GIS and Evidence-Based Policy Making - Chapter 7 doc

It is the high profile of the field of research that motivates this project intoestablishing how geographical information systems GIS may be used toimprove flood warning, and emergency p

Using GIS to Identify Social Vulnerability

in Areas of the United Kingdom That Are

at Risk from Flooding

Tom Kieron Whittington

CONTENTS

7.1 Introduction 133

7.1.1 Background 134

7.1.2 Requirement Study 135

7.1.3 Aims 139

7.2 Methods 140

7.2.1 Index of Social Vulnerability 140

7.2.2 Index of Flood Probability 144

7.2.3 Combined Index of Flood Vulnerability 146

7.3 Results 146

7.3.1 Index of Social Vulnerability 146

7.3.2 Index of Flood Probability 148

7.3.3 Combined Index of Flood Vulnerability 150

7.4 Discussion 151

7.4.1 Evaluation of Results 151

7.4.2 Interface Potential 153

7.4.3 Modifiable Areal Unit Problem (MAUP) 154

7.5 Conclusions 155

7.5.1 Answering the Research Questions Posed 155

7.5.2 Developing the Model for Use within the Flood Industry 156

References 156

Around 5 million people in 2 million properties live in flood-risk areas in England and Wales (Environment Agency, 2000) Property worth over

ß 2007 by Taylor & Francis Group, LLC.

£200 billion and agricultural land worth approximately £7 billion are tially at risk of flooding (HR Wallingford, 2000) The floods of Easter 1998and autumn 2000 gave the United Kingdom an important reminder of ahazard that, though ever present, has been neglected by society in recenttimes Many organizations are encouraged to deal with the problem, which

poten-is predicted to increase in frequency in the future due to climate change andcontinued urbanization of the floodplain (Price and McInally, 2001) There is

a rise in the philosophical approach of ‘‘living with the hazard’’ (Smith andWard, 1998) that focuses on flood warning and emergency planning, thanflood prevention Initiatives to help communities to help themselves aretherefore high on the agenda but require a clear understanding of the socialvariability and different needs of communities at risk

It is the high profile of the field of research that motivates this project intoestablishing how geographical information systems (GIS) may be used toimprove flood warning, and emergency planning and response in theUnited Kingdom To determine how the technology could be best put touse with immediate effect, a requirements study has been accomplishedfrom a literature review and interviews with the main organizationsinvolved in flood warning, planning, and research in the United Kingdom.The conclusion of the requirement study identifies that the spatial distribu-tion of vulnerable groups living within the floodplain is a prime target forresearch, and this group would benefit greatly from GIS investigation.This research attempts to bring together social-vulnerability studies andflood-probability data with GIS technology to produce a high-resolutionindex of flood vulnerability (IFV) A number of applications demonstratehow the index and some of the data layers used in its creation may be used toimprove the efficiency and quality of flood managers’ decision-making.During a flood emergency, planners can quickly identify different groups

of people with different social needs and can thus disseminate resourcesappropriately Alternatively, flood-warning education can be adapted fordifferent communities identified by their postcode The final index produced

is a prototype tool, which requires refinement, but demonstrates how ing studies could be improved with the inclusion of GIS technology

exist-7.1.1 Background

The next century may see apparent increases in CO2due to human activitiesresulting in climate change and consequently flooding from increasedintensity and frequency of rainfall and sea-level rise (Price and McInally,2001) Most flood-prevention schemes can be expected to fail if a high-levelflood scenario occurs, and there are many locations where an engineeringsolution is impractical or could lead to considerable damage to the envir-onment (Borrows, 1999; Environment Agency, 2001) The expansion ofurbanized areas can create the risk of more-frequent flood situationswhere increased precipitation results in greater runoff (EnvironmentAgency, 2001; Price and McInally, 2001) Around 5 million people in

ß 2007 by Taylor & Francis Group, LLC.

2 mi llion pro perties are vulne rable to flood risk in Engla nd and Wal es(Env ironmen t Agenc y, 2000); the flood leve ls during the Oct ober 2000floods were the highest on record in many location s, and 10,000 proper tieswere affecte d (Enviro nment Agenc y, 2001).

