Fire Safety Journal 72 (2015) 25–32

Contents lists available at ScienceDirect

Fire Safety Journal

http://d0379-71

n CorrE-m

rosencra

journal homepage: www.elsevier.com/locate/firesaf

GIS method for characterizing fire flow capacity

Martin M. Kaufman a,n, Troy Rosencrants b

a Department of Earth and Resource Science, 516 Murchie, University of Michigan-Flint, Flint, MI 48502-1950, USAb GIS Center, 504 Murchie, University of Michigan-Flint, Flint, MI 48502-1950, USA

a r t i c l e i n f o

Article history:Received 22 September 2014Received in revised form20 January 2015Accepted 1 February 2015

Keywords:Geographic information systemsHistorical fire incidencesFire flow analysisHydrant availabilityFire infrastructure assessment

x.doi.org/10.1016/j.firesaf.2015.02.00112/& 2015 Elsevier Ltd. All rights reserved.

esponding author.ail addresses: martyk@umflint.edu (M.M. Kau@umflint.edu (T. Rosencrants).

a b s t r a c t

There are numerous methods currently used to calculate required water flow rates for sprinklered andnon-sprinklered buildings. The aim of this study is provide a flexible automated procedure for in-dentifying locations lacking adequate fire flow. To accomplish this objective, this research uses a GISprocedure to determine the spatial relationships between fire hydrants and historical fire incidences, andintegrates the recommended hydrant spacing and building type specifications from the International FireCode. This method was tested in two communities in eastcentral Michigan, USA. The results indicate anability to define clusters of fires, determine the availability of hydrants, and assess the suitability of theavailable fire flow, including areas of potential extra capacity. Using these same data, additional GISanalyses can optimize hydrant location, ascertain the frequencies of different categories of fires, andidentify the patterns of building types prone to fires.

& 2015 Elsevier Ltd. All rights reserved.

1. Introduction

The water requirements for firefighting include the rate of flow,the residual pressure required at that flow, the flow duration, andthe total quantity of water required. Specifically, fire flow is de-fined as “the rate of water flow, at a residual pressure of 20 psi andfor a specified duration that is necessary to control a major fire in aspecific structure” [2,30]. Determining the correct fire flow is ne-cessary to achieve successful fire control efforts. If the fire flow isover-calculated, there could be a negative impact on the waterdistribution system; under-calculation of the fire flow may resultin the loss of the building or lives [4,16].

Several recognized methods are currently used to determinethe required fire flow for a building [5,10]. These methods can bebroken down into two general categories: those for buildingplanning (building and/or fire code requirements), and those foron-scene fire service use. The building planning methods accountfor a range of variables in determining fire flow, such as buildingconstruction, occupancy, and fire size. This method allows forbuilding and community planners to assess current or futurebuildings against the existing or planned water supply and adjustaccordingly. The on-scene fire flow calculation methods consist ofone equation with one variable used to determine the fire flow,which allows the firefighters on scene to assess whether they need

fman),

more hose lines or apparatus to fight the fire [5].

1.1. GIS and fire science

Geographic Information Systems (GIS) have been broadly ap-plied to natural disasters in the areas of risk reduction and riskresponse [8]. With the increasing sophistication of spatial analytictools, GIS has been combined with spatial statistics to investigatefire incidence [19]. For example, Asgary et al. [3] analyzed tem-poral and spatial dimensions of structural fires in Toronto, Canada.Using a classification scheme of fire types, their study found clearspatial and temporal patterns of fire for the different types of fires.In another study using GIS, Schacterle et al. [26] were able to find asignificant association between vacant housing and fire risk. Otherresearchers, such as Corcoran et al. [9] have been able to findlinkages between socioeconomic characteristics with fire risk andspecific types of incidents in space.

