lity 
lite 
are 
and 
er- 
has 
nal 
ing 
vial 
for 
her 
5 in 
the 
ic, 
Che 
[he 
[he 
Patrono, Andrea 
MULTITEMPORAL LANDSAT TM DATA FOR MONITORING THE EFFECTS OF FOREST 
FIRES AND VEGETATION RECOVERY PROCESSES IN MEDITERRANEAN AREAS 
Andrea PATRONO, Adriano BAPTISTA, Jan EBELTJES 
Western European Union Satellite Centre, Madrid, Spain 
pat@weusc.es, bap@ weusc.es, ebe @ weusc.es 
KEY WORDS: Wildfires, Multi-temporal, Multi-spectral data, Modelling, Monitoring. 
ABSTRACT 
Evaluation of forest fire damage is an important issue in Mediterranean areas where the long arid season often creates 
favourable conditions for the occurrence of fires. Such fires are a major concern because of the environmental damage 
they cause. The objective of this study was to establish a methodology to detect and outline the extent of the damage 
caused by forest fires and to monitor the post-fire land recovery in Mediterranean-type vegetated areas. After experiment- 
ing and testing several techniques, a set of methods has been developed and assembled to provide an operational image 
processing flow for fire management. A set of study areas located in Greece and Turkey, heavily affected by fires in the 
last decade, has been used as test sites. Multi-temporal LANDSAT Thematic Mapper (TM) imagery was used to study 
the pre-fire situation, to delineate the immediate extent of the damage after the fire and to monitor and estimate the recov- 
ery of the vegetation. The study demonstrated that via the trend analysis of vegetation indices imagery derived from the 
visible and near infrared channels of TM (in particular the Normalized Difference Vegetation Index (NDVI)), it was pos- 
sible to evaluate the health and rate of change, to monitor the speed of the growth and, with some limitations, to predict 
future growth of vegetation. Under certain constraints (time series, image availability, etc.), the methodology has proven 
reliable and repeatable in most of the considered situations, providing a practical solution for large area monitoring of 
post-fire situations and representing a key input for post-fire environmental management. 
1 INTRODUCTION 
11 Study Objectives 
The first goal of the study was the recognition and exact delineation of areas affected by fire. Their geographical location 
is generally known but there may be a gap of information concerning the accurate definition of the extent and the inten- 
sity of the phenomenon (due to, e.g., problems of access in mountainous areas). For multi-temporal analysis, remotely 
sensed data offers a viable solution, especially if multi-spectral sensors are used (like LANDSAT TM). Once affected 
areas were identified, the study went through a set of steps to evaluate the environmental damage, to assess the health of 
the vegetation and to keep track of its recovery process. 
The sequence of methods applied is the following: (a) evaluation of the effectiveness of different techniques for the iden- 
tification and delineation of the areas affected by fires; (b) identification and classification (when feasible in terms of 
image availability and ground information) of the pre-fire land cover of affected areas; (c) monitoring of the recovery of 
vegetation; this phase was subdivided in several steps: NDVI analysis of affected areas (using post-fire multi-temporal 
data), identification of the areas characterised by common growth and modelling of the growth rates with best-fitting 
techniques; (d) comparison, by mean of cross-tabulation, between the land-cover, areas affected by fire and recovery 
results. 
The study was designed to create output information and products that were optimised for decision makers involved in 
the fire management, e.g. to better establish and plan mitigation measures (like tree replanting) and systematic ground 
check for risk analysis, to evaluate the effectiveness of restoration activity implemented in the post-fire period. The man- 
agement of fire affected areas is necessary and, of course, critical to avoid other natural hazards that affect and further 
degrade burnt areas (as an example, heavy soil exploitation and erosion may decrease slope stability and cause land- 
slides). All analysed areas were mainly characterized by mediterranean forests - some fires also affected anthropic areas 
- for which fire has always constituted a serious threat. Although coastal pine forests, mountainous landscapes covered 
by fir, pine and beech trees and the degraded natural ecosystems of evergreen sclerophyl brush (e.g. maquis vegetation 
etc.) are well adapted to fires, frequent and repeated fires denude the vegetation from the land cover making necessary a 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7. Amsterdam 2000. 1131