Rivista Italiana di Telerilevamento - 2009, 41 (2): 65-78 Italian Journal of Remote Sensing - 2009, 41 (2): 65-78

A multi-year geographic database of fire affected areas derived from satellite images in the National Parks of Italy Pietro Alessandro Brivio1, Bruno Petrucci2, Mirco Boschetti1, Paola Carrara1, Monica Pepe1, Anna Rampini1, Daniela Stroppiana1 and Paolo Zaffaroni1 1

CNR-IREA, Institute for Electromagnetic Sensing of the Environment, via Bassini 15, 20133 Milano, E-mail: [email protected] 2 Ministero dell’Ambiente, Direzione Protezione Natura, via Capitan Bavastro 174, 00147 Roma

Abstract

A geographic data base of burned areas in Italian National Parks was built using satellite images. Around five hundred ASTER and SPOT images were acquired for the period 2001-2005 and processed with photo-interpretation, multiple thresholds of spectral transforms (NBR and BAI) and Maximum Likelihood. In five years the area burned is more than 5000 ha (fire average size is 7.7 ha) and the most affected Parks are in Southern Italy. Satellite maps were compared with field observations from the Corpo Forestale dello Stato showing that remote sensing and ground based data provide complementary information. Burned area maps are delivered as layers of a Web Map Service (WMS) of a Spatial Data Infrastructure (SDI) INSPIRE compilant. Keywords : forest fires, burned areas, National Parks, ASTER images, spectral indices, Web Map Service.

Una base-dati geografica multi annuale delle aree bruciate nei Parchi Nazionali d’Italia ottenute con immagini da satellite Riassunto

Una base-dati geografica delle aree bruciate nei Parchi Nazionali italiani è stata costruita con immagini da satellite. Per il periodo 2001-2005 sono state acquisite circa 500 immagini ASTER e SPOT ed elaborate con fotointerpretazione, soglie multiple di indici (NBR e BAI) e Maximum Likelihood. In cinque anni le aree bruciate sono più di 5000 ha e i Parchi più colpiti sono quelli dell’Italia meridionale. I risultati sono stati confrontati con i rilievi a terra del Corpo Forestale dello Stato mostrando la complementarietà dei due tipi di informazione. Le mappe delle aree bruciate sono servite come piani informativi di un Web Map Service (WMS) di una Infrastruttura di Dati Spaziali (SDI) costruita secondo la direttiva INSPIRE. Parole chiave: incendi boschivi, aree bruciate, Parchi Nazionali, immagini ASTER, indici spettrali, Web Map Service.

Introduction

Land cover and land use dynamics are recognized as being of increasing importance to the Earth System and its human utilization. Their dynamics are relevant to the carbon cycle and

65

Brivio et al.

Multi-year database of burned areas in the National Parks of Italy

have major impacts on climate change; they are also crucial in relation to biodiversity and conservation, and natural resources management. Fire is one of the most potent agents of land cover change and its intensity has implications for subsequent ecological succession. Vegetation fires are a disturbance factor in almost all the ecosystems around the globe [Thonicke et al., 2001] and most of the fire events are human-initiated with only a small proportion due to natural causes [Levine, 1990]. Mediterranean countries of southern Europe are greatly affected by fires especially in summer when the dry and hot weather sets the ideal conditions for fire ignition and spread [Petrucci, 2007]. Most of these countries do not have a proper database of fire events [Paganini et al., 2003]. In Italy, fires seriously damage the forest ecosystem every year and in summer they can get particularly intense especially in the southern regions of the country [Blasi et al., 2004]. More than 11000 fires have been registered since 1970 and every year affected on average 50 000 ha of forest; the worst year in the last decade has been 2007, when the forest surface hit by fire amounted to 115242 hectares [CFS, 2008]. The greatest damage to the ecosystem is within protected areas; in 2007, 27% of the total area burned occurred in protected areas. As stated in Law n. 353/2000 (“Legge-quadro in materia di incendi boschivi”), fire prevention, monitoring and post-fire assessment in protected areas is planned by local managers and approved by the Ministry of the Environment. Since cartography of fire affected areas is a fundamental component of this strategy, Direzione Protezione Natura of the Ministry of the Environment has involved IREA-CNR in an experimental project focusing on the use of satellite images for mapping burned areas within the National Parks for the period 2001-2005. Detailed information about the project and its results can be found in Brivio et al. [2007] and accessed through the site http://milano.irea. cnr.it/WEB_Incendi_Parchi/main.htm. Since the Law n. 353/2000 was issued, cartography of burned areas has been carried out by Corpo Forestale dello Stato (CFS), using field GPS and other traditional methods [Piccoli and Cattoi, 2007]. However, GPS survey is very expensive and it has some technical limitations; for example, it often does not account for small unburned patches within the fire perimeter, nor does it takes into account fire severity level [Garcia and Chuvieco, 2004]. Satellite data can provide a picture of the fire affected areas and represent a cost effective source of information to guide or to be integrated with field GPS surveys and to fill the gaps of historical archives. The use of satellite remote sensing techniques for fire monitoring covers different domains from the detection of active fires to the delimitation of burned areas and the assessment of the impact of fires on the vegetation cover, such as burn severity. Active fires mapping aims to detect very intense and quick events such as the flaming fires and takes advantage from satellite images acquired with a high temporal frequency such as NOAA-Advanced Very High Resolution Radiometer (AVHRR), ESA-Advanced Along-Track Scanning Radiometer (AATSR), NASA-Moderate Resolution Imaging Spectroradiometer (MODIS) and Meteosat [Flasse and Ceccato, 1996; Arino and Plummer, 2001; Giglio et al., 2003; Roberts and Wooster, 2008]. Burned area mapping aims to perimeter fire affected areas and to assess burn severity and algorithms have been developed at different spatial scales. From global to regional scales geostationary satellites as Meteosat, the Japanese GMS and GOES [Prins and Menzel, 1992; Boschetti et al., 2003] and polar orbiting satellite with coarse and medium resolution as NOAA-AVHRR, SPOT-Vegetation or MODIS [Barbosa 66

