DSpace at VNU: Flood vulnerability among rural households in the Red River Delta of Vietnam: implications for future climate change risk and adaptation

O R I G I N A L P A P E RFlood vulnerability among rural households in the Red River Delta of Vietnam: implications for future climate change risk and adaptation Pamela McElwee1 •Tuyen N

O R I G I N A L P A P E R

Flood vulnerability among rural households in the Red

River Delta of Vietnam: implications for future climate change risk and adaptation

Pamela McElwee1 •Tuyen Nghiem2•Hue Le2 •

Huong Vu2

Received: 5 May 2016 / Accepted: 26 November 2016

Ó Springer Science+Business Media Dordrecht 2016

Abstract The Red River Delta (RRD) of Vietnam, one of the world’s most denselypopulated deltas, is already vulnerable to flooding events, and climate change forecastsproject increased exposure to flood risk in coming decades due to changes in rainfall, stormintensity and frequency, and sea-level rise However, there is a relative neglect of thisregion in the literature on natural hazards and climate change, particularly on how floods inthe RRD might affect poor people and different livelihood sectors, how flood risk isunderstood and acted on, and how flood impacts experienced by households influence localadaptation choices This article presents research undertaken in 2009–2010 to understandthe impacts of flooding in a typical rural zone (Thai Binh Province) of the RRD to assessoverall vulnerability, particularly the relationship between poverty, livelihoods, and floodimpacts, as well as to assess the range of adaptation and flood risk reduction optionscurrently used Our findings indicate that while poor households do not appear to be moreexposed to floods than others, their incomes are more sensitive to relative impacts fromfloods Yet poverty alone did not explain flood vulnerability, as age of household andlivelihood sector involvement showed stronger relationships to flood impacts Flood riskperceptions were also uneven, but poor people did not seem to take less proactive flood riskreduction measures than others There are few long-term adaptation actions to floodingbeing taken by households of any income class, and there is a need for better communityand government aid after flood events to help households cope with increased flood risks inthe RRD, rather than relying on improvements in hard infrastructure, as is currently thedominant approach in the region, particularly given future forecasts of increased rainfallfor northern Vietnam under climate change

& Pamela McElwee

Center for Natural Resources and Environmental Studies (CRES), Vietnam National University, 19

Le Thanh Tong, Hanoi, Vietnam

DOI 10.1007/s11069-016-2701-6

Keywords Flooding
Vietnam
Vulnerability
Poverty
Climate-dependent livelihoods
Adaptation

1 Introduction

Climate change is likely to bring a multitude of new regional weather changes, many ofthem water-related, and Asia is particularly vulnerable to these trends Asia alreadyaccounts for the largest number of flood events in recent decades (Douben 2006), andforecasts indicate that parts of the region are likely to see flood events rise in frequency andscale (Chang 2010; Hirabayashi et al 2013) Forecasted exposure to floods globallyindicates that the countries of Southeast Asia face particularly strong increases in risk inthe future due to a combination of development decisions that put people in the path offloods, along with projected future climate change impacts (Neumann et al.2015a; Win-semius et al 2015) Thus, there is an urgent need to assess current vulnerability andadaptation to flood hazards in this part of the world, and to identify ways in which futurerisks can be better anticipated and managed through vulnerability reduction efforts(Jongman et al.2012)

This is particularly true for countries like Vietnam, where there are dense populationsliving in low-lying areas, and who are currently subject to regular flood events even in theabsence of climate change Vietnam was ranked fourth in a 2009 global ranking of totalpopulations exposed to floods (UNISDR2009), given a long coastline and high numbers ofpeople living in these exposed areas, with over 80 major flood events recorded between

1985 and 2010 in the Dartmouth Flood Observatory database (http://www.dartmouth.edu/%7Efloods/Archives/index.html) Floods have been the second most hazardous cli-mate event in Vietnam after hurricanes, with over 5000 killed and 25 million affected inthe past half century (IMHEN and UNDP2015) Regionally downscaled climate changescenarios for Vietnam indicate that floods are likely to worsen in the future, due to changes

in amount and seasonality of rainfall, along with sea-level rise of up to 1 m in low-lyingdeltas (MONRE2009; Bangalore et al.2016) Indeed, higher precipitation events leading

to increased flooding already appear noticeable in climatic patterns in some regions ofVietnam (Souvignet et al 2013) These risks are compounded by poor developmentdecision making in sectors such as forestry, hydroelectric development, and infrastructurethat can increase flood risks (Beckman2011) Therefore, more needs to be known abouthow existing floods impact households, the different types of social vulnerability to floods,and how households are coping with or adapting to floods

