Managed Retreat As A Response To Natural Hazard Risk

NATURE CLIMATE CHANGE DOI: 10.1038/NCLIMATE3252ARTICLESManaged retreat is spatially and economically different frommany other risk management measures. The benefits of protectionand accommodation measures largely accrue where they areimplemented, whereas retreat in one location can benefit otherlinked, exposed areas, such as in the Netherlands (SupplementaryTable 2, Example 6)17 . Unlike engineering measures with ongoingmaintenance costs, retreat once implemented involves minimalrecurring financial costs while permanently reducing natural hazardrisk. The timescale and discount rates used in cost–benefit analysistherefore affect the ranking of coastal management measures;one study found that retreat tends to be favoured over timescalesgreater than 25 years18 .Documenting recent experiences with managed retreatTo compile a comprehensive database of recent efforts to implementmanaged retreat, we searched for ‘managed retreat,’ ‘communityrelocation,’ ‘climate displacement,’ ‘island abandonment,’ and‘planned resettlement’ in the peer-reviewed literature and inreports from governments, development agencies, and researchorganizations. Our definition of managed retreat (see Methods)excludes resettlement driven by mining, dams, or generaldevelopment objectives. The definition also requires at leasttwo parties, an implementing or enabling party and the residentsaffected by the intervention because they are relocating or becausetheir assets are moved or altered in land abandonment. We includeany instance in which one party initiates retreat, regardless ofwhether retreat has taken place. Only sufficiently documentedexamples could be analysed, so smaller-scale and developingcountry examples especially may have been missed.The 27 cases identified, described in Supplementary Table 2,capture the large majority of well-documented managed-retreat examples involving two or more parties. They originate in 22 countriesspanning all major world regions, occur in pre- and post-disastersettings, and address tropical storms, flooding, erosion, earthquakesand tsunamis. The cases are either one-time interventions or programs that have conducted multiple interventions.This diversity of applications includes several distinct clustersof interventions, such as numerous examples of post-disastermandatory relocations and locally driven relocation efforts. Tounderstand what drove these discrepancies, we sought to identifycriteria that could map the full range of cases while reflectingfundamental differences among clusters. Data on many potentialdistinguishing factors were assembled to enable systematic crosscase comparison19 . Our analysis of these various factors revealedthat managed-retreat processes were most fundamentally shaped bythe relationship and interactions of the two parties involved.Conceptual model developmentWe use the two parties’ motivations to structure a conceptual modelencompassing documented experiences with managed retreat. Themodel serves as a point of departure for unpacking case-specificcomplexities. By enabling comparison among vastly differentapplications, it lays the groundwork for deeper investigation of thevarious factors shaping the process of managed retreat, such aspower dynamics between parties.In the conceptual model (Fig. 1), the residents’ initial willingnessto move is reflected in the horizontal axis. Either the residents(those who would be moving) initiate the move, or they do not.On the right-hand side, residents initiate and thus support themove; on the left-hand side, residents do not initiate retreat. Thevertical axis encompasses the implementing party’s motivationto support managed retreat. The implementing party, often agovernment entity, represents a broader group of constituents.It is therefore more likely to support retreat when benefits areperceived for that group. The broader group’s benefits may includereduced expenditure on disaster relief, improved environmental2Broader societybenefits fromrelocationGreaterGoodMutualAgreementResidents donot initiatethe moveResidentsinitiate themoveHunkeredDownSelfRelianceOnly residentsbenefit fromrelocationFigure 1 Conceptual model of managed retreat. The horizontal andvertical axes reflect the perspectives of the residents and implementingparty, respectively.protection, or reduced exposure to natural hazards. Because thevertical axis is defined in our model from the perspective ofthe implementing party, it does not necessarily align with theperspective of constituents who might or might not move. Forexample, an implementing party may define broader benefits on thebasis of environmental justice, or in a way that does not accountfor the importance of a sense of home. The positions of bothparties depend on their values and perceived risk levels. Accordingly,different values and misperceptions can affect where each partylocates within the model and contribute to misaligned perspectives.Both axes are treated as gradients rather than discrete categories toreflect the nuance and complexity associated with each case.These two axes create four quadrants. In the top-right quadrant,Mutual Agreement, residents initiate retreat and the implementingparty likely supports it. In the top-left quadrant, residents donot initiate the move, but broader society would benefit, so theimplementing party is motivated to support retreat. Interventionsin this Greater Good quadrant often resemble exercises of eminentdomain, similar to dam-related resettlements, and may requiresubstantial incentives to persuade residents to relocate. In thebottom-left quadrant, Hunkered Down, residents do not initiateretreat, and broader society benefits little. Managed retreat mayoccur in such a situation due to other motivating factors, differencesbetween the two parties’ valuations of costs and benefits, ormisperceptions by either party. Finally, in the bottom-right SelfReliance quadrant, residents support managed retreat, but theimplementing party has little incentive to do so.Understanding recent experiences with managed retreatRecent experiences demonstrate how each party’s motivations shapethe processes and outcomes of managed retreat. Each case or clusterof cases is mapped onto the conceptual model in Fig. 2, indicatinghow the two parties’ perspectives vary across the landscape ofexamples. Supplementary Table 2 provides details for each case.Post-disaster voluntary relocation programs fall in the MutualAgreement quadrant because both parties are likely to supportimplementation of retreat. For example, the US Federal EmergencyManagement Agency (FEMA) finances property buyouts throughthe Hazard Mitigation Grant Program (HMGP) (SupplementaryTable 2, Example 1). After a disaster, owners of high-risk propertiesare offered buyouts at the house’s pre-disaster market value, andNATURE CLIMATE CHANGE ADVANCE ONLINE PUBLICATION www.nature.com/natureclimatechange 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.

