Geoscience Reference
In-Depth Information
irrespective of the month of the year in question. The main difference that exists
between the correlations that relate to the 30 year normal and to the individual
months of 1986 is the differences in the monthly coefficients, as can be seen in
Table 2.5.
However, the coefficients that relate to daily rainfall totals are more unstable.
These coefficients can change from being positive one day to negative the next. For
example, the maximum value produced by the coefficients may be produced by a
50 m resolution DEM or a 5,000 m resolution DEM; it depends on the day in
question. The differences that exist between the coefficients are increased.
These differences result from the different classes of rainfall amounts that exist.
Monthly rainfall amounts, especially those amounts that correspond to 30 year
normals, will not take any daily event into consideration. Only the strong trends are
taken into consideration as these trends can be easily explained by the topographic
variables that are calculated by large-scale DEM and which describe the
environment on a small scale. Conversely, daily rainfall totals are more erratic and
spontaneous: a rain shower may only affect one small sector, and may not affect any
of the neighboring areas. Precipitation, which is associated with the arrival of poor
weather, is also affected by spatial irregularities. In these conditions, it is not
uncommon to see that large-scale topographic variables, such as slope and
roughness, influence the distribution of daily rainfall. This distribution of daily
rainfall is recorded by fine-resolution DEM (with a resolution of 50 m). This fact
once again highlights that there is a relationship between the spatial and temporal
scales [CAR 03].
The fact that the coefficients vary with the different resolutions of the DEM is
quite interesting because it shows that the scale factors (which are recorded by the
different resolutions that can be applied from one DEM to another) shape the spatial
organization and distribution of precipitation. In these conditions it seems that the
processes of interpolation need to consider the fact that coefficients vary with
resolution in order to improve the quality in estimation. Certain methods of
interpolation are based on this principle [JOL 94; JOL 03]. Candidate independent
variables are broken down into several windows, which are equivalent to multiple
resolution DEMs, and an additional sorting process makes it possible to identify the
best estimation variables and optimal window. The rain amount (or any other
dependent variable as temperature) is estimated by using a multiple regression
method.
2.5. Conclusion
This chapter will be summarized briefly. It is important to mention the
importance of the quality of the spatial information that is available and which, to a
large extent, conditions the quality of the models that are then created. The famous
aphorism “garbage in, garbage out” can be applied here. This quality can be
associated with one of the following characteristics:
