Content uploaded by Nigel C Bennett
Author content
All content in this area was uploaded by Nigel C Bennett
Content may be subject to copyright.
Ecological variables governing habitat suitability
and the distribution of the endangered
Juliana’s golden mole
Craig R. Jackson
1
, Trine Hay Setsaas
2
, Mark P. Robertson
3
& Nigel C. Bennett
1
*
1
Mammal Research Institute, Department of Zoology and Entomology, University of Pretoria, Pretoria, 0002 South Africa
2
Department of Biology, Norwegian University of Science and Technology, Realfagbygget, N-7491 Trondheim, Norway
3
Department of Zoology and Entomology, University of Pretoria, Pretoria, 0002 South Africa
Received 8 October 2007. Accepted 22 August 2008
Juliana’s golden mole (Neamblysomus julianae) occurs in three isolated populations in the
northeastern parts of South Africa. This cryptic species is not evenly distributed throughout its
restricted range and appears to have very specific habitat requirements. Its endangered status
reflects the necessity for a conservation management programme, which to date has not been
comprehensive. A primary hindrance to such initiatives has been the lack of information
pertaining to its habitat requirements. We assessed various soil and vegetation parameters, at
each population site, in areas where the animals were found to be present or absent. A multiple
logistic regression model highlighted the importance of soil hardness (governed by soil
particle size distribution), in combination with the cover provided by trees, as the two
ecological factors that best explained habitat suitability for Juliana’s golden mole at the three
localities. An IndVal analysis failed to identify any plant species that could reliably act as an
indicator of habitat suitability for this fossorial mammal. These results have important
implications for the conservation of the species.
Key words: Neamblysomus julianae, golden moles, small mammals, habitat suitability, habitat
requirements, distribution.
INTRODUCTION
Landscapes are naturally heterogeneous, compris-
ing a mosaic of various habitat types that result in
an uneven distribution of species over space
(Mauritzen et al. 1999; Sanderson et al. 2002).
Through evolutionary processes, organisms may
respond to this variation by becoming either niche
specialists or generalists (Elena & Sanjuán 2003;
Harmon et al. 2005). Highly specialized species are
often limited to a particular type of landscape,
being dependent on the ecological processes
associated with it (Harmon et al. 2005). Effective
conservation planning thus needs to consider the
heterogeneous and dynamic nature of ecosystems
(Huston 1994; Koehler 2000) as well as the under
-
lying processes that shape a species’ distribution
(Caughley & Gunn 1996).
Most golden mole species (Chrysochloridae) are
fossorial and difficult to detect in their subterra
-
nean niche. Furthermore, many species in this
50-million-year-old family (Stanhope et al. 1998)
are habitat-specific and are range restricted on
account of their acute adaptations to specific soil
conditions (Bronner 1997; Bronner & Bennett
2005). These attributes, in combination with
anthropogenic modification of habitats, are the
primary reasons why 10 of the 21 species appear
on the IUCN red list, with an additional three
species listed as data-deficient (Bronner 2006). The
2004 IUCN assessment of southern African mam
-
mals listed five golden mole species within the 10
most endangered mammal species. After De
Winton’s golden mole (Cryptochloris wintoni),
known from only a few specimens collected more
than 50 years ago (Bronner 1997), Juliana’s golden
mole (Neamblysomus julianae) is South Africa’s
second most endangered golden mole (Pretoria
population; Bronner 2006).
The species was only described in 1972 from a
specimen on the Bronberg Ridge (BR) in eastern
Pretoria, and was subsequently found at Nylsvley
Nature Reserve (NNR) 120 km to the north, and
the southwestern Kruger National Park (KNP)
350 km to the east (Meester 1972; Bronner &
Bennett 2005). The dearth of information available
for this species is typical for the majority of golden
mole species (Bronner & Bennett 2005). Juliana’s
golden mole is restricted to sandy soils (Skinner &
Smithers 1990), but individuals are not evenly
African Zoology 43(2): 245–255 (October 2008)
*Author for correspondence. E-mail: ncbennett@zoology.up.ac.za
distributed throughout this species’ geographic
range and appear to be restricted to particular
bushveld habitat types (C.R.J., pers. obs.).
Habitat requirements underlying the patchy
distribution of golden moles are not understood
and this data deficiency compromises conservation
efforts. Adequate habitats need to be identified
and conserved to protect and sustain the ecological
requirements of this species. This is especially
important for theBronberg populationof Juliana’s
golden mole, which probably had a historical
distribution in an area approximately 15 km
long by 1 km wide that is now estimated to have
declinedby 80% (Bronner2006)because the natural
habitat has been transformed by roads and hous
-
ing.The othertwopopulations ofthisspecies have
also seen large reductions in their available habitat
resulting from various land use practices (includ
-
ing cultivation) resulting in extensive habitat loss
and fragmentation (Bronner 2006). Some mammal
species may be well adapted to these highly frag-
mented habitats (Terborgh 1974; Barko et al. 2003)
but golden moles are certainly not because of their
specialized adaptationsfor their subterranean life,
coupled with poor dispersal capabilities.
Dealing with the special habitat requirements,
poor dispersal capabilities, sensitivity to human
disturbance, and low detection probabilities typi-
cally associated with many rare, habitat specific or
cryptic species, pose serious challenges for conser-
vation planning (Piggott & Taylor 2003). Conser
-
vation efforts are desperately required to address
the issue of habitat connectivity and protection for
Juliana’s golden mole. This necessitates a thorough
understanding of the ecological requirements
and distribution of the species, both at local and
regional scales. This paper documents the first
investigation into the habitat characteristics asso
-
ciated with the distribution of Juliana’s golden
mole.
