the forest fire of parnitha mountain, greece
DESCRIPTION
The forest fire of Parnitha mountain , Attica, Greece on June 28th 2007AbstractMount Parnitha is the highest mountain in Attica (one of the 13 administrative regions of Greece) with an elevation of 1,413 m, located 20km northwest of the capital city of Athens. Parnitha is a densely forested mountain with pretty rich flora and fauna (the flora of Parnitha comprises of 1.100 taxa; that is equal to the taxa of whole Scandinavia). Parnitha is part of the ecological network Natura 2000 with a natioTRANSCRIPT
The forest fire of Parnitha mountain , Attica, Greece on June 28th 2007 Christos Kastrisios
GEOG 652 – Final Project
Introduction The main objective of this project is to
identify and measure the area burned and need to
be reforested and evaluate in what extent the
place of exceptional natural beauty (NATURA)
was affected. Additionally I will evaluate the
current condition of the burned area; several
public and private reforestation efforts have been
performed and I indent to find the area that
remains destroyed although the reforestation and
the natural recovery of the forest.
References Jensen, J., “Introductory Digital Image Processing”(2005)
http://www.parnitha-np.gr/index.htm
http://www.parnitha-np.gr/glk_master.pdf
Abstract Mount Parnitha is the highest mountain in
Attica (one of the 13 administrative regions of
Greece) with an elevation of 1,413 m, located
20km northwest of the capital city of Athens.
Parnitha is a densely forested mountain with
pretty rich flora and fauna (the flora of Parnitha
comprises of 1.100 taxa; that is equal to the taxa
of whole Scandinavia). Parnitha is part of the
ecological network Natura 2000 with a national
park and a place of exceptional natural beauty.
On June 28th 2007 a massive forest fire
broke out which within three days claimed a
large proportion of the rare Greek Fir and Aleppo
Pine forest, birds and rare animal species.
According to scientists it would take about a
century for the ecosystem to recover without
intense reforestation efforts.
Data The study area and period are limited by the
fire. The time frame is before and after the fire in
2007 as well as the condition in 2011. The fire
started on June 28th but due to the temporal
resolution of Landsat 5 the imagery used was
acquired on May and July 2007. For 2011 I will
use an August L5 image. Imagery data were
downloaded from http://glovis.usgs.gov/. For
classification purposes additional L7 images
were also utilized. GIS data (administrative areas
and Natura datasets) acquired from the Greek
http://www.geodata.gov.gr. Finally ENVI,
ArcMap, ArcGlobe and Google earth were used
as well as yahoo. maps web site.
Methodology In the flowchart diagram we can see the
procedures implemented in the study. After the
acquisition of data, the images were preprocessed in
ENVI. The atmospheric error was corrected by
calibrating the images and then a subtraction of the
minimum value in each band accounted for the Dark
Object correction. After having stacked bands,
1,2,3,4,5 and 7, the images were spatially subseted to
the area of study using custom ROIs as EVF.
The NDVI transformation couldn‟t be utilized to
determine the burnt area size due to the nearby quarry
area (SW) which reflectance values were similar to
the burnt forest (Fig. 4 and Fig.11 for NDVI values).
Similarly the significance change in agricultural areas
(NW) reflectance characteristics from May to July,
raised difficulties in utilizing the NDVI results for
Change Detection analysis (CD) (Fig. 5 and Fig.11).
Therefore the best delineation of the study area
was achieved using supervised classification methods
(ISODATA created a fuzzy result with regards to the
May image). After classifying the May and July
(Figure 6) images, I ran clumping and combining
classes functions, in order to give spatial coherence to
and bring out the Burnt Area class (Figure 6). Then by
utilizing the higher resolution images (L7) as well as
Google Earth and yahoo.maps applications, the
necessary for the accuracy assessment ground truth
ROIs were created. The overall accuracy of the July
classification was 95.2489% and Khat 0.9445. The
Producer and user accuracies are shown in Figure 9.
Similar accuracies achieved for May and August.
The final step was assessing the Land Cover
change from July 2007 to August 2011 for the
destroyed area. Figure 8 is an ArcGlobe export
showing the CD map superimposed above the August
image. In Fig.11 we can see the NDVI values for
random pixels before, after the fire and in August „11.
Data Acquisition
Pre-Processing
Image Subsetting
Stacking
Calibration
Image Exploration
Image Classification
Change Detection
GIS data and results Utilization
May to July 2007
Accuracy Assesment
Combine Classes
Clumping Function
Classificat.
Transfor- mations
Enhance- ment
July 07 to Aug 2011
ArcGlobe utilization
ArcMap utilization
Re-project GIS Data
GIS data
Landsat L7
Landsat L5
Results and discussion The CD analysis revealed that 46.783.900 m2
(18.1 sq.miles) were destroyed from the fire. That is
approximately one third of Parnitha‟s Natura Area (57
sq.mi.) (Figure 10). Most of the burnt areas where
densely forested (50%), while 37% was medium
dense forest and 13% was classified as sparse/fields.
The 2011 imagery analysis revealed that, due to
reforestation and natural processes, 9.6 sq.mi. had
increased vegetation existence in comparison to July
2007 (yet, only 1 sq.mi. could be considered as dense
forest). Unfortunately about half of the previously
forestall areas are now mostly fields and bare soil.
Fig.11
NDVI
Values
Dense
Forest
(reforested)
Medium
Forest
(Refor.)
Sparse
Vegetat.
Agricult.
common
values
Agricult.
extreme
value
Quarry
Previously
Dense ‘11
destroyed
May 07 0.60 0.47 0.39 0.27 0.36 0.08 0.64
July 07 0.06 0.08 0.09 0.33 0.19 0.08 0.07
Aug 11 0.50 0.42 0.25 0.33 0.31 0.08 0.15
Agriculture
Quarry
Fig.1 – May 2007
Fig.2 – July 2007 Fig.3 – Aug 2011
Fig.8 – CD ‘07-’11
Fig.7 – CD Forest Fire
Fig.4 – July NDVI
Fig.10 – Burnt and NATURA areas
Fig.5 – NDVI Ch.Detection
Agriculture change
Fig.9 – July Accuracies
Fig.6 –Classification