Content uploaded by John G Kie
Author content
All content in this area was uploaded by John G Kie on Oct 11, 2018
Content may be subject to copyright.
F0res~~;olog-j
Management
ELSEVIER
Forest Ecology and
Management 88 (
1996) 157- 165
Modelling habitat suitability for black-tailed deer (OdmcCZem
hemionus cokmbianus) in heterogeneous landscapes
Brian B. Boroski apb7 * , Reginald H. Barrett ‘, Irene C. Timossi ‘, John G. Kie a91
a US Forest Service, Pacijk Southwest Reseurch Station, 2081 East Sierra Avenue, Fresno, CA 93710, USA
’ Depariment of Environmental Science, Policy, and Munagement, University of Calqornia, Berkeley, CA 94720, USA
Abstract
The California Wildlife Habitat Relationships (CWHR) System is a set of standardized procedures including an
information system describing the management status, distribution, life history, and habitat requirements of 643 species of
terrestrial vertebrates regularly occurring in California. The original CWHR Database consisted of a series of non-spatial,
matrix models, including one for black-tailed deer. A series of second-generation models incorporates explicit spatial
relations for use within a geographic information system. These models were designed for landscapes larger than 60000 ha
with a minimum mapping unit of 0.8 ha. As part of our development of a third-generation model applicable to landscape
depictions with a minimum mapping unit of 0.06 ha, we assessed the utility of a measure of landscape composition, and an
interspersion index, both based on feeding and cover habitat suitability classes found within deer home ranges.
General linear model analysis indicated that the percentage of the home range comprised by each class and the
interspersion index were correlated (P < 0.05) with feeding-cover class, site, the interaction between feeding-cover class and
site, and the interaction between home range type (actual or random) and site..
The percentage of home ranges comprised by feeding-cover class and the level of interspersion within home ranges
differed significantly by suitability class. We detected differences between actual deer home ranges and random home ranges
of similar size only within the site that had significant separation in the composition of classes and the interspersion of
classes. This suggests that landscape representations resulting in classes with similar composition levels and similar levels of
interspersion at the scale of the animal’s home range may hinder the assessment of habitat
use
within home ranges by
promoting Type II errors.
We demonstrated that organism-centered specifications for extent and grain can be defined from a generic structural
representation of the landscape using functional-based models. Spatial extent was defined using the adaptive kernel home
range estimator. Within this extent, habitat suitability classes based on feeding and cover produced the grain size for
black-tailed deer (mean 0.44 ha, range 0.06-7.79 ha). Furthermore, this approach portrayed the low feeding-high cover
matrix class exerting the dominant influence within home ranges. Thus, we suggest measures of landscape composition and
interspersion may be useful for characterizing landscapes when assessing habitat use by deer. They may also be of value to
natural resource managers attempting to predict the response of deer to proposed changes in land use.
Keywords: 0docoileu.s hemionus columbiamu; Habitat suitability; Modelhng; GIS
’ Corresponding author at: US Forest Service, Pacitic Southwest Research Station, 208 I E. Sierra Avenue, Fresno, CA 937 10, USA. Tel.
209-487-5589; fax: 209-487-5978.
’ Present address: Pacitic Northwest Research Station, 1401 Gekeler Lane, LaGmade, OR 97850, USA.
0378-I 127/96/$15.00 Copyright 0 1996 Elsevier Science B.V. All rights reserved,
PII SO.J78- 1 127(96)0382 l-2
158 B.B. Bormki et d/Forest Ecology and .Munugement 88 (1996) 157-165
1. Introduction
The impetus for the California Wildlife Habitat
Relationships (CWHR) System was a guild based
information system developed for the Blue Moun-
tains of Oregon and Washington (Thomas, 19791.
The first efforts in California also used a regional
approach. California was divided into five geo-
graphic zones: north coast/Cascades (Marcot, 1979);
northeast interior (Airola, 198Oa; Airola, 198Ob;
Airola, 1980~; Shimamoto and Airola, 1981; Lau-
denslayer, 1982; Laudenslayer et al., 19821; western
Sierra Nevada (Hurley and Asrow, 1980; Vemer and
Boss, 1980); Lake Tahoe Basin (O&i, 1980); and
southern California (Lee and Keeney, unpublished
report). However, the California information systems
modeled the habitat relationships of individual
species as opposed to guilds with the intention of
increasing the precision of environmental impact
analysis by emphasizing the entire wildlife commu-
nity (Grenfell et al., 1982; Airola, 19881. Incomplete
coverage of the state, inconsistencies in classification
schemes at the regional scale, and funding limita-
tions led to the initiation of a state-wide system in
1981. The CWHR System described and modeled
the habitat relationships and requirements, manage-
ment status, geographic distribution, life history, and
responses to habitat changes of 643 species of resi-
dent and migratory terrestrial and aquatic vertebrates
regularly occurring in California (Airola, 1988; Tim-
ossi et al., 19941.