Rivista Italiana di Telerilevamento - 2009, 41 (2): 65-78 Italian Journal of Remote Sensing - 2009, 41 (2): 65-78

et al., 1999; Tansey et al., 2004; Roy et al., 2005] have been used. At local scale Landsat Thematic Mapper (TM) or Enhanced Thematic Mapper (ETM+) [Smith et al., 2007; Bottai et al., 2008; Silva et al., 2005] are frequently used while the commercial IKONOS satellite is now finding application [Mitri and Gitas, 2006; Stow et al., 2007]. Classification approaches used in the literature for burned area mapping with high resolution data range from single date supervised classification of post-fire images, such as Maximum Likelihood [Brivio et al., 2003] or the more recently introduced object-oriented [Mitri and Gitas, 2004], to threshold-based techniques of spectral indices, such as the Burned Area Index (BAI) [Chuvieco et al., 2002] and the Normalized Burn Ratio (NBR) [Key and Benson, 1999], applied to post-fire images or to pre and post-fire difference images. In this project, high resolution images acquired by the Advanced Spaceborne Thermal Emission (ASTER) sensor and SPOT data where ASTER coverage was insufficient were used to produce burned area maps in the Italian National Parks for the period 2001-2005. Very few literature works focused on the use of ASTER images for fire monitoring, although the sensor offers spectral and geometrical characteristics potentially suitable for this purpose. ASTER is an on-demand instrument flying onboard the Terra satellite and it collects data simultaneously with Terra-MODIS [Yamaguchi et al., 1998; Schroeder et al., 2008]. For this reason, ASTER data have been so far exploited for validation of MODISderived products [Morisette et al., 2005; Csiszar et al., 2006; Giglio et al., 2008]. The non operational character of the ASTER mission could be the major drawback that prevents the use of these data for fire monitoring. In this project the final layers of the burned area maps were distributed by adopting methods and standards recommended by the INSPIRE Directive [European Parliament, 2007] for the development of a Spatial Data Infrastructure (SDI) to allow the larger spread and utilization of the results by different users and for subsequent ecological analysis. They are served through a Web Map Service (WMS) and can then be viewed by any WMS client such as the Portale Cartografico Nazionale of the Italian Ministry of the Environment (URL: http:// www.pcn.minambiente.it/ progetto_incendi.htm).

