Liton Chakraborty·Jason Thistlethwaite·Andrea Minano·Daniel Henstra·Daniel Scott

Abstract This study integrates novel data on 100-year f lood hazard extents,exposureof residential properties,and place-based social vulnerability to comprehensively assess and compare f lood risk between Indigenous communities living on 985 reserve lands and other Canadian communities across 3701 census subdivisions.National-scale exposure of residential properties to f luvial,pluvial,and coastal f looding was estimated at the 100-year return period.A social vulnerability index(SVI)was developed and included 49 variables from the national census that represent demographic,social,economic,cultural,and infrastructure/community indicatorsof vulnerability.Geographic information system-based bivariate choropleth mapping of the composite SVI scores and of f lood exposure of residential properties and population was completed to assess thespatial variation of f lood risk.Wefound that about 81%of the 985 Indigenous land reserves had some f lood exposure that impacted either population or residential properties.Our analysis indicates that residential propertylevel f lood exposure issimilar between non-Indigenousand Indigenous communities,but socioeconomic vulnerability is higher on reserve lands,which conf irms that the overall risk of Indigenous communities is higher.Findings suggest the need for more local verif ication of f lood risk in Indigenous communities to address uncertainty in national scale analysis.

Keywords Canada·Flood risk assessment·Indigenous communities·Social vulnerability

1 Introduction

Indigenous communities in Canada,including First Nations,Me´tis,and Inuit peoples living on-and off-reserve,are often identif ied as among the populations most socially vulnerable to climate change(Ford 2012).A growing body of research has shown disparities in socioeconomic status of Indigenous populations relative to benchmark populations and other Canadian communities(Hajizadeh et al.2018).Indigenous disadvantage and marginalization persist in part due to Canada’s colonial legacy and the intergenerational effects of residential schools(Reading and Wein 2009).But other factors,such as environmental and social injustice(Thompson 2015)and inequalities in education,employment,and income opportunities(Anderson et al.2016),are also implicated.Broader scholarship suggests that persistent social and economic inequalitiesexacerbate the social vulnerability of marginalized and socially deprived communities(Cutter 1995),which results in greater susceptibility of vulnerable groups to the impacts of environmental hazards(Cutter et al.2003;Andrey and Jones 2008).Vulnerability to hazards and disasters represents‘‘the degree of potential for loss;the propensity or predisposition to be adversely affected;or circumstances that put people at risk’’(Mavhura et al.2017,p.104).Research has revealed ways to locate and quantify socially vulnerable populations,which have been useful to design targeted risk reduction and awareness building strategies to reduce social vulnerability(Wisner et al.2004;Arias et al.2016;Drakes et al.2021;Tate et al.2021).

Waldram(1988)argued that Indigenous communities around the world are the most vulnerable to the impacts of f looding,becausethey havebeen forced to liveon marginal land and in remote locations to make room for settlers.Indeed,a few studies have documented that Canadian Indigenous communities bear signif icant f inancial,psychological,and social burdens associated with f looding,and they have been disproportionately affected by f loodrelated displacement(Thompson et al.2014;Martin et al.2017).Moreover,these communities are likely to face greater f lood exposurein achanging climate(McNeill et al.2017;Ford et al.2018;Khalafzai et al.2019).To date,however,there has been no comprehensive,national-scale assessment of f lood risk to Indigenous communities in Canada that combines hazard and exposure data with a measure of social vulnerability(Patrick 2017;Fayazi et al.2020).This is unfortunate,because measuring social vulnerability to f looding provides valuable information to support Indigenous emergency management planning,especially under accelerating climate change.Moreover,the geospatial assessment of f lood risk provides foundational inputs for the evaluation of f lood risk management(FRM)strategies(Koks et al.2015)and informs culturally appropriate climate change adaptations(McNeeley and Lazrus 2014).

This study is the f irst to assess place-based social vulnerability and f lood exposure for Canadian Indigenous populationsliving on reserves,including landsclassif ied as‘‘Indian reserves,’’‘‘Indian settlements,’’and‘unorganized territories’’across 3701 census subdivisions(CSDs).1‘Residence on or off reserve refers to whether the person’s usual place of residence is in a CSD that is def ined as‘on reserve’or‘off reserve’’’(Statistics Canada 2018,p.18).The purpose is to assess the extents of f lood risk to Indigenous communities and to compare this risk to non-Indigenous communities at the national scale.Furthermore,the study contributes to knowledge about socioeconomic factors that contribute to f lood risk among First Nations,Me´tis,and Inuit peoples living on-reserve.This knowledge is critical to identifying socially vulnerable neighborhoods where scarce FRM resources are needed most,and to inform a socially equitable approach to FRM policy(Sayers et al.2018).

