森林防火住宅類(lèi)型的特征與制圖外文文獻(xiàn)

Characterization and mapping of dwelling types for forest fi re prevention Corinne Lampin Maillet a Marielle Jappiota Marl ne Longa Denis Morgea Jean Paul Ferrierb aCemagref UR EMAX 3275 Route de C zanne CS40061 13182 Aix en Provence cedex 5 France bAix Marseille I University Emeritus Professor Aix en Provence France a r t i c l ei n f o Article history Received 8 April 2008 Received in revised form 18 July 2008 Accepted 29 July 2008 Keywords Spatial analysis Dwelling typology GIS Forest fi re risk Prevention Wildland urban Interface a b s t r a c t In a context of risk of forest fi re the development of actions concerning wildfi re prevention and land management is necessary and essential particularly in wildland urban interfaces WUI The term WUI always includes components such as human presence and wildland vegetation Both the hazard probability of fi re outbreak distribution and the vulnerability of urban areas can be characterized through the spatial organization of houses and vegetation The fi rst step required is to characterize and map WUI in large areas and at a large scale which in turn requires qualifying different types of dwellings and mapping them With this goal in view the paper presents a brief synthesis of results com ing from an exploratory process for the characterization of dwelling types Lampin C Long M Jappiot M Stewart et al 2007 Urban sprawl observed in or near forests rep resents a signifi cant threat to the environment Johnson 2001 Radeloff et al 2005b Indeed it leads to habitat fragmentation Theobald Miller fax 33 4 42 66 99 71 E mail address corinne lampin cemagref fr C Lampin Maillet Computers Environment and Urban Systems 33 2009 224 232 Contents lists available at ScienceDirect Computers Environment and Urban Systems journal homepage removal of fuel in the immediate vicinity of homes Cohen 2000 and plans for emergency evacuation Church Hammer Stewart Winkler Radeloff Stewart et al 2003 2007 Theobald 2001 Kamp and Sampson 2003 used population density instead of housing density to de fi ne the potential WUI area in the middle of a density spectrum which varies from 40 to 400 people per square mile However according to Alberti et al 2007 these metrics do not discriminate well between different landscape patterns essentially because the number of housing units has increased at a faster rate than that of the human population Lepczyk Stewart et al 2007 Theobald and Romme 2007 developed a method for map ping the WUI at the national scale using housing density but indi cated that for planning and management activities more detailed data are necessary Buildings should be located based on aerial photographs or using fi eld based data but this kind of data is not available in all countries Various methods corresponding to the generalization of spatial data have been developed to charac terize towns the envelop convex method Delaunay triangulation and recently a generalization pattern from the BD TOPO from the French National Geographical Institute database Boffer 2001 Ruas 1999 The latter method characterized towns as areas that contain similar numbers of buildings which are merged before simplifying their shape using the mathematical morphology approach developed by Matheron and Serra 1964 and also con siders recent studies from Edwards and Regnault 2000 Other authors Kalhori Le Corre Guillaume Galaup Zhang Lampin Jappiot Long Mansuy Ruas 1999 identifi es the centre of town by clusters of houses that C Lampin Maillet et al Computers Environment and Urban Systems 33 2009 224 232225 are less than 50 m apart The number of houses can also be taken into account urban rules consider that houses are clustered when a group contains at least two houses separated by less than 100 m The method based on mathematical morphology principles Cani cio Voiron 1995 Fisher Perkins Walker it is dif fi cult to protect houses against fi re without endangering fi re fi ght ers Scattered dwellings SD correspond to clusters of 4 50 houses located less than 100 m apart This number of houses refers to a fi re fi ghting strategy A fi re fi ghter team composed of four trucks can protect up to 10 houses if they are located relatively close to Fig 1 Location of the study area in south eastern France 226C Lampin Maillet et al Computers Environment and Urban Systems 33 2009 224 232 one another less than 100 m apart Clustered dwellings CD refer to clusters of more than 50 houses located less than 100 m apart In this last category two sub classes have been determined a fi rst one identifi es dense clustered dwellings DCD corresponding to clusters of less than 10 houses located less than 30 m apart and a second one identifi es very dense clustered dwellings VDCD corre sponding to clusters of more than 10 houses located less than 30 m apart that being equivalent to well urbanized areas Different operations were carried out with Arcgis 9 1 software Geo treatment enabled us to create buffers with different diame ters 50 and 15 metres The overlaying buffers were merged and the houses within each buffer zone were counted to differentiate housing types according to the rules presented below For each house of a different class isolated scattered or clustered a buffer of 100 m was created To avoid overlaying of buffers a rule of pre dominance was established where the buffer corresponding to the most clustered class took priority 4 4 Calculation of spatial criteria related to the risk of forest fi re and statistical analysis Spatial criteria related to the risk of forest fi re were determined for each type of dwelling Calculations were made for each polygon created in the study area using the buffering process with a 50 m radius around each house to discriminate the three main types of dwellings isolated scattered and clustered and a 15 m radius around each house belonging to clustered dwellings in order to dis criminate the two sub classes of clustered dwellings dense and very dense clustered