Materials and methods

Aster and Spot data ASTER images covering the 21 Italian National Parks (Fig. 1) were obtained through NASA’s EOS Data Gateway (now ECHO WIST, URL: https://wist.echo.nasa.gov) thanks to the Department of Geography, University of Maryland. We used AST07 product, which contains surface reflectance for Visible-Near Infrared (VNIR, bands 1-3) and Short-Wave Infrared (SWIR, band 4-9) sensors. The VIS/NIR spatial resolution is 15 m and SWIR spatial resolution is 30 m, each scene covers an area of 60 km x 60 km. The project relied on ASTER data available in the NASA archive that guaranteed a good coverage of the study area for the entire study period, except for 2005. For this reason we recurred to SPOT 2, 4 and 5 data to fill the gap for Aspromonte (6 SPOT images), Cilento (4), Gargano (3) and Pollino (6) National Parks. SPOT data were obtained in the framework of the OASIS (Optimising Access to Spot Infrastructure for Science) Programme. The SPOT 2 data cover the VNIR spectral region with a 20 m spatial resolution, SPOT 4 with the same spatial resolution has an additional band in the SWIR region and SPOT 5 with bands in the VNIR and SWIR regions has a higher spatial resolution of 10 m. 67

Brivio et al.

Multi-year database of burned areas in the National Parks of Italy

The geo-database of the project contains about 500 ASTER and 20 SPOT satellite scenes (Fig. 1), with attributes such as acquisition date, cloud cover and percentage of the scene covered by the National Parks area.

Figure 1 - Database of the satellite images acquired over Italian National Parks from 2001 to 2005.

Classification approaches and validation ASTER data were pre-processed to resample SWIR bands to the higher spatial resolution of the VNIR bands. The poor geo-location accuracy of the four SPOT 2 scenes required registration to ASTER data previously re-sampled to a spatial resolution of 20 m. SPOT 4 and 5 images were used with their original resolution. All images were visually inspected using different false colour compositions to identify the presence of fire affected areas within National Parks and about 25% of the images were selected and classified. On the basis of preliminary experiments [Zaffaroni et al., 2007], three techniques were identified as suitable for burned area detection and only one of them applied to each scene (post-fire image) based mainly on the distribution and the frequency of the burned patches. Photo-interpretation of RGB composites (832 - 321 for ASTER and 432 – 321 for SPOT data) was applied when few burned areas were identified in the image in which case this approach is faster than processing the entire scene with a supervised automated algorithm. This first approach provided directly the burned polygons for the satellite scenes to which it was applied. ASTER images with many sparse burned areas and few shadows were processed with a multi-index threshold method applied to NBR [1] and BAI [2] spectral indices. These indices, that have been widely applied for fire monitoring, highlight different spectral properties of the post-fire reflectance: after fire occurrence NBR decreases because it is sensitive to vegetation’s consumption and BAI increases because it reveals ash and charcoal deposition.

68

Rivista Italiana di Telerilevamento - 2009, 41 (2): 65-78 Italian Journal of Remote Sensing - 2009, 41 (2): 65-78

t - t SWIR 61@ NBR = NIR t NIR + t SWIR BAI =

1 62 @ 2 _0.1 - t RED i + _0.06 - t NIR i 2

where RED = ASTER band 2, NIR = ASTER band 3 and SWIR = ASTER band 8. This second approach consists of the steps shown in Figure 2. First, thresholds were applied separately to the two indices with values derived empirically image by image (step 1); then the two maps derived from the indices were combined by intersection (AND condition) to produce a binary burned/unburned map (burned=1and unburned=0) (step 2). The intersection of the maps is a conservative approach based on the convergence of the two indices in detecting burns (a pixel is mapped as burned only if it is detected by both indices), thus reducing false alarms. The burned class is converted to vector (step 3). Most of the images were classified with this automated multi-threshold method. Finally, supervised Maximum Likelihood classification was applied to satellite images with many burned areas and a high spectral mixture between burned surfaces and other classes, such as shadows and bare soils. About 150 training pixels were collected for each of seven thematic classes (burned, vegetation, bare soil, urban, cloud, shadow and water). Bands 8, 4, 3, 2, NBR and BAI were used as features for the classification of ASTER images, whereas for SPOT data all the spectral bands were used. After the application of either of the last two automated approaches to derived pixel by pixel classification of the satellite scenes into burned/unburned, a median filter was applied and the raster burned class was converted into vector (burned polygons). Only polygons greater than 1 ha were retained in the final maps. Visual inspection of the operator was still necessary to reduce some residual noise. Each satellite scene therefore provided a final vector layer of burned polygons. The accuracy of the burned area layers derived with the multi-index threshold method was assessed, for about half (20 ASTER scenes) of the images processed with this technique, by comparison with burned perimeters derived by photo-interpretation (reference maps). The error matrix and related accuracy measurements were computed based on Congalton [1991]. The twenty ASTER images were selected for different years and for the most affected parks. Photo-interpretation is widely used for building a reference dataset for accuracy assessment of satellite-derived burned area maps at regional scale [Silva et al., 2005]. Although this approach is biased by the subjectivity involved in photo-interpretation, it was the only feasible one since field data available for the project were not suitable for accuracy assessment of the burned area layers derived over such a large area. Field GPS surveys by CFS started only in the last few years and were still in an experimental phase; vectors of the burned polygons were therefore available only over the Pollino, Gargano and Cilento National Parks for the fire season of 2004 and 2005. Only a qualitative comparison could be carried out. Moreover, in order to reduce the effect of subjectivity on accuracy measures, photo-interpretation was carried out by two operators that had not previously seen neither the satellite scenes nor the output of the classification. 69