This study considered f lood risk to be a product of exposure to f lood hazards,combined with social vulnerability to their impacts as advocated by Byers et al.(2018).It analyzed social vulnerability at the Indigenous community level,where communities were def ined based on the boundaries of CSDsin the 2016 censusadministered by Statistics Canada(2019a).Our social vulnerability index(SVI)was measured by quantifying population subgroups within CSDs that have lower socioeconomic status before f loods occur,based on an index of 49 variables.Flood exposure analysis captured the percentage of the population and number of residential properties within a CSD exposed to f luvial,pluvial,or coastal f looding at the 100-year recurrence interval.Geographic information system(GIS)-based bivariate choropleth mapping indicated spatial hotspots of f lood risk and delineated locationspecif ic f lood disadvantage by revealing neighborhoods that face the greatest f lood risk;that is,where high social vulnerability coincides with f lood exposure.

2 Data and Methods

This study utilized national datasets of f lood hazards,residential properties,census of population,and off icial community boundaries at the CSD level to identify people and property exposed to f looding across Canada.Table 1 presents the datasets used in this study to identity hazards,estimate exposure,construct the SVImeasures,and delineate f lood risk at the CSD level.

Table 1 Social and f lood datasets used to assess f lood risk to Indigenous communities in Canada

2.1 Indigenous Peoples and Communities

Census subdivision is the term used by Statistics Canada to reference‘‘municipalities(as determined by provincial/territorial legislation)or areas treated as municipal equivalents for statistical purposes(e.g.,Indian reserves,Indian settlements,and unorganized territories)’’ (Statistics Canada 2018,p.81).Statistics Canada records a total of 5162 CSDs and categorizes them into 53 types,such as city,rural community,and Indigenous reserve,based on off icial designations adopted by federal,provincial,and territorial authorities.In Canada,‘‘Indigenous peoples’’refers collectively to the original inhabitants of North America and their descendants,who are categorized into three groups of Aboriginal peoples,including Indians(or First Nations),Inuit,and Me´tis(Crown-Indigenous Relations and Northern Affairs Canada 2017).As the purpose of this study was to estimate f lood risk to Indigenous communities and to compare this risk to non-Indigenous communities,the CSDs representing Indigenous communities included‘‘on-reserve’’populations legally aff iliated with First Nationsor Indian bands,living in any of six CSD categories,including Indian reserve/Re´serve indienne(IRI),Indian settlement(S-E´),Indian government district(IGD),Terres re´serve´es aux Cris(TC),Terres re´serve´es aux Naskapis(TK),and Nisga’a land(NL)(Fig.1).

Fig.1 Flood-prone areas in Indigenous reserves as determined by JBA Risk Management(2020).Source The 2016 census subdivisions cartographic boundary from Statistics Canada;and the 2018 f lood hazard data from JBA Risk Management

The study included f lood exposure analysis of population and residential properties for all 985 Indigenous reserve lands of the six CSD types from the total 5162 CSDs.Due to Statistics Canada’s conf identiality regulations concerning geographic and population thresholds for statistical output vetting,the SVIscores were available for only 360 of the 985 Indigenous reserves.Considering a national scale socioeconomic vulnerability assessment of on-reserve Indigenous communities,the unavailability of microdata on remaining reserves may underestimate or overestimate SVI scores.Hence,the results of relative vulnerability scores across Indigenous reserves should be interpreted carefully.

2.2 Flood Hazard Areas

Thestudy analyzed undefended,f lood-proneareasfor three types of f lood hazards:f luvial(riverine and ice-jam overall),pluvial(intense precipitation-caused inundation of lands that are not necessarily proximate to a body of water),and coastal(storm surge),as determined by JBA Risk Management,a global,market-leading f lood catastrophe modeling f irm.JBA Risk Management(JBA)’s 2018 f lood hazard datasets(that is,Canada Flood Maps at 30-meter horizontal resolution)were made available through a research partnership with the University of Waterloo.These Canada Flood Maps,themost widely used in the Canadian(re)insurancemarket,arenational in scope,so they enable f lood hazard assessment at any location in Canada(JBA Risk Management 2020).

