spectral variation with respect to other 
environmental factors (Cohen, 1991). Several 
experiments have shown close relationships 
between leaf reflectance and water content in the 
middle infrared region of the spectrum (Tucker, 
1980; Hunt and Rock, 1989; Carter, 1991). 
However, these observations at leaf level need to 
be adapted to field conditions, where a complete 
canopy is measured and therefore more factors of 
noise (soil, observation geometry, canopy 
architecture, etc.) are-present. 
Based on these experiences, several authors have 
observed good correlations between vegetation 
indices and FMC. Unfortunately, middle infrared 
data, the most sensible to water content, are only 
available in few of the current sensors. Therefore, 
most experiments have been based on ordinary 
vegetation indices (such as the NDVI, which 
include just near infrared and red reflectance). For 
herbaceous species, significant Pearson r values 
have been found between NDVI and moisture 
content in Australia (Paltridge and Barber, 1988) 
and the USA (Burgan and Harford, 1993). For 
shrub species, the correlations were less 
significant, although in some Mediterranean 
broad-leaf plants, good adjustments were also 
found (Alonso et al., 1996). 
An alternative method for FMC estimation is 
based on thermal measurements. The ratio of 
actual and potential latent heat (LE/LEp) has 
found to be a good indicator of canopy water 
status (Jackson, 1986; Moran et al, 1994). 
Potential LE can be obtained from meteorological 
measurements, while actual LE from the 
difference of air and surface temperature. This 
approach has been successfully used for fuel 
moisture estimation from NOAA-AVHRR data 
(Vidal et al., 1994). However, since vegetation 
indices also offer an estimation of plant cover, a 
combination of temperature and vegetation 
indices may improve the estimation of plant water 
content (Moran et al., 1994; Vidal and Devaux, 
1995). This approach has been found clearly 
related to both field data and fire occurrence 
(Prosper-Laget et al., 1994; Vidal et al., 1995; 
Alonso et al., 1996). 
Operational use of satellite data in short-term fire 
prevention requires several questions to be 
addressed which are related to spatial, spectral 
and temporal resolution. The spatial resolution 
affects the degree of mixture in fuel types, which 
may offer very different dynamics in moisture 
status. To what species the satellite signal is more 
sensitive remains unsolved. Spectral resolution 
affects the availability of middle infrared data. 
New vegetation indices need to be derived as 
soon as the hyperspectral sensors are orbiting. 
Finally, temporal resolution needs to match 
operational requirements. NOAA-AVHRR data is 
currently the only sensor providing enough 
frequency for danger estimation. Improvements in 
meteorological satellites (such as Meteosat 
Second Generation) will greatly benefit fire 
danger estimation, since they could provide 
several measurements per day. À better spectral 
resolution is also required to improve current data 
for estimating moisture content 
In any case, satellite information should be 
combined with meteorological danger indices 
because they are better suited to estimate FMC of 
the dead vegetation lying on the understorey. 
Integrated indices with both, satellite and 
meteorological information should most probably 
provide the best improvement over current danger 
indices. The specific procedures for integration 
need further research. 
3. FIRE DETECTION 
Fire detection through remote sensing has been 
based on middle infrared data analysis. 
Considering that forest fires temperatures 
commonly range from 500 to 1,000 K (Robinson, 
1991), according to Wien's displacement law the 
most suitable band for fire detection is located 
between 5.8 and 2.9 um (the emissive part of the 
middle infrared region). The thermal infrared 
region presents the peak of emittance at common 
Earth temperatures (around 300 K), and therefore 
may be used to estimate background temperature 
for false alarms discrimination. 
Operational fire detection from space is obviously 
very much dependent on temporal resolution. The 
Earth resources satellites (such as Landsat or 
SPOT) do not provide enough temporal frequency 
for fire detection. On the contrary, meteorological 
satellites have proven to be very useful for these 
purposes. NOAA-AVHRR images offer adequate 
coverage cycle (12 hours) for some applications. 
Moreover, they include a channel in the middle 
infrared region and two in the thermal infrared, 
640 International Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998 
  
  
  
  
ima 
low 
desi 
at li 
spol 
burr 
freq 
sum 
199 
coul 
or 1 
tend 
(Ma 
dyn: 
goo: 
199 
Ope 
In 
requ