On the infrastructure side of fire analysis, GIS has been used tostudy fire station location and response. An analysis of fire stationsin California, USA demonstrated that short and long term costs canbe saved by optimizing station location with a systematic ap-proach that includes GIS [21]. Other approaches to fire stationsiting have been developed from an operations research perspec-tive, and recent investigations have incorporated GIS into theirmodels [1,11]. One limitation of existing GIS-related research onurban fire infrastructure has been the relative lack of studies ad-dressing fire hydrant location and the implications for providingadequate fire flow. Here, there is the opportunity to combine the

M.M. Kaufman, T. Rosencrants / Fire Safety Journal 72 (2015) 25–3226

temporal characteristics of fire incidence (fire history) with theexisting spatial patterns of hydrants to ascertain whether adequatefire flow is available. Thus, an a priori risk assessment can be madeby fire departments to identify locations within their coveragearea that may lack adequate fire flow. This research is designed toprovide fire personnel the ability to make a risk assessment aboutfire flow capacity at specific locations, with the added ability toidentify fire clusters for further investigation from the maps pro-duced. Fire flow capacities available for each land parcel within thestudy areas are also investigated to help identify locations wherepotential fire flow deficiencies may occur when no historical firedata exists.

The research hypothesis asserts that fire flow capacity is notsufficient in all cases within a fire department's coverage area. Thishypothesis is tested in two communities in east-central Michigan,USA with a flexible automated GIS-based procedure for identifyinglocations lacking adequate fire flow. Through standard GIS op-erations, the spatial relationships between fire hydrants and his-torical fire incidences are constructed and integrated with therecommended hydrant spacing and building type specificationsfrom the International Fire Code [17]. This method is thus com-patible with the building planning methods used to determine fireflow requirements.

2. Methods

A multiple-method case-study approach is employed to in-vestigate fire flow capability in the study areas. In a case study, aprimary objective is to assess the ability to transfer theory, prin-ciples, or techniques existing at the geographic scale of the case toother areas or “cases” [31]. The methods used to characterize fireflow capacity in this study include map analysis, GIS modeling, andfield observation. Being largely data-driven, the use of a GIS pro-cedure mandates a specific sequence of tasks [15–25]. With re-spect to this research, the flow of work included study area se-lection, data collection, GIS map layer preparation and projection,development of the GIS procedure (model) for integrating thenecessary data and arriving at a solution, and field verification of

Fig. 1. Stud

the results.

2.1. Study area selection

The study area for this research encompasses the two com-munities of Grand Blanc and Flint Township located within Gen-esee County, Michigan (Fig. 1).

These communities were selected based upon the availability ofhydrant data in digital format and the desire of the fire chiefs toparticipate in the study. Both communities contain a variety ofland uses ranging from highly urbanized to semi-rural, and thuscontain different patterns of hydrant spacing and density.

2.2. Data collection

GIS data consist of map layers containing spatial and attributedata. In vector-based GIS systems, spatial data are the points, lines,or polygon features constituting the visible map; attribute data arelinked to each spatial feature to provide descriptive and analyticalinformation. For instance a spatial data point feature might re-present a fire hydrant, and the attribute data associated with thathydrant might include its part number, flow rate, and operationalstatus. To perform the tasks of this research, the following datawere collected for each community: fire hydrant locations andtheir flow rates, 3-year fire history from 2010 to 2013, land parcels,digital street maps, the International Fire Code standards for fireflow (Table 1), and hydrant requirements based on fire flow(Table 2).

Referring to Table 1, the columns under the heading “Fire flowCalculation Area” represent a range of building types. The buildingtypes are derived from the International Building Code and arebased on their construction materials, occupancy, and height; withsubtypes denoted by the letters A and B to indicate relative fireresistance [18]. For the sake of brevity, only the construction ma-terials will be presented here. Types I and II buildings are con-structed of noncombustible materials. Type III has exterior wallsmade of noncombustible materials and the interior building ele-ments are of any material permitted by the code. Type IV consistsof heavy timber, with the exterior walls having noncombustible

y area.

Table 1Minimum required fire flow and flow duration for buildings [17].