Indigenous communities are located across the country,which justif ies the use of JBA’s datasets since JBA possesses the only map resource available at a national scale that isindependent of political,jurisdictional,or land-claim considerations.Although a national scale dataset is required to ensure coverage of all Canadian Indigenous communities,it is important to note several limitations.Flood inundation modeling in Canada at a national scale involves tremendous uncertainty given the dynamic range of geography,topography,and available data inputs.Most Canadian f lood hazard assessments are conducted at the provincial,local,or watershed scale where more precise measurement is possible(Faulkner et al.2016).For example,several uncertainties in the f lood model used for the study’s maps limit the accuracy of local scale determinations of f lood risk.These uncertainties include a 30-meter resolution,a lack of data on pluvial f lood defences(for example,storm sewer capacity),and insuff icient gauge and HYDAT(National Water Data Archive-Canada)data proximate to Indigenous communities.The latter uncertainty is a particularly important limitation since Canadahasmany lakesthat remain ungauged relative to other countries and Indigenous communities are often located near these lakes.For this reason,the use of these data is considered an initial assessment that with further local validation can be used alongside socioeconomic vulnerability to improve f lood risk assessment of Indigenous communities.

The geospatial analysis in this study focused on the 100-year return period—a f lood that has a 1-in-100(1%)probability of occurring in any given year—and assumed no defences for f luvial f looding or coastal f looding.The 100-year f lood is the generally accepted regulatory standard for most of Canada,which is also commonly used in f lood hazard research and policy documents(Burton and Cutter 2008;Ludy and Kondolf 2012;Burn et al.2016).Pluvial f looding was modeled using a direct rainfall method whereby Canadian intensity-duration-frequency curves were used to estimate rainfall and then applied to model cells before runoff was calculated using a hydraulic model(Hall 2015).

JBA’s f lood hazard extent datasets(in raster GIS f ile format)were f irst imported into ArcMap 10.7.1 to visualize f lood-prone areas(Fig.1).Statistics Canada’sspatial layers of the 2016 CSD-level boundaries(in polygon shapef ile)were added to visualize the hazard and exposure,and to generatef lood risk mapsfor Indigenousand other Canadian communities measured at the CSD level.Figure 1 shows spatial delineation of f lood-prone areas in Indigenous reserve lands(for example,Nisga’s NL,Mistissini TC,and Blood 148 IRI)that are subject to f luvial,pluvial,and coastal f lood hazards.

2.3 Calculating Flood Exposure of Population and Residential Properties

Physical exposure refersto thepeopleand assets,including residential properties and critical infrastructure,that are likely to be affected by a hazard(UNDP 2004).Flood exposure is typically measured by identifying and quantifying the people and assets that would be affected by a f lood of a particular size,such as the 100-year f lood(Holmes and Dinicola 2010).Following the methods of Qiang(2019),our study estimated f lood exposure using population and residential properties,which involved three phases.First,we calculated the total number of people and residential properties within each dissemination block(DB),the smallest geographic unit used by Statistics Canada to disseminate population and dwelling counts,which refers to‘‘an area bounded on all sides by roads and/or boundaries of standard geographic areas’’(Statistics Canada 2018,p.90).We then aggregated the DB-level totals to the CSD level,and f inally,calculated the percentage of the population and number of residential properties(asdescribed in Eqs.1 and 2)and spatially combined them with the 100-year f lood hazard extents at the CSD level.To calculate the total number of residential properties located within each DB,Statistics Canada’s 2016 DB boundary data were spatially joined with the national address points dataset on‘‘residential properties(count)’’(spatial layer in point shapef ile)from DMTI Spatial Inc.,Canada’s leader with renowned expertise in location analytics and intelligence solutions(DMTISpatial Inc.2018).Using the 2016 census,each DB was also paired with its respective population count.The points dataset contained about 15.9 million addresses in Canada,including industrial,commercial,and residential addresses,but only the approximately 11 million residential address points were included as part of this analysis.A 15-meter buffer was generated for all the residential properties in the absence of building footprint data for residential property point locations.Using the output buffer polygons of residential properties,a binary analysis(that is,yes=1;no=0)was used to indicate whether properties intersected with the f lood hazard extent.

The roughly 11 million residential address points were spatially joined with their respective DBs,and these were aggregated to the CSD level.An address point represented a single unit(for example,apartment,unit),so there were cases in which multiple addresses were present in the same geographic location(for example,a condominium building).The majority of the CSDs had at least one residential property,but 1730 of the 5162 CSDs did not intersect with any residential addresses.We calculated the total number of residential properties on Indigenous reserve lands(that is,24,020 properties located among 260 reserve lands)by creating a binary f lag to distinguish the CSDs that were recorded by Statistics Canada as reserves(that is,the six Indigenous CSD types noted above)with a‘‘1’’and nonreserves(that is,the other 47 CSD types)with a‘‘0.’’