The Bette r Regul ation Task Force (2000) re commen ds that policy make rsconsid er vulne rable peop le at all st ages of their work, with gre atest consi d-eratio n to in clude vulnerabi lity im pact assessm ents Knowl edge of wheresocial differen ces lie with in commun ities poten tially at ris k from floodi ngand the general natur e of their circumst ances is needed to bette r targetpubli c aware ness informati on and to resp ond appropri ately to emerge ncysituat ions (Mor row, 1999; Tuns tall, 199 9; Blyth et al., 2001; Envi ronmentAgenc y, 2001) On e of the mai n reasons for targetin g vuln erable gro ups

is the desire to concen trat e on the worst affec ted areas and popu lation(Jaspar s and Shoham 1999); emerge ncy plan ners need to know who theyare and wh ere they are concen trated (Mor row, 199 9) Perform ance of floodforecast ing and war ning syst ems appear s to be poor in the Un ited Kin gdom(Hag gett, 1998; Horn er, 2000; Penning- Rows ell et al., 2000; Envi ronmentAgenc y, 2001) but wil l contin ue to im prove, pro vided that the rig ht infor-mation can be deli vered in advanc e to the right peop le (FHRC , 200 1).There is a need to devel op accura te flood- hazar d maps and flood- rescueaction plans for hazard -prone areas (Rant akokko, 1999) The Environ mentAgenc y has created the ind icative floodpl ain map usin g historical andrainfal l catchmen t mode ls (e.g., ISIS and MIKE11) The mapped floodplai nboun daries are disaggr egated to unit postc odes to a ssist publ ic iden tifica-tion of risk, but there is a mi smatch bet ween postc ode uni ts and iden tifiedflood- risk areas (Plough er, 2000) Boyle et al (1998) discrimi nate the flood-plain into uni ts def ined by the qua ntifica tion and spatia l variabi lity of fl oodhazard Flood-pr obabili ty contours are create d with hydrol ogical mode ling

of flood flow rates assoc iated with differen t return-pe riods (the flood quency in years) The use of GIS in this typ e of hazar d exposur e pro vides anefficient and accurate assessment for areas prone to flooding (Boyle et al.,1998), but it does not consider the socioeconomic variability that may also beassociated with that location

fre-7.1.2 Requirement Study

It is the intention of this research to enhance the flood-warning and gency-planning industry with GIS technology There are a number oforganizations in the United Kingdom with different responsibilities withinthe flood indu stry (Table 7.1); consultant s and resea rch insti tutes areassigned projects to develop the roles of these organizations As the industrydoes not have a single function, it was necessary to identify one aspect oftechnical research that could reasonably be undertaken with limited timeand data, and yet provide a useful service A GIS requirement study hasbeen administered to evaluate the research needs of flood warning andemergency planning, but these particular needs are difficult to define

emer-ß 2007 by Taylor & Francis Group, LLC.

TABLE 7.1

Flood Hazard Responsibilities in the United Kingdom

Environment, Food and Rural Affairs (DEFRA) have policy responsibility for flood and coastal defense in England and administers the legislation that enables work to be carried out The Flood and Coastal Defence Programme is aimed at reducing risk to people and the developed and natural land by financially supporting, advising, and guiding flood and coastal defense operating authorities, and funding research programs Following the 2000 floods, better definition of flood or erosion risk areas was identified as being of particular necessity to flood planning (HR Wallingford, 2000) Meteorological

office

Continuous monitoring of weather conditions and rainfall patterns from remote sensing and ground-based measurements falls under the authority of the meteorological office (MetOffice) Computer models simulate river discharge based on rain-gauge readings over time-periods in various parts of different catchments If a flood situation arises, the MetOffice alerts the EA and local authorities to the possible threat

flood-responsible for education campaigns prior to any particular event.

EA regional offices deal with more localized issues with respect to flood-alleviation schemes and flood warning There is some variation between EA regions in the scope and sophistication of the facilities available to support operational decision-making (Haggett, 1998)

Police and local

authorities

Local planning and emergency considerations: It is the role of the police

to organize localized planning and response on behalf of LA and EA guidelines Once alerted to the onset of a flood event a command- and-control center is set up by the police, local warnings are issued, and emergency response teams dispatched The police have roles in the evacuation of people at-risk and traffic management (Smith and Ward, 1998) Without detailed information on the social

characteristics of different threatened communities, emergency planners and services can have great difficulty in reaching specific communities in need (Haggett, 1998) The police have endeavored

to improve reliability and speed of flood-warning dissemination and focus on emergency response through the use of technology (Horner, 2000)

ß 2007 by Taylor & Francis Group, LLC.

