Geoscience Reference
In-Depth Information
To sum up, CLC can reliably estimate the biomass that is calculated from the
land cover layer that it produces. However, CLC is not as good a tool as the NDVI,
and as a result, the NDVI should be used whenever possible. The satellite
measurement produced by the NDVI is much more exact and is more geometrically
precise than that produced by CLC.
2.4.3.
Choice of independent variables
The majority of GIS software offers a reasonable number of functionalities,
which make it possible to create a varied selection of information layers. There is
the temptation to create and store a large number of these information layers and
then to correlate these different layers with temperatures to try to explain the spatial
variation of these temperatures in question. This approach is commendable but is
not without risks. These risks will be shown in this section with the help of an
example, by using distance as the explanatory variable.
Calculating a Euclidean distance on a raster image is quite a simple process, and
insofar as distance is considered as one of the founding concepts of geography,
there are very few geographers who can resist carrying out such calculations.
The world of climatology is composed of a large number of areas that have
different characteristics (thermal characteristics, in particular) and which influence
their neighboring areas. For example, there are the glaciers, whose katabatic winds
transfer, by advection, masses of glacial air to their surrounding areas. Another
example includes the oceans, whose maritime influences affect continental climates
over a distance of several thousands of kilometers. Incorporating this notion of
distance into a determinist approach, such as interpolation, stems from the
hypothesis that if a geographic object has a slight influence on temperature, then the
impact of this influence will decrease as distance increases. There are two key issues
(amongst others) raised that need to be examined in order to validate this hypothesis
and the two issues include: what really happens and how do the different variables
function spatially?
The second of these issues was tested on the 1,530 French climatological
stations which make up the Météo-France temperature network. The test consists of
estimating temperature variation by the distance to the center of a town or city (dist-
town). Distance to town is calculated from the barycenter of the “dense town” CLC
polygon (CLC code 1). The hypothesis is first verified because the correlation
coefficients are of a significant value for each of the three climate variables that are
tested (Table 2.2): temperature decreases as the distance from large towns and cities
increases, this phenomenon is also known as the “urban heat island”. This fact then
makes it possible to predict that the distance from a town can be used as a good
estimator as far as the spatial variation of temperature is concerned. With this in
mind, this information could then be stored in the GIS and used as a prediction tool
within the regression models.