Theanalysisdetermined that 4869 of the5162 populated CSDs were exposed to the 100-year f lood hazard.To estimatethe exposed population,we multiplied theexposed land area by the population density(that is,number of people per square kilometer)for each DB and then aggregated that population count at the CSD level.The study determined that about 5.2 million of the total 35.2 million population was exposed to f lood hazard across 4702 CSDs,compared to 56,646 of the total 382,587 population exposed across 787 Indigenous reserve lands.

It is important to note some of the uncertainty associated with thisexposure analysisdueto limitationsin the dataset.Exposure is likely to be overestimated in some areas since the hazard is measured at 30 meters and buildings without footprints were assigned 15-meter buffers.As a result,the assessment captures more property within f lood risk zones than with maps using a higher resolution(for example,5 meter).The decision to include these properties was based on the need to take a conservative and comprehensive assessment of exposure.Recent analysisin Halifax,Canada did conf irm that the 30-meter resolution 100-year f lood map aligned with local maps,but a more rigorous validation is necessary using multiple communities(Thistlethwaite et al.2018).Furthermore,some overestimation occurred because the exposure data do not differentiate whether an address is on the ground f loor of multi-unit buildings.As a result,all addressesin these buildings were included even if they were on different f loors to ensure all potential exposure was captured.

2.4 Measuring Social Vulnerability

Thestudy used microdatafrom the2016 censusto populate an index of 49 context-specif ic variables that have been used frequently in previous empirical studies to measure social vulnerability to f loods and other hazards(Cutter et al.2003;Messer et al.2006;Oulahen,Mortsch et al.2015;Oulahen,Shrubsole et al.2015;Fatemi et al.2017).These variables,which were extracted at the CSD level,represented diverse aspects of socioeconomic,demographic,ethnic,and cultural characteristics(Table 2).Before constructing the index,we removed CSDs that did not comply with Statistics Canada’s census data analysis guidelines,statistical output vetting rules(for example,conf idential homogeneity rule and dominance rule for dollar value variables such as household income and owner estimated home values),and a geographical requirement(for example,CSDs containing less than 250 population and 40 households).

The study used Principal Components Analysis(PCA)to construct the SVI,following the‘‘hazard-of-place’’model to measure neighborhood-level relative social vulnerability(Cutter et al.2003).We also used PCA diagnostics tests such as the Kaiser-Meyer-Olkin(KMO)measure of sampling adequacy(Kaiser 1974),Bartlett’s Test of Sphericity(Bartlett 1954),and Cronbach’s alpha(α)coeff icient(Cronbach 1951)to examine multicollineraity problems,validity,and reliability of the dataset.As our data passed all three diagnostics tests,we applied PCA with varimax rotation to all standardized variables for extracting orthogonal componments.The number of components were then selected using Cattell’s(1966)scree plot(Fig.2).

Fig.2 Scree plot of eigenvalues of components.CI conf idence interval

A directional adjustment(or cardinality)was applied to the components that had positive loading scores on the indicators of income and wealth(such as median per capita income of component 1 and median home value of component 3 in Table 3).Theoretically,a wealth indicator decreases vulnerability(negative cardinality),so the component scoreson that dimension wereinverted(Cutter et al.2003;Cutter et al.2013).The component scores were then

Table 3 Principal Components Analysis results:Rotated component matrix

?

?

weighted and added to generateanon-standardized SVIfor each CSDj(NSSIj),where weights were calculated as follows:

Since the values of theNSSIjindex can be negative or positive,making it diff icult to interpret and compare the scores by CSDs,a standardized SVI was developed.The values of SVIranged on a scale of 0 to 100,and they were calculated using the following formula for each CSDj(Chakraborty et al.2020):

The SVIscores were reversed for ease of interpretation and comparison between CSD-level communities,such that a CSD with higher SVI score represented a better socioeconomic status and lower social vulnerability to environmental hazards(Chan et al.2015).The spatial distribution of social vulnerability was visualized through GIS-based choropleth mapping,by joining the SVIscores to the 2016 CSD level cartographic boundary.

2.5 Assessing Flood Risk

The study assessed f lood risk at the CSD level for Indigenous and other communities across Canada by identifying the spatial intersection of the 100-year f lood hazard extents,exposure of residential properties and population,and social vulnerability.This method is more comprehensive than those typically applied in the hazard and disaster risk management literature(Albano et al.2017;Armenakis et al.2017).By following a GIS-based bivariate choropleth mapping technique(Frigerio et al.2016),we developed a f lood risk assessment matrix(Fig.3).

