Household Dislocation Algorithm 1: A Modified HAZUS Approach* Jing-Chen Lu Walter Gillis Peacock Yang Zhang Yi-Sz Lin. April, 2007

Transcription

1 Household Dslocaton Algorthm : A Modfed HAZUS Approach* Jng-Chen Lu Walter Glls Peacoc Yang Zhang Y-Sz Ln Aprl, 2007 Hazard Reducton and Recovery Center Texas A&M Unversty * Ths document dscusses and provdes detal nstructons for the creaton of the frst dslocaton algorthm created for and mplemented nto the Md-Amercan Earthquae Center s MAEVz program focusng on Shelby County, Tennessee. Ths wor was supported by the Md-Amercan Earthquae Center wth fundng from the Natonal Scence Foundaton. Any opnons, fndngs, and conclusons or recommendatons expressed n ths paper are those of the authors and do not necessarly reflect the vews of the Natonal Scence Foundaton or the Md- Amercan Earthquae Center. Suggested Reference: Lu, J.C., W.G. Peacoc, Y. Zhang and Y.S. Ln (2007) Household Dslocaton Algorthm : A Modfed HAZUS Approach. Hazard Reducton and Recovery Center, Texas A&M Unversty. HRRC Reports: 07-03R. (hrrc.arch.tamu.edu/publcatons/research reports/07-03r Dslocaton Algorthm.pdf ) Dslocaton Algorthm - Modfed HAZUS Approach June 8, 2007

2 Household Dslocaton Algorthm : A Modfed HAZUS Approach After a severe dsaster resdents may leave ther homes wllngly or unwllngly for a varety of reasons such as structure damage, housng repar or remodelng, utlty falures, request or requrements to leave by buldng owners or publc offcals, job loss, suspenson of publc servce (transportaton), or general neghborhood degradaton. Dslocated households can face many dffcultes. For example, there may be extra expenses and stress assocated wth the need to relocate and reestablsh new lvng arrangements. For homeowners, reparng or rebuldng of ther orgnal home can be made more dffcult because of dstance and requre addtonal transportaton expenses as the wor s undertaen and montored. There may also be addtonal transportaton costs as household members travel to jobs or schools located closer to ther orgnal homes. In addton, many dslocated resdents wll have lmted alternatve lvng choces (e.g. home of relatves or frends, rental unts, or hotels etc.), whch may force relance on publc shelters and temporary housng solutons whch wll often, of necessty, be provded by the local, state, or federal government. The loss of populaton wll also have consequences for local busness as they attempt to recover and reestablsh themselves. In lght of these ssues t s mportant for emergency managers, planners, concerned communty organzatons, busnesses, and polcy maers to be able to estmate dslocaton patterns that mght follow a dsaster so that pragmatc emergency response plannng and effcent deployment of response and recovery resources can be undertaen. By nowng the numbers of households lely to be dslocated and the dslocaton pattern wthn an area, polcy maers can tae actons to reduce dsorder durng the emergency and response stages, and potentally enhance restoraton and recovery processes. Basc logc behnd ths approach: The followng algorthm s based on a modfed HAZUS approach for estmatng household dslocaton. HAZUS derves t estmates of dslocated households based on aggregate census tract data and damage estmates. Damage estmates are used to derve the percent of snglefamly dwellng unts n complete damage state and the percent of non-sngle famly structures (multfamly) n both extensve and complete damage states for each census tract. These fgures are then weghted and multpled by the number of sngle and non sngle-famly dwellng unts respectvely to estmate the total number of dwellng unts that wll generate dslocated households, whch n turn s multpled by the average number of household per dwellng unt to derve the number of dslocated households for a census tract. We term the approach specfed n ths document a Modfed HAZUS approach because t employs the basc logc utlzed by HAZUS, however t dffers n a number of mportant ways. Frst, t wll utlze damage state probabltes (P IM ) for each resdental structure followng Ba, Hueste and Gardon (2006). These structure based estmates are lely to be dfferent those utlzed by HAZUS. Second, we wll employ damage states also propose by Ba et al (2006) whch are dfferent but comparable to HAZUS and these damage states wll be weghted by dslocaton factors n a fashon smlar to HAZUS. Our approach also utlzes the rcher structural nventory data for the MTB whch provdes data on the actual number of dwellng unts per resdental structure. Rather than predctng household dslocaton by tract, ths approach wll use census bloc-groups as the base level of aggregaton because bloc-groups are lely to be more meanngful to planners and emergency managers. Fnally, we wll recommend that maps of the spatal dstrbuton of dslocated households by bloc-group also be generated to facltate plannng. Dslocaton Algorthm - Modfed HAZUS Approach June 8,