There is a particular need to understand whether poor households or households withspecific types of livelihoods are more vulnerable to flooding in Vietnam A recent globalreport on poverty and climate impacts has noted that poor people are often more impacted

by floods, as their incomes are more dependent on weather, their housing and assets areless protected, and they are more prone to health problems (Hallegatte et al.2015) Otherstudies confirm that poor people tend to be more exposed to floods as they often settle inpoorer quality and cheaper lands in floodplains; they tend to be sensitive to losses in thatthey lose relatively more when hit by floods than wealthier households; and they have alower capacity to cope with and adapt to flood impacts due to lower access to savings,borrowing, or social protection (Cutter et al 2000; Brouwer et al 2007; Braun andAßheuer2011) Climate-sensitive livelihoods can be another factor in flood vulnerability;

activities like farming and fishing can be particularly damaged by floodwaters (Thomasand Twyman2005; Paavola2008; Gentle and Maraseni2012) These two types of socialvulnerability have linkages, as globally, more poor households tend to be dependent onagriculture for their livelihoods than wealthier ones (Davies et al.2009).

Yet there have been relatively few studies in Vietnam specifically on how floods ferentially affect poor people and different livelihood sectors, such as in terms of exposureand sensitivity, and how flood impacts experienced by households influence local adap-tation choices (Birkmann et al.2012; World Bank and AusAID2013) This is particularlytrue for the Red River Delta (RRD) that dominates northern Vietnam, as this region,despite being densely populated and impacted by major floods in recent years, has beenmostly neglected in the academic literature Much more attention has been paid to theMekong (Nguyen2007; Few and Tran2010; Dun2011; Birkmann et al.2012; Few et al

dif-2013) and Perfume River basins (Tran et al.2008; Tran et al.2009; Tran and Shaw2010),and urban flooding and climate change impacts in major cities like Ho Chi Minh City,Hanoi, Can Tho, Hue, and Danang (Hung et al.2007; Hung et al.2010; Birkmann et al

2010; Bich et al.2011; Tu and Nitivattananon2011; Razafindrabe et al.2012) Much less

is known about vulnerability and adaptation to floods in rural areas of the RRD, as only afew studies are available (Tinh2011; Dang et al.2011) This lack of attention is serious,due to the fact that in terms of total numbers of population forecasted to experience floodhazards in Vietnam as a result of future climate change, the RRD is predicted to be themost affected region of the entire country (Bangalore et al.2016)

Given this dearth of information and the importance of the RRD, this article presentsresearch undertaken in 2009–2010 to understand the impacts of flooding in a typical ruralarea there Our aim was to assess exposure to floods; overall vulnerability, particularly therelationship between poverty, livelihoods, and flood impacts; and to assess the range ofadaptation and flood risk reduction options currently used in the RRD, given the fact thatfloods are predicted to increase in both intensity and frequency in the future Our findingsindicate that while poor households do not appear to be more exposed to floods than others,their incomes are more sensitive to relative impacts from floods Yet poverty alone did notexplain flood vulnerability, as age of household and livelihood sector involvement alsowere correlated with flood impacts, which we explain in terms of settlement histories andinvestment strategies of households Flood risk perceptions were also uneven, but poorpeople did not seem to take less proactive flood risk reduction measures than others, which

is a potentially positive sign for future planning Finally, qualitative work with householdsimpacted by floods revealed strong needs for community and government aid after floodevents, particularly as private insurance is nonexistent in this area, and previous disasteraid has been mostly ineffective in helping households adapt to longer-term flood hazardsthat are projected for this region

