extremes and consequently unable 
to support rich landscapes. 
By definition, diversity is positively 
correlated with richness (Figure 3). 
Of interest however, is the fact that 
diversity more closely approaches 
Fig. 3. The relationship of landscape 
diversity, landscape richness and maximum 
diversity for 423 islands in Penobscot Bay, 
Maine. 
maximum levels when diversity and 
richness are low, and less likely to 
reach maximum levels when 
diversity and richness are high. This 
is because as the number of spectral 
classes on given island increases, 
the likelihood of equitable 
distribution of pixels among those 
classes declines. 
The number of mammal species on 
islands is negatively correlated with 
distance of islands from the 
mainland (Figure 4.). This is 
because distant islands are more 
difficult for mammals to reach than 
island close to shore. Also, islands 
CONCLUSIONS 
Although the study area for this work 
was a sample of islands within a 
single SPOT image scene, the 
methodology could easily be used 
to measure beta diversity for other 
landscapes. For example, nature 
reserves and habitat fragments could 
be measured and monitored through 
a similar approach. Similarly, 
gamma diversity could be measured 
by running the same analysis on 
entire SPOT or Landsat scenes or for 
that matter on suites of images 
covering thousands of square 
kilometers. It is also possible to 
“create” diversity maps that identify 
pockets of high landscape diversity 
(Podolsky in prep.). 
y = -1.2 + 3.1X R A 2 = 0.697 
Log(10) Distance to Mainland (km) 
Fig. 4. Mammal species richness as a 
function of distance to the mainland for 18 
islands in Penobscot Bay, Maine. 
with high landscape diversity support 
richer assemblages of mammals 
than do islands with relatively low 
landscape diversity (Figure 5.) 
y = -1.2 + 3.1X R A 2 = 0.697 
Landscape Diversity 
Fig. 5. Mammal species richness as a 
function of landscape diversity for 18 islands 
in Penobscot Bay, Maine. 
) IMPLICATIONS 
Digital earth imagery represents a 
rich source of information of value to 
ecologists and conservation 
biologists. These data, in concert 
with microprocessors and analytical 
software tools, can allow ecologists 
to ask questions and derive answers 
at the landscape or geographic level. 
In the future, as the capabilities of 
microprocessors improves and 
software environments are created, 
ecologists will routinely reference 
these data. Thus it may be possible 
for analytical tools to keep pace with 
the increasing rate of human impact 
on the earth’s surface. Most 
important however, is that the 
information derived from these data 
46