Transcript No Slide Title

Investigating the environmental cause of
global wilderness and species richness distributions
Crewenna Dymond, School of Geography, University of Leeds, UK
There has been considerable research into the causes of gradients in species richness,
for example latitude (Rohde 1992) and water energy dynamics (O’Brien 1998 ) are just two of
the proposed theories. In the field of wilderness science there has been little attempt to
integrate biodiversity into wilderness investigations, in spite of biophysical naturalness being
frequently used as an attribute of wilderness in inventory or identification procedures. However,
it is possible to identify factors that could be simultaneously responsible for the distribution of
species richness and wilderness (see Figure 2). It is the aim of this research to investigate how
far these factors are responsible for the distributions of wilderness and species richness
observed in the mid 1990’s. Additionally, using the same data it is possible to determine
whether biodiversity and wilderness contribute to the distributions of one another.
Soil
Aspect
Solar energy
BIODIVERSITY
Latitude Altitude Evapotranspiration Precipitation Temperature Population
Naturalness
WILDERNESS
Remoteness
Figure 2. Factors identified as important for the distribution of wilderness and biodiversity,
in purple, while additional factors that define each individually are marked in blue.
Method
1) From published global databases on climate, altitude, latitude and population
summary information were calculated. These variables include mean, minimum, maximum
and range and for human population, density and rural population density were calculated.
Species richness data, for mammals, birds, flowering plants and conifers and cycads (seed
bearing plants) are at the national scale, in order to factor out the effect of area, these data were
regressed against the log10(area) of the country and the residual values used in further analysis.
These residuals indicate how much more than expected the richness of a country is, assuming
a linear relationship between species and area on a logarithmic scale.
2) Principal Component Analysis (PCA) was used to reduce the number of variables and to
ensure independence. The new PCA axes are a product of the summary variables for each
factor. For example, the new latitude axis is the product of the distributions of mean, minimum,
maximum and range in latitude in each country (see Figure 3).
3) Multiple regression models were built to test the contribution of each new axis to the
distribution of species richness of major taxa (mammals, birds, flowering plants and conifers
and cycads) and of a series of wilderness quality proportions. A backwards stepwise procedure
leaves only those factors that fulfil the entry requirements of the model in the final step and
whose contribution is statistically significant.
This process was repeated for each of the
environment
factors.
Temperature,
precipitation
and
evapo-transpiration
(AET, PET and deficit) were incorporated
into the same PCA axis (climate) as these
are closely related. Population density
and rural population density were
calculated and incorporated into a single
population axis.
Axis
1
2
Eigenvalue
2.592
1.072
Variance
64.798%
26.804%
Results
The stepwise regression procedure revealed that different factors are responsible
for the distribution of each group and some are better explained than others. For example,
Figure 4 shows that climate
Normal P-P Plot of Regression
Normal P-P Plot of Regression
and
latitude
positively AdjR = 42.0% Standardized Residual (Mamres) AdjR = 16.6% Standardized Residual (conres)
contribute to the variation in B coefficients
B coefficients
the distribution of mammal Climate = -0.529
Latitude = -0.455
= 0.767
Elevation = 0.410
species
richness
and Latitude
p = 0.00
p = 0.01
explain 42%. For the conifer N = 137
N = 119
and cycad group latitude
and elevation are most
explanatory
but
only
account for 16.6% of the
Observed Cum Prob
Observed Cum Prob
variation. The B coefficients
confirm that conifers and
Figure 4. Probability plots of residuals for mammals and
cycads
are
positively
conifer/cycad groups from multiple linear regression
effected by elevation and
prefer high latitudes.
There is also a fluctuation in the ability of
the regression models to explain wilderness
2
Change in Adjusted R for Wilderness
quality. This can be seen in Figure 5 where
Quality Proportions
the highest Adjusted R2 is found for the
40
mid-range wilderness quality categories;
30
37.9% of the variation is explained for
20
Category 15. For high wilderness quality all
10
of the dependent factors explain some of
0
the variation in the distribution of wilderness
1
2
3
7
11
15
17
19
21
(climate, latitude, elevation and population).
Wilderness quality category
For categories 7 and 11, at mid-quality,
latitude is not longer considered important
Figure 5. Change in Adjusted R2 values with
and at low qualities (19 and 21), only
reduction in wilderness quality (1 = high; 21 = low) elevation is contributory.