Fire-flow calculation area (square feet)

Type 1A and 1B Type IIA and IIIA Type IV and V-A Type IIB and IIIB Type V-B Fire-flow (gal/min) Flow duration (h)

0–22,700 0–12,700 0–8200 0–5900 0–3600 1500 222,701–30,200 12,701–17,000 8201–10,900 5901–7900 3601–4,800 175030,201–38,700 17,001–21,800 10,901–12,900 7901–9800 4801–6200 200038,701–48,300 21,801–24,200 12,901–17,400 9801–12,600 6201–7700 225048,301–59,000 24,201–33,200 17,401–21,300 12,601–15,400 7701–9400 250059,001–70,900 33,201–39,700 21,301–25,500 15,401–18,400 9401–11,300 275070,901–83,700 39,701–47,100 25,501–30,100 18,401–21,800 11,301–13,400 3000 383,701–97,700 47,101–54,900 30,101–35,200 21,801–25,900 13,401–15,600 325097,701–112,700 54,901–63,400 35,201–40,600 25,901–29,300 15,601–18,000 3500112,701–128,700 63,401–72,400 40,601–46,400 29,301–33,500 18,001–20,600 3750128,701–145,900 72,401–82,100 46,401–52,500 33,501–37,900 20,601–23,300 4000 4145,901–164,200 82,101–92,400 52,501–59,100 37,901–42,700 23,301–26,300 4250164,201–183,400 92,401–103,100 59,101–66,000 42,701–47,700 26,301–29,300 4500183,401–203,700 103,101–114,600 66,001–73,300 47,701–53,000 29,301–32,600 4750203,701–225,200 114,601–126,700 73,301–81,100 53,001–58,600 32,601–36,000 5000225,201–247,700 126,701–139,400 81,101–89,200 58,601–65,400 36,001–39,600 5250247,701–271,200 139,401–152,600 89,201–97,700 65,401–70,600 39,601–43,400 5500271,201–295,900 152,601–166,500 97,701–106,500 70,601–77,000 43,401–47,400 5750295,901–Greater 166,501–Greater 106,501–115,800 77,001–83,700 47,401–51,500 6000– – 115,801–125,500 83,701–90,600 51,501–55,700 6250– – 125,501–135,500 90,601–97,900 55,701–60,200 6500– – 135,501–145,800 97,901–106,800 60,201–64,800 6750– – 145,801–156,700 106,801–113,200 64,801–69,600 7000– – 156,701–167,900 113,201–121,300 69,601–74,600 7250– – 167,901–179,400 121,301–129,600 74,601–79,800 7500– – 179,401–191,400 129,601–138,300 79,801–85,100 7750– – 191,401–Greater 138,301–Greater 85,101–Greater 8000

Table 2Number and distribution of fire hydrants [17].

Fire-flow re-quirement (gpm)

Minimumnumber ofhydrants

Average spa-cing betweenhydrants (feet)

Maximum distancefrom any point onstreet or road fron-tage to a hydrant

1750 or less 1 500 2502000–2250 2 450 2252500 3 450 2253000 3 400 2253500–4000 4 350 2104500–5000 5 300 1805500 6 300 1806000 6 250 1506500–7000 7 250 1507500 or more 8 or more 200 120

M.M. Kaufman, T. Rosencrants / Fire Safety Journal 72 (2015) 25–32 27

materials and the interior building elements are of solid or lami-nated wood without concealed spaces. Type V has exterior andinterior walls consisting of any materials permitted by the code,which include wood-frame single-family residences.

Table 2 shows the minimum number of hydrants and theiraverage spacing for each range of minimum fire flow require-ments. These values are parameters in the GIS model used to de-termine adequate fire flow.