2 Background: flood risk and vulnerability in Vietnam

2.1 Flood regimes in Vietnam

The northern, central, and southern regions of Vietnam have traditionally had divergentways of coping with flood hazards, and the lack of information about and attention to floodrisk and vulnerability in the RRD, while other areas of Vietnam have seen much moreresearch, is likely due to the misperception that the RRD has sufficient infrastructure in the

form of dams and dikes to prevent large-scale flooding In the Mekong Delta, slow-onsetyearly floods have prompted an adaptation regime of ‘‘living with the floods’’ (Huu2012;Ehlert 2013), while in central Vietnam short slopes and high risk of flash floods havefocused efforts on preparedness and disaster risk reduction projects (Beckman2006; Tran

et al.2008; Tong et al.2012) In the north of Vietnam in the RRD, floods usually occurprimarily after storms or extreme rainfall events, and the focus has been on infrastructurefor flood prevention, including investments in reservoirs, river and sea dikes, and dredging(Pilarczyk and Nuoi2005) However, recent flood events in the RRD have pointed out thedeficiencies in this approach, as existing infrastructure has proved inadequate for extremerainfall-induced floods, and household losses have been substantial (IFRC2008; UN2015).The future holds more flood risk for Vietnam under climate change, particularly theRRD, as forecasts indicate changes in flood hazard patterns in terms of timing, intensity,and duration can be expected (Ngo-Duc et al.2014; Giang et al.2014) Monsoonal rainshave been tending to come earlier in the season than in previous decades, and increases inflood peaks have been documented for all regions of Vietnam (with the exception of theRed River, due to storage reservoirs built in the upstream) since 1961 (IMHEN and UNDP

2015) The overall predictions for the end of the twenty-first century under mediumgreenhouse gas emissions scenarios are that the dry seasons will get drier, and in the rainyseason, larger volumes will fall in shorter periods, exacerbating floods in places thatalready experience them In the RRD, rainfall is expected to increase at a higher rate thanany other part of the country (MONRE2009) (see Table1) These hazards are combinedwith existing high densities of population in the RRD who are likely to be exposed andpotentially vulnerable to these future flood hazards (Bangalore et al.2016)

2.2 Vulnerability to floods in Vietnam

Floods are the most frequently experienced climate risk in Vietnam, according to thenational Vietnam Household Living Standards Survey (VHLSS), which covers all pro-vinces and regions and assesses household well-being approximately every 2 years.According to VHLSS data, floods cause the most income damage of all climate events atthe household level in Vietnam (Arouri et al.2015) Overall, 10% of all communes (thelowest level of state government) in Vietnam were affected by at least one flood in 2010

Table 1 Changes in annual rainfall (%) relative to period of 1980–1999 projected for Vietnam’s regions under a medium emission scenario (B2) Source MONRE ( 2009 )

Climatic region Decades in the twenty-first century

2020 2030 2040 2050 2060 2070 2080 2090 2100

Southeast and Mekong Delta 0.3 0.4 0.6 0.8 1.0 1.1 1.2 1.4 1.5 Bold indicates Red River Delta

(Kozel2014) Several studies have pointed out, based on the VHLSS data or case studies

of particular floods, that floods have negative impacts on household welfare in Vietnam, aswell as leading to death and injury (Thomas et al 2010; Bich et al.2011; Navrud et al

2012), and the poorest households in Vietnam seem to be disproportionately impacted(Birkmann et al.2012; Rubin2014)

The channels by which poor households in Vietnam experience floods are related toboth exposure and sensitivity Poor households often settle in flood plains or in low-qualityhousing, increasing their exposure to flooding For example, a study in the Mekong Deltafound 38% of the region’s poor but only 29% of the region’s non-poor lived in frequentlyflooded areas (Nguyen Van Kien2011) Another study in Hue city in the central region ofVietnam found that poor households experienced flooding at rates more than 40 cm higher

in their houses than richer homes built with higher elevations (Tran et al 2008) Thisincreased exposure of the poor leads to increased mortality from floods; deaths fromdisasters (including floods and storms) correlate strongly with poverty rates in Vietnam’sprovinces, according to one study (Rubin2014) Evidence from the Mekong Delta and HoChi Minh City’s urban wards indicates that exposure to floods has led to higher health risksamong the poor, including both waterborne and respiratory diseases (Few and Tran2010;World Bank and AusAID2013) Exposure also relates to the ability to prepare in advancefor flooding; a study in Hoi An city in the central region noted that poorer householdsfound it more difficult to protect household assets from floods, as they did not have secondfloor storage areas in their houses, or could not afford to pay workers to move goods tosafety in time (McElwee et al.2010)

