A fuzzy GIS approach to fire risk assessment: a case study of Sydney Olympic Park, Australia

Click here to return to program

Spatial Sciences Conferences 2003

Click here to return to author index

A fuzzy GIS approach to fire risk assessment: a case study of Sydney Olympic Park, Australia Thomas Zeng1), 2), John Hudson1), 2), Susan Kay1) and Edwina Laginestra1) 1)

Sydney Olympic Park Authority 7 Figtree Drive, Sydney Olympic Park, Sydney, NSW 2127, Australia Phone: 61-2-9202-0123; Fax: 61-2-9202-0111; email:[email protected]

2)

Division of Geography, School of Geosciences Faculty of Science, University of Sydney City Road, Sydney, NSW 2006, Australia email: [email protected]

Abstract Bush fire is a natural part of the Australian environment with the potential to devastate rural and urban communities through loss of life, property and infrastructure. While better risk assessment offers some hope in ameliorating the impacts of bushfire, these assessments are challenging due to the complexity of physical and social landscapes, the limited amount of available data describing these landscapes and community’s subjective assessment of the inherent value of natural and built environments. This paper demonstrates how the difficulties of preparing a fire risk assessment can be met through the application of a Fuzzy Geographic Information Systems approach to Parkland areas of Sydney Olympic Park. The factors of fire risk are grouped into four main categories: Physical Factors, Asset Values, Conservation Values and Mitigation. These factors were interpreted into fuzzy terms, based on the thresholds predefined by expert knowledge and past experience of parklands fires. A fire risk model was built with different fuzzy operations and applied to the fuzzificated data layers to map the degree of fire risk across the 420Ha site, resulting in a fire risk gradient map. Historic fire spots were overlaid on the modeled result, showing a high correlation between actual and modeled fire risk. The approach shows that the fuzzy GIS model is a valuable management tool for fire risk mitigation and management planning. Moreover, it’s a cost-effective method for broad scale fire risk assessment with general applicability beyond the limits of the Parklands. Advantages and disadvantages of the approach are discussed, followed by recommendation for further research. Keywords: Fire Risk, Fuzzy Set, GIS and Environmental Hazard Assessment.

1. Introduction Bushfire is one of most serious natural hazards in the Australian environment. The January 2003 fires that spread out of control through suburbs of the nation’s capital, Canberra, were a timely reminder of the devastation caused by bushfires in both rural and urban areas of southeastern Australia. The public outcry following the Canberra

1

Spatial Sciences Conferences 2003

fires highlighted the need for better- informed planning of urban areas, especially where they adjoin fire vulnerable rural lands and national parks. The expectation of better fire risk planning is that it will deliver an informed level of community security while acknowledging the natural role of fire in the Australian environment. Research into bushfire risk includes assessments of fuel load (Salazar and GonzalesCaban, 1987), fire behaviour (McArthur, 1966 and 1967; Buckley, 1992; and Rothermel, 1995), prescribed fire (Mutch, 1994; Paysen, et al., 1998), fire simulation and fire resource allocation (Hof, et al., 2000) with applications of Geographic Information Systems (GIS) in assessing fire risk emerging in recent years (Chuvieco, 1996; Gallun and Cliff, 1996; Miguel- Ayanz, and Schmuck, 1998; Hawkes et al., 1999; Farris, et al., 1999; Rogeau and Ian, 1999; Ross, 1999; Mora and Gilberto, 1999; Bhaskarn, 2000). Despite these effo rts, there remains considerable difficulty in integrating present knowledge of bushfires into reliable fire risk assessments. This problem is confounded in many instances by a lack of detailed information required to make assessments, the cost of acquiring information and the time frames needed to acquire reliable baseline data. In view of these limitations, Fuzzy GIS appears to hold some promise in modeling fire risk in much the same way it has been applied in other fields such as, soil mapping (Zhu, et al., 1996), real estate (Zeng and Zhou, 2000) and coastal hazard mapping (Zeng et al., 2001). This paper describes an application of a Fuzzy GIS approach in assessing fire risk across the parkland area of Sydney Olympic Park. A description of the study area and brief overview of the principles of Fuzzy GIS modeling precedes a detailed account of the model development, application and limitations. The paper concludes with a discussion of the results and recommendations for further research.