2.3. GIS map layer preparation and projection

For a specified geographic extent (study area), all of the GISlayers must have internal consistency and the ability to overlay oneach other with an acceptable amount of spatial error Internalconsistency refers to the correct correspondence between thespatial features and their attribute data. For example, each flowrate assigned to a hydrant should represent its actual flow Theability to overlay layers is dependent upon several factors, in-cluding: cartographic errors in map preparation, incorrect

projection methods, and data compiled from multiple sourceshaving different spatial resolutions [29]. To minimize attributeerror, statistical range checks were performed to help detect dataentry errors in the hydrant flow data, as these values were derivedfrom paper maps. After their entry, the hydrant flow data weredisplayed and compared to the paper source maps for accuracy.Cartographic errors were minimized through consistent mapprojection, which are attempts to portray the surface of the earthor a portion of the earth on a flat surface. Some distortions ofconformality, distance, direction, scale, and area always resultfrom this process. Conformality refers to the preservation of scalein any direction, with the lines of latitude and longitude inter-secting at right angles. Some projections minimize distortions insome of these properties at the expense of maximizing errors inothers, whereas other projections attempt to only moderatelydistort all of these properties [24]. For this application, a StatePlane projection was applied for all map layers to facilitate spatialoverlay operations [27]. Fire history data were geocoded and ad-ded as a point layer to the GIS application. Geocoding is the pro-cess of transforming a description of a location such as an addressto a location on the earth's surface [12]. In this application, theaddress of each fire in the 3-year history file was matched to streetreference files containing geographic centerline coordinates, streetnumbers, street names and postal codes. The 3-year histories werecompiled from the fire incident reports input to the National FireIncident Reporting System [23]. Data available from the fire in-cident reports also enables fire flow characterizations for specificcategories of fire, such as those related to hazardous materials orvehicles. In this application, the only types of fires investigated arebuilding fires. Land parcels, streets, and hydrants were obtained indigital format and comprised the remaining three layers of data inthe GIS. Boundaries were inspected for consistency and alignmentafter each layer was projected and overlaid on the others.

Fig. 2. Model to determine fire flow capacity.

M.M. Kaufman, T. Rosencrants / Fire Safety Journal 72 (2015) 25–3228

2.4. Model development

Models are descriptions of systems, where systems consist ofdiscrete and interacting components with inputs and outputs, asequence of processes, and feedback mechanisms [6]. A GIS has athese elements, with feedback pertaining to the human actionstaken in response to the solution presented. Given this definition,GIS models specify the operations performed on each data layer toproduce the desired output [20]. Fig. 2 depicts the sequence ofoperations required to characterize fire flow capacity.

Starting on the left side of the model, the fire history locationsand hydrants are input into a spatial join. A spatial join creates ajoined table in which fields from two layers' attribute tables aremerged based on the relative locations of the features in thoselayers. In a spatial join, the option is available to have statisticsperformed on the fields, along with the ability to delete any un-necessary fields. For this case, the flow of each hydrant within thesearch radius is summed. The search radius can be determinedfrom the building type in Table 1; however, building types are notexplicitly input on the Fire Incident Report [23], so the largestaverage spacing between hydrants is used (500 ft or 152.4 m; Ta-ble 2). Employing the maximum radius of 152.4 m is the mostrobust way to test the model, since the use of shorter distancesmight miss hydrants that should be included and yield a falsepositive for deficient fire flow.

After the spatial join, the minimum required fire flow is cal-culated in row 1 on the right side of the model based on the fireflow requirement shown in column 1 of Table 2. If the buildingtype or area is not available, the lowest amount of required fireflow (1500 gal/min) is input from Table 2.Once again, this is themost robust way to test the model, since any deficiency discoveredusing the least amount of required fire flow indicates deficienciesare more likely to exist for all other building types and squarefootages. In row 2, a threshold for potential excess fire flow ca-pacity is calculated by doubling the minimum required fire flowfor any building type, and the authors recognize this is somewhatarbitrary. Since there is no precedent for defining excess fire flow,we selected double the lowest flow rate that would apply to largeareas of residential structures. It is, however, quite possible that adifferent threshold value may be more appropriate, with the actualvalue accounting for local conditions. This variance in identifying asuitable threshold may occur because system water pressure de-creases as more hydrants are used, specific water system config-urations exist, and the presence of vacant properties may reducethe fire flow requirements. The GIS model employed here has thecapability to change the threshold value easily.