METHODS
Habitat requirements were inferred by the assess
-
ment of soil and vegetation properties at study
plots where signs of Juliana’s golden mole were
present and in plots where they were absent.
Fresh signs of golden mole burrowing activity in
the form of newly pushed ridges are only observed
after periods of rain, which softens the soil and
makes it more cohesive. During the summer rain
-
fall months of February and March 2005, soil and
vegetation properties were recorded in 48 study
plots (NNR: 12, BR: 13, KNP: 23) and incorporated
both presence and absence of the Juliana’s golden
mole. Study plots consisted of randomly selected
5×5mplots, with a minimum distance of 100 m
between plots. We ensured that the vegetation in
the plots was representative of the surrounding
vegetation by visually assessing the dominant
species and vegetation structure in and around
the plot.
Determination of presence/absence status
Species presence/absence surveys are commonly
used in ecology and conservation management,
yet they can never be used to confirm that a
species is absent from a given location (MacKenzie
et al. 2002; Guisan & Zimmerman 2000; MacKenzie
2005). Failure to detect the presence of a species in
an occupied habitat patch is a common sampling
problem when a) the population size is small, b)
individuals are difficult to sample, or c) sampling
effort is limited (Gu & Swihart 2004). Detecting
the presence of Juliana’s golden mole is difficult
because of the animal’s cryptic ways. Another
difficulty is that not all suitable habitat is always
occupied, which may result in false negative
observations (Fielding & Bell 1997).
Juliana’s golden mole does not usually move
aboveground or produce conspicuous soil mounds
on the soil surface and it is very difficult to catch.
This means that one has to observe foraging
tunnels to be certain that the species is present
and that the habitat is suitable. Ecological data
collectedfor presence plotsin this studywillthere
-
fore accurately describe the habitat of the species.
Certain plots where the species is considered to be
absent may have been patches of unoccupied but
suitable habitat or apatch in which the presence of
the animal went undetected. However, all plots
were carefully searched to ensure that if the animal
was present that it was detected. Despite certain
limitations, this method suffices to highlight
habitat preferences of the species thus aiding
conservation planning. The likely consequence of
false negative errors is to reduce the magnitude
of difference observed between habitat that is
considered to be suitable and habitat that is
considered to be unsuitable.
Vegetation survey
In each plot, the cover of each plant species was
assessed using the Braun-Blanquet scale (Kent &
Coker 1995). Total vegetation cover for a plot was
assessed by calculating the total surface area of the
25 m
2
quadrat, expressed as a percentage, that had
246 African Zoology Vol. 43, No. 2, October 2008
some form of vegetation cover emanating from it.
Furthermore, the relative cover provided by the
tree, shrub and herbaceous layers was evaluated.
Tree density (individuals greater than2min
height), average height (to the nearest half-
metre) and canopy cover (percentage of total
quadrat) was assessed within a greater 10 × 10 m
area that incorporated the original 25 m
2
plot. The
larger area provided a more realistic representa
-
tion of these properties within the general land
-
scape.
Soil properties
At allfour corners of aquadrat, soilsamples were
augured (auger bucket 80 mm in diameter), soil
depth evaluated to a maximum depth of one
metre (measuring maximum auger depth), and a
static cone-penetrometer (Herrick & Jones 2002)
assessment was undertaken. Soil samples were
transferred directly into plastic bags. Soil texture
was ascertained after drying samples in an oven at
40°C to constant mass.
Soil texture provides an indication of the relative
proportions of the various separates in the soil
(Van der Watt & Van Rooyen 1995). The particle
size distribution or range of particle sizes in a
sample influenceseveral soil properties, including
compaction and soil permeability to water (Brady
& Weil 1999). A well-graded soil refers to the
constituent particles being distributed over a wide
range of sizes and, conversely, a uniformly or
poorly graded soil refers to the size of particles
being distributed over a narrow size range. A
representative sub-sample (approximately 500 g)
was putthrough a series of nine sieves that ranged
in size from 8 mm to 0.063 mm and were vibrated
on an electronic shaker for 10 minutes (Briggs
1977). Individual sieve contents were weighed
and the relative proportion of the sample calcu
-
lated, thereby giving the particle-size distribution
of each sample.Shannon’s Diversityindex(H) was
used to assess the degree of heterogeneity in soil
particle-size distributions for presence and absence
plots (Smith & Smith 2003).
A penetrometer (Eikelkamp Agricultural Instru
-
ments, Netherlands), consisting of a rigid cone-
tipped rod attached to a pressure-measuring
device (proving ring), was used to determine the
hardness of the soil. During testing the penetrom
-
eter was pushed into the ground at a slow, steady
speed and the soil resistance was measured at
50-mm intervals, beginning at the soil surface. The
measure of soil strength (or resistance) was taken
directly from the dial on the proving ring and the
units of measurement are presented in kN/50 mm.
These measurements facilitate the quantitative
characterization of soil hardness, which is thought
to be one of the major factors governing soil suit
-
ability for golden moles.
Statistical analysis
IndVal Indicator Analysis. The association of
specific plant species with golden mole presence
or absence was investigated using IndVal-indica
-
tor analysis (Dûfrene & Legendre 1997) pro
-
grammed in M
ATLAB
®
R2006a (The MathWorks,
Inc., U.S.A.). The indicator value method facili
-
tates the identificationof indicator speciesfor a pri
-
ori established groups of samples (in this case
presence and absence). The indicator value of a
species i in group j is calculated by multiplying its
group specificity (A
ij
) with its group fidelity (B
ij
).