2. Study Area The study area is Sydney Olympic Park, the principal venue of the 2000 Olympic Games, located some 14 km west of the Sydney CBD on the southern shores of the Parramatta River (Figure 1). The site occupies an area of around 680 ha of which 430 ha are parkland, one of the largest urban parks in the world. The Parklands include a range of remediated, constructed and remnant landscapes in the form of woodlands, fresh and saltwater wetlands and grasslands. Set within these environments are areas of heritage significance including pre-European aboriginal sites as well as heritage buildings related to prior agricultural, military and industrial uses dating back to the late 1800’s. The Parklands also contain habitats of threatened and locally rare faunal and floral species, notably the Green and Gold Bell Frog and the saltmarsh plant Wilsonia backhousei. It is a complex and sensitive area with the potential for bushfires to cause significant and lasting damage.

3. Context and Risk Definition The production and maintenance of a Fire Management Plan is a requirement of the Parklands Plan of Management (2003). The plan identifies the level of bushfire risk across the Parklands and establishes strategies to avoid unreasonable risk to life, property and the environment. A component of this plan is the production of a Fire

2

Spatial Sciences Conferences 2003

Risk Model which will identify areas of high fire risk within the Parklands to be used as an aid in creating strategies to minimise the potential for ignition and reduce the severity of any fires that may occur. Risk is prioritised on a sliding scale from high to low. These classifications are based on characteristics of the Parklands environment that are likely to influence the chance of a fire igniting, spreading and causing damage to the community, environmental or heritage assets. The concept of fire risk embraces three aspects: •

Hazard: The potential severity of the fire that is influenced primarily by the vegetation type, slope and weather conditions. • Threat: How close the fire hazard is to an asset. The closer the asset is to a fire hazard the more likely it is to be damaged or destroyed by fire. • Vulnerability: The capacity of the asset to withstand or recover from a bush fire incident. The factors used to determine risk were: a) Vegetation Type The moisture content of fuels affects the ignition potential and spreading characteristics of a fire. Moisture affects both these parameters by increasing the specific heat and thermal conductivity of the fuel so that more heat must be adsorbed in order for the fuel to reach ignition temperature. Once ignition has occurred, moisture will act as heat sink and interfere with the ignition of surrounding vegetation. As plant moisture decreases, fire intensity and rates of spread increase. Grasslands – The native grasses within the Parklands generally have low moisture content. It is likely that they are cured between 60%-80%. Limited maintenance has meant that native grasses retain much of their dry biomass from one season to another, and do not have the opportunity to regenerate. Native grasslands are therefore the most vulnerable to ignition. Woodlands – Native woodlands within the Parklands have generally been excluded from burning and are likely to have a relatively high level of combustible fuels. This vegetation type is likely to be a major fire risk. Casuarina Forest – Casuarina glauca forests within the Parklands have a higher moisture content and lower levels of combustible dry fuels then the sclerophyllous woodlands. Casuarina forests have therefore been classified as a moderate fire risk. Mixed Plantings – Within the Parklands newly planted areas of vegetation typical of the Cumberland Plains region of Sydney exist. Presently these areas pose only a moderate fire risk, due to the small fuel loads. In the future, as the plantings mature, it is likely that this classification may require revision.