In row 3, the summed rated fire flow of the hydrants is thencompared to the minimum and excess thresholds to determine ifthat location of a fire had potential deficiencies in fire flow or

potential excess fire flow. The output file from the model is a pointfile of fire history locations with an attribute of summed fire flow.Each of these features has a field denoting whether that particularlocation has a potential fire flow deficiency, an excess of potentialfire flow, or is within the adequate range. Thus, specific areasneeding further investigation are identified. To supplement thehistorical fire flow assessment, a parcel-level fire flow computa-tion is also produced. Thus, areas potentially lacking fire flow canbe identified in cases where there has been no significant fireincidence.

2.5. Field verification

Whenever possible, all solutions derived from a GIS procedurewith environmental elements should be verified in the field. Forthis application, all of the locations identified with potential fireflow deficiencies and surpluses were visited by the authors andlocal fire department personnel to confirm the location of thehydrant(s) and the building types.

3. Results and discussion

Execution of the GIS model using historical fire data identifiedseveral locations within Flint Township with potential fire flowdeficiencies and numerous areas of potentially excessive fire flowin Grand Blanc. The composite overlay of the GIS layers for FlintTownship is shown in Fig. 3a with two of the areas of potential fireflow deficiency circled and labeled with a capital letter. Most ofthese areas lie along the commercial zones near the interstatehighways and main arterial roads.

In circled area “A” there were 15 fires over the 3-year periodfrom 2010 to 2013. Each historical fire is represented either by asmall circle with an internal “x”, a star, or a small square with aninternal “x”. As shown on the map legend, these symbol shapesindicate whether the fire flow was deficient, sufficient, or poten-tially excessive. Upon inspection, three of the fires within “A” hadpotentially deficient fire flow capacity, while circled area “B”contains 24 historical fires, 18 of which had potentially deficientfire flow. The spatial distribution of hydrants within the searchradii correlate well to the fire deficiencies identified. Within “A”there were 19 hydrants with a combined fire flow capacity of20,400 gallons per minute. Hydrants here were within a searchradius of a historical fire 52 times, with 8 instances of 4 or morehydrants in the same search radius of a fire. The absence of hy-drants in the southern portion of circle “A” accounts for the ma-jority of deficient fire flows. In circled area “B”, where there weremore fire flow deficiencies, there are 29 hydrants with a fire flowcapacity of 35,200 gallons per minute, yet a hydrant was within

Fig. 3. (a) Flint Township fire flow capacity and (b) Flint Township: potential areas of fire flow deficiency and a larger-scale map showing the two areas of deficient fire flow.

M.M. Kaufman, T. Rosencrants / Fire Safety Journal 72 (2015) 25–32 29

the search radius of a fire only 7 times. The absence of hydrants inthe northeast quadrant of circle “B” accounts for this result.

Fig. 4a indicates no areas of fire flow deficiency in Grand Blanc.This outcome, however, demonstrates the capability of the modelto identify areas with a potential excess of fire flow capacity andunneeded hydrants. In many urbanized regions – especially thosewith an industrial legacy – vacant land parcels (brownfields) mayoccupy significant portions of the total land area [22]. For example,in nearby industrialized Flint, Michigan, 19 percent of the land iscategorized as vacant brownfield [13]. A land parcel layer with thevacancy status of each parcel can be incorporated within the

model to help identify areas with excess fire infrastructure, asdemonstrated in Fig. 4b.