Brivio et al.

Multi-year database of burned areas in the National Parks of Italy

Figure 2 -���������������������������������������������������������������������������� Method for detecting burned surfaces based on NBR and BAI spectral indices.

Also field data from the Foglio Notizie (AIB/FN, 2003-2005) were made available by the CFS for validation and a comparison of the number of events and the total burned forest area was carried out for each park; however, geographical information of these field observations was limited to a reference points to which the area burned is referred. Spatial Data Infrastructure One of the issues in the project is the compatibility of results with the available layers produced and diffused by the Italian Ministry of the Environment through the “Portale Cartografico Nazionale” which allows to access and view maps of interest for both external and internal usage. This portal adopts methods and standards recommended by the INSPIRE Directive for the development of a Spatial Data Infrastructure (SDI). In particular, maps that are not included in the portal archives, but are served by any other body through a Web Map Service (WMS), can be added and visualized in an interoperable way. Following the INSPIRE philosophy, and in particular, the principle that geo-data should be hosted and distributed by the subject who produced them and is able to update and maintain them at their best level, we decided to produce burned area maps as layers of a WMS created at CNR-IREA premises for purposes of the European Project IDEUnivers (Infrastructure de données spatiales entre Universités et Centres de recherche dans le Méditerranée Occidentale; URL: http://www.ideunivers.eu) [Barea et al., 2008; Bucci and Carrara, 2008]: this way they can be viewed by the Ministry geoportal, and also by all WMS clients able to add WMS addresses.

Results and discussion

Burned area polygons The final product of the project is a geographic database of burned areas polygons inside the borders of the Italian National Parks for the period 2001-2005. We identified 652 fire

70

Rivista Italiana di Telerilevamento - 2009, 41 (2): 65-78 Italian Journal of Remote Sensing - 2009, 41 (2): 65-78

polygons and estimated a total of 5038 hectares burned in five years. The number of polygons that can be extracted from our database is somehow correlated to the number of fire events although more than one polygon can correspond to the same fire event or the same polygon may be produced by different events occurring over the same area. The average size of the burned polygons is 7.7 hectares. Year by year details of the number of available images, detected polygons and the estimated area burned are given in Table 1. Table 1 - Annual number of satellite images, detected polygons and the extent of burned surface during the period 2001-2005. ASTER (442) and SPOT (19) Year

N. of images

N. of burned polygons

Total burned surface (ha)

Average size (ha)

2001

65

140

1179

8.42

2002

92

49

228

4.65

2003

140

177

1172

6.62

2004

90

157

1441

9.18

2005

74

129

1018

7.89

Total

461

652

5038

7.73

According to our database the worst year was 2004 when about 24% of the 5-year total number of events burned about 1441 ha; the second worst year was 2001 (1182 hectares) immediately followed by 2003 (1172 hectares). We estimated an exceptionally low fire activity in 2002 with only 228 ha burned and 49 fire polygons. The number of images available for mapping burned area least influences the total area detected as long as at least one clear satellite scene is available over the area of interest towards the end of the fire season. The more scenes are available the most accurate mapping will be since persistency of fire scar spectral signal decreases with time. As expected, National Parks of Southern Italy, where very high temperatures and dry conditions make vegetation very susceptible to fire, are the most affected by fire (Tab. 2): Cilento Park with 1159 hectares records the greatest burned surface, followed by Pollino (1150 hectares), Aspromonte (785 hectares) and Gargano (717 hectares). About 75% of the total area burned in the 2001-2005 period was detected in these parks. The multi-temporal dataset shows that the fire affected area can vary significantly from year to year; although globally 2004 was the worst year, not all major Parks registered the most severe season in the same year (e.g. Gargano). Although not located in Southern Italy, the Arcipelago Toscano National Park has been severely hit by fires that burned 627 hectares in five years (data not shown) that is a significant proportion of its area. The satellite-derived geographic database aims to fill the gap in historical archives of fire affected areas when other data cannot be collected anymore. An example of the geographic data that we provided is given in Figure 3b where burned area maps in the Cilento National Park are shown for each year. From this multi-year map most of the burned areas appear distributed in the coastal region, that is the most accessible and populated area of the park. In Figure 3a are given some examples of the major burned areas that we mapped from

71

Brivio et al.