3

Spatial Sciences Conferences 2003

Recreational Turf – Recreational turf within the Parklands is well watered and kept short for use as active and passive recreationa l pursuits. This vegetation type will pose only a minor fire risk. Wetlands – Due to the inherent dampness of the wetland environment, these areas present only a negligible fire risk. Figure 2 shows the distribution of main vegetation types within the parklands. b) Slope With all other conditions being equal, fires travel faster up slopes than on flat ground. The general rule is that the speed of the fire doubles with every 10 degrees increase in slope. Therefore, the steeper the slope the greater the fire risk. c) Aspect In the southern hemisphere, hillsides that face towards the north or west will receive more direct and longer hours of exposure to the sun, This means that fuel moisture content will be lower than southerly or easterly aspects under the same general conditions. Areas of the Parklands with north or west aspect are classified as posing a greater fire risk. d) Wind Apart from the drying effects of wind on fuel moisture content, wind speed affects the rate of spread of fires by drying the fuel ahead of the fire and allowing it to burn at higher temperature and allowing the fire front to move more quickly. A wind rose diagram based on local weather station data was used to define wind speed and directions. e) Assets The Parklands contains a variety of built and natural assets that require protection from fire. High-risk assets were determined by identifying the threat (i.e. how close the fire hazard is to an asset) and the vulnerability of the asset (i.e. the capacity of the asset to withstand or recover from bushfire). The threat to an asset was determined to be extreme if an asset was positioned within 100m from a fire hazard. The risk to an asset decreases as the distance between the fire hazard and the asset increases. The vulnerability of an asset was determined by its value, not only in monetary terms but also in heritage and environmental values. Generally assets that have only monetary value are replaceable. However those that are regionally significant due to their high environmental or heritage value will not be so easily replaced and are therefore pose a greater fire risk.

4

Spatial Sciences Conferences 2003

f) Neighbours The proximity to neighbouring residential, commercial or industrial facilities was incorporated into the Fire Risk Model. Neighbouring properties adjoining the Parklands may be a source of ignition, and present a risk to life and property. a) Fire Fighting Infrastructure In a fire situation, access to hydrants and other water sources can reduce the risk of the fire spreading and causing further risk to life, property or the environment. Areas isolated from fire fighting infrastructure therefore pose a greater risk.

4. Fuzzy GIS Model The term fuzzy set was coined by Zadeh (1965) as a generalised form of set theory. Unlike traditional Boolean logic which defines whether or not an element belongs to a crisp set (1 or 0), a fuzzy set defines a degree of belonging through a membership function. In effect, fuzzy set theory deals with sources of uncertainty that are vague or non-statistical in nature such as operational definitions based on “rules of thumb”, estimations of natural processes, classification of environment types and the like. Unlike conventional fire models that aim at modeling fire behaviour and fire frequency, the Fuzzy GIS approach evaluates the spatial distribution of contributing factors of fire and the risks they would cause with consideration on approximation of asset locations. Basically, the Fuzzy GIS approach is to apply different fuzzy membership functions to data layers and the proximity to a type of object, such as heritage buildings. This process is also called ‘fuzzification’. Fuzzy operations are then applied, using either map algebra or user-defined mathematical algorithm, along with weighting functions, to combine the fuzzified data to produce fire risk maps Thresholds for the assignment of fuzzy membership functions for the different risk factors are shown in Table 1. Note that for the thresholds set out in Table 1, “1” represents the highest risk while “0” represents the lowest risk. Intermediate risk determined in consultation with Parklands management. Using an analysis of the factors listed in Table 1, a Fuzzy GIS model (Figure 3) is built that consists of two sub- models, Natural Hazard and Asset Value s, respectively (Figure 3a and 3b). The modeling is implemented in Grid module of ArcInfo GIS software, with Arc Macro Language (AML).

5

Spatial Sciences Conferences 2003

Table 1. Sydney Olympic Parklands fire risk factors – fuzzy membership functions

Indicator No 1. Conservation Value

Data Themes Threatened species

Thresholds x i < 30 m, xi = 1 30 > x i >300 m , x i = F(x i) x i > 300 m, xi = 0

Water body and distance to threatened x i < 30 m, xi = 1 species habitats 30 > x i >250 m, x i = F(x i) x i > 250m, xi = 0 2. Physical Constraints