3 I. Base data requrements. The Modfed HAZUS Dslocaton Algorthm. Census data at the bloc group level: In the HAZUS pacage, the data at Census tract level of aggregaton are used to estmate possble dslocaton household. Bloc group data are used here to estmate more detaled nformaton thereby facltatng plannng wthn local communtes and countes. The followng are the data needed for the dslocaton algorthm that are part of the Shelby County US Census data provded by French and Muthuumar. These data are also avalable n the downloadable zp fle created for the socal vulnerablty algorthm at They are n the fle called: shelby_sv_tnsp.dbf. Varable name Varable defnton! TOT_HH " Total No. of Households! TOT_HU " Total Housng Unts The above data are employed to calculate the average number of households per dwellng unt. 2. From the nventory data: The modfed HAZUS algorthm wll requre data from the Shelby County Inventory data (v.0) produced by French and Muthuumar. The algorthm wll be executed for resdental structures only. It s therefore crtcal that MAEVz be able to clearly dentfy resdental structures and these structures must be clustered nto ther respectve census bloc-group areas. In the nventory data (v.0) structure type s recorded under the varable: OCC_TYPE. Whle there are a varety of types of structures, the algorthm should only be run usng sngle famly resdental structures (RES) and mult famly structures (RES3). The dslocaton algorthm wll also need the number of dwellng unts per structure from the nventory data. Followng the nventory data names the dwellng unts for structure wll be desgnated NO_DU. So, the varables needed from the Inventory data (v.0) are: Varable name Varable defnton! OCC_TYPE " Structure occupaton type. The algorthm needs only sngle famly structures (RES) and mult-famly structures (RES3)! NO_DU " No. of dwellng unts n the structure. NOTE f ths s mssng for RES, assume the value s. 3. Damage State Probabltes (P IM ): The fnal crtcal data necessary for these calculatons wll be the Damage State Probabltes for each resdental structure gven the ntensty measures (P IM ) for an earthquae event or scenaro. The damage state probabltes (P IM ) are those dscussed by Ba, Hueste and Gardon (2006). These wll be combned wth Dslocaton Factors (see Table below) for sngle famly and mult-famly (non-sngle famly) resdental structures to determne the dslocaton probablty for each resdental structure gven the damage state probabltes for a gven ntensty measure. Dslocaton Algorthm - Modfed HAZUS Approach June 8,

4 II. Dslocaton assumptons: The HAZUS model, assumes that structure damage s the only factor drvng household dslocaton and t dfferentates resdental housng nto two forms sngle famly and multfamly (.e., non-sngle famly) resdental structures. The possblty of household dslocaton s based on the damage states of these two forms of housng. Specfcally the expectaton s that sngle-famly housng n slght, moderate, and extensve damage states and multfamly housng n slght and moderate damage states wll not result n household dslocaton. On the other hand, t s assumed that00% of the household n completely damaged sngle famly and multfamly housng wll be dslocated, and that 90% of the households n extensvely damaged multfamly housng are dslocated. We wll employ the same basc logc, however the damage state categores proposed by Ba, Hueste and Gardon (2006) Insgnfcant (I), Moderate (M), Heavy (H), and Complete (C) wll be employed. The dslocaton factors (DsF) for each state and for each type of resdental structure are presented below n Table. Table. Dslocaton Factors by Damage States Dslocaton factors Proposed MAE Sngle Famly Mult-famly Damage States DsF DsF ( ) ( ) Insgnfcant (I) Moderate (M) Heavy (H) Complete (C).0.0 III. Process for estmatng dslocaton household for bloc group:. Calculate average number of households per dwellng unt by bloc-group, Aveu : By calculatng the average number of households per dwellng unts we get some noton of the number of households adjustng for occupancy rates. Ths adjusted mean wll be used to estmate the number of dslocated households.! AveU = TOT _ HH TOT _ HU 2. Calculatng the dslocated households: The followng assumes that these calculatons wll ) be produced for resdental structures [OCC_TYPE=RES or RES3] and 2) that structures can be dentfed as sngle-famly [RES] or mult-famly (non-sngle famly) [RES3] structures. NOTE the P IM value s generated followng Ba, Hueste and Gardon (2006). In other words, P IM s the probablty assocated wth each damage state (I, M, H, and C) for a partcular structure gven a specfed ntensty measure (S a ). a. Calculatng the number of households dslocated for each structure based on whether t s a sngle famly (RES) or mult-famly structure (RES3):. Calculatng the number of dslocated households for each sngle famly [OCC_TYPE=RES] structure. Ths formula does not nclude NO_DU as wll be ncluded n the next formula because the number of dwellng unts s assumed to be one for sngle famly: Dslocaton Algorthm - Modfed HAZUS Approach June 8, 2007