Multi-year database of burned areas in the National Parks of Italy

ASTER data with the case of the extraordinary fire event that occurred in 2001 in the Arcipelago Toscano that burned alone 232 ha. These RGB composites clearly show how satellite images can depict and therefore allow quantification of the unburned patches within the burned polygons. These unburned area can significantly change the total area estimate and are very difficult to identify even when fire perimeters are collected with GPS from airplane [Garcia and Chuvieco, 2004]. Table 2 – Total burned surface and number of burned polygons per year for the most affected Italian National Parks.

CILENTO POLLINO ASPROMONTE GARGANO

2001

2002

2003

2004

2005

Total

n. of burned polygons

31

13

55

53

70

222

burned surface (ha)

146

36

279

278.8

419.3

1159.1

n. of burned polygons

19

18

54

47

9

147

burned surface (ha)

239.8

57.3

313.7

440

99.2

1150

n. of burned polygons

31

2

24

2

32

91

burned surface (ha)

240

6.5

244.4

17.8

276.9

785.6

n. of burned polygons

33

6

30

37

12

118

burned surface (ha)

158.5

43.1

216.2

135.4

163.7

716.9

Figure 3 – a) Some of the largest burned areas detected from ASTER data; b) ��������������������� Spatial and temporal distribution of burned areas from satellite in the Cilento National Park during 2001-2005.

Results validation According to the statistical validation, overall accuracy was more than 99% with an average omission error of 50% and a low commission error (< 15% on average). As expected the 72

Rivista Italiana di Telerilevamento - 2009, 41 (2): 65-78 Italian Journal of Remote Sensing - 2009, 41 (2): 65-78

adopted approach is very conservative and drastically reduces false alarms. However, these results quantify the average scene accuracy (error measurements are computed on an image by image basis) and do not quantify the accuracy of the final product of annual burned area maps such as the one shown in Figure 3b. The availability of frequent satellite images over the same area during the season can fill the gap of the burned areas omitted in each scene by exploiting the high accuracy of detected burns (low commission errors) that increases when image acquisition is closer to the fire event. Accuracy of the burned area maps is a function of complex interactions of several elements such as site and fire characteristics (e.g. pre-fire vegetation type and conditions, background soil, fire severity, post-fire processes) and of the time of imagery acquisition (e.g. atmospheric conditions, geometry of acquisition, time between image acquisition and fire date). Despite the large number of satellite images, the limited availability of ground reference data in the study area prevented the systematic analysis of the influence of the time lag between the fire event and image acquisition on the accuracy of the results. However, previous analysis conducted in an area outside the boundaries of the National Parks [Zaffaroni et al., 2007] highlighted that larger underestimation occurred for events further in time from satellite acquisition, although no statistical significant relation was found between omission error and time lag. The comparison between satellite perimeters and CFS field data provided results such as those summarised in Table 3 for the Cilento Park and shown in Figure 4. It is important to remark that at the stage of project assignment GPS surveys were still in an experimental phase (http://www2.minambiente.it/pdf_www2/dpn/aib/ supporto_eg/ sperimentazione_pi_satellite/rapporto_finale_progetto_luglio_2007.pdf). Table 3 – Comparison between CFS data (AIB/FN and vectors) and satellite results for the Cilento National Park in 2005. CILENTO 2005

N. of events

Surface (ha)

CFS vectors

54

286.4

CFS AIB/FN

83

490.0

ASTER and SPOT

70

419.3

This comparison exercise highlighted that in some cases the remote view of satellite images allows a better delimitation of burned perimeters and unburned patches within the areas with respect to GPS field surveys that are very difficult where accessibility becomes an obstacle to field surveys (Fig. 4a). Moreover, satellite images provided information on burned polygons in non forested areas whereas CFS data are only restricted to the forest ecosystem. On the other hand, satellite-based maps can have gaps due to cloud cover or weak burned signal due to the time between fire event and image acquisition such as in the case of Figure 4b, where the ground survey took place two days after the fire event (26 August). Omission could be increased by the conservative character of the algorithm adopted. In these cases only field observations can fill gaps. Geometric accuracy of satellite observations can be poor for wider viewing angles thus leading to geo-location errors. It therefore appears clear that the two approaches, ground and remote observations, are complementary and need to be integrated for an exhaustive monitoring of fire affected areas at regional scale. 73

Brivio et al.