3. Mitigation

Vegetation Type: • Native grass • Woodland • Casuarina • Mixed planting • Turf • Wetland • Not coded Slope

x i < 5o , xi = 0 5o > x i > 25o , x i = F(x i) x i > 25o , xi = 1

Aspect

Based on Diagram

Community: 15 -150m

x i < 15 m, xi = 1 15 > x i >150 m , x i = F(x i) x i > 150 m, xi = 0

Fire Hydrants 30-300m

x i < 15 m, xi = 0 15 > x i >250 m , x i = F(x i) x i > 250 m, xi = 1

Water resources 30 –250m 4. Assets

1 0.75 0.6 0.5 0.3 0.1 0.02

Culture Heritage Buildings

Distance to Cultural Heritage Buildings: 30 –150m

Wind

Rose

x i < 30 m, xi = 1 30 > x i >200 m , x i = F(x i) x i > 200 m, xi = 0 Exceptional High = 1 High = 0.75 Medium = 0.5 Contributory = 0.25 (As above)

Vegetation Regeneration Costs

6

Spatial Sciences Conferences 2003

The data is initially in ArcInfo Coverage format, then converted into grid. The modeling was undertaken in Grid Module, an extension of ArcInfo. The proximity to physical entities, such as frog ponds or buildings, is calculated using the ‘Costdistance’ function of Grid. The resultant data layer is fuzzificated with a fuzzy membership function. The fuzzification processing is illustrated in Figure 4. Fuzzy operations were used to combine the fuzzificated data.

5. Results The results of Nature Hazard and Asset Value submodels are shown in Figure 5a and 5b respectively. The outputs of each submodel are combined using a ‘plus’ (aggregation) fuzzy operation to produce the final map (Figure 6). High fire risk areas are shown in upper middle part of the study area where there is a coincidence of high fuel load woodlands, cultural heritage buildings and threatened species (Green and Gold Belt Frog) habitats. The result shown in Figure 6 can be used to reclassify the site according to the management needs, for example reclassifying the parklands into the highest 5%, 10%, 15%, and 20% of the highest fire risk areas, equating to areas of 17.2, 34.9, 54.7 and 70.7 ha respectively (Figure 7). These figures can then be used in resource planning for mitigation of the fire risk, such as, thinning/residue disposal in the urban interface (Kalabokidis and Omi, 1998) and use of prescribed fire to reduce fire risk (Martin, et al., 1989; Mutch, 1994; Paysen, et al.,1998). A sensitivity analysis of the weightings used for model inputs was run with examples of the modeled fire risk shown in Figure 8. To evaluate the reliability of the modeled results, historical fire spots were digitised into a coverage and overlaid on top of the fire risk map. All spots are located in the high fire risk area (Figure 9), providing some confidence in the modeled results.

5. Discussion There are many GIS applications in fire risk assessment reported in the literature however they either use Boolean logic overlay or demand extensive and detailed input data, making them difficult to apply where data are either limited or absent or difficult to interpret when the outputs include multiple maps. (Hawkes et al., 1999; Chuvieco, 1996; Farris, et al.,1999; Finney, 1998). The Fuzzy GIS model described here takes a different approach, compensating for data gaps by incorporating, or codifying, expert knowledge. In this way an assessment of risk can be developed using the available relevant datasets (eg. terrain, generalised vegetation classifications etc.) in combination with expert knowledge of the factors likely to contribute to bushfire risk. The approach is inherently inclusive, giving the field experts access to a technology which allows them to explore the implications of their knowledge of fire risk, and the uncertainty of this knowledge, while retaining the ability to incorporate new information as it comes to hand. The model developed for Sydney Olympic Park highlighted some limitations of the approach. Firstly, there is a clear need for careful consideration of the fuzzy

7

Spatial Sciences Conferences 2003

membership function and fuzzy operator for each input data layer. A danger is that arbitrary assignment of membership functions and operators may invalidate model outputs where the bases of the functions and operators have not been informed with the best available information or expert knowledge. Uncritical acceptance of model runs must be guarded against. Secondly, there are many detailed aspects of fire modeling which have not been considered in this example such as fire frequency, fire behaviour (including fire intensity, McArthur Index, rate of spread for fuel quantities) due to a basic lack of this type of information for the site. Further modeling of fire risk should take advantage of this information as it becomes available. Further development of the model would benefit from: •