A
ij
is the mean abundance of the species i in the
sites of group j relative to its abundance in all
groups considered. B
ij
is the relative frequency of
occurrence of species i in the sites of group j:
A
ij
= (n individuals
ij
)/(n individuals
i
)
B
ij
= (n sites
ij
)/(n sites
j
)
IndVa
L = (A
ij
) × (B
ij
) × 100
The resulting indicator value is expressed as
percentage of perfect indication (i.e. when all
individuals of a species are found only in one of
the a priori established groups and when the
species occurs in all sites of that group). A species
with high specificity and high fidelity will have a
high indicator value (McGeoch et al. 2002).
Multiple logistic regression analysis
The presence of the golden mole in a plot can
be considered as a binominal process; it is either
present with a probability of P or absent with a
probability of 1 – p, where var (p)=p (1 p). Such
data are therefore suitable for logistic regression
analysis by means of generalized linear mixed
effects models. Population was included as a ran
-
dom effect to incorporate spatial variability. A
global model was produced to investigate the rela
-
tive importance of each explanatory factor (ecologi
-
cal variables).This model applieda multiple regres-
sion analysis to the explanatory factors. Eleven
alternative and biologically relevant candidate
models, all derived from the global model, were
analysed. The models incorporated the fact that
data were collected from three different popula
-
Jackson
et al.
: Habitat and distribution of Juliana’s golden mole 247
tions. Akaike’s Information Criterion (AIC;
Burnham & Anderson 2002), based on the princi
-
ple of parsimony, was employed to select the most
appropriate model. In particular, the equation
AIC log
ce
LK
KK
nK
=− + +
+
−−
22
21
1
(
$
))
()
()
θ
was used and can account for low sample sizes. K
denotes the number of parameters estimated in
the model, and n is the sample size. In addition,
Akaike weights were calculated as:
ω
i
i
r
R
=
−
=
∑
exp(
exp(–
∆
∆
/)
/)
2
2
1
r
.
In the AIC sense, this can be interpreted as the
probability that a specific model is the best, in that
it minimizes the expected Kullback-Leibler (KL)
discrepancy, given the data and the set of candi
-
date models (Burnham & Anderson 2002).
The model with the highest Akaike weights (ω)
consequently explains the most variation using
the fewest parameters. At the same time the bio-
logical relevance of the models must always be
carefully considered (Burnham & Anderson 2002).
Because all the models had a similar structure and
sample size, the respective AIC
C
values from the
different models were comparable. In order to
calculate the confidence intervals for the coeffi-
cients of the best model, the Markov Chain Monte
Carlo method wasused, whichgenerates asample
from the posterior distribution of the parameters
of the fitted model (R Development Core Team
2006). Coefficients were considered significant
when confidence intervals did not overlap with
zero (Burnham & Anderson 2002).
Owing to the problem of multi-collinearity in
multiple regression analyses, the pairwise correla
-
tion between all the variables was evaluated. In
particular, there was strong inter-correlation among
the various tree variables and soil variables. Thus,
to avoid multi-collinearity among the explanatory
variables in the multiple regression analysis as
well as to reduce the number of explanatory
factors, the first component from two principal
component analyses (PC1-factor; SPSS 2005) was
used. For the vegetation variables, the PC1-factor
was based on the measurements of tree layer,
average tree height, canopy cover and tree den
-
sity, while the Shannon index (particle size distri
-
bution) and soil hardness were included for the
soil variables. For the tree measurements, the PC1
accounted for 63% of the total variance of the
variables included in the analysis (Eigenvalue =
2.53), and was positively correlated with canopy
cover (r = 0.875, P < 0.001), tree layer (r = 0.841,
P < 0.001), average tree height (r = 0.802, P <
0.001) and tree density (r = 0.640, P < 0.001). The
PC1-factor can thus be interpreted as the physio
-
gnomic properties of trees based on the variables
included in the principal component analysis, and
will hereafter be referred to as the ‘tree factor’. For
the soil measurements, thePC1 accounted for 80%
of the total variance of the variables included
in the analysis (eigenvalue = 1.59), and was
positively correlated with Shannon’s index (r =
0.892, P < 0.001) and soil hardness (r = 0.892, P <
0.001). The PC1-factor in this principal component
analysis can therefore be interpreted as the soil
variables governing and describing soil hardness
and compaction, and will hereafter be referred to
as the ‘soil factor’.
RESULTS
Plant species
A total of 245 plant species were recorded from
all of the study plots (Appendix A). The indicator
analysis did not yield any species that had both a
high specificity and fidelity to presence sites. A
subjective benchmark value of 70% IndVal was
used to denotean indicatorspecies (McGeoch etal.
2002). Only three species emerged at each popula
-
tion that had significant probabilities, but these all
had low IndVal values, ranging from 20.00% to
55.56%(Table 1).Thehighestrankedspeciesacross
all plots had an IndVal value of 55.56% and was a
grass species, Eragrostis trichophora, found in BR.
Multiple logistic regression
The logistic regression model investigated the
relative importance of all the environmental
explanatory factors on the probability of finding
Juliana’s golden mole present in the study plots.
Initially, a global multiple regression model was
constructed that included all explanatory variables,
namely: soil factor, tree factor, soil depth, vegeta
-
tion cover andshrublayer. Then anumber of alter
-
native models wererun(Table 2).The AIC
C
criteria
favoured a model that included a significant effect
of tree factor and soil factor. The tree factor was
positively associated with the presence of the
mole, indicating that the probability of golden
moleoccurrenceincreasedas tree factorincreased.
In light of the variables comprising the ‘tree
248 African Zoology Vol. 43, No. 2, October 2008
factor’, this indicates that an increase in tree
height, layer, density, and canopy cover increases
the habitat suitability for Juliana’s golden mole.
The soil factor was negatively associated with the
presenceof thegolden mole, indicatingthat occur
-
rence of the mole decreased as soil factor increased.
Thus, as soil particle size heterogeneity (Shannon’s
index) increases, which in turn results in a corre
-
sponding increase in soil hardness, the probability
of finding Juliana’s golden mole decreases.