Sensitivity and vulnerability are often used interchangeably, and refer to the fact thatloss and damage may relate more to the socioeconomic or cultural situation of a householdrather than their physical exposure, including indicators like poverty status or the types oflivelihood activities they pursue, such as agriculture and other high-risk and climate-dependent activities Previous research from Vietnam confirms the links between poverty,climate-sensitive livelihoods, and vulnerability One study in the central region of Vietnamnoted that more poor households reported a ‘‘high impact’’ from flood hazards than themiddle class or rich (Casse et al.2015), while another study in the same region found thatflood damage made up a larger portion of annual household income for the poor (26% ofincome) versus only 14% of the non-poor (Navrud et al.2012) A study from the MekongDelta found that poor households were more likely to experience submerged homes, joblosses, reduced income from fishing, and temporary evacuation during floods than thosewho were better off (Nguyen and James 2013) Sensitivity also relates to informationaccess, as poor households may receive less early warning information in order to preparefor flood risks if they cannot afford TVs or mobile phones (Bruun and Casse 2013).Further, while floods are often indiscriminate, and can hit both poor and wealthy in thesame area, it is often the poor who have fewer reserves to aid in recovery or in longer-termadaptation (Arouri et al.2015; Casse et al.2015) Floods can therefore be one factor inpushing near-poor people into poverty if there is not sufficient safety-net and livelihoodsupport to aid in coping responses (Beckman 2006; Miller 2006) Local institutions,informal safety nets, and local social capital are enormously important in responding todisasters and climate events, and poor households may have trouble accessing these forms

of assistance, as well as access to more formal government disaster support, whose erage in Vietnam can be spotty (Casse2013; Chau et al.2014b)

cov-Yet all too often, reports from Vietnam do not focus on these social factors of nerability and risk Much literature on flooding focused on Vietnam uses methods thatconflate vulnerability simply with physical exposure (Hien et al 2005), or uses overly

vul-general indexes which have very few indicators related to social vulnerability, particularly

at household levels (Dinh et al.2012; Ho et al.2013; Chau et al.2013,2014a) Thus, there

is a need for a closer look at social vulnerability to flooding in Vietnam, particularly in theRRD

3 Study area and methods

The Red River dominates northern Vietnam, with a total catchment area of 86,660 km2,fed by tributaries including the Thao, Lo, and Da Rivers The lower delta consists of sixdifferent mouths (the Thai Binh, Ninh Co, Tra Ly, Ba Lat, Van Uc, and Day Rivers) thatfan out from the main Red River and empty into the Gulf of Tonkin in the East China Seaover 200 km of coastline (see Fig.1) The lower RRD lies south and east of the capital ofHanoi, covering an area of nearly 1.3 million hectares over nine provinces (4% of the totalarea of Vietnam), and is characterized by mostly agriculture and aquaculture production.58% of the delta is less than 2 m below sea level (Tinh 2011) The RRD is the mostdensely populated area of the country, with a total population of over 17 million people,and a population density that can reach 1225 persons per km2(Devienne2006) Much ofthis population is engaged in climate-sensitive livelihoods, particularly farming and fishing(Adger1999)

The RRD has a strong monsoonal climate; around 80% of the total annual rainfalloccurs during the 6-month season from approximately late May to early November, with

an average rainfall of 1900 mm/yr (Tinh2011) Typhoons from the South China Sea whichbring sudden and torrential rains and high winds to coastal areas are also a factor in the fallmonths, and on average, the RRD is hit by six typhoons a season (Kelly2001) Changes inthe frequency and severity of flood events have been documented for the RRD, with a rise

in the average number of days of flooding per event in the past half century (Khanh and Le