The matrix depicts a spatial relationship between social vulnerability(that is,proportion of CSD-level inverted SVI scores)and f lood exposure of residential properties and population.To represent the spatial distribution of f lood risk,the inverted SVI scores,exposure of residential properties(in percentage),and exposure of population(in percentage)were classif ied into f ive categories,including Very Low,Low,Moderate,High,and Very High.The SVI scores were divided using the Jenks natural breaks classif ication method,which is also known as the goodness of variance f it.By minimizing the squared deviations of the class means,the method seeks to reduce the variance within classes and maximize the variance between them(Jenks 1967).The exposure variables were classif ied using the‘‘quantile(equal count)’’scheme.Finally,f lood risk maps for all CSDs were generated to identify‘‘hotspots’’across Canadian provinces,meaning those CSDs with both a high level of social vulnerability and high exposure of residential properties and population to 100-year f lood hazards.

3 Results

A statistical summary of GIS-based f lood exposure results was reported in Sect.3.1,whereas PCA-based statistical summary results of SVIindicators and geographic patterns of social vulnerability were evaluated and compared by communities in Sect.3.2.Finally,the results of social vulnerability and f lood exposure of residential properties and population were spatially combined at the CSD level using GIS-based classif ication(described in Fig.3)to comprehensively assess f lood risk in Sect.3.3.

Fig.3 Flood risk matrixcomparing social vulnerability and f lood exposure extents

3.1 Flood Exposure

The study found that 80.7%of the 985 Indigenous reserves have some f lood exposure in terms of either population or residential properties at the 100-year return period.Indigenous communities in the 985 reserve CSDs have a slightly higher percentage of population exposed to 100-year f lood hazards(14.8%)than other Canadian communities(14.7%).In terms of population exposure,we found that almost 98.3%of the 809 populated Indigenous reserve CSDs were exposed to some of the three types of f lood hazards.The study also estimated 66.5%of the total 260 Indigenous reserve CSDs have some f lood exposure of residential properties.

At the provincial scale,Indigenous peoples in Prince Edward Island(PE)have the highest f lood exposure(Fig.4).Manitoba has the highest percentage of totalpopulation(49.8% of the total 1,278,365)exposed to 100-year f lood hazards.Manitoba also has the highest proportion of non-Indigenous population(51.8%)exposed to f looding,but this includes all of its largest city,Winnipeg,which is protected by the Red River Floodway,a 47-kilometer diversion channel that was built to accommodate 100-year riverine f looding.British Columbia has the second highest percentage of population exposed to f looding,in both cases of Indigenous and non-Indigenous communities.

Fig.4 Population at-risk of f lood hazards by province/territory in Canada.AB Alberta,BC British Columbia,MB Manitoba,NB New Brunswick,NL Newfoundland and Labrador,NS Nova Scotia,NT Northwest Territories,NU Nunavut,ON Ontario,PE Prince Edward Island,QC Quebec,SK Saskatchewan,YT Yukon

The analysis revealed that residential f lood hazard exposure is similar between Indigenous communities(21.5%of 24,020 properties)and other Canadian communities(19.1%of 11,025,746 properties).Comparing provinces,more than 80%of on-reserve residential properties in Alberta are exposed to f looding,which is the highest proportion among all Canadian provinces and territories(Fig.5).Manitoba has the highest percentage of total residential properties(62.7%of the total 323,926)exposed to f looding.Although Prince Edward Island has the lowest percentage(11.7%of the total 30,210)of its residential properties exposed to f looding across Canada,it has the second highest percentage of residential properties(41.9%of the total 31)exposed to f looding considering only Indigenous reserve communities.

Fig.5 Residential properties at-risk of f lood hazards by province/territory in Canada.AB Alberta,BC British Columbia,MB Manitoba,NB New Brunswick,NL Newfoundland and Labrador,NS Nova Scotia,NT Northwest Territories,NU Nunavut,ON Ontario,PE Prince Edward Island,QC Quebec,SK Saskatchewan,YT Yukon

To visualize spatial variation of f lood exposure of residential properties through geospatial mapping,we categorized the total number of residential properties exposed to f luvial,pluvial,and coastal f lood hazards into f ive categories based on an‘‘equal count quantile’’classif ication scheme.These categories included Very Low(less than 3.1%of residential properties exposed at the CSD level),Low(between 3.1 and 8.3%),Moderate(between 8.4 and 14.1%),High(between 14.2 and 22.2%),and Very High(more than 22.2%).The same‘‘equal count quantile’’classif ication scheme was applied for mapping population exposure to f looding,and the total population exposed to 100-year f lood hazards was divided into f ive categories,including:Very Low(less than 4.7%),Low(between 4.7 and 8.4%),Moderate(between 8.5 and 11.8%),High(between 11.9 and 16.5%),and Very High(more than 16.5%).