High quality wilderness (category 3) was found to contribute to the species richness of
mammals (a further 5.9% Adj. R2) and flowering plants (4.8%). This was a negative contribution,
meaning that low species richness was important for wilderness. Conifers and cycad richness
added between 7.0% and 17.6% to the success of the low quality wilderness models, again this
was a negative contribution indicating that low numbers of this group are associated with low
quality wilderness.
2
2
1.00
1.00
.75
.75
Expected Cum Prob
Aim
Figure 3. Example of PCA graph of latitude to derive a
Expected Cum Prob
Figure 1. Global
wilderness
quality
continuum, at a
resolution of 0.5
decimal
degrees, each
cell
has
a
wilderness
quality from high
(22 - green) to
low (1 - pink)
(WCMC 2000;
Lesslie 2000).
Using PCA to summarize variation
The top PCA graph for latitude shows the new axis explaining variation within summary variables
ordination of each country based on the
four summary latitude variables of mean,
minimum, maximum and range. The table
of eigenvalues and variance below
demonstrates that Axis 1 (x) is better at
explaining the variation within these data
than axis 2 (y). The second chart indicates
how well mean latitude correlates with the
new Axis 1 (r = 0.887).
.50
.25
0.00
0.00
Percentage
The environmental factors which affect biodiversity, specifically species richness,
and wilderness quality were investigated at the global scale using national species richness
data (Groombridge 1994) and a continuous wilderness quality grid (WCMC/Lesslie 2000). At a
high wilderness quality (category 3 and those above) a combination of climate, elevation and
population explained one fifth of the wilderness distribution, whilst at low quality (category 15
and those above) latitude, elevation and population explain 37.9% of the distribution. Latitude
and climate explained nearly half of the variation in mammalian species richness, whilst climate
alone explained 16.7% of the variation in the distribution of flowering plants. It was found that
high elevation and latitude were key to the distribution of high wilderness quality and the conifer
and cycad group were also determined by these characteristics. The most important
determinants of species richness were found to be low latitude and ‘good’ climate with
precipitation and temperature being most influential. Understanding the factors defining patterns
of wilderness today will help plan for their protection on a large scale. Appreciating how the
same factors effect the distribution of species richness will aid in conservation of biodiversity,
particularly that in protected wilderness or that requiring pristine habitat. This research is part of
a Ph.D. to investigate species richness and wilderness interactions at multiple scales, including
a study in Tongass National Forest, Southeast Alaska.
.25
.50
.75
1.00
.50
.25
0.00
0.00
.25
.50
.75
1.00
Adjusted R2
Abstract
Discussion
Results indicate that there is a fluctuation in the ability of the models to
explain the variation in species richness and wilderness quality. For mammals low latitude and
‘good’ climate (high precipitation and constant warm temperatures) were important
determinants. For conifers and cycads, high latitudes and elevation were found to be
contributory. High wilderness quality is determined by a combination of all factors, reflecting the
variation in locations in which wilderness currently persists. However it was determined that
high latitudes and high elevation were particularly important. The negative contribution of
conifer and cycad species richness to the distribution of low quality wilderness indicates that
this group may also be dependent on environments with wilderness characteristics. A difference
in the environmental factors that determine the species richness of different groups has been
found, whilst wilderness quality appears to respond to the same conditions. Further research is
needed to determine how far these findings are true at smaller scales.
References
Groombridge, Brian (Ed); 1994, Global Biodiversity Data Sourcebook, WCMC Biodiversity Series, WCMC,
Cambridge
O’Brien, Eileen, M; 1998, Water-energy-dynamics, climate and prediction of woody plant species richness: and
interim general model, Journal of Biogeography, 25, 379-398
Rohde, Klaus; 1992, Latitudinal gradients in species richness: the search for the primary cause, Oikos, 65, 514-527
World Conservation and Monitoring Centre; 2000, Global Continuous Wilderness Grid, personal communication,
May 2000
Lesslie, Rob;2000, Creation of the Global Continuous Wilderness Grid, personal communication, June 2000
Research supervisors: Dr. S. Carver and Dr. O. Phillips
Research Sponsored by NERC GT04/98/130, Congress attendance sponsored by Anglo American
Corporation, organized by the Wilderness Trust. Email - [email protected].