FIREGUARD : Mapping wildland fuels and infrastructure at

Transcrição

FIREGUARD : Mapping wildland fuels and infrastructure at
Forest Fire Research & Wildland Fire Safety, Viegas (ed.)
© 2002 Millpress, Rotterdam, ISBN 90-77017-72-0
FIREGUARD : Mapping wildland fuels and infrastructure at the
management unit level with very high spatial resolution satellite imagery
for fire prevention and control in Mediterranean-type landscapes
Cliff Banninger (Joanneum Research, Austria, [email protected]), Alexander
Almer, Hannes Ragam, Andreas Wimmer (Joanneum Research, Austria), Jim Hogg (University of
Leeds, UK), Gavriil Xanthopoulos (NARF, Athens, Greece), Ioannis Gitas and Giorgos Kazakis
(MAICh, Chania, Greece), Kostas Kalabokidis (University of the Aegean, Lesvos, Greece), Celeste
Coelho, Antonio Ferreira, & Carla Domingues (Universidade de Aveiro, Portugal), Jose Rodrigues
& Miguel Galante (Direccao Geral das Florestas, Lisboa, Portugal), Francisco Rego & Maria
Maia (Instituto Superior de Agronomia, Lisboa, Portugal), (Herminio Botelho, Carlos Loureiro, &
Paulo Fernandes Universidade de Tras-os-Montes e Alto Douro, Vila Real, Portugal)
Keywords: wildland fires, fuel and infrastructure mapping, remote sensing, IKONOS, QuickBird
ABSTRACT: FIREGUARD will provide detailed and up-to-date quantitative information on fuels,
terrain, and infrastructure of wildland areas in Mediterranean-type regions, using Ikonos, QuickBird, and similar imagery. The information will be stored on an integrated geographical information system and accessible by fixed and mobile devices over the Internet, using terrestrial and satellite telecommunication networks. FIREGUARD will provide forest and fire-fighting services at the
district and sector management unit levels with more accurate and cost-effective information on
fuel accumulation for improved fire prevention and control than is achievable by currently practiced methods, thereby allowing more precise fire hazard and risk assessment for better fuel and
fire management.
1 INTRODUCTION
1.1 The Mediterranean fire regime
Europe experiences, on average, 60,000 wildfires annually, which destroy approximately 700,000
ha of wildland at an estimated 1.5 billion euros in fire-fighting and restoration costs. The Mediterranean region is particularly susceptible to wildfires, where approximately 500,000 ha of wildland
are burnt each year, with the last 25 years having seen nearly a tripling in area burnt. Of the five
European Mediterranean countries, Spain and Portugal suffer the most in the number, frequency,
and size of wildfires, followed by Greece, Italy, and France.
The Mediterranean ecosystem comprises a heterogeneous and highly fragmented mosaic of
grass, shrub, wood, and forest lands that form a complex and highly variable vegetation cover
within an equally complex and highly variable land use and terrain setting. Together, these factors
strongly influence the region's fire regime. As in most semi-arid climatic regions, the rate of fuel
accumulation is high and the natural reduction through decomposition is low, making the region
extremely vulnerable to wildfires.
The vast majority of Mediterranean wildfires are small in size (< 30 ha), with most being less
than 5 ha. Although only initially affecting a small area, small fires have the potential to grow to
larger and far more destructive fires, if not quickly and properly brought under control and put out.
Because most fires are small in size, it is at the fire management unit level (i.e., district and especially sector levels) that most fires occur and are fought, rather than at the department or regional
level, as for large-scale fires.
1
1.2 Wildland fire controlling factors
The three primary factors influencing wildfire behaviour are vegetation (fuel properties), topography (slope, aspect, and elevation), and meteorological or weather conditions (temperature, relative
humidity, and wind speed and direction). These three factors play crucial roles in all aspects of fire
management strategy and planning, including fuel ignition, combustibility, and flammability determination, fire behaviour prediction, fire hazard, risk, and danger assessment, and in fire management systems. Whereas weather is a major factor in a fuel's susceptibility to ignition and flammability, and topography for the direction and rate of spread of a fire, fuel controls ignition,
combustibility and flammability, and strongly influences fire intensity and spread. Fuel is, therefore, the underlying factor as to whether a fire can occur, as without combustible material there
cannot be a fire.
The amount of fuel (its load) governs a fire's intensity, the spatial distribution of fuel (both vertical and horizontal) its rate of spread and length of front, and the fuel vertical structure a fire's
flame length within a fuel complex. As high intensity fires occur in heavy fuel loads or accumulations and are the most difficult to control, fuel is one of the most important factors in fire control.
Of fuel, topography, and weather, fuel is the only one that can be modified to help prevent wildfires, or to control them once they have started. Fuel is, therefore, the key element in any fire prevention and control strategy and management for a wildland area.
Topography also exerts a strong control over the behaviour of a wildfire, through the slope and
aspect of a terrain on the exposure of fuels to solar isolation and their loss of moisture and subsequent drying out. Slope also exerts a control on the propagation of a fire, by its influence on wind
speed. Aspect, on the other hand, exercises a control on fire spread direction, through its influence
on wind direction, whereas elevation affects fuel flammability by its influence on air humidity. The
exposure of fuels to ventilation and the drying effects of wind is also dependent upon its topographic position. These factors make topography an important factor in fire management.
1.3 Hindrances to effective fire prevention and control
Information on fuel complexes and topography are required by forest and fire managers to help them in
their decision making for the management of wildland areas, and in determining wildfire hazard, risk,
danger, and behaviour, and fuel hazard. This information is also needed for field work related to the carrying out of fuel management activities and in the modelling of fire behaviour for prediction purposes.