3.2 Social Vulnerability Index Indicators

Principal Components Analysis with orthogonal varimax rotation and eigenvalues of greater than 1 have identif ied ten multidimensional componentsbased on our data.These components explain 72.9%of the total variation.Table 3 reports component loading scores on individual variables.Findings suggest that 33 of the 49 sociodemographic indicators represent close membership of social vulnerability.

The f irst component accounts for 18.1%of the total variation in the data,where the proportion of population with Aboriginal Peoples(ZPABORIGIN),North American Indian/Inuit/Me´tis(ZPINDINUTM～S),and median per capita income of census family(ZPERCAPINC)show positive loadings and the proportion of white population(ZPWHITE)show negative loading scores.Hence,the f irst component represents a combination of cultural and economic factors with racial/ethnic characteristics of the Canadian population,including Indigenous peoples and communities,which is a strong indicator of social vulnerability.Variable loading scores in the second component,with 9.6%variation in the data,represent both demographic and economic factors of social vulnerability,whereas the third component,with 6.8%variation in the data,represents a combination of social and economic factors of vulnerability.

We interpret each component listed in Table 3 based on the SVIindicators and factorsof social vulnerability group membership as described in Table 2.For example,component seven can be interpreted as a pure demographic factor,describing the SVI indictors of a special needs population;component eight as a mixed factor of cultural and infrastructural/community membership,listing racial/ethnic and built environment characteristics of populations;and component nine as a pure cultural factor,reporting racial/ethnic indicators of vulnerability and so on.Overall,we f ind that the components of SVI are multidimensional in Canada.

3.2.1 Patterns of SVI

By joining the SVI scores with the 2016 census subdivisions and provincial/territorial cartographic boundary f iles,we produced a GIS-based choropleth map to visualize national-scale socioeconomic disparities and the extents of social vulnerability across CSDs.Using the equal count(quantile)classif ication method,we divided the SVIscores into f ive categories,including Very Low,Low,Moderate,High,and Very High.We used graduated classif ication along with spectral color ramping to display spatial patterns of social vulnerability in an inverted color ramp such that a higher SVIscore indicates lower social vulnerability in a CSD(as indicated in the map legend of Fig.6).For more visual clarity,we mapped social vulnerability and f lood exposure,and combined these variables to produce f lood risk maps for three adjacent provinces in Canada.

By ranking and comparing mean SVI scores,we determined that Indigenous peoples living on reserve lands are more socially vulnerable than inhabitants of other Canadian communities(Fig.7;Table 4).This difference is ref lected vialower mean SVIscoresfor Indigenousreserve communities in all Canadian provinces and territories,except for Yukon and Nunavut territories in which SVI scores are unavailable(Fig.7).The SVI scores for all CSDs were also ranked according to their level of social vulnerability,and Table 4 displays the 10 most socially vulnerable CSDs.The ranking value of‘‘1’’in Table 4 suggests highest vulnerability(that is,lowest SVI score)for a CSD.

Fig.7 Comparison of the mean SVI scores by communities(SES=socioeconomic status).AB Alberta,BC British Columbia,MB Manitoba,NB New Brunswick,NL Newfoundland and Labrador,NS Nova Scotia,NT Northwest Territories,NU Nunavut,ON Ontario,PE Prince Edward Island,QC Quebec,SK Saskatchewan,YT Yukon

3.3 Combining Social Vulnerability and Flood Exposure to Assess Flood Risk

We combined high f lood exposure of residential properties and population with the high vulnerability of the human environment,resulting in a critical assessment of the placebased‘‘social risk’’of f looding(Roder et al.2017).The most socially vulnerable areas(that is,red colored geographical areas in Fig.6)were spatially combined with the areas with Very High f lood exposure of residential properties(that is,red colored geographical areas in Fig.8a)and Very High f lood exposure of population(that is,red colored geographical areas in Fig.8b)to delineate hotspots of f lood risk across Indigenous communities and other Canadian communities(for example,dark red and black colored geographical areas in Fig.8c,d).Due to limited space,we created f lood risk maps for three adjoining Canadian provinces only(Ontario,Manitoba,and Saskatchewan).This GIS-based bivariate choropleth mapping procedure permitted us to assess f lood risk by visualizing the relationship between social vulnerability and f lood hazard exposure across 3701 CSDs.The spatial integration method also allowed for detection of the hotspots of f lood risk in Canada(Table 5)—CSDs with elevated 100-year f lood hazard exposure and elevated social vulnerability—which highlights the distribution of the highest-risk CSDs and provinces/territories.

