Inadequacies of indices
Redundancy
Transfer of indices
Indeterminate boundaries
Many indices may be necessary to cope with the different
data types and formats that are commonly used in landscape studies but
there are still inadequacies.
Riitters
et al. (1995) show that a lot of the indices are highly intercorrelated.
This is because there are basically only a few measures that can be made
from land cover maps,
so what we end up with is redundancy
and replication in some landscape indices.
Most of the indices widely used today were developed
on, and applied to data sets in North America and there are problems in
applying them to other areas.
Hulsoff
(1995) looked at pattern indices such as land use type, patch number,
mean patch size, patch shape and dominance, and change indices. She found
that not all the indices widely used in North America were suitable for
application in Dutch landscapes.
This means that different indices or measures may be necessary for different landscape types.
Most of the indices so far are assessments of patch
geometry. These are useful metrics of landscape structure provided boundaries
between landscape elements can be clearly identified, like Kellerberrin.
Problems arise however, when the landscape takes the form of a continuum
with landscape features grading into one another, eg in the New
England tablelands.
Defining objects in this situation is difficult. One
solution is to treat the landscape as a series of gradients
rather than discrete objects and to apply a range of texture
measures derived from image processing. These can be used with remotely
sensed data or data that exists as surfaces. This constitutes a move away
from patch theory and could be a good way of measuring landscape structure
in landscapes, such as those found in Northern
Australia.
Texture refers to 'the spatial variation in brightness
values within a region of an image' (Haines-Young
and Chopping, 1996). Descriptors of texture include properties such
as:
- Uniformity
- Coarseness
- Regularity
- Frequency
- Linearity
Other measures that could be used are ones based on
spatial statistical analysis, for example spatial autocorrelation
(Pearson,
1998).
 |
Identify appropriate measures
|
 |
| Would the same measures be appropriate in studying Kellerberrin and New England? Identify what would be best in each location. | ||
 |
A good article providing more information on indices is Reading 1.5 written by Haines-Young and Chopping. | |