Fuel management practices are an integral part of land and fire management strategies for controlling fuels. Information on fuel properties is essential for deciding short, medium, and long-term
management options in deciding on the most effective fuel treatment and fire suppression measures. Forest and fire managers face a multitude of problems and uncertainties with regards to fuel
reduction, as the task is complex and quantitative information on fuels is generally not available or
only of a general nature, as well as often not easily and economically attainable. This hinders effective fire prevention and pre-suppression activities, ranging from the planning to the operations
phase. Problems and uncertainties related to the lack of sufficient information on fuel properties are
particularly evident in the application of models that have been developed to help explain, understand, and predict fire behaviour under varying fuel, terrain, and weather conditions.
Many mathematical models for predicting fire behaviour and determining fire danger are currently available for forest and fire management support (e.g., BEHAVE, FARSITE, FIRE!,
CFFDRS). These models are used by forest and fire-fighting organisations for analysing wildfire
scenarios and suppression alternatives, by predicting the behaviour of active fires (e.g., rate and direction of spread, fire intensity and perimeter, and potential fire paths and corridors) for planning
pre-suppression (particularly prescribed burning and fuel-breaks) and suppression strategies and
measures. They are used widely for fire prevention planning and fuel modelling, as well as for general contingency planning purposes. Aspects of these models are also used in fire management decision support systems (e.g., AFFIRM, PROETHEUS, FOMFIS-FMIS, AIOLOS, E-FIS). All the
models require spatial information on vegetation cover characteristics (fuel models) and terrain (to-
2
© 2002 Millpress, Rotterdam, ISBN 90-77017-72-0
pography) as input parameters, in addition to weather (wind speed and direction, air temperature,
and air humidity), but fuel information is, however, by far the most important of the three.
2 FIRE MANAGEMENT FUEL INFORMATION NEEDS
The biggest problem facing forest and fire-fighting services in fuel management for fire prevention and
control is the high diversity and fragmentation of the Mediterranean plant community and the great
variability shown by it over short distances. The patchy, quilt-like distribution of vegetation arising
from the discontinuous canopy closure that is characteristic of the Mediterranean landscape is a result,
in part, of the region's land being approximately 65% privately owed, with most holdings smaller than 5
ha. This makes it difficult to adequately inventory and monitor the Mediterranean landscape on a regular and up-to-date basis for fire pre-suppression planning purposes by traditional field surveys. There is
a pressing need for comprehensive and routine inventorying and monitoring of wildland fuel properties
and fire-fighting infrastructure, with a focus on high fire risk areas, where frequent monitoring is necessary to identify when fuel loads approach a critical stage. Although fire policy and strategy are often
formulated at the national, regional, and department or provincial level within a country, it is at the district and sector forest management unit levels that most fires start and are fought, and where information on fuels, infrastructure, and terrain is needed to prevent and combat fires.
Accurate information on fuel complexes, such as accumulation, stratification, distribution, and
dynamics is required to properly assess potential fire behaviour and to be able to undertake effective proactive fire prevention and control measures. However, there is an acute lack of detailed
quantitative information on fuel complexes and especially their dynamics over time (seasonally and
yearly), especially at scales that can effectively support fire behaviour modelling and fire prevention and control measures at the forest management unit level. This is also true for infrastructure
and topographic information, which is needed for both fire behaviour modelling and presuppression and suppression purposes. The monitoring of fuel complexes is particularly important
for overseeing forest and fire management strategies, and ensuring that their implementation has
been properly carried out. Fire managers are sorely in need of such information at the right time
and at the right scale, to be able to undertake timely and accurate assessments of fire hazard situations. It is only in this way that they can take specific and effective actions to reduce fire hazard
and risk through appropriate fire prevention and control measures.
One of the chief constraints to collecting information on fuel complexes is the limited available
resources (manpower, funds, and infrastructure) of forest and fire-fighting organisations. For example, the traditional field survey approach to collecting data is time-consuming, laborious, and
costly, as well as selective in application. It consists predominately of point, plot, and transect acquired data, which are difficult to characterise in a broader context, especially in a highly heterogeneous landscapes, such as the Mediterranean region.
2.1 Current wildland fuel mapping practices
Forest and fire managers, for the most part, lack adequate fuel maps of their forest districts and sectors, both in thematic content (e.g., detailed information on fuel type, amount, structure) and scale
(scales larger than 1:50 000). Even when available, they are slow and costly to produce by traditional methods and difficult to update, as most site evaluations for wildfire planning and management purposes are still based on traditional methods of ground surveys (e.g., transects) and field
sampling (e.g., point and small plots), supplemented possibly by conventional aerial photography.
These survey methods are time consuming, tedious, and man-power intensive, and, hence, costly.
In addition, they generally provide only limited and often selective coverage of a site and the analogue nature of their data makes them awkward and inconvenient to work with and update. Site inventories derived from these surveys are often far out-of-date, as the updating of maps by these
methods is normally undertaken only every 5–10 years, owing to the time-consuming and expensive preparation of fuel maps by conventional methods, including aerial photography. Because of
Forest Fire Research & Wildland Fire Safety, Viegas (ed.)
3
the high cost of treating fuels, most fire prevention and control programmes have time and budget
constraints on the gathering of information from a site that are not adequately served by field surveys, especially for large, difficult to access, or remote areas.
Most fire behaviour models and management systems provide useful insight and greater understanding into how wildland fires develop and spread under different sets of initial assumptions;
however, most models and systems lack precision, owing to both the generalities or uncertainties in
the quantification of the fuel, topography and weather input characteristics, and the small scale of
the output maps (normally 1:50 000 to 1: 400 000). Because of these limitations, fuel complexes
that are dominated by a single continuous stratum, such as grass or shrubs, are more suited for fire
modelling purposes than those that are discontinuous and comprise several strata, such as mixed
forest-shrublands. However, most Mediterranean wildlands have a complex fuel arrangement that
often consist of several fuel strata, such as a litter layer, a grass-forb layer, a shrub layer, an understorey layer, and an overstorey layer, which makes them difficult for fire modelling. Also, fuel
loading and continuity often vary considerably over short distances, even within a stand, and the
homogeneity and continuity of fuels differ among vegetation types.
