GÁBOR MEZŐSI and GÁBOR DORMÁNY Department o f Physical Geography, University o f Szeged,
H-6722 Szeged, Egyetem u. 2.
INTRODUCTION
Problems with forest fires have been aggravated in the recent decades. They appeared primarily as an ecological issue and not so much as an economic problem.
When examining intensities of forest fires in Hungary we mostly find forest-litter fires, and only smaller territories are struck by a total destruction of forest stand. This is why the main question arising is how the damaged vegetation can be replaced in a natural way and what ecological patterns are adequate to be created during sylvicultural activity.
These ecological patches (e.g. non-forest belts) should serve as barriers against fires, and they also help in re-establishing vegetation with ecological corridors. To set up efficient management, information is needed on the expected intensity of forest fires at the different parts of the woodland. The following paper presents the results o f a pilot analysis made on a drainage basin at Bodony, in the Mátra Mountains where mainly man induced forest fires occurred (which is a usual case in Hungary), and natural ones were less frequent. (Another question arising could be the relationship between these fires and the global aridification.)
PRECEDENTS
The first mapping of forest fire hazard were performed over extensive territo
ries using remote sensing methods in the 1960s and 70s. It turned out that this method is quite efficient but without detailed data from „training sites” it is not too reliable, and therefore it can only be used with reservations. The geoecological mapping of the drain
age basin of Kataréti stream at Bodony in the Mátra Mountains together with GIS appli
cation allowed the combination of many variables. To make the analysis easier, TM data were available for several seasons and a forest fire has already happened in this area, so
the problem could be reduced to a kind of similarity analysis, but at least it had made the checking of the results possible (Figure 1).
METHODS
During the analysis with the combined and integrated approach mentioned above we used three parameters according to Chuviecco, E. and Congalton, R.G.
(1989):
A map was compiled, showing vegetation in general and woodland in detail at a scale of 1:25 000, and with the digital analysis of a Landsat TM image (taken July 1, 1987) the following categories were created considered important from the viewpoint of the survey (see Figure 2):
The Landsat TM images available were first geometrically corrected, and then according to the training sites determined during field work, and by using the maximum likelihood method the following land use categories were determined:
Fig. 1 - Location of the test area.
Fig. 2. The ecological state o f the vegetation. - 1= degraded; 2 = initial; 3 = optimal; 4 = agricultural use; 5 = settlement (Bodony)
- dense and moderately dense coniferous forest, - dense mixed forest,
- dense and moderately dense broad-leaved forest, - willow and acacia grove, and
- other patches of forest.
To determine forest density the above described remote sensing image was used, i.e. the vegetation density index (NDVI value) was calculated with the combina
tion of bands 3 and 4 (4-3/4+3). The analysis was carried out using ERDAS software (version 8.2) on SUN workstation. After processing the remote sensing data it was also shown, that certain ecological differences could be identified on the basis of digital information alone with a rather high accuracy (80-85%), which opens perspectives for the further investigation of the problem of forest fires.
The other parameter analysed from the viewpoint of forest fire hazard was to
pography. No serious differences in precipitation were detected in the test area between the forest sections, which might be attributed to the eastern aspect and the moderate angle of slopes. The map of hillslopes were composed of digitised contours made by Arclnfo 7.03 based on the logic, that the steeper the slopes are, the more serious the hazard is (e.g. through the effect of burning parts rolling down the slope)(FigMre 3)
As the third factor, the forest paths and paved roads were involved in the re
search. Their role is very complex. On the one hand they reduce the potential hazard as barriers, on the other hand they could be the corridors of fire spreading through their intensive use. The main reason why the buffer zones were constructed along these routes
Fig. 3. Map o f slope categories o f the test area - 1 = 0-5%; 2 = 5-12%; 3 = 12-17%; 4 = 17-25%;
5 = 25% and more
using GIS was to specify the increased hazard in the 150 meter zone along forest paths, and in the 50 meter zone along paved roads.
We calculated fire hazard from the above mentioned factors, according to the following method. We categorised vegetation, the angles of slopes and the buffer zones along the roads according to the hazard posed by them in a way that vegetation was given a tenfold multiplier, slope had received a triple one and buffer zones had a multi
plier of 0.5. We divided the vegetation and slope angles into the following categories;
Type o f vegetation Fire hazard Coefficient
dense and moderately dense coniferous forest serious 2
dense mixed forest serious 2
dense and moderately dense broad-leaved forest medium 1
willow and acacia grove low 0
other patches o f forest low 0
Angles o f slope Fire hazard Coefficient
0- 12% low 0
12-25% medium 1
above 25% serious 2
Areas belonging to the buffer zones along roads and paths received a value 1 while the other surfaces 0.
RESULTS
After all, we calculated the topologic sum of the weighted values and coeffi
cients. The higher the values were acquired, the more serious the hazard is expected. As Figure 4 shows, we put the results into three hazard categories. These results were com
pared analogically to the remote sensing data, and it figures out the visible correlation between the wetness of vegetation (band 4, TM) and the ecologically explained fire hazard, which is less related to land use. Nevertheless, these results were compared — with the intention of verification - to the forest-litter fires, that were observed in the summer seasons during 1992-95. The agreement was evident in the middle part of the drainage basin (around village Bodony), and it was the same west of the settlement. On the western edge of the test area with the highest orographic position where the most serious hazard was predicted no forest litter fires and forest fires were observed.
Fig. 4. Map o f forest fire hazard. - 1 = no hazard; 2 = moderate hazard; 3 = serious hazard
REFERENCES
Chuvieco, E. and Congalton, R. (1989). Application o f remote sensing and GIS to forest fire hazard mapping. Remote Sensing Environ. 29. pp. 147-159
Csorba, P. (1989). A tájstabilitás és ökogeográfiai stabilitás. Földrajzi Értesítő 38. pp. 395-410.
Leser, H. and Klink, H. J. (1988). Handbuch und Kartieranleitung Geoökologische Karte 1:25 000 (KA GÖK 25). Trier, 349 p.
Mezősi, G., Bárány-Kevei, I., Mucsi, L. and Balogh, I. (1993). First results o f GIS based geoecological mapping. Földrajzi Közlemények, pp. 132-155.
Naveh, Z. and Liebman, A. S. (1984). Landscape Ecology. Theory and Applications. Springer Verlag, New York-Berlin-Tokyo
Zólyomi, B. et al. (1967). Einreichung von 1400 Arten der ungarischen Flora in ökologischen Gruppen nach TWR-Zahlen. Fragmenta Botanica. 4. pp. 1 01-142.
L. Bassa-A. Kertész (eds): Windows on Hungarian Geography
Studies in Geography in Hungary 28, Geographical Research Inst. HAS, Budapest 1998, pp. 91-98