owing to the dearth of a ppropria te litera ture To overcome this problem , aseries of intervie ws has been condu cted with man y of the key organizat ionsfrom the flood ind ustry in the United Kingdom to supp lement findingsfrom releva nt litera ture The object of the inte rviews are twof old: to collate

an audit of curren t GIS to co nsider how the techn ology is used in fl oodwarning and emerge ncy planning, and iden tify areas where GIS are no tmeeting their potenti al; a nd to conside r the opi nions of some of the leadi ngpractitio ners in these areas, wh ich would significant ly ben efit the flood-warning and emerge ncy-plan ning pro cess

The audit shows that most system s deal with flood war ning and pre ventio n, or emergency planning, but very few syst ems are used for em er-gency resp onse that re quires pre cise and up- to-date inform ation for effici entdecisio n-maki ng on how to respon d to chang es in ci rcumst ances (Table 7.2)

-No system s-in-use consi der all a spects of the flood indu stry, and ne arly allthe use of GIS is onl y involv ed with research , possibl y indicating thedeficien cy of actual running syst ems Over all the system s are spar se inutilizi ng soc ioeconom ic data and as sessing fl oods accordi ng to popu lationand commodity risk Advan ces of GIS may invo lve more soc ial consid er-ations of floodin g as the nece ssity for monitorin g flood- risk comes from theeffect it may have on soc iety It may also be fruitf ul to conside r a syst em thatnot on ly provides flood war ning but a lso assists in the emerge ncy planni ngand respons e of a flood in real time, anoth er use of GIS which is not beingoptim ally exploited Real-tim e GIS may be useful in modify ing emerge ncyplans appr opriate ly duri ng a flood event as new da ta bec omes avai lable.Interviews included advances for research in the flood industry thatorganizations felt were important, and these were found to generally concurwith ideas expressed in the literature The so-called ‘‘living with hazards’’philosophy (Alexander, 1997; Smith and Ward, 1998) has created the needfor more effective initiatives to be researched within the roles of warning,

ß 2007 by Taylor & Francis Group, LLC.

Context Used

Name

Flood Warning Flood Prevention Emergency Planning Emergency Response Research Assistance Customised System STandard GIS Software Real-Time Strategic Database Querying Mapping Map Querying Report Writing Mulit-Task Windows Model Integration Data Integration Used by Organization Used by other Organizations Public Meteorological Hydrological Topographical Land Use Social Economical Address Transport

Climate Patterns Water Depth Water Extent Damage Population

CEH Wallingford General use

Halcrow General use

Home Office HAZMOD

Notes: a Environment Agency, b National Flood-warning Centre, and c Flood Hazard Research Centre, University of Middlesex

planning, and response Emergency planners, police, and local authoritieswould benefit from GIS, but the technology does not currently play animportant role Traditionally, GIS technology has been used in supportingsurface-water modeling and flood-hazard exposure analysis by providingthe ability to integrate modeling results with other layers of information toenhance the decision-making process (Boyle et al., 1998).

However, GIS provides the means of integrating different phenomenonsuch as social and geographical data, in order to increase the overall under-standing of the relationship between society and disaster (Dash, 1997).Technological solutions could be sought to improve the disseminationprocess with decision-support tools (Haggett, 1998) but for many floodplanners the advantages of GIS are not considered or fully understood.For instance, the City of Edinburgh Council stores all information forresponding to a flood within one paper file binder During a hazard situ-ation, members of a committee sit around a table and discuss appropriatecourses of action Benefits of GIS and automated information are seen asnegligible because flooding seldom occurs twice The Thames police alsoresolve response management without the use of GIS, where and when it

is necessary; a system that the users consider sufficient (Whittington, 2001).Conclusions of the requirements study identify communicating theflood message (education prior to and alerting during an emergency) andimproving disaster response as two important advances in flood research.Both of these initiatives are common to the need for discriminating betweendifferent groups of people living within the floodplain