2.1.1 Ground surveys
The traditional and current manner that forest and fire-fighting services gather information on wildland fuel properties and infrastructure is by field investigation, which is manpower intensive, time
consuming, laborious, expensive, and limited in the amount of information it can provide for an
area. Field surveys are normally affected by a shortage of personnel and funds to undertake and
support the work that is required to properly map an area, especially where wildland areas that are
remotely situated and not easily accessible by field crews. The collection of data by in situ field
methods, involving point and plot sampling and transects and allows for only small and limited
coverage of a selected area and a non-comprehensive inventorying of fuel characteristics, which
can be extremely difficult to properly characterise for a heterogeneous environment, such as Mediterranean wildlands. Because of the slow nature of this work, updating is not frequent enough to
map the seasonal and even yearly changes that occur in fuel complexes and land use, which results
in incomplete, generalised, inaccurate, or out-of-date maps often being used for forest and fire
management planning.
2.1.2 Aerial photography
Aerial photography is used to overcome some of the deficiencies of ground surveys, although most
aerial photographs that are available are out-of-date or even not available, of inconsistent or poor
quality, or generally restricted to black and white photographs, because of the expense and military
restrictions of acquiring up-to-date and colour aerial photographs. In many Mediterranean countries
(e.g., Greece), access to aerial photography is often restricted by the military, with only low resolution black and white aerial photos being made available of an area. Aerial photography is also very
expensive, especially for large areal coverage and for colour imagery, time-consuming to work
with, and labour intensive, and its use in the production of maps requires a long lead time. Aerial
photography campaigns involve extensive preparation time and logistic support, high stand-by time
costs in inclement weather, long delivery time (weeks to months) from time of acquisition, and an
infrequent re-flying of an area, with an average interval between surveys of close to 10 years. resulting, as with maps produced from ground surveys, in them often being out-of-date by the time
they are available.
Aerial photographs have the additional limitation of a lack of discrete spectral bands that are often required for detailed fuel properties’ mapping, plus the problem in maintaining colour balance
between emulsion lots and processing batches and, thereby, consistency of tonal or colour rendition
between flight strips. Aerial photography is also plagued by low dynamic range (5–6 bits or 32–64
grey levels) and poor contrast, especially in vegetated areas. Although aerial photographs can be
scanned and digitised, this incurs a loss in spatial resolution and dynamic range. Aerial photographs
are also only provided in a fixed product size (dependant on scale), and expensive and time-
4
© 2002 Millpress, Rotterdam, ISBN 90-77017-72-0
consuming mosaicking is required for large area seamless coverage. Aerial photographs, however,
is the only remote sensing data that can currently provide below 0.6 m data of an area.
Although severely limited in their ability to provide forest and fire managers with needed information for fire prevention and control, field surveys and aerial photography have been and still
remain the primary method of assessing and monitoring wildland areas with respect to fire hazard
and risk.
2.1.3 Aerial digital cameras and remote sensing systems
The advantages of acquiring imagery in digital format over that of analogue or film format are the
former's faster, more efficient, and hence more convenient storing, retrieval, processing, manipulation and transformation (enhancement), and annotation capabilities. The digital format also allows
for easy integration with other digitised data. A few digital aerial cameras are currently available,
but their data are much more expensive than film aerial photography. Although digital cameras offer the flexibility of discrete and selectable spectral bands, a major drawback of digital cameras is a
lack of on-board calibration, making imagery difficult to calibrate. Also, frame-based imagery is
hard to register accurately.
2.1.4 Satellite remote sensing systems
Satellite remote sensing is able to provide a synoptic overview and complete coverage of a site
anywhere in the world that is unable to be performed or easily achieved with field surveys and aerial photography (except, perhaps, high altitude photography). It can provide cartographic quality
data, comprehensive and repetitive coverage, and rapid surveillance for site inventorying, monitoring, and updating in an efficient, cost-effective, and frequent manner that is better than by field investigations and aerial photography. Satellite remote sensing can be particularly effective in the detection, identification, classification, and evaluation of wildlands situated in difficult to access or
remote areas, and it is often the only feasible way of obtaining data from such areas. Satellite remote sensing data are able to provide fast and efficient analysis and interpretation of site features
and conditions and, because of their digital format, are compatible with geographical information
and visualisation systems.
Spaceborne remote sensing data provide site information in digital form and at a fraction of the
cost associated with the traditional methods (ground and aerial photography surveys) currently in
common use. When combined with geographical information systems, fuel complexes can be rapidly assessed and regularly monitored, thereby allowing appropriate fuel treatment to be undertaken. One of the major advantages of remote sensing data over aerial photography is that the latter
is film-based, whereas, except for Russian data, almost all satellite imagery is acquired digitally.
Compared to aerial photography and field surveys, a combined satellite remote sensing, GIS,
and visualisation approach to fire prevention and control is able to provide information required for
fire planning and management in a rapid, digital, and cost-effective manner, thereby avoiding the
need of scanning analogue data before they can be stored in a GIS database. Remote sensing, therefore, permits fast and efficient updating of databases compared to traditional data acquisition methods and mapping techniques, and the digital nature of their data allows for their ease of manipulation and the efficient extraction and interpretation of information at accuracy's comparable to aerial
photogrammetry.
A major advantage of satellite remote sensing data is that they can acquired information on fuel
complexes quicker than by and for a fraction of the cost of traditional methods, particularly field
surveys, thereby providing significant savings in manpower and time, and hence cost to fire prevention and control programmes. Data acquisition costs are approximately a tenth of that of conventional field surveys, and, as the cost of creating a database is normally approximately 80% of a
project's cost, the cost benefits of a combined remote sensing and GIS's approach to forest and fire
management planning and implementation will be significant. Because of the lower costs of satellite data, more current information of a site can be obtained for more frequent updating of changes
at a site. Satellite remote sensing can substantially reduce the requirements for extensive and costly
Forest Fire Research & Wildland Fire Safety, Viegas (ed.)