5 = DsF P AveU! ( IM ) 2. Calculatng the number of dslocated households for each mult-famly [OCC_TYPE=RES3] structure : = DsF P NO DU AveU _! ( IM ) b. Calculate the number of dslocated households: The followng recommends generatng a total number of dslocated households by bloc group. Calculaton of the total number of dsplaced household n the bloc group, DsHh, s smply the sum of dslocated sngle and mult famly households n each bloc-group. It mght also mae sense to calculate the percentage of households dslocated from the bloc-group, PDsHh. In these formulas, the stands for the total number of resdental structures (buldngs) of each type (sngle [ ] and multfamly [ ]).! DsHh = + = =! PDsHh = ( DsHh TOT _ HH )00 ; where and are the number of sngle and mult famly structures respectvely. c. Aggregate the total number of dslocated households for a jursdcton by smply summng across bloc-groups n the jursdcton, TotDh j. The default should be the County (.e., Shelby) but the user should be able to defne areas (wth the caveat/warnng that blocgroups may not conform to the jursdctonal boundares one mght be nterested n).! TotDh j = DsHh n IV. Expected output:. Frst there should be a report of dslocated household by bloc group and the total number of dslocated household at county level. See Appendx. 2. Second, there should also be a map of number of dsplaced household by bloc group (usng DsHh ). See Appendx Thrd maps of percent of dsplaced household wthn the bloc group (usng PDsHh ). See Appendx 6. V. A note on uncertantes: It should be noted that the modfed HAZUS approach presented above reles heavly on Ba, Hueste and Gardon s (2006). Drawng upon the logc of ther wor, t would be possble to consder as the mean dslocaton, ˆµ, for structure gven a certan ntensty IM Note: Snce the total number of households s taen from U.S. Census data and the dsplaced households wll be estmated based on structures n a bloc-group from the nventory data, t s possble that these percentages may be problematc, partcularly n communtes experencng rapd development snce the census data were collected. Dslocaton Algorthm - Modfed HAZUS Approach June 8,

6 measure. Furthermore, contnung wth ther logc, t would appear that one mght be able to calculate the standard devaton, σ, and hence develop confdence ntervals and predcton ˆ IM ntervals, by followng the procedure they outlned. However, that would requre some degree of confdence n the dslocaton factors (DsF and DsF or, employng ther symbology, µ and µ ) treatng them as mean values for dslocaton probabltes gven specfc DsF DsF damage states (I, M, H, and C). Unfortunately there are no systematcally collected data from whch the dslocaton factors suggested by HAZUS were based other than expert opnon (founded on qualtatve ntervews of dslocated households). Hence extendng the logc they suggest for the estmaton of dslocated households may well be questonable. In addton, the procedure Ba et. al., (2006) suggest derves confdence ntervals and predcton bands for a partcular structure, whle the goal here s to develop estmates for a blocgroup and ultmately some jursdcton, such as a county or muncpalty. I suppose we could consder DsHh a random varable and thereby calculate a mean ( µˆ ) and standard devaton ( σˆ DsHh j DsHh j ) for a gven jursdcton (such as a county) to derve confdence ntervals. However, t would be dffcult to gnore the ssue that the estmates themselves (partcularly the dslocaton factors) are derved from a paucty of emprcal evdence. Addtonal concerns would be that we are dealng here wth multple error sources (e.g., n the dslocaton factors, n the applcatons of multple fraglty curves across a varety of resdental structure types) that would undoubtedly propagated through the process of dervng these estmates. Furthermore, I would doubt that the error s randomly dstrbuted throughout an area. For example, the algorthm s lely to generate error somewhat proportonal to the dstance from areas of hghest damage (.e., wor better near areas of hgher levels of damage generatng hgher errors as one moves away from those areas). All of these factors should be consdered as we attempt to develop some noton of the uncertantes n estmaton. The smple fact s that there are almost no systematc studes that have attempted to document actual dslocaton resultng from any form of dsaster, for any perod of tme. Dslocaton Algorthm - Modfed HAZUS Approach June 8,