Statistical support for the chosen model was not
as pronounced when compared to the second best
model, as it was compared to the third, fourth and
fifth best models, as the two highest ranked
models had both a ∆AIC
C
less than 2 (Table 2). The
Akaike weight criteria suggested that the highest
ranked model was 1.66 times more likely to be the
KL best model compared to the second ranked
model (Table 2), which also included the effect of
shrub layer. Accordingly, this indicates that shrub
layer, which was positively correlated with the
occurrence of the mole, should be considered as a
contributing component of the characteristics that
determine the habitat suitability for Juliana’s
golden mole. Table 3 shows the statistical attrib
-
utes for components incorporated in the best
model.
DISCUSSION
IndVal
Climate and substrate are the two determining
factors for vegetation growth and each ecological
region has its own set of unique variables that
Jackson
et al.
: Habitat and distribution of Juliana’s golden mole 249
Table 1. Plant species associated with presence plots at each population that had significant
probability values. The corresponding IndVal values are presented.
Population Species IndVal Probability
KNP
Ascolepis capensis
20.00 0.010
Helichrysum acutatum
20.00 0.008
Tristachya leucothrix
20.00 0.006
NNR
Acacia toritillis
28.57 0.044
Dichrostachys cinerea
28.57 0.049
Justicia flava
28.57 0.043
BR
Eragrostis trichophora
55.56 <0.001
B8-64 (unidentified) 44.44 <0.001
Combretum molle
33.33 0.015
Table 2. The candidate models explaining the probability of finding Juliana’s golden mole present in study plots. The
models are based on multiple logistic regression analyses and ranked according to descending values of the Akaike
weights (ω
i
). According to the principle of parsimony, the model with the highest ω
i
explains most of the variation
using the fewest parameters.
K
indicates the number of model terms plus one for intercept and error term, AIC
C
repre
-
sents Akaike information criterion corrected for small sample size, and ∆AIC
C
denotes the deviance in AIC
C
from the
model with the lowest AIC
C
. The table lists the five best candidate models out of 11 potential models.
Model
K
AIC
C
∆AIC
C
ω
i
Soil factor + Tree factor 3 50.33 0.00 0.335
Soil factor + Tree factor + Shrub layer 4 51.34 1.01 0.202
Soil factor 2 52.46 2.13 0.116
Soil factor + Tree factor + Soil depth 4 52.66 2.33 0.105
Soil factor + Tree factor + Veg. cover 4 52.76 2.43 0.099
Table 3. Factors affecting the probability of finding
Juliana’s golden mole present in study plots. β, S.E.,
t
and CI
min
,CI
max
denote the regression coefficient,
standard error,
t
-value and 97.5% confidence interval for
the coefficients, respectively. Coefficients are consid
-
ered significant when confidence intervals do not overlap
with zero.
Coefficients: S.E. T CI
min
, CI
max
(Intercept) 1.527 0.088 17.421 1.362, 1.694
Soil factor –0.334 0.065 –5.174 –0.449, –0.200
Tree factor 0.112 0.057 1.946 0.009, 0.230
influence the plant community structure (Breden
-
kamp & Brown 2001). The IndVal analysis investi
-
gated the potential use of sympatric plant taxa as
indicator species, but did not yield any associa
-
tions of great affinity. An important consideration
is the size andnumber of sample plots. A relatively
low sample size incorporating 5 × 5 m quadrats
may not have been sufficiently large to capture all
plant species occurring withinthe suitablehabitat.
Species that occur at lower densities within this
region may thus have only been recorded in a
small percentage of sample plots. Their signifi
-
cance could potentially be overlooked or the
projected probability underestimated when using
the IndVal analysis. It is thus difficult to interpret
the present results of the IndVal analysis. Although
IndVal values were significant for some species
all species showed low indicator values (ranging
between 20.00 and 55.56%), whereas values
greater than 70% would typically be required be
-
fore a species is considered to have a meaningful
indicator value (McGeogh et al. 2002). This means
that none of the species sampled were present in
most of the plots where Juliana’s golden mole was
recorded as present (presence plots) and absent
from most of the absence plots.
The results from this study suggest that none
of the species recorded can be considered to be
reliable indicator species for Juliana’s golden mole
habitat.
Dense stands of large (>3 m) Terminalia sericea
are very common in both the NNR and KNP, but
are absent from the BR. These species are good
indicators of sandy soils and often highlight areas
that are inhabited by Juliana’s golden mole (C.R.J.,
pers. obs.). This species did not feature strongly in
the IndVal analysis, most likely because it does not
occur at the BR population. When combining the
plant species from all three populations, this
would directly affect the IndVal analysis. Addi
-
tionally, the subjective placement of the relatively
small survey quadrats may not have adequately
captured the occurrence of these trees. Since this
paper aims to describe the habitat utilized by
Juliana’s golden mole, it is important to state the
nature of this association. In areas of mixed
bushveld, Terminalia sericea often occurs in a
shrub-like form and does not form dense stands of
tall conspecifics. It should be noted that in these
instances the presence of the species is not usually
associated with loose sandy deposits, and would
consequently not indicate potentially suitable
habitat (C.R.J., pers. obs.).
Multiple logistic regression
The multiple logistic regression revealed that
the model including the soil factor (derived from
particle size distribution (PSD) and soil hardness)
in combination with the tree factor (derived from
tree density, layer, height, and canopy cover) best
explained the probability of habitat being suitable
for Juliana’s golden mole. The model indicated a
positive relationship between the presence of
Juliana’s golden mole and the tree factor, but an
inverse association between the soil factor and
the presence of the golden mole. The soil factor
accounts for more variation than tree factor, and
this variable (on its own) is ranked as the third best
model overall. Based on the above, we can con
-
clude that a poorly graded soil largely determines
habitat suitability for Juliana’s golden mole, but
with an increasing amount of shade provided by
trees, the habitat suitability greatly increases.