As shown in Fig. 4b, there are some areas within Grand Blancpotentially having excess capacity, despite vacant parcels occupy-ing only a small percentage of the land area (9 percent). The de-finition of excess capacity is problematic, since fire situations varyover time due to property development and variations in buildingsize and use, so an excess of fire flow today may become a defi-ciency tomorrow. The objective here is to allow for the in-vestigation of the possibility there may be excess capacity andinfrastructure in certain areas. Within this framework, excess fire

Fig. 4. (a). Grand Blanc fire flow capacity and (b) Grand Blanc: potential areas of excess hydrant capacity.

M.M. Kaufman, T. Rosencrants / Fire Safety Journal 72 (2015) 25–3230

flow capacity was defined in Section 2.4 as one or more hydrantswithin a 152 m search radius of a historical fire that create anavailable fire flow over twice the minimum value of 1500 gal/min.Fire professionals should evaluate each potential case of excesscapacity, with a more rigorous analysis made possible by the ad-dition of tenure data available from the U.S. Census Bureau at theblock level [28] and current utility usage.

Points A, B, and C in Fig. 4b refer to fire locations where there isa potential excess in fire-flow. Point A has a fire flow rating withinthe search radius of 29,500 gallons per minute. Point B has a fireflow rating within the search radius of 20,000 gallons per minuteand point C had a fire flow of 24,000 gallons per minute. All threepoints were significantly higher than the 3000 gallons per minutethat was set as the threshold.

3.1. Parcel-based fire flow analysis

The coding of the GIS model shown in Fig. 2 was revised toperform a parcel-level fire flow assessment. In the interests ofbrevity this analysis is shown for only one of the study

communities. Fig. 5a depicts the fire flow deficiencies for FlintTownship at the parcel-level, and Fig. 5b illustrates the differencebetween these areas and those depicted by the historical fire flowdeficiency map in Fig. 3a.

This difference map shows those areas within Flint Townshipthat did not have historical fires, but may possess a potential de-ficiency in fire flow. As indicated by this figure, there are someclusters (especially in the northern part of the township) of po-tential fire flow deficiency not revealed by the historical fire data.In urban areas where there has been new development and littlehistory of fire, using the parcel-based GIS model can help firepersonnel identify areas of potential inadequate (or excess) fireflow capacities.

3.2. Future implications of this research

With this GIS model, the existing spatial and attribute data canenable additional GIS analyses. For instance, future hydrant loca-tions can be optimized by simulating the required distances andcorrelating these to the anticipated land uses and building types. It

Fig. 5. (a) Parcel-level fire flow capacity for Flint Township and (b) fire flow deficiency difference map.

M.M. Kaufman, T. Rosencrants / Fire Safety Journal 72 (2015) 25–32 31

is also possible to ascertain the frequencies of different categoriesof fires such as hazardous materials available from the fire codesavailable in the attribute table, and to identify the patterns ofbuilding types prone to fires.

The temporal aspects of fire flow availability can also be in-vestigated. If the times of fire occurrence are input from the fireincidence report, inclusion of these data would allow for the as-sessment of current fire flow availability. Since water usage varieswidely in urban areas, the quantity of water and system pressureavailable for fire flow varies with the existing demand on thewater supply system [14]. Integration of fire temporal data into

Supervisory Control and Data Acquisition, or SCADA systemswould permit a real-time (instantaneous) transfer of availablewater to a fire scene. SCADA systems are specialized computernetworks and devices that work in concert to monitor and controlkey processes involved in the management of machinery, equip-ment and facilities. Measurements taken from a variety of sensors(e.g., temperature, pressure, flow) are used to make decisions, forexample; to open a valve and release water from a tank when itfills up [7].