Fig.6 Spatial distribution of the Social Vulnerability Index at the census subdivisions level.Source The 2016 census of population microdata;Census subdivisions cartographic boundary;Statistics Canada

Fig.8 Delineation of f lood risk maps for Manitoba,Ontario,and Saskatchewan.Source The 2016 census of population microdata;Census subdivisions cartographic boundary;Statistics Canada;The 2018 Canadaf lood map and pricing data;JBA Risk Management;The 2018 DMTI CanMap Route Logistics;DMTI Spatial Inc

Table 4 The ten most socially vulnerable communities

The study found that 41 of 3701 CSDsare located in the range of High to Very High risk of f looding(Table 5).There are only f ive CSDs in areas with Very High f lood risk(with 20.8–29.4%of population and 30.9–66.7%of residential properties exposed to f looding,and a score of 7.6–18.4 on the SVI).Most importantly,all high-risk CSDs are located in Indian reserve/Re´serve indienne(IRI)areas except the Carmacks village(VL)in the Yukon.At the national scale,40 hotspot areas among the 360 Indigenous reserves are geographically stratif ied across eight provinces,including Alberta,British Columbia,Manitoba,New Brunswick,Nova Scotia,Ontario,Prince Edward Island,and Quebec.Among all provinces,the highest number of hotspots were found in British Columbia(13 IRI)and in Ontario(10 IRI).Although f lood risk and its extentsappear to be distributed unequally in termsof f lood exposure and social vulnerability at the national scale,no hotspotswere located in Saskatchewan,Newfoundland and Labrador,the Northwest Territories,and Nunavut.Overall,f lood exposure is similar between non-Indigenous and Indigenous communities,but social vulnerability is higher on reserve lands,which conf irms that the overall risk of Indigenous communities is higher.

Table 5 Hotspots of f lood risk in Manitoba,Alberta,Nova Scotia,Quebec,and British Columbia

4 Implications for Flood Risk Management Policy in Canada

The results of this study emphasize that both f lood exposure and social vulnerability have considerable spatial variation,which contributes to a heterogenous pattern of f lood risk across Canada.Our f indings show that Indigenous peoples living on reserve face a higher f lood risk than the general population,due to higher social vulnerability.This suggests that policymakers must consider the spatial pattern of socioeconomic vulnerability in the design of f lood risk management(FRM)strategies to optimize the allocation of scarce resources(Koks et al.2015).Manyscholars have argued that development of comprehensive risk maps by combining measures of hazard exposure and social vulnerability is an essential step towards eff icient and effective resource allocation,which will ensure f lood mitigation measures are implemented where they are needed most(Tapsell et al.2010).

The SVI developed in this study can be used to locate geographical areas with high concentrations of socially vulnerable groups and their relevant socioeconomic,demographic,cultural,and built environment characteristics.This method can further guide f lood risk communication to spur household preparedness,risk awareness,and risk reduction(Cutter et al.2013).Creating a knowledge base of place-based social vulnerability would be a valuable policy and planning tool to increase f lood risk awareness and f lood preparedness programs,because those who areaware of therisksaremorelikely to invest in f lood mitigation measures(Kuhlicke et al.2011).By using the SVI,FRM policiescould betailored to enhancecommunity resilience that ultimately increases people’s capacity to resist,cope with,and recover from the impacts of f looding(Thieken et al.2007).

Tailored f lood risk reduction policies based on the SVI offer greater fairness in line with principles of environmental justice,and can guide practical and economically eff icient investments in integrated FRM(Sayers et al.2018).Although our f indings,along with the risk mapping approach developed in the current study,have identif ied high-risk communities and the most socially vulnerable places across Canada,it is critically important to adopt a portfolio of policy instruments,such as f lood mitigation measures,regulations,economic incentives,and comprehensive geospatial data analysis,in order to reduce f lood risk to Indigenous peoples(Hegger et al.2018).