Poorly graded soil, with a more uniform PSD, is
more resistant tocompaction while theavailability
of various particles sizes in well-graded soils makes
such substrates prone to compaction (Brady &
Weil 1999). The significant positive correlation
between soil PSD and soil hardness confirms that
as the heterogeneity in the PSD increases, soil
hardness also increases (Jackson et al. 2008).
The PSD regulates the ease of tunnelling and
consequent energy expenditure, with the small,
35 g golden moles being confined to softer soils.
The high energetic costs of underground locomo
-
tion by subterranean mammals have been well
documented (Nevo 1979; Bennett & Faulkes 2000;
Luna et al. 2002; Luna & Antinuchi 2006) and
burrowing efficiency is affected primarily by the
relationship between soil hardness and the cost of
tunnelling (Luna & Antinuchi 2006). The Namib
Desert golden mole (Eremitalpa granti namibensis)
occurs in exceptionally loose dune sands. In this
species, the gross energy cost of sand swimming
was found to be 26 times more expensive than
running on the surface (Seymour et al. 1998). Sand
swimming through soft sands, however, only
incurred energetic costs amounting to less than
one tenth of the energyrequired bymammals tun
-
nelling through compact soils (Seymour et al.
1998). The physical and energetic implications of
harder soils would consequently make occupying
such soils unfeasible.
The roleof the physiognomic vegetation proper
-
ties in the two best models emphasize the contribu
-
tion of vegetation structure to habitat suitability.
Increased vegetation cover would be expected to
250 African Zoology Vol. 43, No. 2, October 2008
have a major influence on microclimatic condi
-
tions such as soil temperature and soil moisture. A
combination of these two factors, specifically in
the relativelyhot and dry conditions characteristic
of all three populations, would directly affect soil
hardness and consequently the optimal micro
-
habitatconditions for the species(Jackson2007). In
addition, soil moisture and temperature affect the
biology of soil micro-arthropods (Choi et al. 2002)
withmoister rather thandrier conditions expected
to be more favourable for these organisms (Ivask
et al. 2006). Such conditions would most likely
support a greater abundance and diversity of soil
invertebrates thus making food acquisition easier
for golden moles. Subterranean mammals occupy
an energy-restricted environment, and when
locomotion and food acquisition is energetically
demanding,a reliablefoodsource is ofvital impor
-
tance.
The effect of PSD on vegetation structure chiefly
concerns soil permeability andassociated nutrient
status. Characteristic tree species that are frequently
associated with well-drained, nutrient poor sandy
soils in the north eastern parts of South Africa
include Terminalia sericea and Burkea africana (Low
& Rebelo 1996). These broadleaved species have
large spreading canopies that provide shade and
are frequently encountered in Juliana’s golden
mole habitat (C.R.J., pers. obs.).Leaves thatare lost
by the trees also accumulate on the soil surface,
further buffering the substrate from direct sun
-
light. Limiting the amount of sunlight on the soil
surface would keep the soil temperature lower
and the soil moist for longer periods. These vari
-
ables have been shown to be important for both
daily and seasonal activity patterns in Juliana’s
golden mole (Jackson 2007). Poorly graded sandy
soils with a very sparse vegetation cover may thus
not provide microclimatic conditions essential to
Juliana’s golden mole, as soils would be hotter and
harden more rapidly after rainfall events, limiting
the window period during which foraging would
be possible (Jackson 2007).
Although the IndVal analysis failed to highlight
the importance of any one species as an indicator
of the presence of N. julianae, the multiple logistic
regression highlighted the importance of a well-
established layer of vegetation for this species. It
appears that the role of the vegetation in shaping
microclimatic conditions is more important than
the presence of any particular plant species.
This study suggests that conserving this species
wouldentail more thanjust protecting sandysoils.
The structural properties of the sandy soils are
important, as well as the amount of vegetation
cover. Considering vegetation structure in conser
-
vation programmes would be of great importance
since many areas in southern African have seen
indigenous trees removed for firewoodand natural
vegetation replaced by monocultures. This would
leave the soil surface more exposed and alter
components of the natural ecosystem such as soil
temperature and moisture. These changes may be
very detrimental to this highly habitat-specific
golden mole species. Consequently, the effect of
land use is also of importance for conserving the
species as it may not survive successfully on
private land should the land use be of an incom
-
patible nature.
ACKNOWLEDGEMENTS
For assistance in the field we would like to thank
Bernard Coetzee, Marna Broekman, and Natalie
Callaghan. The guidance and assistance provided
by Magda Nel, whilst attempting to identify plant
specimens in the herbarium, is greatly appreci-
ated. Steven Khoza and Samuel Nkuna provided
security during data collection in the KNP, while
Thembi Khoza and Andrew Deacon provided
logistical assistance in the park. Thor-Harold
Ringsby kindly providedadvice onstatistical anal-
yses. C.R.J. was supported by a National Research
Foundation grant holders bursary to N.C.B., while
all fieldwork was funded by the WWF Green Trust
(South Africa).
REFERENCES
BARKO, V.A., FELDHAMER, G.A., NICHOLSON, M.C.
& DAVIE, D.K. 2003. Urban habitat: a determinant of
white-footed mouse (Peromyscus leucopus) abundance
insouthern Illinois.Southeastern Naturalist 2:369–376.
BENNETT, N.C. & FAULKES, C.G. 2000. African Mole-
rats: Ecology and Eusociality. Cambridge University
Press, Cambridge, U.K.