dwellings Three indicators were calculated house density mean area to be cleared per house The French For est Orientation Law of July 11 2001 makes clearing brush obliga tory within a 50 m perimeter around each house although the local mayor can extend this limit to 100 m located at a distance of less than 200 m from forest or scrubland and mean perimeter per house to be protected by fi re fi ghters in the case of forest fi re Then a statistical analysis was performed on the whole of polygons obtained by studying the entire study site The aim was to assess whether there was any statistical difference not only between the three main dwelling types based on spatial criteria such as housing density area to be cleared and perimeter to be protected but also inside the dwelling cluster variable between dense and very dense clustered dwelling types This calculation was made with Statgraphics software Comparisons between the three popu lations corresponding to isolated scattered and clustered dwell ings were made as well as comparisons between the two sub populations of clustered dwellings dense and very dense clustered dwellings 4 5 Relationships between the distribution of fi re ignition points and dwelling types Regarding the distribution of fi re ignition points in wildland urban interfaces WUIs as a function of different types of dwelling an indicator of fi re ignition density was calculated using the fi re ignition point database This was defi ned as the ratio of the number of ignition points located within a given area to the surface of this Fig 2 Dwelling type map in the study area C Lampin Maillet et al Computers Environment and Urban Systems 33 2009 224 232227 area and is expressed in the number of fi res per 1000 hectares We considered different types of territory within the study area iso lated WUIs scattered WUIs and clustered WUIs and the remainder of the study area The surface area of the different types was calcu lated and the number of fi re ignition points located within each part was determined 5 Results 5 1 Maps of dwelling types based on distance plus counting criteria using GIS The process of creating buffer zones based on both the spatial juxtaposition of houses and on the number of houses enabled us to create a map of dwelling types using Arcgis 9 1 software This is an effi cient method for the characterization of dwelling types isolated scattered and clustered dwellings and has the advantage of being easy to use The method is based on distance and house counting criteria using GIS and can thus be applied to large areas Fig 2 Each house belongs to a single dwelling type as we can see if we zoom in on the map Fig 3 The repartition of polygons according to dwelling types is presented in Table 1 Two sub clas ses dense and very dense were then defi ned in the clustered dwelling class by buffering the houses included in the 72 polygons of this class Fig 4 compares the result of mapping dwellings with a photo corresponding to the extract of the mapped area Maps of dwellings were produced easily and can be used in the fi rst step in characterizing and mapping wildland urban interfaces 5 2 Dwelling types statistically different according to spatial criteria related to the risk of forest fi re Over the whole of the study area the three populations corre sponding to isolated dwellings ID scattered dwellings SD and clustered dwellings CD have been compared for each spatial cri terion mean housing density mean area to be cleared and mean perimeter to be protected The number of observations corre sponding to polygons is 979 for the isolated dwelling population 524 for the scattered dwelling population and 72 for the clustered dwelling population A box plot comparing the three dwelling types for mean housing density Fig 5 1 mean area to be cleared Fig 5 3 and mean perimeter to be protected Fig 5 5 shows that the type of dwelling has a signifi cant effect on each response var iable Isolated dwellings have the highest mean area to be cleared and mean perimeter to be protected and the lowest mean housing density In contrast clustered dwellings have the lowest mean area to be cleared and mean perimeter to be protected and the highest mean housing density Fig 5 1 to 5 6 presents box plots from left Fig 3 Details of dwelling type map Table 1 Number of identifi ed polygons in the study area Isolated dwelling ID 979 Scattered dwelling SD 524 Clustered dwelling CD 72 Dense clustered dwelling DCD 411 Very dense clustered dwelling VDCD 142 228C Lampin Maillet et al Computers Environment and Urban Systems 33 2009 224 232 Fig 4 Extract of mapped area corresponding to the photo Fig 5 1 6 Box plots left to right top to bottom C Lampin Maillet et al Computers Environment and Urban Systems 33 2009 224 232229 to right and from top to bottom calculated for spatial criteria and different dwelling types ID SD CD and DCD VDCD The two populations corresponding to the sub classes of clus tered dwellings dense clustered dwellings DCD and very dense clustered dwellings VDCD have been compared for each spatial criterion mean housing density mean area to be cleared and mean perimeter to be protected The number of observations polygons is 411 for the dense clustered dwelling population and 142 for the very dense clustered dwelling population As before the box plots comparing the two clustered dwelling types dense and very dense for mean housing density Fig 5 2 for mean area to be cleared Fig 5 4 or for mean perimeter to be protected Fig 5 6 show that the type of dwelling has a signifi cant effect on response variable Dense clustered dwellings have the highest mean area to be cleared and mean perimeter to be protected and the lowest mean housing density In contrast very dense clustered dwellings have the lowest mean area to be cleared and mean perimeter to be protected and the highest mean housing density Results confi rmed that there were highly signifi cant differences between the different types of dwellings The non