There are two ways that communities could be classified within thefloodplain First, since disaster vulnerability is partially socially constructed(Morrow, 1999), identifying different levels of social vulnerability couldimprove efficiency of warning and emergency response For example,awareness that 75% of a community are non-English speakers could result

in more appropriately tailored education and warning, whereas the ledge that 50% of a community are aged over 75 years could help allocatethe dispatch of sufficient help

know-Secondly, flood return periods (or scenarios) are effectively an estimatedprobability of a flood occurring at any year The flood risk map for Englandand Wales (Morris and Flavin, 1996) and the indicative floodplain map(Environment Agency, 2000) consider the maximum predicted flood but

do not distinguish degrees of risk Flood risk can vary within the area of acommunity and a detailed understanding of flood-risk variability could beextremely valuable to flood planners Flood-vulnerability mapping couldhelp prioritizing emergency dissemination where accurate flood data iden-tifies where and when different flood scenarios are going to occur

calcu lated from socioeco nomic and flood- depth da ta The broad appro ach is

to merge the physica l ass essment of flood risk with the social as sessm ent ofvuln erability to produce an IFV to the unit postc ode level The index is used

to create a pro totype decision- support tool that utilize s GIS func tionality toimpr ove the efficiency of flood- war ning and emerge ncy planning, as well ashavin g scope for em ergency resp onse From this research , it is hoped toconvi nce em ergency planners of few GIS benefits The vuln erabi lity-inde x isnot re quired in this research to be highl y accura te; it is the concep ts that areund er scrutiny It is thought that the typ e dec ision-s upport tool is useable ,given that sufficie ntly accurate da ta is av ailable

The mai n resea rch question s consid ered are

. How can GIS be used to ass ist the identificati on of vuln erablegroups wi thin commun ities in areas prone to floodin g?

. What data provides the best indic ators of soc ial vuln erability?

. What data would best ser ve the definiti on of flood- risk spat ialvariabil ity?

. Does the inte gration of the two datase ts improv e the vulne rabilit yindex?

7.2 Methods

7.2 1 Index of Social Vulnerabi lity

The mode l is tested on a 30-km stretch of the River Tham es betwee n Etonand Walto n-on-Th ames Socio economi c da ta is used to map an in dex ofsoc ial vulnerabi lity (ISV) at the level of enume ration districts (EDs) Socialvuln erability to a disaster may not only be affec ted by pover ty but is theover all ability to resp ond to a hazar d situat ion (Tu nstall, 1999) Certainphy sical and soc ial attr ibutes (age, race, and gend er) and living arr ange-ments (si ngle pare nt hou seho lds), wher e the relati onship wi th soc ial cl ass isnot so well-de fined, a re likel y to have as mu ch, if not mo re effect onvuln erability as pove rty Far from bein g mutual ly exclus ive, these factor stend to occu r in combi nations that intensify risk exp onentiall y (Mor row,1999) Data was obt ained to refle ct a wide ran ge of attributes that mayincrease (or dec rease) social vuln erability of a smal l commun ity Fromdiscussion with experts, criteria expressed in the literature (e.g., Morrow,1999; Tapsell, 1999, 2000; Tunstall, 1999; Dralsek, 2000; EnvironmentAgency, 2000, 2001) and information that can reasonably be thought todetermine social vulnerability, a restricted set of 10 domains have beenchosen The 1991 U.K Census provides nine domains and the 10th onefrom the index of multiple deprivation; the calculations of a domain areconstr ained from the avai lable source da ta (Tabl e 7.3)

ß 2007 by Taylor & Francis Group, LLC.

TABLE 7.3

Domains Used for Calculating the Index of Social Vulnerability

% of long-term illnesses to total ED population

Assumes all ill persons are equally dependent

on others and thus equally as vulnerable

None, all people with long-term illness are equally vulnerable

(1991)

their ability to respond effectively (Tunstall, 1999)

% of OAPs that are:

(a) ill, (b) >75 years, and (c) live alone

All OAPs are potentially vulnerable but being

>75 years, ill, or living alone will increase the vulnerability

(a)–(c) are given equal weighting (as cannot distinguish further variability) and multiplied together

Results standardized

Census (1991)

Dralsek, 2000)

Number of parents with children aged:

(a) 0–4 (b) 0–4 and 5–15 (c) 5–15

Different aged children may have different abilities and having more than one child may affect how easily parents can cope in a hazard situation.