BRADY,N.C.&WEIL,R.R.1999.TheNature and Properties
of Soils, 12th edn. Prentice Hall, New Jersey, U.S.A.
BREDENKAMP, G.J & BROWN, L.R. 2001. Vegetation – a
reliable ecological basis for environmental planning.
Urban Greenfile, Nov/Dec 2001.
BRIGGS, D. 1977. Sediments. Butterworths, Boston.
BRONNER, G. 1997. Golden moles (Chrysochloridae).
In: The Complete Book of Southern African Mammals,
(eds)G.Mills&L.Hes. StruikPublishers,Cape Town.
BRONNER, G.N. 2006. Neamblysomus julianae. In: IUCN
2006. 2006 IUCN Red List of Threatened Species. Online
at:www.iucnredlist.org (accessed 30 December2006).
BRONNER, G.N. & BENNETT, N.C. 2005. Order
Afrosoricida: In The Mammals of the Southern African
Subregion, (eds) J.D. Skinner & C.T. Chimimba, 3rd
edn. Cambridge University Press, Cambridge.
Jackson
et al.
: Habitat and distribution of Juliana’s golden mole 251
BURNHAM, K.P. & ANDERSON, D.R. 2002. Model Selec
-
tion and Multi-model Inference. A Practical Informa
-
tion-theoretic Approach, 2nd edn. Springer, New York.
CAUGHLEY,G.C. & GUNN, A. 1996. ConservationBiology
inTheoryandPractice. Blackwell Science,Cambridge.
CHOI, W.I., RYOO, M.I. & KIM, J. 2002. Biology of
Paronychiurus kimi (Collembola: Onychiuridae)
under the influences of temperature, humidity and
nutrition. Pedobiologia 46: 548–557.
DUFRÊNE, M. & LEGENDRE, P. 1997. Species assem
-
blages and indicator species: the need for a flexible
asymmetrical approach. Ecological Monographs 67:
345–366.
ELENA, S.F. & SANJUÁN, R. 2003. Climb every moun
-
tain? Science 302: 2074–2075.
FIELDING, A.H. & BELL, J.F. 1997. A review of methods
for the assessment of prediction errors in conserva
-
tion presence/absence models. Environmental Conser
-
vation 24: 38–49.
GU, W. & SWIHART, R.K. 2004. Absent or undetected?
Effects of non-detection of species occurrence on
wildlifeh – habitat models. Biological Conservation,
116: 195–203.
GUISAN, A. & ZIMMERMAN, N.E. 2000. Predictive
habitat distribution models in ecology. Ecological
Modelling 135: 147–186
HARMON, L.J., KOLBE, J.J., CHEVERUD, J.M. &
LOSOS, J.B. 2005. Convergence and the multidimen-
sional niche. Evolution 59: 409–421.
HERRICK, J.E. & JONES, T.L. 2002. A dynamic cone pen-
etrometer for measuring soil penetration resistance.
Soil Science Society of America Journal 66: 1320–1324.
HUSTON, M.A. 1994. Biological Diversity: the Coexistence
of Species on Changing Landscapes. Cambridge Univer-
sity Press, Cambridge.
IVASK, M., KUU, A., TRUU, M. & TRUU, J. 2006. The
effect of soil type and soil moisture on earthworm
communities. Journal of Agricultural Science 17: 7–16.
JACKSON, C.R., LUBBE, N.R., ROBERTSON, M.P.,
SETSAAS, T.H., VAN DER WAALS, J. & BENNETT,
N.C. 2008. Soil properties and the distribution of the
endangered Juliana’s golden mole. Journal of Zoology,
London 274: 13–17.
JACKSON, C.R. 2007. The ecology and conservation of
Juliana’s golden mole (Neamblysomus julianae). M.Sc.
thesis, University of Pretoria, Pretoria.
KENT, M. & COKER, P. 1995. Vegetation Description and
Analysis, a Practical Approach. John Wiley & Sons,
Chishester.
KOEHLER, H.H. 2000. Natural regeneration and succes
-
sion – results from a 13 years study with reference to
mesofauna and vegetation, and implications for
management. Landscape and Urban Planning 51:
123–130.
LOW, A.B. & REBELO, A.G. (eds) 1996 Vegetation of South
Africa, Lesotho and Swaziland. Department of Environ
-
mental Affairs and Tourism, Pretoria.
LUNA, F., ANTINUCHI, C.D. & BUSCH, C. 2002.
Digging energetics in the South American rodent
Ctenomys talarum (Rodentia, Ctenomyidae). Canadian
Journal of Zoology 80: 2144–2149.
LUNA, F. & ANTINUCHI, C.D. 2006. Cost of foraging in
the subterranean rodent Ctenomys talarum: effect of
soilhardness. Canadian Journalof Zoology84:661–667.
MACKENZIE, D.I. 2005. Was it there? Dealing with
imperfect detection for species presence/absence
data.Australianand NewZealandJournal ofStatistics47:
65–74.
MACKENZIE, D.I., NICHOLS, J.D., LACHMAN, G.B.,
DROEGE, S., ROYLE, J.A. & LANGTIMM, C.A. 2002.
Estimating site occupancy rates when detection
probabilities are less than one. Ecology 8: 2248–2255.
MAURITZEN, M., BERGERS, S.J.M., ANDREASSEN,
H.P., BUSSINK, H. & BARENDSE, R. 1999. Root vole
movement patterns: do ditches function as habitat
corridors? Journal of Applied Ecology 36: 409–421.
McGEOCH, M.A., VAN RENSBURG, B.J. & BOTES, A.
2002. The verification and application of bioindi
-
cators: a case study of dung beetles in a savanna eco
-
system. Journal of Applied Ecology 39: 661–672.