parametric Kruskall Wallis test was used to test both the difference between the three main classes of dwelling isolated scattered and clus tered dwellings and between the two sub classes of clustered dwelling dense and very dense clustered dwelling In each case the probability value is less than 0 05 As a result there is a statis tically signifi cant difference for each criterion not only between the three main types of dwelling but also between the two sub classes of clustered dwelling Corresponding to highly statistically differ ent types of dwelling each of them has been characterized by their statistical values Table 2 summarizes the main statistical values for housing density area to be cleared and the perimeter to be pro tected for each type of dwelling The actual values are presented in Table2as median mean standarddeviation minimum maximum 5 3 Localisation of fi re ignition points as a function of type of dwelling The mean fi re ignition density value was calculated for the whole Aix en Provence study area for the period 1996 2007 This value reached 3 5 fi re ignition points per 1000 hectares The same value was calculated specifi cally for wildland urban interfaces and proved to be 1 5 times higher in wildland urban interfaces than elsewhere with 5 3 fi re ignition points per 1000 hectares Fig 6 Our results show that fi re frequency is very much higher in WUI areas than in natural areas In addition the fi re ignition density was calculated for each of the dwelling types Within dwelling types we observed that the density value was two times higher for isolated dwellings than for clustered dwellings Specifi cally fi re ignition density values increased greatly from clustered dwellings 4 2 fi re ignition points per 1000 ha to scattered dwell ings 5 2 fi re ignition points per 1000 ha and fi nally to isolated dwellings 9 5 fi re ignition points per 1000 ha This suggests that the spatial pattern of dwellings has a real impact on fi re occur rence Humans and their spatial distribution explain a part of the variability in the number of ignition points 6 Discussion and conclusions The method developed to characterize dwelling types seems to be effi cient using GIS treatment based on the two following crite ria distance between houses and counting houses located close to gether Moreover this method is easy to use can be transferred to land agencies or managers and can be applied to large areas on the scale of forest stands or several communities The maps that can be produced using this method fulfi l the needs of the main end users concerned with forest fi re prevention and forest fi re fi ghting In deed end users such as forest and land planning managers or forest fi re fi ghters are interested in the location of wildland urban inter faces that are directly related to different types of dwellings in con tact with vegetation Maps of different dwelling types can be interpreted for use in developing fi re fi ghting strategies or fi re pre vention measures Main criteria were defi ned housing density area to be cleared and the perimeter to be protected for each housing category These criteria allow managers and fi re fi ghters to measure the effects of the dwelling structure on prevention measures or fi re fi ghting ac tions It will be more diffi cult to protect isolated dwellings whose protection perimeter is almost 10 times larger than the perimeters for clustered dwellings It will be more diffi cult for a home owner to keep a large area around an isolated dwelling clear where this is three times more important than in a clustered dwelling The mean density values that discriminate the three dwelling types proposed Table 2 Characteristics of dwelling types Dwelling typesNumber of polygons Number of housesHousing density number of houses per km2 Area to be cleared per house ha Perimeter to be protected per house m Isolated dwelling ID 9791 1 4 1 1 3 99 104 23 53 199 1 1 0 2 0 5 18 9 353 328 67 178 507 Scattered dwelling SD 5246 9 9 3 50 159 163 35 81 256 0 63 0 64 0 14 0 4 1 2 149 149 39 62 260 Clustered dwelling CD 72170 332 482 51 2614 313 346 144 141 700 0 32 0 34 0 1 0 1 0 7 38 46 27 14 119 Dense clustered dwelling DCD 4115 22 70 1 973 212 233 92 86 560 0 47 0 49 0 18 0 18 1 1 148 176 96 47 446 Very dense clustereddwelling VDCD 14253 153 288 10 2135 479 507 150 267 995 0 20 0 21 0 06 0 1 0 37 29 32 15 9 74 0246810 Clustered dwelling Scatterred dwelling Isolated dwelling Dwellingtypes Fire ignition density Nb of points per km2 Mean fire ignition density on the whole study area Fig 6 Fire ignition density value 230C Lampin Maillet et al Computers Environment and Urban Systems 33 2009 224 232 are approximately 100 houses per square kilometre for isolated dwellings 160 houses per square kilometre for scattered dwell ings and 345 houses per square kilometre for clustered dwellings These density values can be compared to thresholds defi ned in pre vious studies concerned with wildland urban interface density values Theobald 2001 and Stewart et al 2003 2007 fi xed two thresholds to differentiate wildland urban areas with a minimum housing density of six houses per square kilometre for WUI from urban areas with a housing density over 148 houses per square kilometre Other thresholds were used by Hammer et al 2004 for whom the lowest density rural clusters remain below two housing units per square kilometre while the highest density cat egory was more than 128 units per square kilometre These thresh olds can be combined with thresholds obtained with the density function Lampin et al 2007 less than 48 houses per square kilo metre for isolated dwellings around 48 160 houses per square kilometre for scattered dwellings and more than 160 houses per square kilometre for clustered dwellings where the highest values are relatively similar and the lowest values are different This re sult is congruent with the low values obtained with the mathemat ical morphology approach Lampin et al 2007 Values obtained with our pres