The original CWHR Database consisted of a se-
ries of non-spatial models based on a matrix of
habitat types and seral stages with a minimum map-
ping unit of 16 ha (Mayer and Laudenslayer, 19881.
Suitability ratings for feeding, cover and reproduc-
tion derived from scientific literature and profes-
sional judgement were given to each cell in the
matrix (Timossi et al., 1989).
A series of second-generation models incorporate
explicit spatial relations for use within a geographic
information system (GIS) (Davis and Barrett, 19931.
These models were designed for landscapes larger
than 60000 ha with a minimum mapping unit of 0.8
ha. These spatially explicit models refined the non-
spatial suitability estimates based on the geometric
mean of the cover, feeding, and reproductive require-
ments. Also, the spatial models eliminate small, dis-
continuous habitats of low value, sites too far from
water, or too close to roadways.
When modeling habitat suitability, habitat stmc-
ture is best defined relative to a particular organism’s
use of the environment (Wiens, 19761. This implies
defining habitat at different scales. Thus poses ii
serious problem for natural resource managers at-
tempting to predict the impacts from various land
uses on a wide variety of species. The expense of
developing and validating landscape models makes it
unrealistic to develop different representations of the
same area at many scales. What is needed is a single
representation that facilitates multi-scale organism-
centered modeling.
We suggest one way this may be accomplished is
to construct the landscape representation on the dis-
tribution of energy, materials, and species at the
smallest scale reasonable for land management and
planning activities, and model wildlife based on the
interactions among the spatial elements. From a
landscape ecology perspective ~Forman and Godron
19861, we are suggesting to represent the configura-
tions of ecosystems based on their structure and
model wildlife based on the function of the structural
elements. In our efforts to develop a third-generatiun
model for black-tailed deer (0&c~&ti.r /iem&u.T
columbiunus)
applicable to landscape depictions with
a minimum mapping unit of 0.06 ha, we assessed the
utility of (1) a measure of landscape composition
and (21 an interspersion index based on feeding and
cover habitat suitability classes within deer home
ranges.
I .I. Study area
The study was carried out around Clan Angle
Reservoir within the boundaries of the Weaverville
deer herd in Trinity County, California (Fig. 11. The
Weaverville deer herd includes the former New River
and Weaverville herds (Dunaway, 1964; Dunaway.
19651, occupying approximately 2070 km’ along the
drainage of the mainstream of the Trinity River
(Burton and Monroe, 1983). Areas occupied by deer
range in elevation from about 450 m along the
Trinity River to about 2400 m in the Trinity Alps
Wilderness. Portions of the deer herd are resident.
although most are migratory (United States Fish and
Wildlife Service, 1975; Burton and Monroe: 1983;
Loft et al.. 1984).
B.B. Bormki et d/Forest Ecology und Munugement 88 (19915) 157-165
I59
Site 2
Fairview
Ridge
I : i : I
4km
Fig. I. Location of study sites along the shore of Clair En@
Reservoir in Trinity County, CA.
Common conifer tree species include ponderosa
pine
(Pinus
~0&~-0s~), grey pine
(Pinus subiniunu),
sugar pine
(Pinus lambertianu),
Douglas fir
(Pseu-
dotsuga menziesii),
white fir
(Abies concolor),
and
incense cedar
(Calocedrus decurrens).
Deciduous
tree species include California black oak
(Quercus
kelloggii),
Oregon white oak
(Quercus garryunu),
interior live
oak (Quercus wislizenii),
and mountain
dogwood
(Cornus nuftullii).
Shrub species include
deer brush
(Ceunothus integerrimus),
common
snowberry
(Symphoricarpus rivufaris),
buckbrush
(Ceanothus cuneatus),
and poison oak
(Toxico-
dendron diversiloba).
Based on previous work (Dunaway, 196% Loft et
al., 1984) and the potential bias associated with
artificial landscape boundaries, three distinct sites of
winter deer use: (1) Bowerman Ridge (1600 ha), (2)
Fairview Ridge (1200 ha), and (3) Rackerby Ridge
(700 ha), were selected for intensive sampling (Fig.
1). All three sites are peninsulas along Clair Engle
Reservoir, reducing the potential bias associated with
artificially imposed boundaries. Although deer do
traverse the reservoir, particularly during migration
(Loft et al., 198% Boroski and McLaughlin, 19941,
the reservoir represents a natural landscape boundary
between habitable and uninhabitable environments
for deer.