2001) Further, while July and August were traditionally the flood months in the monsoon,flood events are now falling throughout early summer to winter Storm surges and sea-levelrise are also predicted to increase due to climate change, with severe consequences forflood risk, especially for the coastal populations (Neumann et al.2015b)

The RRD is protected by two dike systems: river dikes of 3000 km length that controlfloods from the Red River and its tributaries; and a sea dike of 1500 km in length Some ofthese dikes were first built more than a thousand years ago and upgraded during the Frenchcolonial period (Smith2002) A survey in 12 northern provinces found 14% of local dikes

to be in bad condition, 73% in average condition and only 13% in good condition (Tinh

2011) This can be attributed to the fact that collective contributions to the maintenance oflocal dikes, channels, and canals have been in decline in the post-socialist era (Adger

1999) Although irrigation and drainage in many polders of the RRD were upgraded in the1990s with international development aid, floods still cause damage, especially whenpumping stations cannot keep up with excess water (Ritzema et al.2008) Other problemsinclude poor land-use planning, as currently much land which is designed for agriculturalproduction has been converted into industrial or urban uses, without flood preventionmeasures Unlike many other flood-prone areas of the world (Brown and Damery2002),there are no systematic hazard maps for the RRD to guide zoning and planning Thus, thereare reasons to think flood vulnerability in the RRD might differ from other areas ofVietnam, due to the lack of hazard maps, unregulated development, and overreliance oninfrastructural and technical means of flood risk reduction, such as dikes

Therefore, to determine the impact of flooding on rural areas of the RRD, we undertookresearch in 2008–2009 in one typical area (Thai Binh Province) that had experiencedflooding in the last 10 years Thai Binh is a mostly rural province approximately 70 kmsoutheast of Hanoi, and has been the leading province of the RRD in rice productivity,which can reach 8–10 tons/year in some areas Aquaculture in fresh and saltwater areas hasdeveloped in recent years, but agriculture remains the dominant income source Both theTra Ly and Red Rivers run through Thai Binh Province, and much of the province liesbelow river level and thus relies heavily on river dikes and pumping for drainage We

Fig 1 Map of Red River Delta system and site of field research Source Base map from (Tuan and Shannon

2010 )

selected Kien Xuong district for field research as it had experienced a major flood caused

by excess monsoon rains in recent memory (September 2003) at the time of the research,and because Kien Xuong is predicted to be one of the top 20 districts in the whole country

in terms of population exposed to flooding under future climate change scenarios galore et al 2016) The total number of people who would be exposed in Kien Xuongunder 200-year flood risks would be over 200,000 people, or close to 90% of the entirecurrent population of the district (around 235,000 people)

(Ban-We conducted interviews with 151 households in four villages of Kien Xuong insummer of 2009 after pretesting of questionnaires in fall 2008 We choose two floodaffected communes (Tra Giang and Quoc Tuan) and chose two villages in each site (TrucTam and Duc Duong villages in Tra Giang and Dac Chung Bac and Dac Chung Trung inQuoc Tuan) We excluded non-flood-affected households out of the village census roles, inconsultation with key village informants, as we were interested in flood impacts and socialvulnerability of those affected We selected households to be interviewed based on theproportion of population size of the village by randomly selected every kth householdremaining on the village census lists for interviews with a standardized survey, which wasconducted with either the head of household or the spouse The authors guided the surveydata collection, along with five university students working as surveyors, with all inter-views conducted in Vietnamese Our overall questions related to channels of vulnerability

by which households were effected by floods—did these risks relate to location within thevillage, type and amount of assets, characteristics of livelihoods, or other social factors,which we tested by collecting detailed income measures for before and after flood eventswithin surveyed households