MEESTER, J. 1972. A new golden mole from the
Transvaal (Mammalia: Chrysochloridae). Ann. Trans
-
vaal Mus. 28: 35–46.
NEVO, E. 1979.Adaptive convergence and divergence of
subterranean mammals. Annual Review of Ecology and
Systematics 10: 269–308.
PIGGOTT, M.P. & TAYLOR, A.C. 2003. Remote collection
of animal DNA and its applications in conservation
management and understanding population biology
of rare and cryptic species. Wildlife Research 30: 1–13.
R DEVELOPMENT CORE TEAM. 2006. R: A Language
and Environment for Statistical Computing, R Foun-
dation for Statistical Computing, Vienna, Austria.
SANDERSON, E.W., REDFORD, K.H., VEDDER, A.,
COPPOLILLO, P.B. & WARD, S.E. 2002. A conceptual
model for conservation planning based on landscape
species requirements. Landscape and Urban Planning
58: 41–56.
SEYMOUR, R.S., WITHERS, P.C. & WEATHERS, W.W.
1998. Energetics of burrowing, running, and free-
living in the Namib Desert golden mole (Eremitalpa
mamibensis). Journal of Zoology, London 244: 107–117.
SKINNER, J.D. & SMITHERS, R.H.N. 1990. The Mammals
of the Southern African Subregion, 2nd edn. University
of Pretoria, Pretoria
SMITH, R.L. & SMITH, T.M. 2003. Elements of Ecology, 5th
edn. Benjamin Cummings, San Francisco.
SPSS 2005. SPSS Inc., Chicago.
STANHOPE, M.J., WADDELL, V.G., MADSEN, O., DE
JONG, W., HEDGES, S. B., CLEVEN, G.C., KAO, D.
& SPRINGER, M.S. 1998. Molecular evidence for
multiple origins of Insectivora and for a new order of
endemic African insectivore mammals. Proceedings of
the National Academy of Sciences U.S.A. 95: 9967–9972.
TERBORGH, J. 1974. Preservation of natural diversity:
The problem of extinction prone species. BioScience,
24: 715–722.
VAN DER WATT, H.V.H. & VAN ROOYEN, T.H., 1995: A
Glossary of Soil Science, 2nd edn, The Soil Science soci
-
ety of South Africa, Pretoria.
ZAR, J.H. 1984. Biostatistical Analysis.
Prentice-Hall,
Englewood Cliffs, NJ.
Responsible Editor: G.N. Bronner
252 African Zoology Vol. 43, No. 2, October 2008
Jackson
et al.
: Habitat and distribution of Juliana’s golden mole 253
Table A1. Plant species recorded in presence and/or
absence study quadrats on the Bronberg Ridge,
Gauteng Province, South Africa.
Species name Presence Absence
Acacia
sp. †
Acalypha angustata
†
Acalypha villicaulus
†
Adenia glauca
†
Aloe
sp. † †
Argyrolobium pauciflorum
†
Aristida adscensionis
†
Aristida congesta
†
Aristida transvallensis
†
Asparagus africanus
†
Athrixia elata
†
Barleria
sp. † †
Berkheya seminivera
†
Bidens pilosa
†
Brachiaria serrata
†
Burkea africana
†
Canthium gilfillanii
†
Canthium mundianum
†
Cheilanthus hirta
†
Cleome maculata
††
Combretum molle
†
Commelina africana
†
Cryptolepis oblongifolia
†
Cymbopogon plurinodis
†
Dichoma zeyheri
†
Diheteropogon amplectens
††
Elephantorrhiza burkei
†
Elephantorrhiza elephantina
†
Eragrostis curvula
†
Eragrostis gummiflua
†
Eragrostis trichophora
††
Eustachys paspaloides
†
Foeniculum vulgare
†
Hebenstretia angolensis
†
Helichrysum rugulosum
††
Hibiscus trionum
†
Hypharrenia filipendula
†
Indigofera hilarus
††
Justicia anagalloides
††
Leonotis
sp. 1 †
Loudetia simplex
††
Melinis nerviglumis
†
Melinis repens
††
Ochna pulchra
††
Oldenlandia corymbosa
††
Opuntia ficus-indica
†
Parinari capensis
†
Pearsonia sessilifolia
†
Species name Presence Absence
Pellaea calomelanos
††
Rhynchosia caribaea
†
Rhynchosia minima
Rhynchosia nitens
††
Rhynchosia totta
†
Senecio inornatus
††
Senecio macrocephalus
†
Senecio
sp. 1 †
Setaria sphacelata
††
Strychnos pungens
†
Tagetes minuta
†
Tephrosia
sp. †
Themeda triandra
††
Tristachya biserata
†
Urelytrum agropyroides
††
Vernonia staehelinoides
†
Xerophyta retinervis
††
B11-77 †
B11-79 †
B11-81 †
B12-86 †
B1-3 † †
B13-88 †
B1-4 †
B16-94 †
B2-28 † †
B4-38 †
B5-44 †
B5-46 †
B5-48 †
B8-64 †
Table A2. Plant species recorded in presence and/or
absence study quadrats in Nylsvley Nature Reserve,
Limpopo Province, South Africa.
Species name Presence Absence
Abutilon angulatum
†
Acacia karroo
†
Acacia toritillis
†
Achyranthes aspera
†
Aerva leucura
†
Aristida congesta
††
Aristida diffusa
††
Asparagus africanus
††
Blepharis integrifolia
†
Burkea africana
†
Carrisa bispinosa
†
Appendix A. Plant species lists for the Bronberg Ridge, Nylsvley Nature Reserve and Kruger National Park study
areas.The occurrence of a plant species in a presence or absence study quadrat is designated by †.Specimen codes
are provided forspecies that were not successfully identified in the herbarium, and are listed at the bottom of each list.