M.M. Kaufman, T. Rosencrants / Fire Safety Journal 72 (2015) 25–3232

4. Conclusions

The main hypothesis was confirmed by the model test; fire flowcapacity is not sufficient in all cases within a fire department'scoverage area. Moreover, the converse is also true, this model hasthe capability to identify areas where excess fire infrastructuremight exist. As a result, the efficacy of the model is likely to in-crease when a land parcel layer is included and applied to largerurban areas having significant percentages of vacant land. Giventhe coding mechanism used to identify fire types in fire incidentreports, this model also has the ability to characterize fire flowcapacity for specific classes of fire, such as those associated withhazardous substances. This aspect of the model's flexibility may beof some benefit when deciding on the purchase and use of firefighting foams and the equipment needed for their application.The inclusion of building type on the fire incidence reports wouldincrease the power of this GIS application by enabling an easierspatial characterization of fire incidences by building type. In ad-dition to the identification of the spatial patterns related to fireevents, fire flow, and building types, perhaps the most promisinguse of the model will come with its application to fire's temporalaspects. Here, the model might provide a variety of results rangingfrom basic frequency information about the incidences of specificcategories of fire type, to providing input into real-time SCADAsystems for fire flow assessment and water transfers before firecrews arrive on site.

This research contributes to the fire science literature by pro-viding a flexible method for the assessment of deficient, adequate,and surplus fire flows. Using this model with historical fire flowsalso enables further analysis of fire clusters by specific fire types;whereas running the model at the parcel-based scale can identifyareas of potential fire flow deficiency (or surplus) where no his-torical fires have occurred. The primary objective of case studyresearch is accomplished, as this model is applicable to any urbanarea or fire district. Additional research is needed that investigatesthe clusters of fires and their correlation with various fire flowcapacities and local socio-demographic characteristics. Using GISto merge fire response times to specific fire flow capacities mayalso help fire districts evaluate their current and future fire stationlocations.

Acknowledgments

Content of this publication is based upon work supported bythe Research and Creative Activity funding from the University ofMichigan–Flint's Office of Research and Sponsored Programs(Grant no. U042928). The authors would like to thank John Ring-welski and James Harmes, Fire Chiefs of Flint Township and GrandBlanc, respectively, for their assistance with fire data, and theGenesee County staff; specifically, Jenifier Boyer (EmergencyManagement) and Ken Koleda (GIS) for granting access to theirdata and information. We also thank Sergeant Dan Kelly of theFlint Township Fire Department who provided water flow data fortheir hydrants, Matt Wurtz and Matt Gilbert of Grand Blanc fortheir digital hydrant location and flow data, and Karyn Miller, FlintTownship Supervisor for her efforts in securing data.

References

[1] C. Alexandris, J. Giannikos, A new model for maximal coverage exploiting GIS

capabilities, Eur. J. Oper. Res. 202 (2010) 328–338.[2] American Water Works Association, M 31: Distribution System Requirements

for Fire Protection, 4th ed., AWWA, Denver, CO, 2008.[3] A. Asgary, A. Ghaffari, J. Levy, Spatial and temporal analyses of structural fire

incidents and their causes: a case of Toronto, Canada, Fire Saf. J. 45 (2010)44–57.

[4] R.C. Barr, J.A. Bennett, N. Brunacini, R.J. Coleman, J.M. Eversole, The fire Chief'sHandbook, 6th ed, Fire Engineering, Tulsa, OK, 2003.

[5] M.E. Benfer, J.L. Scheffey, Evaluation of Fire flow Methodologies, Hughes As-sociates, Baltimore, 2014.

[6] K. Boulding, General systems theory: the skeleton of science, Manag. Sci. 2(1956) 2197–2208.

[7] S.A. Boyer, Scada: Supervisory Control and Data Acquisition, 4th ed., Instru-ment Society of America, Research Triangle Park, NC, 2009.

[8] J. Terry Coppock, GIS and natural hazards: an overview from a GIS perspective,in: A. Carrara, F. Guzzettti (Eds.), Geographical Information Systems in As-sessing Natural Hazards, Springer, The Netherlands, 1995, pp. 21–34.

[9] J. Corcoran, G. Higgs, D. Rohde, P. Chhetri, Investigating the association be-tween weather conditions, calendar events and socio-economic patterns withtrends in fire incidence: an Australian case study, J. Geogr. Syst. 13 (2) (2011)193–226.