5
ground data collection, and is able to provide information on where field surveys and ground sampling, where required, should be focused, thereby providing for significant cost savings.
The fragmented nature and spatial complexity of most Mediterranean wildland areas have, however, inhibited up until now the use of currently available satellite data, with their much coarser
spatial resolution than conventional aerial photography, for most forest and fire management uses
at the district or sector forest management unit levels, including the extraction of detailed information on fuel properties and infrastructure.
2.1.5 Low to high spatial resolution satellite systems
Current Earth observation satellites have spatial resolutions that are generally not sufficient to capture the high spatial variability common for Mediterranean fuel complexes, and the use of multispectral satellite imagery, except for Russian film-based satellite imagery. Where the forest and fire
management services have made use of spaceborne remote sensing data, it has been with coarse
resolution multi-spectral (MS) and panchromatic (Pan) imagery, such as NOAA AVHRR, Landsat
Multispectral Scanner (MSS) and Thematic Mapper (TM), SPOT, and IRS. The rather coarse spatial resolutions of these imagery (AVHRR = 1.1 km, MSS = 80 m, TM MS = 30 m, SPOT MS = 20
m, IRS MS = 24 m, TM Pan = 15 m, SPOT Pan = 10 m, IRS Pan = 6 m) have limited their use
primarily to land cover and land use mapping and monitoring (normally only to Corine Level II) at
the regional and pan-European scale.
SPOT and IRS panchromatic imagery, at 10 m and 6 m spatial resolution, respectively, are only
6 bit and have non-optimised gain settings that result in a low dynamic range for many surface features, especially forested areas. Landsat TM 7 panchromatic data (15 m, 8 bit) have yet to be extensively evaluated for forest and fire-related information content, as have also SPOT-4 Vegetation
(1.1 km), Resurs-03 (170 m), and IRS-C & D WiFS (188 m) multi-spectral data, other than in a few
studies. For wildland fire applications, AVHRR data have, by far, been the most extensively used,
followed by a distant TM, with the others hardly being employed at all.
The temporal resolution (revisit period) of SPOT is 26 days (in theory, 13 days with SPOT 2
and 4), IRS 24 days (in theory, 12 days with IRS-C and D), Landsat TM 16 days (in theory, 8 days
with Landsat 5 and 7), and AVHRR twice daily (four times with two satellites). Except for
AVHRR, imagery of an area is acquired normally only on a request or demand basis, which can
prove a hindrance when high frequency data need to be acquired of an area, especially in times of
the year when cloudy weather conditions commonly prevail. The delivery period from the time of
image acquisition to receipt by a customer is usually between 4–6 weeks, except where local facilities that have installed their own AVHRR receiving station to collect imagery in real time.
Digital elevation models can be generated from SPOT panchromatic stereo images, although,
because they are acquired by across-track pointing from different orbits (and, hence, dates and under different viewing and atmospheric conditions) often are not able to provide suitable data quality
to achieve the expected height and positional accuracy of 10 m.
Because of the large instantaneous field of view (IFOV) or pixel size of most satellite remote
sensing systems, automated classification techniques have commonly been used for mapping fuel
types. Spectral indices (NDVI, SAVI, ratios, PCA, Tasselled Cap) and classification techniques
(supervise, unsupervised, maximum likelihood) have been commonly applied to AVHRR, TM,
SPOT, and IRS imagery, to determine or estimate canopy features (species type, canopy cover,
LAI, biomass). Classification accuracy's achieved with TM for Mediterranean wildlands have been
only moderately successful, even with very generalised vegetation categories. For example, Vasconcelos et al. (Proceedings 3rd International Conference on Forest Fire Research, Luso, Portugal,
1998) obtained classification accuracy's with TM data for a test area in Portugal of 62% for agricultural fields, 75% for shrublands, 33% for pine forests, and 75% for other forest types. The coarseness of the spatial resolution of the TM data with respect to the high degree of heterogeneity and
fragmentation of the Mediterranean landscape contributed significantly to the mixed and only moderately successful classification results.
Traditionally, planimetric and thematic maps of a site are derived by aerial photogrammetric
techniques or field survey, the latter using standard terrestrial methods. Where such data are lack-
6
© 2002 Millpress, Rotterdam, ISBN 90-77017-72-0
ing, multi-spectral satellite imagery, such as Landsat and SPOT, have been used, but their spatial
resolutions (20-80 m) are more suitable for regional, rather than forest district or sector management unit level mapping. Maps generated from TM, SPOT, and IRS multispectral imagery generally are at scales of between 1:25 000 and 1:50 000.
There is a clear need for improved spatial resolution information, along with new techniques
and methods of analysing data for forest and fire management planning, especially in the mapping
of Mediterranean wildland fuel properties, infrastructure, and terrain characteristics, than is currently available or in use. Up until now, the best spatial resolution available from satellite imagery
is in the 6–15 m range for panchromatic and 20–30 m for multi-spectral. What is needed is very
high spatial and temporal resolution remote sensing data that would be able to provide most of the
requisite information required by forest and fire managers in a efficient, rapid, and cost-effective
manner, and at a scale and accuracy that can meet most of their needs and requirements. Whereas
ground survey techniques and aerial photography are currently the traditional methods used to acquire this information from a site, a new generation of very high spatial resolution spaceborne sensors (e.g., Ikonos, QuickBird, EROS) can provide comparable information easier, faster, and
cheaper, and in a digital format that is more convenient and efficient to use for handling and analysis purposes, especially with geographical information systems (GIS).
2.2 Very high spatial resolution satellite systems
The Ikonos, QuickBird, and EROS sensor systems currently in orbit offer two metre or better panchromatic and between 2.5 and 3.5 multispectral (Ikonos and QuickBird), plus stereo, imagery.
Both Ikonos and QuickBird have four multi-spectral bands in the blue, green, red, and near-infrared
wavelength regions that are equivalent to the first four Landsat Thematic Mapper bands. Stereo imagery can be acquired in both along and across track pointing for three-dimensional viewing and
analysis and digital elevation model (DEM) generation with height accuracy’s of between 2–5 m.
