1. Pre-evaluation of wild fires in the Extremadura's Regional ParkUsing CHRIS images for wildfires surveillance
EOID: 3152 ; PI: Martinez;
Site Name: Montfrague (Spain)
Montfrague
Long (E): -6,0296, Lat(N) :39,8469, Height (m) 310
Mode :3

2. Status
Montfrague (Mont Fragosum ) has been one of the most important sites for the preservation of the Mediterraneans ecosystem.
Located in Spanish central-west (Extremadura) on the confluence of the Tajo and Tietar rivers have got a rich variety of big size birds (vultures, eagles...) and vegetation.
Fire risk is specially high in this environment by dry warm weather and human impact.

3. Status II
In order to preserve this ecological environment some experiments has been done:
Montfrague Chris Proba Image (M.C.P.I.) acquisition.
Image processing for fire risk estimation using M.C.P.I.:
Endmember extraction, abundance maps generation and endmember identification on this maps.
Image classification using SOM neural networks.
Airborne Hyperex camera image acquisition.
Simultaneous field campaign measurements.
Field measurements.

4. Endmember extraction.
The first image processing was the identification of the endmember present on the scene, in order to perform this task, AMEE algorithm (R) was used.
Six endmembers were obtained and labeled from 1 to 6 (the spectral radiance of these signatures are show in Fig 1.
Endmember 1 can be identified as water, endmembers 2,3 and 5 should be different types of vegetation, and finally soils are represented by endmembers 3 and 6.
The highest abundance for endmember 1 (Fig 2) strongly correspond with Tajo and Tietar reservoirs and other water covers.
Endmember 3 is more abundant next to the rivers and can be identified with soil.
The white shapes in Fig 3 are circular irrigated crop this allow us to identify endmember 2 as vigorous vegetation, highest gray levels can be identified with vigorous vegetation in the north-oriented face of the mountains. The areas with the highest fire risk are those that appear with higher grey levels in figure 5.

5. Fig 1 Endmember,s spectral signature for (M.C.P.I.) image.

6. Fig 2 Abundance map for endmember 1 (water)

7. Fig 3-Abundance map for endmember 2 (crops)

8. Fig 4 Abundance map for endmember 4 (soil)

9. Fig 5 Abundance map for endmember 5 (senesc. veg.)

10. Fig 6 Abundance map for endmember 6 soil.

11. Hyperspectral image processing using som neural network: Monfragüe natural park
MC.P image has been classified using HYPERSPECTRAL-SOM non- supervised neural network. This SOM neural network is a modification of the original one made by our research group (GRNPS).
The scene can be interpreted according to the soil composition, and the results provide us with valuable information. The yellow coloured areas match with still waters, the black ones match with marsh, green areas match with vegetation, and blue and brown match senescent vegetation (the ones with more fire risk).

12. Fig 7- SOM classification map for MCPI

13. Conclusions The areas with higher fire risk has been identified:
Automated endmember extraction and mixture analysis.
Neural Network classification (SOM).
The areas identified with senescent vegetation are the same using both methods, and has been proved by means of ground campaigns