7 Appendx. Varable Lst Varable Name Descrpton Note TOT_HH Total No. of Households 2000 Census (from Dr. French) TOT_HU Total Housng Unts 2000 Census (from Dr. French) Dslocaton Factor for Sngle DsF Famly structures by damage state See Table, based on HAZUS DsF AveU P IM NO_DU OCC_TYPE DsHh PDsHh TotDh j Dslocaton Factor for Multfamly structures by damage state Average number of household per mult-famly structure Probablty of each damage state for a structure (resdental n ths case) gven IM Number of dwellng unts n a partcular resdental structure Occupancy type (type of structure) Ths algorthm only needs RES and RES3 structures Estmated dslocated households for each sngle famly structure n a gven bloc group. Note, NO-DU s not ncluded because the number of dwellng unts s assumed to be. Estmated dslocated households for each multfamly structure n a gven bloc group. Estmated number of dslocated households for a gven bloc. Percent of bloc group See Table, based on HAZUS TOT _ HH TOT _ HU Based on Ba, Hueste and Gardon (2006) From Buldng Inventory Data for Shelby County (v.0) From Buldng Inventory Data for Shelby County (v.0) ( DsF P IM ) AveU ( DsF P IM ) NO _ DU AveU = + = ( DsHh _ ) 00 households dslocated TOT HH Estmated total households n a jursdcton dslocated n DsHh Dslocaton Algorthm - Modfed HAZUS Approach June 8,

8 Appendx 2. Example Calculatons for a sngle bloc-group wth only Four Structures. The frst two structures are mult-famly resdental structures (OCC_TYPE=RES3) and the fnal two are sngle famly resdences (OCC_TYPE=RES). Mult-famly structures (OCC_TYPE = RES3) sngle-famly structures (OCC_TYPE = RES) Mult-famly structure wth 0 dwellng unts (NO_DU=0) Probablty (P IM ) Dslocaton Factor (DsF ) Damage State (S a = 0.88g) Insgnfcant (I) Moderate (M) Heavy (H) Complete (C) Mult-famly structure wth 30 dwellng unts (NO_DU=30) Damage State (S a = 0.277g) ( DsF P M )=.87 Insgnfcant (I) Moderate (M) Heavy (H) Complete (C) ( )= DsF.356 P M Dslocaton Factor (DsF ) Sngle-famly Probablty (NO_DU=) (P IM Damage State (S a = 0.88g) Insgnfcant (I) Moderate (M) Heavy (H) Complete (C) = ( DsF P M )=.587 Sngle-famly 2 (NO_DU = ) Damage State (S a = 0.277g) Insgnfcant (I) Moderate (M) Heavy (H) Complete (C) ( DsF P M )= 2 P m X DsF AveU =.9 =.87(0)(.9) = =.356(30)(.9) = 0. 0 P m X DsF =.587()(.9) = =.53()(.9) = DsHh = + = = = = Dslocaton Algorthm - Modfed HAZUS Approach June 8,

9 Appendx 3. Example of a fctons report of dsplaced household by jursdcton (Shelby county) and by census bloc group. Number of Dsplaced Household Percent of Dsplaced Household Shelby County, TN % Bloc Group 757XXXXXXXX % 757XXXXXXXX 53 68% 757XXXXXXXX 9 59% 757XXXXXXXX 23 7% 757XXXXXXXX % 757XXXXXXXX %.. 757XXXXXXXX % 757XXXXXXXX 92 59% 757XXXXXXXX % Dslocaton Algorthm - Modfed HAZUS Approach June 8,

10 Appendx Example Number of Dslocated Household Map Dslocaton Algorthm - Modfed HAZUS Approach June 8,

11 Appendx 5 Example of Percent of Dslocated Household Map Dslocaton Algorthm - Modfed HAZUS Approach June 8, 2007