Both the panchromatic and multi-spectral data have a high signal-to-noise ratio and a dynamic
range of 11 bits (2048 levels), versus 6 bits for aerial photography and Landsat, SPOT, and IRS
panchromatic data, and 8 bits for multispectral TM, SPOT, and IRS data, for more information extraction capability, especially for imagery acquired under bright and low lighting conditions. Ikonos and QuickBird have both relative and absolute radiometric calibration of the data, thereby providing consistent high spectral and radiometric (dynamic) fidelity, which is only matched by the
Landsat TM system.
Because of the optical configuration of the sensor systems and their placement in space, the imagery acquired contains little geometric distortion, thereby requiring minimal rectification to produce geo-referenced and geo-coded ortho-images, thereby allowing consistently accurate image
maps to be produced. The use of star-tracking and the global positioning system (GPS), along with
the high geometric integrity of the data, permits the accurate location of ground features in the imagery with a positional accuracy of a few metres for precision mapping with a minimal amount of
ground control required. Positional accuracy of image features is approximately 2 m horizontal and
3 m vertical for imagery with ground control points (GCP's), which is accurate for maps to be generated down to a scale of 1:2500, or 12 m horizontal and 10 m vertical without GCP's (GPS referenced) that is accurate for maps to be produced at a scale of 1:25 000. This high level of spatial accuracy greatly increases the usefulness of the fuel, infrastructure, and terrain information derived
from such imagery for wildland and fire management purposes.
The high temporal resolution of the new generation satellite systems allows repetitive acquisition of imagery from any area in the world at a revisit frequency of 13 days by a single satellite,
depending on latitude (compared to 5–10 year intervals between aerial photography campaigns), or,
with a constellation of satellites, every few hours, thereby ensuring the availability of a continuous
source of time critical data. Fast image processing and delivery are another hallmark that sets the
new data apart from the other types of satellite data, with product receipt from the time of data acquisition to its delivery to the user being 48 hours or less by the internet, or no more than 72 hours
by courier service. Because of the frequent revisit capacity and near real-time data acquisition pos-
Forest Fire Research & Wildland Fire Safety, Viegas (ed.)
7
sibilities of the sensor systems, the de facto real-time receipt of the data via the internet from the
remote sensing data service companies, and the similar real-time geo-processing and analysis of the
data in a geographical information system (GIS) and product generation, ensures that timely information for an area is readily at hand.
Very high resolution satellite data allow both photointerpretation and photogrammetric, as well
as semi-automated and automated information extraction techniques, to be used to ascertain stand
compositional and structural features, versus the use of classification algorithms normally applied
with coarser resolution satellite data, such as Landsat, SPOT, and AVHRR data. One of the major
advantages of very high spatial resolution imagery over the coarser resolution imagery is that they
display features at the human or local scale, rather than the regional or continental scale, so that
ground features and details are readily identifiable and comprehensible by forest and fire personnel.
Canopy architecture, Individual trees and their structure, understorey, trails, small roads and
streams, lookout towers, water storage tanks, and vehicles can easily be discerned and analysed.
The familiarity with the use and interpretation of medium to high altitude aerial photography in the
forest services that has existed for over 40 years ensures that Ikonos-type data are equally readily
comprehensible by forest and fire-fighting service personnel.
Almost all applications of AVHRR, Landsat, and SPOT multi-spectral data for vegetation mapping have involved the use of various vegetation indices (e.g., ratios, normalised difference vegetation index), principle component analysis and its tasselled cap derivative, or unsupervised and supervised classification in conjunction with the maximum likelihood or similar classifiers. The
vegetation classes that are normally derived are usually quite limited (e.g., coniferous, deciduous,
and mixed forests), as are canopy cover (sparse, moderate, and dense), crown closure (open, partial,
and closed) estimates, and age (juvenile or thicket, mature, old). The level of accuracy normally attainable with these classifiers, even for broadly defined vegetation categories, is in the order of 6575%.
The use of very high spatial resolution satellite remote sensing data, especially stereo imagery
for three-dimensional viewing, will provide more quantitative and deterministic information on a
greater number of vegetation properties, such as vegetation types and species, canopy architecture
and structure, tree and canopy height, biovolume, and biomass, and the understorey and overstorey
layers of a fuel complex than previously obtainable from coarser spatial resolution satellite data.
Sub-metre panchromatic data can be used to provide information on stand composition and architecture, tree and stand structure, and, with stereo imagery, tree and canopy heights, which can be
used for biomass estimations. In addition, panchromatic data is able to look through gaps in a canopy at the understorey, using information derived from advanced texture algorithms, such as the
Fourier Transform, Wavelets, and Gabor Features, for use in a rule-based classifier. The advantage
of these type of algorithms compared with the vegetation indices and classifiers (e.g., maximum
likelihood) commonly in use with coarse spatial resolution satellite remote sensing data is their
flexibility and robustness in extracting information from imagery, with classification accuracies in
the region of 90% for an expanded range of vegetation properties.
Very high spatial resolution satellite data are very cost competitive with respect to aerial photography on both a large and unit area basis commensurate with the scale of imagery. The unit area
cost of maps and other information products derived from very high spatial resolution data will be
much less than comparable aerial photography derived products, especially for updating purposes,
with orthorectified satellite imagery costing less than 20% for that of ortho-rectified aerial photography, owing to greater convenience, rapidity, and efficiency in processing digital versus analogue
data. Black and white aerial photography costs approximately 5 euros per hectare, or 500 euros per
square kilometre, whereas very high spatial resolution panchromatic and multispectral imagery
costs about 15 euros per square kilometre for the same areal coverage, and are expected to drop to
in price in the coming years. Although current satellite imagery already costs a fraction of that of
aerial photography, the need to purchase only the ground coverage required for a study, instead of a
fixed and pre-defined areal coverage provided by aerial photography and coarser resolution satellites, makes very high spatial resolution satellite data even more cost-effective.