Multi-year database of burned areas in the National Parks of Italy

Figure 4 – Examples of post-fire ASTER image (14/09/2004) with the comparison between CFS vector data (black polygons) and results from satellite (white polygons) in the Pollino National Park.

Results distribution and visualisation Burned area maps produced from satellite images are distributed, as previously mentioned, in the Project IDE-Univers WMS (URL: �������������������������������������������������������� http://geoportal.irea.cnr.it:8080/geoportal/local_ it.jsp)������������������������������������������������������������������������������������� . The service has been implemented by an open source, free map server implementation tool, i.e. Minnesota Mapserver. Services of an SDI should allow users to discover and possibly access spatial or geographical information from a wide range of sources, from the local to the global level, in an interoperable way for a variety of uses. Advantages of this choice for diffusion and delivery of results consist in its compatibility with international standards as the Ministry geo-portal “Portale Cartografico Nazionale”, which follows the principles of SDIs. Maps distributed in the project’s WMS are viewed both from the Ministry geo-portal (Fig. 5a) and from the viewer of IDE-Univers (Fig. 5b), without requiring any intervention on the WMS: both viewers simply access the WMS URL address that let them know the type of service delivered, the layers distributed and the visualisation properties. Being both maps and map service at the same premises where information is produced, independence and autonomy in all operations of revision and updating are assured. At the same time the burned area geo-database can be viewed by other Web map viewers compliant with the standards and protocols of the INSPIRE Directive, allowing their exploitation by a wide range of users.

Conclusions

This paper presents a multi-year geographic database of fire affected areas obtained from classification of ASTER and SPOT satellite images over all the National Parks of Italy for the period 2001-2005. Burned areas detected from satellite in this five years amount to 5038 hectares and 652 fire polygons, whose average size is 7.7 hectares. Our results confirmed that the most affected parks are located in Southern Italy. The final vector layers for each National Park and year were compiled in a geo-portal in compliance with INSPIRE standards and implementation rules, so that these spatial 74

Rivista Italiana di Telerilevamento - 2009, 41 (2): 65-78 Italian Journal of Remote Sensing - 2009, 41 (2): 65-78

information can be further exploited by a wide range of users. In particular, the product could be a useful source of information for ecological studies of fire regimes and the assessment of vegetation damage and regrowth.

Figure 5 – Burned areas within the Pollino National Park visualized through a) the “Portale Cartografico Nazionale” of the Italian Ministry of the Environment (URL: http://www.pcn. minambiente.it/PCN/progetto_incendi.htm) and b) the CNR-IREA WMS created for the European Project IDE-Univers (URL: http://geoportal.irea.cnr.it:8080/geoportal/local_it.jsp).

Excluding the cases of Gargano and Pollino (2004 and 2005) and Cilento (2005), fire affected area maps derived from satellite data are the only cartographic information available for the period 2001-2005 and such spatial explicit information can not be derived from any other source for these years. A comparison with data made available from CFS (AIB/FN database and GPS polygons) indicated the validity of satellite maps and confirmed the importance of the integration of remotely and ground based techniques for an efficient forest fires monitoring. This integration could be planned for future systematic monitoring of fire affected areas.

Acknowledgments

The valuable contribution from Corpo Forestale dello Stato (CFS) is acknowledged. ASTER images were acquired thanks to the collaboration of the Department of Geography, University of Maryland, USA. SPOT data were obtained thanks to OASIS Programme (Optimising Access to Spot Infrastructure for Science).

References

Arino O., Plummer S. (2001) - The Along Track Scanning Radiometer World Fire Atlas Detection of night-time fire activity. IGBP-DIS Working Paper 23. Postdam, Germany, 2001. Barbosa P.M., Grégoire J.M., Pereira J.M. (1999) - An algorithm for extracting burned 75

Brivio et al.