254 African Zoology Vol. 43, No. 2, October 2008
Species name Presence Absence
Chaelanthus costatus
†
Cleome maculata
†
Combretum molle
†
Commelina africana
††
Crabbea acaulis
†
Dactyloctenium aegyptium
†
Dichrostachys cinerea
††
Digitaria eriantha
†
Diheteropogon amplectens
†
Diospyros lycioides
†
Eragrostis curvula
††
Eragrostis gummiflua
†
Eragrostis
sp. 2 † †
Eragrostis trichophora
†
Euclea crispa
†
Euclea natalensis
†
Eustachys paspaloides
†
Grewia flava
†
Grewia monticola
†
Gymnosporia
sp. †
Hermannia sp.
†
Hermannia
sp. A †
Hermannia
sp. B †
Hermannia
sp. C †
Ipomoea sinensis
†
Isoglossa grantii
†
Justicia flava
†
Kohautia virgata
†
Lannea discolor
†
Lobelia
sp. †
Melinis repens
†
Oldenlandia corymbosa
†
Panicum maximum
††
Perotis patens
†
Rhus dentata
†
Rhynchosia totta
†
Sida cordifolia
†
Solanum delagoensis
††
Strychnos madagascariensis
†
Strychnos pungens
†
Tephrosia polystacha
††
Terminalia sericea
††
Themeda triandra
†
Urelytrum agropyroides
†
Waltheria indica
†
Ziziphus mucronata
†
N10-74 †
N10-78 †
N10-80 † †
N1-1 †
N1-10 †
N1-16 †
N11-84 †
N12-88 †
N14-100 †
Species name Presence Absence
N14-99 †
N1-9 †
N2-19 †
N2-20 †
N2-22 †
N2-25 †
N2-26 †
N3-34 † †
N3-36 †
N3-38 †
N7-55 †
N7-59 †
N7-61 †
N8-63 †
N8-64 †
N8-67 †
N9-73 †
Table A3. Plant species recorded in presence and/or ab-
sence study quadrats in the Kruger National Park,
Mpumalanga Province, South Africa.
Species name Presence Absence
Acacia swazica
†
Achyropsis avicularis
†
Aeschynomene micrantha
†
Agathesanthemum bojeri
††
Alysicarpus rugosus
subsp. † †
perennirufus
Andropogon schirensis
††
Argyrolobium
sp. †
Argyrolobium speciosum
†
Argyrolobium transvaalense
†
Aristida congesta
†
Ascolepis capensis
†
Balanites maughamii
†
Barleria gueinzii
††
Barleria
sp. † †
Bidens bipinata
†
Bothriochloa insculpta
†
Brachycorythis pleistophylla
†
Cassia abbreviata
††
Chamaechrista plumosa
††
Combretum apiculatum
††
Combretum collinum
††
Combretum molle
††
Combretum zeyheri
†
Commelina africana
††
Commelina benghalensis
†
Commelina eckloniana
†
Cucumis zeyheri
††
Dalbergia melanoxylon
†
Jackson
et al.
: Habitat and distribution of Juliana’s golden mole 255
Species name Presence Absence
Dicerocorym seneliodes
†
Dichrostachys cinerea
††
Digitaria eriantha
††
Dioscorea sylvatica
†
Eragrostis superba
††
Euclea natalensis
†
Evolvulus alsinoides
†
Gladiolus dalenii
†
Gloriosa superba
var.
superba
†
Grewia monticola
†
Gymnosporia buxifolia
†
Helichrysum acutatum
†
Hermannia
sp. A †
Hermannia
sp. B † †
Heteropogon contortus
††
Hibiscus engleri
††
Hibiscus micranthes
†
Hyperthelia dissoluta
††
Indigofera delagoensis
††
Jasminium fluminense
††
Justicia flava
†
Lipocarpha chinensis
†
Lotononis laxa
†
Melhania prostrata
†
Melinis repens
††
Panicum maximum
††
Pellaea calomelanos
†
Peltophorum africanum
†
Pentanisia angustifolia
†
Perotis patens
††
Philenoptera violacea
†
Phyllanthus cedrelifolius
††
Phyllanthus incurvus
†
Phyllanthus reticulatus
†
Pogonarthria squarrosa
††
Pterocarpus angolensis
†
Pterocarpus rotundifolia
†
Rhoicissus tridentata
††
Rhus dentata
†
Rhus gueinzii
†
Rhus pyroides
††
Rhus
sp. †
Rhynchosia minima
†
Rhynchosia totta
††
Schmidta pappophoroides
†
Schotia brachypetala
†
Setaria sphacelata
††
Setaria
16 † †
Species name Presence Absence
Sida acuta
†
Solanum delagoensis
††
Solanum panduriforme
†
Sphedamnocarpus pariens
†
Sporobolus afrincanus
†
Stachys natalensis
†
Strychnos gerrardii
†
Strychnos henningsii
††
Stylochiton natalense
††
Tephrosia polystacha
††
Tephrosia
sp. † †
Terminalia sericea
††
Themeda triandra
††
Tristachya leucothrix
†
Urochloa mosambicensis
††
Vernonia natalensis
†
Vigna unguiculata
†
Vitex harveyana
†
Zanthoxylum capense
†
K1-10 †
K11-103 †
K13-110 †
K13-112 †
K14-119 †
K17-143 †
K18-145 †
K18-147 †
K19-150 †
K20-160 † †
K20-163 †
K22-180 †
K22-181 †
K22-185 †
K23-188 †
K23-189 †
K23-190 †
K23-194 †
K23-195 †
K25-206 †
K25-210 †
K25-212 †
K25-213 † †
K25-214 †
K26-218 †
K26-219 †
K26-221 †
K6-51 †