8
© 2002 Millpress, Rotterdam, ISBN 90-77017-72-0
In summary, the resolution and scale of very high spatial resolution satellite imagery for use in
fuel inventories overlaps and complements that of aerial photography, which is normally in the
0.15–1.2 m resolution and 1:2 000 to 1:40 000 scale range. With most Mediterranean forest stands
being less than 1 ha in size and rarely larger than a few 10's of hectares, the resolution of Ikonos,
QuickBird, and EROS imagery can provide comparable information as is now obtainable by medium-scale aerial photography, but in a faster, less costly, and more convenient manner. The ability
of these sensor systems to provide stereo imagery at the aerial photography resolutions and scales
will permit conventional photo-interpretation techniques to be employed in their analysis, and,
thereby, allow, in addition to automatic and semi-automatic retrieval procedures, the extraction of
information in a manner that is familiar to most foresters and fire-fighting personnel. Being able to
display and view a landscape with recognisable objects and features in three-dimension in a realistic as well as at the human scale, will take a great deal of uncertainty out of the feature interpretation and map generation.
3 THE FIREGUARD APPROACH
FIREGUARD offers a rapid and efficient approach to acquire information on fuel complexes and
infrastructure required by forest and fire-fighting services at the forest management unit level. Very
high spatial resolution satellite remote sensing data, in combination with geographical information
systems (GIS) and information technology (IT), offers a cost-effective approach to providing accurate and timely information at a scale and format that is easy to comprehend and readily usable by
forest and fire-fighting personnel, whether for fire behaviour modelling or fuel assessment. When
used in conjunction with selective field survey data, remote sensing offers significant potential to
increase the knowledge base on an area's fuel complex and thereby provide the information required to implement appropriate pre-suppression measures for effective forest fire prevention and
control.
The control of fuel accumulation by forest and fire-fighting services is mainly directed at strategically important or high asset value places, and in areas or sectors within a forest district that are
of high risk. Most fuel treatments are targeted at specific sites within a forest management sector
and comprise near, medium, and long-term applications, thereby making maximum use of human,
equipment, and financial resources, and eliminating or reducing the need to obtain detailed fuel information for an entire forest district.
3.1
FIREGUARD objectives and expected achievements
The overall aim of FIREGUARD is to provide detailed information on fuels, terrain, and infrastructure to forest and fire-fighting services at the district and sector levels for use in fire prevention and
control. FIREGUARD will demonstrate how newly available very high spatial resolution satellite
remote sensing data in combination with an integrated geographical information system (IGIS) can
be used to better target fuel treatment and fire suppressing activities within the framework of fire
prevention and control planning and management..
FIREGUARD will both directly and indirectly address fire prevention and control in its three
stages:
Ɣ Long-term fire prevention
Ɣ Medium-term fire pre-suppression
Ɣ Short-term fire suppression
Long-term fire prevention requires managing fuels for the reduction of ignition risk through fuel
treatment, whereas medium-term fire pre-suppression is primarily concerned with fire readiness
and the pre-positioning of fire-fighting equipment and aids in high fire risk areas, and short-term
fire suppression pertains to fire control and the active measures taken to combat a fire. The main
focus of FIREGUARD is on long-term fire prevention.
The objectives of FIREGUARD are:
Forest Fire Research & Wildland Fire Safety, Viegas (ed.)
9
1) To develop improved tools and methodologies for providing more accurate information for identifying, quantifying, and evaluating fire hazards and risk than is currently obtained by traditional
measures.
2) To develop a proto-type integrated geographical information system (IGIS) for providing information on fuel, infrastructure, and terrain in the Mediterranean region at the district and sector
forest management unit levels, using information technology (IT).
3) To derive information products from remote sensing data on fuel properties, terrain, and infrastructure for efficient fire prevention and control planning and management.
4) To demonstrate the ability of the system to provide comprehensive, accurate, and up-to-date information on fuels and their spatial and temporal variability for enhanced and more efficient fire
prevention and control purposes.
5) To access the IGIS via fixed and mobile devices over the Internet, using terrestrial and satellite
telecommunication networks.
A multi-tier information system will be developed that can be integrated into forest and land
management systems and which is specifically designed for fire prevention and control planning
and implementation. Fuel property, infrastructure, and terrain information will be offered through a
variety of type and format options:
1) Parameters for fuel, fire behaviour, hazard, risk, danger, and fire and fuel management models.
2) Two and three-dimensional annotated black and white and colour hardcopy image-based maps.
3) Data and information accessibility from the IGIS over the Internet, using terrestrial and satellite
communication satellites.
The information will be available over the Internet for down-loading and importing onto desk
and laptop computers, for examination in the office and the field via fixed and mobile devices. Experienced judgement is an import means of appraising fuel complexes and local forest and firefighting service personnel will be able to quickly and easily integrate both objective and subjective
factors related to fuel management requirements from the information products derived from the
system. It will help to define the likely type of fire (ground, surface, or crown) to be expected in an
area, which will provide for better organisation and deployment of fire-fighting equipment and personnel, and the establishment of appropriate infrastructures (such as fire access roads) in fire-prone
areas. By providing information on the fuel status of an area, the system will improve fire hazard
and risk assessments and help establish fire protection priorities and the implementation of fire preventive actions, such as pre-scribed burning. It will also improve fire preventive siliviculture practices, particularly with the planting of fast growing, but highly imflamable trees, such as pines and
eucalyptus.
The information generated by the system will also enhance the reliability and accuracy of fire
behaviour simulation models for predicting the dynamics of a fire, by providing more frequent and
accurate information on fuel accumulation in an area for input into the modes and fire management
systems employed by forest and fire-fighting services. This will allow a better understanding and
assessment of the multi-dimensionality of the fuel and terrain conditions that predispose fuel build
up and influence fire patterns. From this, a better understanding of the linkage between various fuel
properties will be forthcoming.