Multi-year database of burned areas in the National Parks of Italy

areas from time series of AVHRR GAC data applied at a continental scale. Remote Sensing of Environment, 69: 253-263. Barea M., Brovelli M., Bucci F., Carrara P., Guimet J., Koukourouvli N., Pascual V., Redondo M., Simos D., Vaitis M.N. (2008) - A trans-national SDI for academic geoinformation: Lessons learned in the IDE-Univers project. ������������������������� INSPIRE Conference 2008, Maribor (Slovenia), June 23-25, 2008, pp. 54-56 Blasi C., Bovio G., Corona P., Marchetti M., Maturani A. (2004) - Incendi e complessità ecosistemica: dalla pianificazione forestale al recupero ambientale. Ministero dell’Ambiente – DPN p. 353. Boschetti L., Brivio P.A., Grégoire J.M. (2003) - The use of Meteosat and GMS imagery to detect burned areas in tropical environments. ��������������������������������������� Remote Sensing of Environment, 85 (1): 78-91. Bottai L., Montaghi A., Maselli F. (2008) - Il telerilevamento per il monitoraggio degli effetti degli incendi forestali. Italian Journal of Remote Sensing, 40 (1): 75-87. Brivio P.A., Maggi M., Binaghi E., Gallo I. (2003) - Mapping burned surfaces in SubSaharan Africa based on multi-temporal neural classification. International ������������������������� Journal of Remote Sensing, 24(20): 4003-4018. Brivio P.A., Zaffaroni P., Stroppiana D., Boschetti M. (2007) - Un sistema integrato per il monitoraggio e la mappatura delle aree percorse da incendio nei Parchi Nazionali attraverso l’utilizzo dei dati da satellite. Rapporto Finale. Convenzione Ministero dell’Ambiente, Direzione Protezione Natura, pp. 82. Bucci F., Carrara P. (2008) - Mappatura e visualizzazione di dati geografici eterogenei in un Web Map Server: un esempio applicativo nel progetto Ide-Univers. ��������������� 12a Conferenza Nazionale ASITA, L’Aquila, 21-24 ottobre 2008, pp. 563-568. Chuvieco E., Martin M.P., Palacios A. (2002) - Assessment of different spectral indices in the red-near-infrared spectral domain for burned land discrimination. International �������������� Journal of Remote Sensing, 23: 5103-5110. Congalton R.G. (1991) - A review of assessing the accuracy of classifications of remotely sensed data. ����������������������������������������� Remote Sensing of Environment, 37: 35-46. CFS - Corpo Forestale dello Stato (2008) - Dossier Incendi Boschivi 2007. Csizar I., Morisette J., Giglio L. (2006) - Validation of active fire detection from moderate resolution satellite sensors: The MODIS example in Northern Eurasia. IEEE Transactions on Geoscience and Remote Sensing, 44, 1757−1764. European Parliament (2007) - Directive 2007/2/EC (INSPIRE) Flasse S., Ceccato P. (1996) - A contextual algorithm for AVHRR fire detection. Int. Journal of Remote Sensing, 17 (2): 419-424. Garcia M., Chuvieco E. (2004) - Assesment of the potential of SAC-C/MMRS imagery for mapping burned areas in Spain. Remote Sensing of Environment, 92: 414-423. Giglio������������������������������������������������������� L., De������������������������������������������������ scloitres J., Justice C.O., Kaufman������������� Y. (2003) - An enhanced contextual fire detection algorithm for MODIS. Remote Sensing of Environment, 87: 273-282. Giglio L., Csiszar I., Restás A., Morisette J.T., Schroeder W., Morton D., Justice C.O. (2008) - Active fire detection and characterization with the advanced spaceborne thermal emission and reflection radiometer (ASTER). Remote Sensing of Environment, 112: 3055-3063. Key C.H., Benson N.C. (1999) - Measuring and remote sensing of burn severity. In Proc. Joint Fire Science Conference and Workshop, L.F. ��������������� Neuenschwander �� & ���������� K.C. Ryan (Eds.), vol. II, ���������������������������������������������������������������������� Moscow, ID: Univ. of Idaho and Int. Ass. of Wildland Fire, p. 284. 76

Rivista Italiana di Telerilevamento - 2009, 41 (2): 65-78 Italian Journal of Remote Sensing - 2009, 41 (2): 65-78