2. Methods
2.1. Landscape representations
Landsat Thematic Mapper (TM) Satellite imagery
classified by Geographic Resource Solutions (Arcata,
CA) (Brown and Fox, 1992; Stumpf and Koltun,
1992) was used as the data source to identify associ-
ated vegetation types based on species composition
and dominance, tree size class, and tree canopy
closure class for each geo-corrected 25 m
X
25 m
TM pixel. These attributes were used to derive vege-
tation types defined by the CWHR System (Mayer
and Laudenslayer, 1988). The CWHR habitat classi-
fication first divides habitat into the major subdivi-
sions: tree-dominated, shrub-dominated.
herbaceous-dominated, aquatic, and developed. Once
a habitat has been classified into its appropriate
major subdivision, a specific habitat is determined
based on species composition and distribution, and
physiographic criteria such as elevation and soil.
Once a vegetative complex has been classified, struc-
tural condition and seral stage determine the final
CWHR type (Mayer and Laudenslayer, 1988). For
example, tree-dominated habitats are based on canopy
layering, diameter at breast height, and canopy clo-
sure.
Habitat suitability ratings of high, medium, low,
or unsuitable for feeding and cover were assigned to
each vegetation type according to the CWHR
Database, Version 5.0 (Airola, 1988; Timossi et al.,
1994). Suitability ratings for feeding and cover were
combined into 16 possible feeding-cover classes and
used to portray the function of the landscape for deer
(Table 1).
2.2. Deer capture
Twenty-one does were captured and radio-col-
lared, eight in site 1, nine in site 2, and four in site 3.
Deer were captured using a vertically collapsing
I60 B.B. Bowski et d/Forest Ecology und Manugentent 88 (1996) 157-165
Table 1
Feeding-cover classes used to portray the suitability of the Trinity River Basin for deer
Feeding-cover class Label Presence (P) or absence (A)
in study area
Unsuitable feeding and unsuitable covet
Unsuitable feeding and low cover
Unsuitable feeding and medium cover
Unsuitable feeding and high cover
Low feeding and unsuitable cover
Low feeding and low cover
Low feeding and medium cover
Low feeding and high cover
Medium feeding and unsuitable cover
Medium feeding and low cover
Medium feeding and medium cover
Medium feeding and high cover
High feeding and unsuitable cover
High feeding and low cover
High feeding and medium cover
High feeding and high cover
Background
LC
MC
HC
LF
LFLC
LFMC
LFHC
MF
MFLC
MFMC
MFHC
HF
HFLC
HFMC
HFHC
P
A
A
A
A
P
P
P
A
P
P
P
A
P
P
A
version (McCullough, 1975) of the Clover trap and
while they swam Clair Engle Reservoir (Boroski and
McLaughlin, 1994). Radio-collars carried a metal
encased transmitter (Model AT-2, Communications
Specialists, Inc. Orange, CA, or Model LKRT-3,
Lotek Engineering, Inc., New Market, Ont., Canada,
or Model-500 or Model-400, Telonics Inc., Mesa,
AZ) at frequencies between 159.171 and 159.8 10
MHz.
2.3. Deer locations
Locations of deer in the three sites were measured
in 8-h sessions from December 1993 through Febru-
ary 1994. The starting time of each session, within
sites, alternated between three predetermined time
blocks to ensure coverage of the 24-h day. A Scan-
ner/Programmer with a receiver (Models TS- 1, TS-
2, Telonics Inc.) attached to a three-element, hand-
held antenna was used to detect radios. Azimuths
were based on the strongest signal from two or more
locations at known points or by visual sightings. The
location of all antenna sites and visual sightings were
determined using a differential global positioning
system (Pathfinder Professional System, Trimble
Navigation, Sunnyvale, CA).
2.4. Home range estimates
Using program CALHOME (available from J.
Kie; see Larkin and Halkin, 1994 for review) we
defined a 3-month home range for each radia-col-
hired deer. Home range estimaEes were derived using
an adaptive kernel analysis (Worton, I9891 based on
95% of the locations. Ninety-five percent was used
to reduce the influence of inaccurate bearing loca-
tions on home range size (Schoener, 1981; Whi&e
and Garrott, 1990). A 50 X 50 cell grid and the
CALHOME program-calculated optimum bandwidth
(Worton, 1989) were used in analyzing each data set.
To assess the minimum number of locations to
use in estimating home range size, we ptotted the
adaptive kernel estimates as a function of the number
of locations and then fitted a locally-weighted, robust
regression line (Cleveland, 1979) using the LOWESS
smoothing algorithm in SYSTAT (Wilkinson et al.,
1992). We concluded that 20 locations were suffi-
cient to standardize home range size and variance.