As part of the assessment of the information products developed in FIREGUARD, an information and cost-benefit analysis will be performed on the use of very high spatial resolution satellite
remote sensing data in comparison with traditional information gathering methods employed by
forest and fire-fighting services.
3.2 Forest and fire-fighting service information requirements
A proper evaluation of a site is crucial to the planning and carrying out of a successful fire prevention and control programme and requires detailed and accurate information on, as well as an understanding of, fuel, infrastructure, and terrain characteristics of an area. This information is used by
forest and fire-fighting managers for fuel management planning (e.g., fuel treatment) and fire presuppression activities (e.g., pre-positioning of fire-fighting equipment), and is most useful at the lo-
10
© 2002 Millpress, Rotterdam, ISBN 90-77017-72-0
cal forest management unit level. Meetings have been held with forest and fire-fighting service personnel of the forest districts and sectors in which the test areas are situated, to ascertain their information needs and requirements with regard to the above aspects of fire management and the type,
format, and scale at which they prefer the information to be provided.
Of particular importance for planning and implementing fuel treatment measures within a fire
prevention programme is information on the composition, structure, and other biophysical properties of the vegetation components comprising the ground, surface, understorey, and overstorey fuels and their horizontal and vertical structural characteristics and distribution, as well as seasonal
and yearly changes.
3.3 Field investigations
Test areas have been selected in Portugal and Greece that encompass a diversity and mixture of
shrubland and forest covers, as well as terrain types and climatic conditions that are representative
of the Mediterranean region. Vegetation types include pine tree stands of the species Pinus pinaster,
Pinus sylvestris, Pinus pinea, Pinus negra, and Pinus brutia (“Pinus halepensis”), as well as Cupressus sempervirens (cypress), and Eucalyptus globules. A wide variety of woody shrub species going
under the collective names of marquis, garrigue, and phrygana are included within the test areas
and, in some cases, form the dominant vegetation cover. Terrain settings encompass coastal, upland
plain, and mountainous areas, and climatic conditions range from maritime to Mediterranean.
3.4 Field investigation of fuel properties
Field information is currently being acquired on the fuel properties at each of the selected test areas
and comprise a detailed inventory and description of the fuel, infrastructure, and terrain properties
found at each of them, in accordance with the information requirements of the forest and firefighting services. Biometric measurements entail species composition, architecture, structure, and
vertical and horizontal spatial distribution and continuity from representative sites within each test
area. Stand architecture and structure play a critical role in fire management planning, as the vertical structure of elevated fuels influences wind flow through a stand and hence its drying
(evapotranspiration) and fire dynamics and canopy density controls the presence of understorey and
surface fuels through fuel shading. Stand structure is a combination of a vegetation stand's height,
density, crown closure, crown basal diametre, crown base height above ground, and biomass.
The mapping of fuel layers for each complex comprises a major aspect of the field investigations at each of the test sites. Multi-layered stands can consist of any combination of upper (overstorey), middle (understorey), lower (surface), and ground stratum and it is not uncommon for
most, if not all, of these layers to be present in the test areas.
Allometric equations will be established through destructive sampling that relate canopy fuel
properties to discernible and measurable canopy properties by remote sensing means, through standard field measurements and laboratory processing and analysis procedures used in forestry. These
will include determinations of biovolume and biomass and crown bulk density (i.e., biomass per
unit volume) of the dominant and secondary plant species, by calculating stand area and average
plant density, weight (fine fuels), and height. This will give a measure of the aeration state of a fuel
and its flammability, and an estimate of litter layer loading as a function of crown bulk density. The
proportion of dead to live fuels is an important biometric in fire behaviour modelling and this will
be estimated by standard field methods. Positional control of the field sampled areas will be acquired through the use of global positional system (GPS) measurements, which will be able to be
directly incorporated into a IGIS database.
3.5 Field investigation of infrastructure
Information will be acquired for the planning of fire suppression measures and the allocation of fire
suppression assets by fire-fighting organisations, to provide fire crews with effective logistic sup-
Forest Fire Research & Wildland Fire Safety, Viegas (ed.)
11
port and to ensure the most efficient response for the containment and extinguishing of a fire. This
normally entails the pre-positioning of fire-fighting equipment and the identification of support resources that can be called upon to fight a fire. Information on water supply sources (points), natural
fire-breaks, and fire buffer zones, and the status of fire access roads, tracks, and trails and artificial
fire-breaks with respect to their condition and maintenance will be also be acquired. In addition, the
identification of emergency escape routes and safety zones for suppression crews will be noted.
Land use and management practices and especially the interface between wildlands and urban area,
and even rural areas, will be identified for fire pre-suppression planning purposes.
3.6 Remote sensing data
Very high spatial resolution satellite panchromatic stereo and multi-spectral imagery (Ikonos,
QuickBird) will be procured for the selected test areas. Image processing will include the removal
of radiometric variations in the response of detectors (radiometric calibration), removal of the effects of the atmosphere on the spectral response of ground features (atmospheric correction), removal of the effects of topography on the spectral response of ground features (topographic normalisation), removal of geometric distortions related to terrain displacement, using a DEM-based
ortho-rectification procedure (geometric correction and geo-referencing), and conversion to userdefined cartographic projection. This will result in image maps having absolute map co-ordinates
and permit the imagery to be accurately co-registered with the field data sets. Digital mosaicking of
rectified images will be performed for large area coverage requirements, where necessary. Definition and enhancement of image features will be carried out by the application of standard and specially constructed filters and other image enhancement procedures.
Highly accurate digital elevation (DEM’s) and terrain (DTM’s) models will be generated from
either Ikonos or QuickBird stereo imagery for each of the selected test areas. These models will be
used to geometrically correct and geo-reference the remote sensing imagery in support of the
analysis and interpretation of site features and for visualisation purposes.
