J 
1S 
  
jt 
of 
Patrono, Andrea 
  
Figure 3. NDVI images: (a) pre-fire, (b) post-fire, (c)-(d)-(e) respectively 1-2-3 years after fire, (f) 4th predicted year 
the NDVI differences had been computed to identify and group zones with homogeneous growth rates. When detected, 
anomalies, random variations, etc. in trends, which resulted in poor model fitting, highlighted situations previously men- 
tioned, such as anthropic activities, recurrent fires, overgrazing, erosion, etc. or, in general, excessively spoiled/damaged 
plots. For the rest of the areas the modelling emphasized and confirmed the vegetation recovery (that is possible to quan- 
tify using the fitted NDVI curves) after the fire damage (see Figure 3 (f)). In some areas the calculated activity was higher 
than in the pre-fire situation, which is supports the ecological theory/expectation that the rates of production/activity are 
higher in full recovery phases than in equilibrium ones (where part of the effort is focused in maintaining the biomass). 
The results of the method were verified, wherever feasible, by collateral data (e.g. the correlation between known 
replanted areas and extrapolated NDVI trends). 
4 CONCLUSIONS 
The present NDVI interpreting/modelling approach, when supported by a sufficient multi-temporal data set, makes it 
possible to evaluate the initial phase of vegetation recovery processes in fire affected areas. In terms of forecasting full 
vegetation recovery, some limitations have to be mentioned: data availability for the present study covered a range of 
maximum 9 years; this means that only the first steps of the recovery could be effectively/efficiently modelled, such as 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7. Amsterdam 2000. 1135