5 Discussion and Conclusion

The aim of this study was to explore the attributes and spatial patterns of social vulnerability and f lood exposure for on-reserve Indigenous communities versus other Canadian communities at the municipal(CSD)scale in order to support community-based f lood risk analysis.The study conf irmed that using novel data about the f lood risk exposure and vulnerability of Indigenous communities can be measured at a national scale in Canada.This f inding justif ies a focus on Indigenous communities in Canada’s approach to FRM and providesamethod for assessing how resources should be prioritized according to risk.

Our f indings indicate that the factors inf luencing social vulnerability in Canadian communitiesincludeattributesof race and ethnicity,income,built environment,elderly populations,education,occupation,family structure,and access to resources.Our f indings are congruent with the conclusion of previous studies,government reports,and brief ings,and aff irm community-based case studies on f lood risk and exposure,which show Indigenous people living in vulnerable communities experience greater environmental risk(Thompson 2015;McNeill et al.2017).This study also f inds an unequal spatial pattern of social vulnerability and f lood exposure across communities,which contributes to systemic f lood disadvantages,the heterogenous nature of f lood risk,and the differential impact of f looding on Canadian society as a whole.The spatial analysis method developed here and the results that it produced are useful for systemic spatial planning,identify suitable f lood mitigation measures,and facilitate community-level risk f inancing(Linnerooth-Bayer and Hochrainer-Stigler 2015).

Thef lood risk analysispresented in thisarticlehasafew limitations that should be acknowledged.First,the study used a f lood hazards dataset from JBA Risk Management because it is the only national-scale f lood hazards dataset available for Canada.There is bound to be tremendous uncertainty in f lood exposure analysis,particularly when scaling to local levels.A potential disagreement between the f lood risk exposure identif ied through the JBA model and f lood exposures measured through other means or based on a local dataset could justify more future research.Theexposureanalysisdid not include estimatesof physical vulnerabilities consistent with actual f lood damage and depth analysis.As a result,some sites of cultural signif icance without physical structures in Indigenous communities are excluded.

Second,the study was unable to‘‘ground-truth’’exposure and vulnerability due to a lack of pre-event and postevent data.Absence of f ield work limited access to local information related to exposure and vulnerability that might be collected only through site visits and qualitative survey methods(Oulahen,Shrubsole et al.2015).It is particularly important to address data gaps for ungauged waterbodies.Thisgap contributes to uncertainty in the JBA dataset assumptions given the lack of a developed methodology for measuring how ungauged lakescontribute to 100-year f lood plain predictions.For this reason,an important recommendation for future work involves improving assessment of f lood risk associated with ungauged lakes near Indigenous communities,and how proximity to these lakesinf luencesexposure.Wehave also explored the feasibility of mapping and measuring Indigenous proximity to lakes.Due to time and the additional analysis required,this area warrants a future paper focusing on uncertainty associated with Canada’s nationalscale f lood models.

Finally,JBA’s Canada Flood Map hydrology datasets are based on historical data that do not incorporate future climate change projections.Further research should assess future f lood risk by integrating climate change scenariosin f lood exposure analysis and climate change vulnerability index development(Debortoli et al.2019).Moreover,developing community-based f lood mitigation strategies that incorporate socioeconomic and cultural considerations should bethe key agendafor future research,and becomea means to promote fairness and social equity in f lood risk management.

AcknowledgementsSocial vulnerability analysis as presented in this article was conducted at the South-Western Ontario Research Data Centre(SWORDC)–a part of the Canadian Research Data Centre Network(CRDCN).The services and activities of SWORDC are made possible by the f inancial or in-kind support of the Social Sciences and Humanities Research Council of Canada,the Canadian Institutes of Health Research,the Canadian Foundation for Innovation,Statistics Canada,and the University of Waterloo.The views expressed in this article do not necessarily represent those of the CRDCN or its partners.Flood hazard data©JBA Risk Management 2018.Residential housing stock data©DMTI Spatial Inc.2018.

Open AccessThis article is licensed under a Creative Commons Attribution 4.0 International License,which permits use,sharing,adaptation,distribution and reproduction in any medium or format,as long as you give appropriate credit to the original author(s)and the source,provide a link to the Creative Commons licence,and indicate if changes were made.The images or other third party material in this article are included in the article’s Creative Commons licence,unless indicated otherwise in a credit line to the material.If material is not included in the article’s Creative Commonslicence and your intended use is not permitted by statutory regulation or exceeds the permitted use,you will need to obtain permission directly from the copyright holder.To view a copy of this licence,visit http://creativecommons.org/licenses/by/4.0/.