3.7 Data analysis and interpretation
Information requirements as defined by the forest and fire-fighting services will be extracted from
panchromatic and multi-spectral imagery for a detailed analysis of site fuel and infrastructure characteristics. Ground data collected will be used in support of this analysis, as will terrain and topographic information provided by the DTM's and DEM’s. Ground data from the test sites will be used
to ‘calibrate’ and help refine the interpretation of the remote sensing data. Information on fuel
complexes and infrastructure will be acquired through both traditional photo-interpretation (visual)
methods, and by semi-automated, automated, and hybrid information extraction procedures of the
remote sensing imagery, based on spectrally and spatially related algorithms and empirical and
semi-empirical derived relationships. High level image-based fuel property discriminators, based
on both spectral and structural characteristics of vegetation complexes in the remote sensing data
and involving both texture and shape information, will be employed, and, where necessary, developed.
The goal is to derive an automated procedure for the extraction of information within an image.
Rule-based fuel attribute classifiers will be formulated and evaluated, and their performance refined through an iterative process that will utilise field data acquired from the test sites. The use of
advanced segmentation techniques in conjunction with sub-metre panchromatic and 2–4 metre
multi-spectral data for improved feature information extraction will be explored.
3.8 Integrated Geographical Information System (IGIS)
An integrated geographical information system (IGIS) will form the core of a information management and display system and will be based upon remote sensing and field-derived data. These
data will comprise spatially contiguous information from large areas acquired by remote sensing
12
© 2002 Millpress, Rotterdam, ISBN 90-77017-72-0
and point source data collected by field surveys. The IGIS will provide the linkage between remote
sensing and other types and sources of geo-spatial information and their associated non-spatial attributes that will be held in the database system. The database will consist of physical attribute data
in the form of thematic raster and vector (polygon) information layers, as well as behavioural aspects related to it, that describe the fuel properties, terrain, and infrastructure of an area. The creation of an information database on the fuel properties, terrain, and infrastructure at selected sites
will be used to help define fuel hazards and provide input to fire prevention and control planning
and implementation.
The primary role of the GIS will be the storing, retrieving, processing, geo-referencing, analysing, and displaying of both geo-spatial and non-spatial data and information from the selected area
and to synthesise maps and images depicting the fuel, terrain, and infrastructure aspects of a test
area. The IGIS will be designed to facilitate and ensure ease of data retrieval and processing and to
meet the needs and requirements of both the forest and fire-fighting services. The system will be
generic in construction and adaptable to handling a wide range of applications, as well as being
able to operate on both desk and mobile devices.
Simplicity of use will be the underlying factor in the development of the system. Although advanced processing software will be incorporated into it, a simple and straightforward graphical user
interface will be employed that does not require a high level of technical know-how by users. The
system will be able to search for feature attributes in the database, as well as querying database objects, and to analyse their relationships by means of a variety of functions. The relational database
structures will handle environmental, infrastructure, and terrain spatially based attributes for use in
the analysis of an area. Visualisation tools designed to reduce and display data to more easily comprehend their information content will be included. This will permit a more holistic view and intuitive understanding of the interrelationships between vegetation properties, infrastructure, and terrain of an area to be achieved.
3.9 Data visualisation
Advanced visualisation software will be used to reduce and present complex data to readily interpretable information, based on forest and fire-fighting service requirements. It will make use of
data and image fusion techniques through an interfacing with the IGIS database. This will allow
raster, vector, and point data sets of the sites to be quickly accessed and displayed as two and threedimensional graphic representations, or as charts, text, and images. When thematic data are rendered as three-dimensional perspectives, they are able to provides far more information on site
conditions than is derivable from conventional two-dimensional displays of data. Real-time processing will allow interactive analysis of inter-relationships of site features to be examined in the
context of their surroundings. This will be particularly helpful in fire suppression and control.
Photo-realistic rendering and visualisation of the terrain and environment will be achieved with
the use of stereo imagery. The actual appearance of a ground surface can be more realistically portrayed from the generation of three-dimensional views produce from such satellite imagery and derived DEM's. Visualisation tools and techniques will be used to superimpose (drape) imagery or
thematic data over a three-dimensional rendition of a DEM of an area and to allow the resulting
display to be interactively viewed from any perspective and in real time for a better understanding
of the interrelationships between the various fuel properties.
3.10 IGIS products
A wide range of image-derived products will be generated for use in assessing fuel complexes of an
area and be displayed in various combinations and formats, to better visualise the relationships between two or more entities. The types of products generated will be dependent upon forest and firefighting service needs and requirements. It is expected, though, that the product types will be based
both on standard, as well as customer-defined, specifications, although new ways to present, visualise, and access information will be explored as part of product development. These will include
Forest Fire Research & Wildland Fire Safety, Viegas (ed.)
13
ways of allowing users to down-load fuel, infrastructure, and terrain information at an office or to a
mobile device in the field via the Internet, using terrestrial and satellite telecommunication networks.
Products will be made available in both analogue and digital format and at scales commensurate
with their particular use, and will likely range in scale from 1:2500 to 1:25:000. It is expected that
the products will be used primarily in the derivation of fuel and fire hazard maps for areas for planning and implementing fuel reduction and fire management strategy.
4 CONCLUSIONS
The detailed information on fuel properties, terrain, and infrastructure of an area that will be forthcoming from the techniques, methodologies, and products developed in FIREGUARDis expected
to lead to a more pro-active approach by forest and fire-fighting services in fuel management and
fire pre-suppression activities, such as better prescribed fire planning and implementation, fuel and
fire behaviour modelling, and preventive silviculture policies and practices.
14
© 2002 Millpress, Rotterdam, ISBN 90-77017-72-0

Documentos relacionados

Fire Behaviour Simulation in Mediterranean Areas using FARSITE

Fire Behaviour Simulation in Mediterranean Areas using FARSITE relationships among fuels, topography and weather conditions. The use of FARSITE on areas different from those where the simulator was originally developed requires a local calibration in order to ...

Leia mais