ibul 2004 eernationad Archives af He Phorognamnettm Remote Sensing and Spatial Information Sciences. Vol XXXV. Pan Bd, Ianfd 2004 
(pixel multispectral: 2,8m) dated 24 June 2003 at 09:51 GMT information. We have to point out that, unlike the other tablet, 
  
pue centred at 46,28? N and 12,88? E (Spheroid WGS84). So we the last has been realized using a lot of samples in all Italy, but 
age had a large number of pixels to study the vegetation index and especially in Friuli-Venezia Giulia. 
8 biophysical variables correlations. This image presented a cloud 
3 cover around 5%, but these clouds didn’t cover our test areas. 5.2. Comparison between the foliage and LAI 
2 
) 4.6. Pre-processing of the Quickbird scene Even if the number of sample is very small, the comparison 
me. ; between foliage and LAI index points out the good quality of 
à | The acquired image has been already geometric processed this index to study the leaf crown (and so the maturity of 
(method: nearest neighbour) by reseller forest). In fact, the R? correlation index reaches a value around 
sd (www.digitalglobe.com); instead of this, we checked different 0,81(fig. 7). In this case we used Kg/m^ to normalize the arca 
CU. 
location problems. In order to achieve a good result, we again extension of test area 36 and 26 subdivided in four subplots. 
rectified a small portion of the satellite data with our regional The same is for fig.8. 
vectorial map (scale 1:5.000) and different ground targets 
(fig.6) as reference. 
LAI - Foliage 
  
  
  
  
  
dA II UU E EE e 
SHADOW CLOUD S . 
: i | R? - 0,8164 
; 4.3 x 
424 : 
< 
- Í 
4,1 d 
CLOUD | 
4 A — Er a 
  
5 20 2£ 30 35 4p 45 50 
Foliage (Kg/m2) 
  
Figure 7. Correlation between LAI and Foliage. 
  
5.3. Analysis of the correlation between LAI, Wood 
e different Biomass and NDVI values 
Figure 6. The large piece of withe cloth from satellite (RGB : : : : i 
composite: 432). Correlations between variables of flux process (from satellite 
1mage to forestall biomass) seem to be different. Indeed the LAI 
; Regional Moreover we applied the dark-objects method ? as atmospheric seem most sensitive to wood biomass (R^ = 0,75) (fig.8) than to 
d Emilia correction. Finally, we applied the NDVI index as vegetation vegetation index (R^ = 0,58) (fig.9). Obviously, all these 
> — “Istituto index (fig.6). correlations would need a larger number of samples to be 
yicoltura”). 7 checked. 
5. RESULTS 
the wood 
LAI - Wood Biomass 
5.1. Comparison of the biomass weighted data with the 
tablet estimated values 
  
  
  
  
  
  
  
  
The biomass weighted has been compared with the estimated R^ 7 0,7515 i 
(derived from dendrometric tablets) biomass. Unfortunately the i Hs 
sample number is very little, but it was necessary to have a low- < 
impact on the vegetation. Previously this way of research (with 
real ground data by cutting operation) has been experienced M 7 = 
tric tablets. only on grass sample areas (Borfecchia et al., 2001; Balzarolo : 
et al., 2003), Cutting grass is clearly simpler than cutting trees, : 
also for ecological effects. Having a small number of sampling, dé M D. 2 = 2 
lowing two Our comparisons have a very limited statistical significance, but Wood Biomass (Kg/m2) 
they can provide an idea of relationship trend. 
After these proper remarks, we can made some considerations 
from comparison between real and estimated wood biomass. In 
general, the dendrometric tablets tend to undervalue the real 
Figure 8. Correlation between LAI and wood biomass. 
od a high Wood biomass. However, from table 3 and 4 results that the 
‘information Alpine (ISAFA) dendrometric table provides a good amount 
you want to 
you need à MM | 
TM- image 3 € classic dark-objects method assumes that an image has 
least 900 m. Pixels whose surface reflectance is negligible (eg, in a 
kbird image Complete shadow), and the image pixel values of each band are 
Subtracted by its minimum value (Liang, 2004). 
1077