NATURE CLIMATE CHANGE DOI: 10.1038/NCLIMATE3252ARTICLESBroader societybenefits fromrelocationGreaterGoodMutualAgreementRoom for the River 100 peoplePost-disaster voluntary relocation5 cases, 100,000 peopleManaged realignment2 cases, 0 peopleResidents donot initiatethe tlement13 cases, 1.2 M peopleUK Coastal ChangePathfinder 100 peopleResidentsinitiate themoveIsolated settlements5 cases, 100 peopleHunkeredDownSelfRelianceOnly residentsbenefit fromrelocationFigure 2 Managed-retreat conceptual model with recent examples mapped into their respective quadrants. Circles denote sets of cases, and crossesrepresent single cases. Circle sizes are proportional to the number of cases except for autonomous migration, which falls outside our analysis. The numberof people is the approximate number resettled to date. For additional detail on the cases included and the estimated number resettled, see SupplementaryTables 1 and 2.the property is restored to open space. Benefit–cost ratios for theseinterventions range between 2 and 5 depending on the types ofbenefits monetized20,21 . In some cases (for example, Lockyer Valley,Australia), relocation decisions have been made at the community,rather than household, level (Supplementary Table 2, Example 3)22 .Most of these interventions have resettled tens to hundreds ofhouseholds at a time; their voluntary nature can limit their scale.In Mutual Agreement, aligned interests between parties helpovercome regulatory obstacles. For example, in Lockyer Valley, landuse change that ‘would normally have taken between two andthree years to plan and permit was accomplished in four months’23 .Although widespread agreement advances implementation for thisquadrant, resident perspectives can vary within interventions24 .Some households may initially oppose relocation but concede uponrealizing much of their community is moving. Time requiredto process resettlement can be an obstacle; hesitant residentsmay find it easier to rebuild and return home than to wait forfinalization of the buyout or relocation. Although residents mustchoose to relocate, the implementing party’s support facilitates theretreat process.Greater Good managed-retreat interventions are driven by amotivated implementing party, which often must overcome initialresident opposition by incentivizing or compelling relocation. Aspart of the Netherlands’ Room for the River program, years of debatebetween residents, scientists, and government agencies culminatedin the decision to lower dykes around the De Noordwaardcommunity, creating a floodplain for high river flows and protectingdownstream settlements (Supplementary Table 2, Example 6). Thegovernment negotiated with each of 75 affected households, offeringto buy out or elevate their homes17 . In this case, the community’sdisplacement reduced risk for a much larger population.Similar compensation for affected residents features in managedrealignment projects, in which embankments are removed orshifted inland to restore wetlands and reduce spending on coastaldefences. The UK and Germany have implemented dozens of theseprojects over the past several decades (Supplementary Table 2,Examples 7 and 8)25 . Residents are not displaced but may lose land torealignment. Comprehensive cost and benefit data are lacking, butone such project has been shown to be cost-effective given avoidedcoastal defence spending and environmental benefits18 .In Hunkered Down, residents do not initially support retreat,and broader society benefits little from its implementation. Forexample, in the UK Coastal Change Pathfinder Programme,five local councils relocated or bought out households at high riskfrom coastal hazards (Supplementary Table 2, Example 22). Theresidents were offered financial and regulatory relocation supportto incentivize their move. However, an ex post analysis concludedthat societal costs exceeded benefits for several of the interventions(although it did not consider the benefits of reduced urban blightor other local social and environmental improvements)26 . Theimplementing party’s decision to implement retreat may have beendriven by mistaken perceptions about its costs and benefits or byother non-economic benefits.Almost half of the cases are categorized as mandatory resettlement projects. The benefits of such interventions are often hotly debated, so their location on the vertical axis is uncertain. Resident reactions are typically negative, varying with resident circumstances,relocation destination, and retreat process27 . These mandatory resettlements are generally very large in scale, sometimes aiming tomove hundreds of thousands of people (see Supplementary Table 1).Because of resident opposition and scale, these interventions haveencountered numerous logistical and political challenges. After thetsunami in 2004, the Sri Lankan government prohibited rebuildingin the coastal zone (Supplementary Table 2, Example 19). However,an opaque process, redefinitions of the coastal zone, and lack of trustin the government hampered implementation. Some ultimatelymoved back to the coastal zone while others lacked permanenthousing for many months or years28 .NATURE CLIMATE CHANGE ADVANCE ONLINE PUBLICATION www.nature.com/natureclimatechange 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.3

ARTICLESNATURE CLIMATE CHANGE DOI: 10.1038/NCLIMATE3252For Hunkered Down and Greater Good cases, the enabling driveris the implementing party’s commitment and capacity to relocateresidents despite public resistance, complexities of mobilizing anddisbursing funds, and the challenges of determining who moveswhere. These projects often require coordination across numerousinstitutions and door-to-door engagement. Nonetheless, in all casesin these two quadrants, much of the target population was relocateddue to the implementing party’s political will and authority.The Self Reliance quadrant comprises several examples ofresidents struggling to persuade an implementing party to supportrelocation. The Alaskan villages of Newtok, Shishmaref andKivalina have long sought relocation assistance from the nationalgovernment (Supplementary Table 2, Example 24). Althoughpreviously seasonally migratory, these communities began to settlein the late nineteenth and early twentieth centuries to meetgovernment requirements that their children attend formal schools.The school sites, selected by the US Department of the Interior,led to the tribes’ settlement locations. Since then, erosion andflooding have plagued the villages, and each community has decidedcollectively to relocate. But the villages can only access funds on acompetitive ba

Risk management approaches are needed to support the millions of people exposed to potential displacement. Accommodating such risks has limits, and structural protective measures (for example, levees)involvehighmaintenanc