This reduced our sample size from 2 I to 12; four in
site 1, five in site 2, and three in site 3. For each deer
home range, we generated a random circle with the
same area as the home range and randomly located it
within the same site, 1, 2, or 3 used by that &er.
B.B. Boroski et al./Forest Ecdqy atul Munugement 88 (1996) 157-165
161
We converted each deer and random home range
into an ARC/INFO (Version 6.1, ESRI, Redlands,
CA) coverage and clipped them from the feeding-
cover coverage to derive individual feeding-cover
class coverages for each deer and random home
range. These coverages were converted to IDRISI
(Version 4.0, Clark University, Worcester, MA) im-
ages and processed in FRAGSTATS (McGarigal and
Marks, 1994) to describe the percentage of the deer
and random home ranges comprised by each feed-
ing-cover class and to calculate an interspersion
index for each class. The feeding-cover class com-
positions for the home ranges were calculated by
summing the areas of all patches of the correspond-
ing class type and dividing this sum by the total area
of the home range. The values were then multiplied
by 100 to convert them to percentages (McGarigal
and Marks, 1994). The interspersion index reflects
the observed interspersion over the maximum possi-
ble interspersion for a given number of class types.
McGarigal and Marks (1994) define the index as
Interspersion index
zz
ln(&- 1) 1Otl
where eik is total length (m) of edge in the landscape
between patch types (classes) i and J?; this includes
landscape boundary segments representing true edge
only involving patch type
i,
and
m’
is the number of
patch types (classes) present in the landscape, includ-
ing the landscape border if present. The interspersion
index ranges from 0 when a class type is adjacent to
only one other class type up to a maximum value of
100 when a class is equally adjacent to all other class
types (McGarigal and Marks, 1994).
We used SAS (Statistical Analysis Systems Insti-
tute Inc., 1988) general linear model (GLM) proce-
dures to test for differences in the percentage of the
home range comprised by each class and the inter-
spersion of classes within home ranges as a function
of: feeding-cover class (eight levels); home range
type (two levels); and site (three levels); plus the
first order interactions. Type III sums of squares,
which adjust the effects of each of the main factors
for all other factors in the model were used in model
hypothesis testing. Significant (F’ 5 0.05) main fac-
tors and interactions were used in a second GLM
model to estimate least-square mean (lsmean) values.
Using a Shapiro-Wilk W test (Zar, 1984), we dis-
covered the values for the percentage of the home
range comprised by each class were not normally
distributed (P < 0.05). Therefore, we transformed
these values using a natural-logarithm function to
satisfy the assumption of normality. The back trans-
formation of these values was done using a bias
corrected, transformation function (Bradu and Mund-
lak, 1970).
3. Results
GLM analysis indicated that the percentage of the
home ,range comprised by each class (Table 2) and
the interspersion of classes (Table 3) were correlated
Table 2
General linear model analysis of covariance for nahwal-log transformed levels of the Ixrcentage of a home range comprised by various
feeding-cover classes
Source
Feeding-cover class (FCC)
Home type (HRT)
range
Site
FCC X HRT
FCC X Site
HRT X Site
Error
d.f. Type III SS Mean square
I 334.0025972 47.8560853
1 0.0699741 0.0699741
2 2.6089197 I .3044599
7 0.3 I63596 0.045 1942
14 l4.llOl344 I .0078667
2 I .294329 I 0.647 1646
I57 3 I .9324540 0.2033914
F
value
Pr > F
235.29 O.OOOl
0.34 0.5584
6.41 0.002l
0.22 0.9797
4.96 O.OOOl
3.18 0.0442
162
B.B. Boroski et ul./ Forest Ecology und Management 88 (1996) 157-165
Table 3
General linear model analysis of covariance for levels of interspersion in 12 actual and random deer home ranges in Trinity County, CA
Source d.f. Type 111 SS Mean square
F value Pr > F
-- Feeding-cover class (FCC) 7 4264 1.35067 6091.62152 122.38 0.ooo1
Home type (HRT)
range I 68.32852 68.32852 I.37 0.243 1
Site 2 1366.35367 683. I7684 13.73 0.000 1
FCC X HRT 7 134.18408 19.16915 0.39 0.9 I (!O
FCC X Site 14 1337.31133 95.52224 1.92 0.0280
HRT X Site 2 969.3 1238 484.65619 9.73 o.ooo I
Error 157 7814.82048 49.77593 -~-
(P < 0.05) with feeding-cover class, site, the inter-
action between feeding-cover class and site, and the
interaction between home range type and site. On
average, the percentage of a home range comprised
by a class was greater
(P < 0.05)
in site 2 (In
lsmean = 1.89) than in sites 1 (In lsmean = 1.73) and
3 (In lsmean = 1.601, which were similar
(P = 0.14).
