Geoscience Reference
In-Depth Information
The diagram on the left in Figure 8.2 shows a particular point X, whose wind
speed (or temperature, etc) we would like to work out by using the values of other
neighboring points, whose values are known (points A to D).
Figure 8.2.
The risks associated with spatial interpolation
The diagram on the right shows the exact same information, except this time a
major piece of geographical information has been added, relief. The relief is
represented by the contour lines seen on the diagram. The relief of an area will
dramatically change the values of the points. For example, altitude (as we can see,
point X is at the top a mountain ridge) will produce values that are different from
the values produced by the interpolation process that was carried out on the
neighboring points in the box on the left where altitude was not a factor. This is
because wind speed is probably much greater on a mountain ridge than it is on flat
and low land. Temperature also varies, with the temperature that is recorded on the
mountain ridge being much lower during the day in comparison to what the
temperature on the flat land is.
If this method is used, it is assumed that there is a high level of homogenity that
exists in the area we want to study for fear of producing any important errors.
- Environmental analysis works differently even if the space (distance) is
integrated within the analysis. The basic idea is to search for laws that statistically
link the value of a specific climate variable (wind, temperature, etc) to its
surrounding area. The surrounding area needs to be represented in terms of
quantitative data, and independent variables, which will be correlated to the
dependent variable by using the process of linear regression. The first piece of
information that tends to be used relates to altitude levels. There has been a lot of
research into environmental analysis taking place since the 1970s, in which simple
regressions in the form Y= ax+b have been used a lot.
It is thus possible to highlight a particular description variable that can be used
to measure any preferential influence that a particular climate variable might have.
However, the variations that exist between the different climate variables are multi-
causal and additional information needs to be added to the model. One solution
consists identifying clusters that gather all points responding to a particular
criterion, which then considers all of the points as being homogenous. The
