Porres, Maru
the behaviour of the atmosphere, it was decided just to compare the radiometric properties of the two images used in the
study. For this a correction was carried out to adjust the average and standard deviation of one image in relation to the
average and standard deviation of the other, so that the histograms of the two images were comparable.
In order to work only on areas with spontaneous natural vegetation, the rest of the areas (urban, agricultural lands, water
bodies, etc.) were masked out by means of a supervised classification using a Landsat TM image. The training areas to
be used by the classifier were extracted with the support of an exhaustive interpretation of the aerial photographs. After
a separability analysis, five bands were used, and a trained maximum likelihood classifier considering a priori
probabilities extracted from previous works on the area (Pardo, et al., 1999). The result was evaluated using a confusion
matrix, achieving a overall accuracy rate of the 97,1%.
Other two masks were created, one of them for non burned areas, and the other for non studied lithologies. After
application of the three masks over the total set of data (multispectral images, NDVI images and topo-climatic
variables), twelve study areas with spontaneous vegetation that had suffered several forest fires at different dates were
finally obtained.
3.2. MDT analysis
The way in which vegetation is regenerated after a fire is unavoidably related to the topography ofthe terrain it
occupies. Based on digital elevation models it is possible to deduce in a quantitative way some topographic
characteristics —with very clear topo-climatic significance—such as the slope and the curvature, catchment areas, as
well as other factors directly linked to the topography of the area such as the potential solar radiation.
From the digital elevation model of the Spanish Army Geographic Service (SGE), with a 25m x 25m grid size, maps of
slope, orientation and catchment specific areas were calculated. This DEM was also used as the basis for calculating
solar radiation.
3.2.1. Curvature and slope map. For each cell of the DEM, the variation in altitude around the point is calculated in
the directions X and Y, using the following expressions. (Felicísimo, 1994):
Considering the 3x3 neighbourhood:
292 Cia ja Zi jn + Zim, jo) = it ja * 2,2 * Zi-,j-1)
ox 6.h
(02 GiLja * Zi4,j * Zi jd 7 Ci, ja * Zistj * Zi, j1)
q II ee
dy 6h
Considering a 5x5 neighbourhood:
292 Giu + Zict jet + Zig) jel * Zi2,j4) = (Zing, jo + Zi-LjA + Zin, 1 * Zi92,j-1)
pe 20h
ds _ (Zi-1,7-2 * Zi-LjÀ * Zi, jd * Zi, ja2) - Gis,j-2 * Zi, jA * Zi, jd * Zi j2)
dy 20h
where z-f(x,y) is the elevation stored in the DEM and h is the cell size, in this case 25 metres.
By increasing the size of the neighbourhood, the values of the heights of the pixels around the central point are not
considered although they are the points for which the parameters are calculated, from which it can be concluded that
increasing the number of neighbours and thereby the amount of data, does not improve the calculation and it is
necessary to look for other methods in which we can be sure that all the elevations are used with a significance or
weight in function of their distance from the central point.
From these values it is possible to calculate the slope of the terrain a, 1
: : ; ; : ia ; o=
creating an image with the same dimensions as the original, in which each t Fo? e
point stores the value of the slope calculated with the following expression EP "I
(Felicisimo, 1994):
To calculate the curvature the variation in the slope is calculated around a point (or in this case grid cell). It is possible
to calculate what is called the curvature profile K, (curvature of a normal section of land surface by a plane which
includes an external normal vector and a gravity acceleration vector in a given point of the land surface) and curvature
surface K, (curvature of a normal section of the land surface, this normal section is orthogonal to the section with K, in
a given point of the land surface (Florinsky and Kuryakova,1996). Each of these parameters, calculated using the
expressions given below, is created in the form of an image:
1172 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7. Amsterdam 2000.