For some classes, the percentage of the home range
comprised by each class differed significantly, but
the extent of the differences between classes varied
by site (Table 4). The extent of class differences was
greatest in site 1, less so in site 2, and least in site 3
(Table 4).
Mean class size ranged from 0.06 to 7.79 ha.
Mean class size for all classes across all sites was
0.44 ha. In all sites, the dominant matrix was the low
feeding-high cover class (Table 4). Low feeding-low
cover and high feeding-low cover classes consis-
tently comprised the lowest percentage of the home
ranges (Table 4). In site 1, the mean percentage of
the home range comprised by a class was greater
(P = 0.02) for deer home ranges (In lsmean = 1.87)
than for random home ranges (In lsmean = 1.60).
The mean percentage of the home range comprised
by a class was similar for deer and random home
ranges in sites 2 (deer, In lsmean = 1.89; random, In
lsmean = 1.89,
P
= 1.0) and site 3 (deer, In lsmean
= 1.53; random, In lsmean = 1.68,
P = 0.24).
The mean interspersion for classes was signifi-
cantly different
(P
< 0.01) between sites (site 1.
lsmean = 70, site 2 lsmean = 74, site 3 lsmean = 67).
Interspersion values differed
(P
I 0.05) between
feeding-cover classes with the extent varying by site
(Table 5). Class differences were greatest in site I
and least in site 3 (Table 5) as they were with the
analysis of landscape composition (Table 4).
Across all sites, classes with a medium or high
feeding component had greater interspersion values
(Table 5). Low feeding-low cover and low
Table 4
Bias corrected, back-transformed, least-square means of the per-
centage comprised by each deer habitat suitability class. Least-
square means for each site by feeding-cover class were derived
from the general linear model analysis of covariance (see Table 2)
-
Site 1 Bowerman Ridge
LFLC HFLC MFMC HFMC MFLC MFHC LFMC LFHC
0.3 2.2 5.9 7.0 7.8 125 20.3 SO.8
Site 2 Farview Ridge
LFLC HFLC MFMC MFLC HFMC MFHC LFMC LFHC
0.8’ 2.7 5.1 8.4 9.2 11.1 18.3 52.4
Site 3 Rackerby Ridse
LFLC HFLC MFLC MFMC HFMC LFMC MFHC LFHC
0.4 05 7.0 7.2’ 9.6 11.9’ 15.6 45.9
All Sites
LFLC HFLC MFMC MFLC HFMC MFHC LFMC LFHC
05 1.5 5.5 7.8 8.5 12.9 16.4 49.3
Underlined classes represent no detectable difference
(P > OUS)
based on natural-log transformed least-square means. Asterisks
indicate differences
(P
5 0.05) by class between sites.
B.B. Bomski et ul./ Forest Ecology und Munagement 88 (1996) 157-165
163
Table 5
Least-square means of interspersion for feeding-cover habitat
suitability classes for deer found within sites. Least-square means
for each site by feeding-cover class were derived from the
general linear model analysis of covariance (see Table 3)
Site 1 Bowerman Ridge
LFLC LFMC MFHC LFHC MFMC HFLC MFLC HFMC
37.9 49.5 72.3 73.2 82.8 83.2 83.6 85.2
Site 2 Fairview Ridge
LFLC LFMC LFHC MFHC HFMC MFMC MFLC HFLC
42.1 50.8 76.2 81.3’ 84.1 84.2 86.5 87.7
. . .._. ..__. ._. . . .
__________.._._._._...~.~~...~.~.~.. _ . . ..--. _
.___.____ .._._.. . . _....._.. . . ..-.. . . .-. .-.. .
Site 3 Rackerby Ridge
LFLC LFMC HFLC MFHC LFHC MFMC HFMC MFLC
36.8 53.8 69.1’ 72.3 73.6’ 75.9’ 77.5 80.2
___....._.._._..__._~~.~.~~~. _ .-_. __ --_--.---
___._..._..._.________ _ ._.__..._. _.___ -.-..-
______ __.__ __ _....._. .._ _. . . .._. ._._. .._..
All Sites
LFLC LFMC LFHC MFHC HFLC MFMC HFMC MFLC
38.9 51.4 74.4 75.3 79.9 80.9 82.3 83.4
____ ___ __ _______._. .__... .__ . . . . .____ .__. ._.__ ._. .__.. . ._. . . . . .
Underlined classes represent no detectable difference (4’ > 0.05).
Asterisks indicate differences (P < 0.05) by class between sites.
feeding-medium cover classes had significantly
lower interspersion values in all sites (Table 5).
Interspersion values differed between sites for the
following classes: medium feeding-high cover;
medium feeding-medium cover; high feeding-low
cover; and high feeding-medium cover. The inter-
spersion of medium feeding-high cover was signifi-
cantly greater in site 2 (Table 5), while the intersper-
sion was significantly less in the classes medium
feeding-medium cover, high feeding-low cover, and
high feeding-medium cover in site 3 (Table 5). In
site 1, mean interspersion was significantly greater
( P = 0.000 1 I for deer home ranges (lsmean = 74.79)
than for random home ranges (lsmean = 67.15).
Mean interspersion values for deer and random home
ranges were similar (P > 0.20) in sites 2 and 3.
4. Discussion
Deer have been referred to as an edge species
because they use ecotones between forest, shrub, and
grassland or meadow habitats to meet feeding, cover,
and breeding requirements (Wallmo, 1981; Loft and
Menke, 1984). Deer use on low elevation range in
Trinity County, California was correlated with the
juxtaposition of abundant, available browse and good
hiding cover (Peterson, 1983; Loft and Menke, 1984).
It follows that the pattern of ecotonal boundaries
should influence use of areas by deer. This is of
particular interest as land management decisions can
strongly influence these patterns.
The percentage of home ranges comprised by
each feeding-cover class and the level of intersper-
sion within home ranges differed by class. We de-
tected differences between actual deer home ranges
and random home ranges of similar size only within
the site that had significant separation in the compo-
sition of classes and the interspersion of classes. This
suggests that landscape representations resulting in
classes with similar composition levels and similar
levels of interspersion at the scale of the animal’s
home range may hinder the assessment of habitat use
within home ranges by promoting Type II errors
(Manly, 1992).
Our results indicate that lsmean interspersion val-
ues
(site 2 = 74, site 1 = 70, and site 3 = 67) may be
useful as an index to the suitability of winter range
for deer. Interspersion values differed between sites
in some of the most important classes for deer;
medium feeding-high cover, medium feeding-
medium cover, high feeding-low cover, and high
feeding-medium cover. The interspersion of the
medium feeding-high cover class was greater in site
2, the site with the highest lsmean interspersion
value, and less in the classes medium feeding-
medium cover, high feeding-low cover, and high
feeding-medium cover in site 3, the site with the
lowest lsmean interspersion value. The interspersion
index has the qualities of a good habitat suitability
164 B.B. Bomski et cd./ Forest Ecology md Muntlgement 88 (1996) 157- I65
index, as it is directly comparable between sites and
is sensitive to relatively small differences, three units
in our case.
The class composition and interspersion patterns
of feeding-cover classes in the three sites in our
study suggest these measures may be of use in
making land management decisions. Within site 2.
interspersion was high in a variety of medium and
high feeding-cover classes making up relatively sim-
ilar percentages of the landscape. This is favorable
for deer and indicates that small changes in some
classes through timber management activities should
not be detrimental to deer.
In site 1, measures of class composition and
interspersion were separated along feeding and cover
vectors. Management activities might be directed at
manipulating forest stands to decrease class differ-
ences while maintaining or increasing interspersion.
Site 3 differed from the other two sites by lower
measures of class composition for several medium
cover associated classes and lower levels of inter-
spersion in feeding-cover classes with high and
medium feeding values. Land management activities
that increase the extent of habitat classes with
medium or high cover value while increasing the
interspersion of classes with high and medium feed-
ing value should benefit deer.
The concept of spatial scale encompasses both
extent and grain (Forman and Godron, 1986; Wiens,
1989; Turner et al., 1989). Too often, extent and
grain are dictated by technical limitations and not by
the organism under study. In this study, we demon-
strated that organism-centered specifications for ex-
tent and grain can be defined using functional-based
models applied to a generic structural representation
of the landscape having a minimum mapping unit of
0.06 ha. Spatial extent was defined using the adap-
tive-kernel home range estimator. Within a home
range, habitat suitability classes based on feeding
and cover produced the grain size for black-tailed
deer (mean 0.44 ha, range 0.06-7.79 ha). Further-
more, this approach portrayed the low feeding-high
cover habitat suitability class exerting the dominant
influence within the 12 deer home ranges sampled.
Thus, we suggest measures of landscape composition
and interspersion may be useful for characterizing
landscapes when assessing habitat use by deer. They
may also be of value to natural resource managers
attempting to predict the response of deer to pro-
posed changes in land use.
