.20*
1.01
10.11
LSOG_1170
71.28
33.68
5.27
137.30
35.31*
1.30
958.07
3
Ψ (Elev_50*LSOG_1170)
Ψ_intercept
27.12
15.68
-3.61
57.84
5.97e11
0.03
1.32e25
Elev_50
-0.05
0.02
-0.09
0.003
0.01*
>0.001
1.34
LSOG_1170
-48.61
31.15
-109.66
12.44
0.09*
0.004
1.86
Elev_50*LSOG_1170
0.09
0.05
-0.01
0.18
1.09
0.99
1.20
Dur = duration, Elev_50 = elevation at the 50 m scale, SC_Med_50 = size class medium at the 50 m scale, Stream_1170 = stream at the 1,170 m
763
scale, and LSOG_1170 = late-seral old-growth at the 1,170 m scale.
764
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a recently discovered population of Humboldt
martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest Science. Copy-editing may lead to
differences between this version and the final published version.
39
*Indicates OR has been adjusted to reflect a scale appropriate to the variable data range: SC_Med_50 and LSOG_1170 OR = exp(Beta*0.05),
765
Elev_50 OR = exp(Beta*100), Stream_1170 OR = exp(Beta*0.10).
766
Gamblin HEL, Slauson KM, Szykman Gunther M. 2024. Habitat use and distribution of a
recently discovered population of Humboldt martens. Northwest Science 97(4): in press.
Note: This comment has been peer reviewed and accepted for publication in Northwest
Science. Copy-editing may lead to differences between this version and the final published
version.
40
Table 2. Adjusted variable importance weights for variables in the occupancy model set for
767
Humboldt martens monitored in northern California, USA, 2017–2018. Variable weights were
768
calculated as the sum of Akaike’s Information Criterion weights (AICc) for models containing
769
the variable relative to the number of models the variable appeared in, and listed in decreasing
770
order of importance. The average (x
Y) values for each scale-optimized variable at detection and
771
non-detection productive forest habitat survey units are reported along with associated standard
772
error (SE).
773
Variable
Weight
Scale (m)
Detection x
Y ± SE
Non-detection x
Y ± SE
LSOG
0.16
1170
46.0 ± 1.8 %
35.8 ± 2.5 %
Elev
0.15
50
582.0 ± 36.9 m
964.0 ± 67.3 m
SC_Med
0.09
50
17.0 ± 5.7 %
13.2 ± 3.4 %
Stream
0.08
1170
1.6 ± 0.1 km/km2
1.2 ± 0.1 km/km2
CanCov
0.07
3000
74.9 ± 0.7 %
67.8 ± 1.2 %
QMDC
0.03
50
54.4 ± 4.9 cm
46.5 ± 3.2 cm
OGSI
0.02
50
34.7 ± 3.5
31.3 ± 3.0
SC_Lar
0.02
750
18.3 ± 2.7 %
25.3 ± 3.1%
GASH
0.01
3000
36.2 ± 1.9 %
49.4 ± 2.5 %
HardMast
>0.01
3000
13.9 ± 1.0 %
9.7 ± 1.0 %
VAOV
>0.01
3000
17.7 ± 0.4 %
19.1 ± 0.5 %
LSOG = late-seral old growth, Elev = elevation, SC_Med = size class medium trees, Stream = stream
774
habitat, CanCov = canopy cover, QMDC = quadratic mean diameter of conifers, OGSI = old-growth
775
structure index, SC_Lar = size class large trees, GASH = salal, HardMast = trees producing hard mast,
776
and VAOV = evergreen huckleberry.
777