An index of forest fire risk has been determined by using the vegetation in
dex NDVI and the surface temperature Ts, computed from NOAA-AVHRR over 21 M
editerranean French forests. Those 2 satellite parameters can be interprete
d in terms of soil water deficit and vegetation stress in summer. An invers
e linear correlation between their values for each forest pixel of 10 dates
in 1990 was used to establish the index which has been divided into 5 equa
l classes. Those classes correspond with 5 risk classes of forest fire occu
rrence which were mapped for several forests. Periods and areas in the high
est risk class correspond with those where the most important number of fir
es appeared in that year for the studied forests. A statistical model of th
e period of highest fire risk has also been constructed for each forest.