• • • •

Fined tuning of risk factors (long-term wind statistics, soil and vegetation moisture contents in response to ENSO-related climatic cycles, building type and resistance to fire, fire risk activities on adjoining properties and associated with events in the parklands) Customisation of the model for the different landscapes, infrastructure and activities encountered in each parklands precinct Incorporation of CSIRO slope models Comparative study with FIREMODEL Applying other fuzzy GIS methods (eg. Fuzzy Similarity Matrix, Fuzzy Rule Based and Fuzzy Screening

6. Summary The Fuzzy GIS approach set out in this paper demonstrates a cost-effective and inclusive way for rapidly assessing fire risk in a broad sense, particularly in the face of limited data. The approach is necessarily inclusive wherein expert knowledge is sought, codified and merged with available data to achieve the best possible assessment of fire hazard. Moreover, the modeling is reproducible, dynamic and transparent in that successive runs can be completed based new information as it comes to hand and in discussion with stakeholders as the importance of selected hazard risk factors is tested. As the modeling process matures, the reliability of assumptions can be tested in a both a collaborative and systematic fashion with the GIS acting as an enabling technology for the visualization of fire assessment and planning decisions.

8

Spatial Sciences Conferences 2003

Figure 1 Location of Sydney Olympic Park study area on the Parramatta River in western Sydney

9

Spatial Sciences Conferences 2003

Parramatta R.

Ermington

Parramatta R. Homebush Bay

Silverwater

Homebush Bay

Newington

Olympic Venues

Lidcombe

Figure 2 Vegetation types in the parklands area of Sydney Olympic Park. Note that the vegetation mapping is for the Parklands precincts only and does not does not extend into the urban precinct hosting the Olympic sporting facilities nor the surrounding urban areas of Auburn Local Government Area.

10

Spatial Sciences Conferences 2003

cont

bldg

Arc: TOPOGRID dem 5

dem Grid: SLP = slope t ( dem,degree)

bldg

Arc: POLYGRId

veg Arc: additem Table: cal Arc: buffer

Arc: POLYGRId

bldg

bldg

FZM

Grid: ASP = aspect (dem) Grid: Costdistance

SLP

ASP

PXB

FZ SLP

FZ ASP

FZ BDG

PXB

wat

Arc: POLYGRId

veg

FUL

FZ CMB

Cmty

FZ FUL

Legend

NAT HZD

Sub-model-1: Nature Hazard

hydr

Grid: Costdistance

coverage

heri bldg

Rare veg

heri bldg

Rar veg

grid

bldg

veg

Arc: POLYGRId

hydr

wat

CMV

AST

Grid: Costdistance Arc: buffer

FRI RES

PRX

prx heri

prx Rar

AST

Grid: Costdistance

PRX

Sub-model -2: Asset Value

PRX

PRX

AST VAL

Legend coverage

grid

Figure 3. Sub Model 1 Nature Hazard (3a) and Sub Model 2 Asset Value (3b). Model shows input layers (coverage) and their conversion to fuzzified layers for combination at model output. Nature Hazard Model acronyms: bldg=building; cont=contour; dem=digital elevation model; veg=vegetation; ASP=Aspect; FUL=Fuel; PXB=Proximity to building; SLP=Slope; Fuzzified layers shown as FZASP, FZBDG, FZCMB, FZFUL, FZSLP. Resultant Natural Hazard layer = NATHZD. Asset Value Model acronyms: bldg=building; cmty=community; heribldg=heritage building; hydr=hydrant; prxheri=proximity to heritage building; prx rar=proximity to rare and endangered species; rareveg=rare vegetation; wat=water body; AST=Asset; CMV=; FIRRES=Fire fighting resources; PRX=Proximity. Resultant Asset Va lue layer = ASTVAL.

11

Spatial Sciences Conferences 2003

Parkland:

Distance to

Fuzzy set:

yes = 1,

parkland:

range of [0 - 1]

no = 0

0 - 1450 m

(bad - good)

Example of FZM: S-function µ(x)=

1 2

−

1 2

sin [ π

 a + b ] x −  2 b −a  

(a < x < b )

Figure 4 Illustration of fuzzification processing for Parklands proximity.