Acknowledgements
Funding was provided by the [JS Department ot
Interior Bureau of Reclamation, Trinity River Fish
and Wildlife Restoration Program, California Depart-
ment of Fish and Game, US Forest Service Pacific
Southwest Research Station, California Agricultural
Experiment Station Project 5410-MS. and the Golden
Gate Chapter of Safari Club International. B. Stans-
berry. J. Jones, M. Penhollow, and R. Bridgman
collected the home range data. We thank J. Pastor
and an anonymous reviewer for their comments on
the manuscript. Mention of trade names or products
is for information only and does not imply endorse-
ment by the US Department of Agriculture.
References
Airola, D.A., 1980a. California wildlife habitat relationships prcj-
gram, northeast interior zone. Reptiles and amphibians. Vol.
II. USDA, Forest Service, Lassen National Forest, Susanviile.
CA, 137 pp.
Airola, D.A., 1980b. California wildlife habitat reiationships pro-
gram, northeast interior zone. Birds. Vol. III. USDA. Forest
Service, Lassen National Forest, Susanville, CA. 590 pp.
Airola, D.A., 198oC. California wildlife habitat relationships pro-
gram, northeast interior zone. Mammals. Vol. IV. USDA.
Forest Service, Lassen National Forest, Susanviile, CA, 25.5
PP.
Airola. D.A., 1988. A guide to the California Wildlife Habitat
Relationships System. State of California, The Resource\
Agency, Department of Fish and Game, Sacramento, CA, 74
PP.
Boroski, B.B. and McLaughlin. P.L.. 1994. An efticient technique
for capturing swimming deer. Calif. Fish Game, 80: 36-41.
Bradu, D. and Mundlak, Y., 1970. Estimation in lognormal trnear
models. J. Am. Stat. Assoc., 65: 198-21 I.
Brown, G.K. and Fox, L., 1992. Digital classification of thematic
mapper imagery for recognition of wildlife habitat character--
istics. Proc. 1992 ASPRS/ACSM Convention, American Sc-
ciety for Photogrammetry and Remote Sensing, Bethesda,
MD.
Burton, T.S. and Monroe, G.S., 1983. Weaverville deer herd
management plan. California Department of Fish and Game.
Redding, CA, 83 pp.
Cleveland, W.S., 1979. Robust locally weighted regression and
smoothing scatterplots. J. Am. Stat. Assoc., 74: 829-836.
Davis. L.S. and Barrett, R.H.. 1993. Spatial integration of wildlife
B.B. Boroski et d/Forest Edogy und Manugement 88 (1996) 1.57-165
165
habitat analysis with long-term forest plamting over multiple-
owner landscapes. In: D.C. Le Master and R.A. Sedjo (Edi-
tors), Proc. Modeling Sustainahle Forest Ecosystems, Wash-
ington DC, pp. 143-154.
Dunaway, D., 1%4. A habitat management plan for the
Weaverville deer herd unit. USDA Forest Service, Shasta
Ttinity National Forest, 12 pp.
Dunaway, D., 1965. A habitat management plan for the New
River deer herd unit. USDA Forest Service, Shasta Trinity
National Forest, 17 pp.
Forman, R.T.T. and Godron, M., 1986. Landscape Ecology. 3ohn
Wiley, New York.
Grenfell, W.E., Jr., Salwasser, H. and Laudenslayer, W.F., Jr.,
1982. The California wildlife/fish habitat relationships sys-
tem. Cal-Nev Wildl. Trans. 1982, pp. 27-33. .
Hurley, J.F. and Asrow, E.S., 1980. Western Sierra
wildlife/habitat relationships computer user manual. WHR
Appl. Note 80-1, USDA, Forest Service, Pacific SW Region,
San Francisco, CA, 155 pp.
Larkm, R.P. and Halkin, D., 1994. A review of software packages
for estimating anima1 home ranges. Wildl. Sot. Buli., 22:
274-287.
Laudenslayer, W.F., Jr., 1982. Wildlife habitat relationships pro-
gram, northeast interior zone. Introduction and species-habitat
relationships matrix, Vol. I. USDA Forest Service, Pacific SW
Region, San Francisco, CA, 161 pp.
Laudenslayer, W.F., Jr., &don, M.F. and Airola, D.A., 1982.
California wildlife habitat relationships program: northeast
interior zone. Computer user’s manual, Vol. V. USDA, Forest
Service, Pacific SW Region, San Francisco, CA, 25 pp.
Loft, E.R. and Menke, J.W., 1984. Deer use and habitat character-
istics of transmission-line corridors in a Douglas-tir forest. J.
Wildl. Manage., 48: 1311-1316.
Loft, E.R., Menke, J.W. and Burton, T.S., 1984. Seasonal move-
ments and summer habitats of female black-tailed deer. J.