Levine J.S. (1990) - Global biomass burning: a case study of the gaseous and particulates emissions released to the atmosphere during the 1997 fires in Kalimantan and Sumatra, Indonesia. In Biomass burning and its relationships with the climate system, J.L. Innes, M. Beniston, and M.M. Verstraete (eds.), Advances in global change research, Kluwer Academic Publishers, pp. 15-31. Mitri G., Gitas J. (2004) - A semi-automated object-oriented model for burned area mapping in the Mediterranean region using Landsat-TM imagery. ������������������������������� Int. Journal of Wildland Fire, 13: 367-376. Mitri G., Gitas J. (2006) - Fire type mapping using object-based classification of Ikonos imagery. International Journal of Wildland Fire, 15: 457–462. Morisette J.T., Giglio L., Csiszar I., Justice C.O. (2005) - Validation of the MODIS active fire product over Southern Africa with ASTER data. International Journal of Remote Sensing, 26 : 4239−4264. Paganini M., Arino O., Benvenuti M., Cristaldi M., Bordin M., Coretti C., Musone A. (2003) - ITALSCAR, a regional burned forest mapping demonstration project in Italy. Int. Geoscience and Remote Sensing Symposium, 2003. IGARSS ‘03. Proc. 2003 IEEE Volume 2, 21-25 July 2003: 1290 - 1292. Petrucci B. (2007) - Azioni del Ministero dell’Ambiente e della Tutela del Territorio e del Mare nel settore degli incendi boschivi. L’Italia Forestale e Montana, 1: 27-38. Piccoli D., Cattoi M. (2004) - I rilievi delle aree percorse dal fuoco effettuati dal Corpo Forestale dello Stato per le attività di polizia giudiziaria. Presentazione del seminario “Telerilevamento e spazializzazione nel monitoraggio forestale e ambientale per la difesa degli ecosistemi” Università degli Studio del Molise, Pesche (IS), 24 January 2007. Prins E.M., Menzel W.P. (1992) - Geostationary satellite detection of biomass burning in South America, International Journal of Remote Sensing, 13: 2783-2799. Roberts G.J., Wooster, M.J. (2008) - Fire Detection and Fire Characterization Over Africa Using Meteosat SEVIRI. IEEE Trans. on Geoscience and Remote Sensing, 46 (4) : 1200-1218. Roy D., Jin Y., Lewis P.E., Justice C.O. (2005) - Prototyping a global algorithm for systematic fire affected area mapping using MODIS time series data. Remote Sensing of Environment, 97: 137-162. Schroeder W., Prins E., Giglio L., Csiszar I., Schmidt C., Morisette J., Morton D. (2008) Validation of GOES and MODIS active fire detection products using ASTER and ETM+ data, Remote Sensing of Environment, Vol. 112 (5): 2711-2726. Silva L.M.N., Sà A.C.L., Pereira J.M.C. (2005) - Comparasion of burned area estimates derived from SPOT-VEGETATION and Landsat ETM+ data in Africa: influence of spatial pattern and vegetation type. Remote Sensing of Environment, 96: 188-201. Smith A.M.S., Drake N.A., Wooster M.J., Hudak A.T., Holden Z.A. Gibbons C.J. (2007) - Production of Landsat ETM+ reference imagery of burned areas within Southern African savannahs: comparison of methods and application to MODIS. Int. Journal of Remote Sensing, 28 (12): 2753–2775. Stow D., Petersen A., Rogan J., Franklin J. (2007) - Mapping burn severity of Mediterranean type vegetation using satellite multispectral data. GIScience and Remote Sensing, 44: 1-23. Tansey K., Grégoire J.M., Binaghi E., Boschetti L., Brivio P.A., Ershov D., Flasse S., Fraser 77

Brivio et al.

Multi-year database of burned areas in the National Parks of Italy

R., Graetz D., Maggi M., Peduzzi P., Pereira J.M., Silva J., Sousa A., Stroppiana D. (2004) - A global inventory of burned areas at 1km resolution for the year 2000 derived from SPOT Vegetation data. Climatic Change, 67 (2): 345–377. Thonicke K., Venevsky S., Sitch S., Cramer W. (2001) - The role of fire disturbance for global vegetation dynamics: coupling fire into a Dynamic Global Vegetation Model. Global Ecology & Biogeography 10 (6): 661–677. Yamaguchi Y., Kahle A.B., Tsu H., Kawakami T., Pniel M. (1998) - Overview of Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER). ������������������ IEEE Transactions on Geoscience and Remote Sensing, 46: 1062-1071. Zaffaroni P., Stroppiana D., Brivio P.A., Boschetti M. (2007) - Utilizzo di immagini ASTER per la delimitazione di aree percorse da incendio, Italian Journal of Remote Sensing, 39: 93–101. Received 19/03/2009, accepted 30/04/2009.

78