12

Spatial Sciences Conferences 2003

5a. Nature Hazard Model

5b. Asset Value Model

Fgure 5 a / 5b Final result map by combining Nature Hazard submodel and Asset Value submodel with ‘Plus’ fuzzy operator.

13

Spatial Sciences Conferences 2003

Figure 6 Distribution of highest fire risk areas.

14

Spatial Sciences Conferences 2003

0

500

1,000

2,000

Meters

0

500

0

500

1,000

2,000

Meters

Highest 10% area

Highest 5% area

0

1,000

2,000

Highest 15% area

500

1,000

2,000

Meters

Meters

Highest 20% area

Figure 7 Summary of highest fire risk areas.

15

Spatial Sciences Conferences 2003

a)

c)

b)

d)

Figure 8 Sensitivity analysis on weightings for fire risk modeling.

16

Spatial Sciences Conferences 2003

Figure 9 Historical fire spots plotted on the high fire risk areas.

17

Spatial Sciences Conferences 2003

References Andrews, P.L.; Williams, J.T. 1998, Fire potential evaluation in support of prescribed fire risk assessment. In: Fire in ecosystem management: shifting the paradigm from suppression to prescription. Tall Timbers Fire Ecology Conference Proceedings, No. 20. Tall Timbers Research Station, Tallahassee, FL: 64-68. Bhaskaran, S. 2000, Integrating remote sensing and GIS for fire hazard categorization, Proceedings of the IEEE/IGARSS conference, 24-28, July, 2000, Hawaii, Honolulu, UAS, 2102-2104. Bhaskaran, S., Forster, B., Paramesh, N., Neal,T., 2001, Integrating remote sensing and GIS for fire hazard categorization, Proceedings of the IGARSS symposium, 9-13, July, Sydney, Australia. Buckley, A.J. 1992. Fire behavior and fuel reduction burning: Bemm River wildfire, October 1988. Aust. For. 55: 135-147. Cheney, N.P. 1994. The effectiveness of fuel reduction burning for fire management. In: Fire and biodiversity: the effects and effectiveness of fire management. Proceedings of the Conference held Oct. 8-9, 1994, Footscray, Melbourne, Australia. Chuvieco, E., 1996, Mapping the spatial distribution of forest fire danger using GIS. Geographical Information Systems 10(3): 333. Cleaves, D.A.; Martinez, J.; Haines, T.K. 2000. Influences on prescribed burning activity and costs in the National Forest System. Gen. Tech. Rep. SRS-37. Asheville, NC: USDA, Forest Service, Southern Research Station. 34p. Deville and Harding (1997) Applying the Precautionaty Principle, Sydney, The Federation Press, 79p. Farris, C. A., Pezeshki, C. and Neuenschwander, L. F., 1999, A comparison of fire probability maps derived from GIS modeling and direct simulation techniques. In Proceedings from the Joint Fire Science Conference and Workshop, vol. 2. Moscow, Idaho: University of Idaho. pp. 130-137. Finney, M.A. 1999. Some effects of fuel treatment patterns on fire growth: a simulation analysis of alternative spatial arrangements. 13 p. http://www.r5.fs.fed.us/hlglg/ProjectFiles/fire/fire1/Fire1.html . Gallun, H. and Cliff, B, 1996, Monitoring fire-affected wildlands in the Mediterranean region by applying a remote sensing and GIS approach. International Archives of Photogrammetry and Remote Sensing 31(B7): 240245.