Wildi. Manage., 48: 1317-1325.
Manly, B.F.J., 1992. The Design and Analysis of Research Stud-
ies. Cambridge University Press, Cambridge, UK.
Marcot, B.G., 1979. California wildlife relationships program:
north coast/Cascades zone. USDA, Forest Service, Six Rivers
National Forest, Eureka, CA, 4 ~01s..
Mayer, K.E. and Laudenslayer, W.F., Jr., 1988. A guide to the
wildlife habitats of California. State of California, The Re-
sources Agency, Department of Fish and Game, Sacramento,
CA, 166 pp.
McCullough, D.R., 1975. Modification of the Clover deer trap.
Calif. Fish Game, 61: 242-244.
McGarigal, K. and Marks, B.J., 1994. Fragstats: Spatial pattern
analysis program for quantifying landscape structure, Version
2.0, Oregon State University, Corvalhs, OR, 67 pp.
Osaki, S.K., 1980. Wildlife habitat relationships: Lake Tahoe
Basin Management Unit. USDA, Forest Service, Lake Tahoe
Basin Management Unit, South Lake Tahoe, CA, 150 pp.
Peterson, G.E., 1983. Structural and vegetative components of
deer summer range in the Klamath Mountains. M.S. Thesis,
University of California, Davis, 44 pp.
Schoener, T.W., 1981. An empirically based estimate of home
range. Theor. Popul. Biol., 20: 281-325.
Shimamoto, K. and Airola, D.A., 1981. Fish and wildlife habitat
capability models and special habitat criteria for the northeast
interior zone forests. USDA, Forest Service, Modoc National
Forest, Alturas, CA, 260 pp.
Statistical Analysis Systems Institute Inc., 1988. SAS/STAT
User’s Guide, release 6.03. SAS Institute, Inc., Gary, NC,
1028 pp.
Stumpf, K.A. and Koltun, J.M., 1992. Rule based aggregation of
raster image classifications into vector GIS databases with
five- and forty-acre minimum size mapping units. Proc. 1992
ASPRS/ACSM Convention, American Society for Pho-
togrammetry and Remote Sensing, Bethesda, MD.
Thomas, J.W., 1979. Wildlife habitats in managed forests: the
Blue Mountains of Oregon and Washington. USDA, Forest
Service, Agricultural Handhook No. 553, Washington, DC,
512 pp.
Timossi, 1.C.. Sweet, A. and Dedon, M., 1989. Microcomputer
user’s manual for the California Wildlife Habitat Relationships
Database System. State of California, The Resources Agency,
Department of Fish and Game, Sacramento, CA, I3 pp.
Timossi, 1.C.. Sweet, A., Dedon, M. and Barrett, R.H., 1994.
User’s manual for the California Habitat Relationships micro-
computer database, Version 5.0. State of California, The Re-
sources Agency, Department of Fish and Game, Sacramento,
CA, 77 pp.
Turner, M.G., O’Neill, R.V., Gardner, R.H. and Milne, B.T.,
1989. Effects of changing spatial scale on the analysis of
landscape pattern. Landscape Ecol., 3: 153- 162.
United States Fish and WiIdIife Service, 1975. Deer loss compen-
sation program resulting from Trinity River Division, Central
Valley Project, California: A report to the Trinity River Basin
Fish and Wildlife Task Force. USDA Fish and Wildlife Ser-
vice, Portland, OR, 30 pp.
Vemer, J. and Boss, AS., 1980. California wildlife and their
habitats: western Sierra Nevada. Gen. Tech. Rep. PSW-37,
USDA, Forest Service, 439 pp.
Wallmo, O.C., 1991. Mule and Black-tailed De.er of North Amer-
ica. University of Nebraska Press, Lincoln, NE.
White, G.C. and Garrott, R.A., 1990. Analysis of Wildlife Radio-
tracking Data. Academic Press, New York.
Wiens, J.A., 1976. Population responses to patchy environments.
Amt. Rev. Ecol. Syst. 7: 81- 129.
Wiens J.A. Spatial scaling in ecology Funct. Ecol. 3 1989 385-
397.
Wilkinson, L., Hill, M.A., Wehta, J.P. and Birkenbeuel, G.K.,
1992. SYSTAT for Windows: Statistics, Ver. 5. SYSTAT,
Inc., Evanston, IL, 750 pp.
Worton, B.J., 1989. Kernel methods for estimating the utilization
distribution in home-range studies Ecology 7@ l64- 168.
Zar J.H., 1984. Biostatistical Analysis. Prentice-Hall, Engiewood
Cliffs. NJ.