[10] A.E. Cote, C.C. Grant, J.R. Hall, P.A. Powell, R.E. Solomon, Fire ProtectionHandbook, 20th ed., National Fire Protection Association, Quincy, MA, 2008.

[11] R.G. Cromley, J. Lin, D.A. Merwin, Evaluating representation and scale error inthe maximal covering location problem using GIS and intelligent areal inter-polation, Int. J. Geogr. Inf. Sci. 26 (2013) 495–517.

[12] K.J. Dueker, Urban geocoding, Ann. Assoc. Am. Geogr. 64 (2) (1974) 318–325.[13] Flint, City of Flint Master Plan, 2013, (adopted).[14] L.A. House-Peters, H. Chang, Urban water demand modeling: review of con-

cepts, methods, and organizing principles, Water Resour. Res. 47 (2011)W05401. http://dx.doi.org/10.1029/2010WR009624.

[15] W.E. Huxhold, An Introduction to Urban Geographic Information Systems,Oxford University Press, Oxford, UK, 1991.

[16] IAFC, International Association of Fire Chiefs, Fundamentals of Fire FighterSkills, 2nd ed., Jones and Bartlett Publishers, Sudbury, MA, 2009.

[17] ICC, International Code Council, International Fire Code 2012, Country ClubHills, IL, 2011.

[18] ICC, International Code Council, International Building Code 2012, CountryClub Hills, IL, 2011.

[19] C.R. Jennings, Social and economic characteristics as determinants of re-sidential fire risk in urban neighborhoods: a review of the literature, Fire Saf. J.62 (2013) 13–19.

[20] M.M. Kaufman, M. Wurtz, Small system maintenance management using GIS,J. Am. Water Works Assoc. 90 (1999) 70–79.

[21] A.T. Murray, Optimising the spatial location of urban fire stations, Fire Saf. J. 62(2013) 64–71.

[22] J.I. Nassauer, J. Raskin, Urban vacancy and land use legacies: a frontier forurban ecological research, design, and planning, Landsc. Urban Plan. 125(2014) 245–253.

[23] NFIRS, National Fire Incident Reporting System, Department of HomelandSecurity, Federal Emergency Management Agency, U.S. Fire Administration,2014, ⟨https://www.nfirs.fema.gov/⟩. (accessed 08.22.14).

[24] A. Robinson, R. Sale, J. Morrison, Elements of Cartography, 4th ed, John Wiley& Sons, NY, 1978.

[25] J. Rowley, P. Gilbert, The market for land information services, systems andsupport”, in: P.J. Shand, R. Moore (Eds.), The Association for Geographic In-formation Yearbook-1989, Taylor and Francis, London, England, 1989,pp. 85–91.

[26] S.E. Schachterle, D. Bishai, W. Shields, R. Stepnitz, A.C. Gielen, Proximity tovacant buildings is associated with increased fire risk in Baltimore, Maryland,homes, Inj. Prev. 18 (2012) 98–102.

[27] J.E. Stem, State Plane Coordinate System of 1983, NOAA Manual NOS NGS 5,, U.S. Department of Commerce, Charting and Geodetic Services, Rockville, MD,1989.

[28] U.S. Census Bureau, 2010 Census of Population and Housing, Population andHousing Unit Counts CPH-2-1, United States Summary U.S. GovernmentPrinting Office, Washington, DC, 2012.

[29] H. Veregin, Error modeling for the map overlay operation,” in: M. Goodchild,S. Gopal (Eds.), Accuracy of Spatial Databases, Taylor and Francis, London,England, 1989, pp. 2–11.

[30] L.J. Wenzel, Water Supply Requirements for Public Supply Systems, 20thEdition, NFPA Handbook, Quincy, MA, 2008 (Section 15, Chapter 2).

[31] R.K. Yin, Case Study Research: Design and Methods, 2nd ed, Sage Publications,Thousand Oaks, CA, 1994.