18

Spatial Sciences Conferences 2003

Hawkes, Brad, and Judi Beck. 1999. A wildfire threat rating system. Fire Management Notes 55(2): 25-30. Helms, J.A. 1979. Positive effects of prescribed burning on wildfire intensities. Fire Mgmt. Notes, 40: 10-13. Hesseln, H.; Rideout, D.B.; Omi, P.N. 1998. Using catastrophe theory to model wildfire behavior and control. Can. J. For. Res. 28: 852-862. Hof, J.; Omi, P.N.; Bevers, M.; Laven, R.D. 2000. A timing-oriented approach to spatial allocation of fire management effort. Forest Science 46(3): 442-451. Kalabokidis, K.D.; Omi, P.N. 1998. Reduction of fire hazard through thinning/residue disposal in the urban interface. Int. J. Wildland Fire 8(1): 29-35. Kallander, H.R. 1969. Controlled burning on the Fort Apache Indian Reservation, Arizona. Tall Timbers Fire Ecol. Conf. Proc. 9: 241-249. Mariel, A. and Marielle J., 1996, Wildland fire risk mapping using a geographic information system and including satellite data: Example of Les Maures forest, south east of France. EARSeL 4(4): 49-56. Martin, R.E.; Kaufmann, J.B.; Landsberg, J.D. 1989. Use of prescribed fire to reduce wildfire potential. In: Proceedings of the symposium on fire and watershed management. GTR-PSW-109. Albany, CA:USDA Forest Service, Pacif ic Southwest Research Station: 17-22. McArthur A. G., 1967, Fire Behavior in Eucalypt Forests. Forestry and Timber Bureau, Canberra, Leaet 107. McArthur, A.G., 1966, Weather and Grassland Fire Behavior, Forestry and Timber Bureau, Canberra, Leaet 103. Miguel-Ayanz, A. A. and Schmuck, G., 1998, Forest fire monitoring and damage assessment in Mediterranean landscapes through the integration of remote sensing and GIS techniques. International Archives of Photogrammetry and Remote Sensing 32(7): 170. Miller, C.; Urban, D.L. 2000. Modeling the effects of fire management alternatives on Sierra Nevada mixed-conifer forests. Ecol. Appl. 10(1): 85-94. Mora, F. and Gilberto H-C., 1999, Modeling and mapping wildfire potential in Mexico based on vegetation and drought conditions using remote sensing and GIS technology. In Proceedings from the Joint Fire Science Conference and Workshop, vol. 2. Moscow, Idaho: University of Idaho. pp. 25-38. Mutch, R.W. 1994. Fighting fire with prescribed fire: a return to ecosystem health. J. For., 92(11): 31-33.

19

Spatial Sciences Conferences 2003

Paysen, T.E.; Narog, M.G; Cohen, J.D. 1998. The science of prescribed fire: to enable a different kind of control. In: Fire in ecosystem management: shifting the paradigm from suppression to prescription. Tall Timbers Fire Ecology Conference Proceedings, No. 20. Tall Timbers Research Station, Tallahassee, FL: 31-36 Parklands Plan of Management, 2002, Plan of Management for the Parklands at Sydney Olympic Park. Sydney Olympic Park Authority. http://www.sydneyolympicpark.nsw.gov.au/index.asp?DocID=2788&CatID=67 Rogeau, M-P. and Ian P., 1999, Estimating risks of burning in mountain landscapes: A GIS application. In Proceedings from the Joint Fire Science Conference and Workshop, vol. 2. Moscow, Idaho: University of Idaho. pp. 228-232 Ross, C. 1999, GIS blazes a trail in fire management-In Nevada's Lake Tahoe Basin, satellite technology is helping officials control natural fires and effectively use prescribed burns. American City & County 114(2): 34-39. Rothermel, R.C. 1995. Characterizing severe fire behavior. In: Symposium Proceedings: Fire and Wilderness and Park Management. Gen Tech Rep. INTGTR-320, Ogden, UT: USDA Forest Service, Intermountain Forest and Range Exp. Station: 262-264. Salazar, L.A., and Gonzales-Caban, A., 1987, Spatial relationships of a wildfire, fuelbreaks, and a recently burned area. West. J. Appl. For. 2: 55-58. Zadeh, L. A., 1965, Fuzzy sets. Information and Control, 8, 338-353. Zeng, T. Q. and Zhou, Q., Optimal spatial decision making using GIS: a prototype of a real estate geographical information system (REGIS). Int. J. Geographical Information Science, 15(4), 307-321 Zeng, T. Q., Zhou, Q. Cowell, P. J., and Huang, H. J., 2001, Coastal GIS: functionalities and applications. Journal of Geospatial Engineering, 3(2), 109126.

20