Younger children are considered the most vulnerable, especially those that cannot walk

When added, (a)–(c) makes up number of single parents The weightings are applied

by multiplying (a)–(c)

by individual significance and totalling value; e.g., youngest children are

Census (1991)

TABLE 7.3 (continued)

Domains Used for Calculating the Index of Social Vulnerability

understanding of flood warnings or methods of response (Morrow, 1999)

% of non-U.K born to total ED population

Assumes all people born outside United Kingdom are equally vulnerable and just considers the proportion of the total population (Data does not identify particular English-speaking nations)

None, all people born outside the United Kingdom are potentially equally vulnerable

% of business properties to total

ED properties

Has a negative impact on social vulnerability due to smaller number

of residential properties

None, business properties have an equal effect on reducing social vulnerability

% is standardized and multiplies

consider negative impact

Census (1991)

Children

[ED]

As dependants are vulnerable themselves and increase the vulnerability of others (Morrow, 1999), they may not understand how to respond

% of children to total

ED population

Assumes all persons are all dependent on others and therefore equally vulnerable.

Further breakdown of ages is considered for single parents

None, all children are equally vulnerable (conflicts with method for calculating single parent vulnerability)

May inhibit flood recovery

as houses become damp and is also a further indication of poverty

% of ‘‘no central heating’’ to total ED properties

Assumes properties without central heating are equally vulnerable.

No data to indicate the contrary

None, all properties without central heating are equally vulnerable

% of ‘‘no car’’

properties to total

ED properties

Assumes properties without cars are equally vulnerable No data to indicate the contrary

None, all properties without cars are equally vulnerable

Income, employment, health and disability, education and training, access to service, and housing data are all used as separate domains

to create index at ward level

None, the deprivation index is already calculated

Values calculated from wards to EDs and standardized

DETR (2001) a

All domain values are standardized to a range of 0 to 100.

Calculated domains are combined to form an overall index with eration of how much of an impact each domain has on socioeconomicvulnerability Simply summing the different domains together does notallow for the fact that some domains may have more of an impact onvulnerability than the others The index of multiple deprivation (DETR,2001) combines different domains by attributing each domain a percentageweight according to importance To be consistent with the DETR study, thedomains are subjectively ranked, attributed a percentage weight (Table 7.4),and the value for each ED is adjusted accordingly Finally, the domainvalues are summed to create the index and adjusted to values between 0and 100 to create the ISV, which is calculated from the equation:

7.2.2 Index of Flood Probability

The desired index of flood probability (IFP) is a set of contours that definethe extent of different flood return-periods The functionality of ArcInfo GISrequires a continuous data surface to be able to calculate the contoursaccurately An interpolation method is thus required, which models theentire extent of a return-period based on actual flood-depth data recordedfrom historic events Flood-depth data and a digital elevation model (DEM;with resolution 50 m horizontal and 1 m vertical) are used as model inputdata to map the predicted extent of flooding caused by different floodreturn-periods and combining these maps produces an IFP The flooding

TABLE 7.4Weight Assigned to EachVulnerability Domain

process is simplified by an assumption that states: flood depth is reduced asthe elevation above the river increases The method requires the calculation

of the height above the river (HAR) for the whole of the floodplain.River height is calculated by assigning elevation values to each pointalong a digitized river-line using a DEM Converted to a raster grid, theriver appears as a series of cells with different elevations A cost-allocationcommand allocates cells in a new grid the elevation value of the nearestriver cell HAR is calculated by subtracting the height of river grid from theDEM (Figure 7.1)

Actual flood data used (Source: FHRC) consists of actual water-depths forfive flood return-periods (10, 25, 50, 100, and 200 years) recorded at thelocation of 10,000 properties in the study area A method was required tointerpolate flood depth by incorporating the sample data with the HARgrid Statistical interpolation techniques were considered (e.g., kriging andinverse distance weighting) but thought inappropriate due to the sporadicand clustered spatial distribution of the sample data Regression analysisattempted to correlate flood-depth and HAR as a model of flood extent butproduced inconclusive results owing to an inadequate resolution of DEM(Whittington, 2001)

An alternative approach that produces a satisfactory result uses the mum depth of flood for each scenario identified from the flood data andassumes that, given an isotropic surface, the flood depth would be the sameeverywhere By subtracting the maximum flood depth from HAR, anypositive values in the resulting grid indicate a flood The positive valuesare converted to 1 (for a flood) and negative to 0 (no flood), repeatingthe process for all scenarios The five scenario grids are literally added

maxi-0–1 m

No data 8–70 m 7–8 m 6–7 m 5–6 m 4–5 m 3–4 m 2–3 m 1–2 m

FIGURE 7.1 (See color insert following page 328.)

Modeled height above the river (HAR) (ß Crown Copyright= database right 2007 An Ord- nance Survey=EDINA supplied service.)

ß 2007 by Taylor & Francis Group, LLC.