We also conducted focus group discussions with over 50 participants Topics discussed

in focus groups included histories of climate events in the local area and different tation measures to floods undertaken by the community and their efficacy Provincial,district, and commune government officials (27 total) in different sectors dealing withclimate and water management issues were also interviewed These tools were all aimed atoutlining the differential types of vulnerabilities and responses to flood events that existed

adap-in this representative site of the RRD

4 Results

4.1 Household demographic data

Households in Kien Xuong district that were interviewed tended to reflect demographictrends in the RRD region: namely, an aging population continuing to reside in the coun-tryside, while children often migrated to Hanoi or other areas for work The average age ofthe household survey respondent was 55, and households tended to be fairly small (3.55members), with less than one child under 16 still residing in the home per household Allrespondents were literate, with an average educational level of 9 years of schooling(equivalent to lower secondary school training) The residents were primarily farmers, with100% of households reporting at least some farm income from the year before the surveywas administered The large majority of households owned their own rice lands, while asmaller percentage also owned or had access to fishponds as a source of income (seeTable2) Small livestock was also an important source of income (from pigs and chickens,primarily, although a few wealthier households had buffalo or cattle)

Housing stock varied somewhat in the researched area; 72% were concrete or brick,while the rest were a mix of brick and wood (considered less permanent and stable) Only5% of households had homes of more than one story, either two story or one story with roofmezzanine, known as gac lung, which could be used to store assets during floods, whileother households had to make do with less stable ‘‘attics’’ of bamboo floors in the rafters ofone-story homes (see Fig.2) All households owned their own homes and most hadelectricity and tubewell water Other common household assets included motorbikes andboats, both of which can be useful in transporting people, livestock, and household goods

to safety during floods

4.2 Experience of flood events

Kien Xuong district had experienced several severe flood events in living memory (seeTable3) Focus group discussions revealed key differences in the types of floods, most ofwhich were caused by excess rainfall or storms or were ‘‘out of season’’ from the tradi-tional flood months of the early fall, as well as the historical severity of floods experienced,such as in fatalities and damage Community discussions confirmed that most householdsperceived a 2003 flood to have been the most serious in recent memory, so we used thisevent as our key research topic The 2003 flood was experienced as a series of heavy rainsfrom the 10th to the 14th of September, leading local rivers (the Thai Binh and Tra Ly) to

Table 2 Basic demographic characteristics of the respondents Source Household survey, 2009

Average number of children still living at home per HH 5 73

(1132 USD)

20,011,435 VND (1140 US) Percentage HH with farm income at the time of survey 100%

Percentage HH owning some irrigated rice land 99%

Average landholding of irrigated rice land 0.21 ha 13 ha

Average landholding size (residential, agriculture and other,

e.g., forest or aquaculture)

Percentage HH with livestock income 88%

Percentage HH owning or renting a fish pond 26%

Percentage homes made of concrete or brick 72%

Percentage homes with only one story 95%

Percentage HH with domestic tubewell water 98%

Percentage HH owning at least one motorbike 72%

Percentage HH owning at least one boat 30%

overtop river dikes, and the dikes were broken in several places The local dike was said to

be weaker than in the past due to a number of migrant families who had moved into landright next to the dike, building houses too close to the structure Pumping stations wereunable to keep up with the floodwaters, and the drainage system canals functioned verypoorly, preventing water from draining out of fields back to the river Fifteen people were

Fig 2 House types in Kien Xuong (two story with mezzanine, right; one story with mezzanine, center; and one story with no mezzanine and traditional tile roof, far left) The two houses on the right also have raised foundations, while the traditional one on the left does not

Table 3 Past flood events in Thai Binh in past decades Source Key informant interviews in the field, 2009 Year Impacts in Kien Xuong, Thai Binh

1968 The back dike was broken causing a severe flood, and then the sea dike was broken Much flooding and waterlogged land for a long time At least seven people died

1971 Severe storm and broken sea dike, leading to extensive evacuation and damage, and multiple deaths

1976 Flood caused multiple deaths

1978 Heavy rains caused severe flood

1980 Heavy rain Lost two-thirds of rice crop to flood

1986 Heavy rain Severe flood, some house damage

2003 Heavy September rains leading Tra Ly river to overflow; back dike and small dike for production broken, causing 9 days of waterlogging All rice fields were flooded and destroyed, some houses as well

2008 Sudden severe rain in September (Tropical Storm Hagupit) led to loss of some crops

reported dead in the province from the 2003 flood, and 31,000 km2of land were affected;the total area of rice cultivation affected in the province by inundation was 60,000 ha Thedamage caused to aquatic production was also large, and $38 million in property losseswere reported from the province.

Households in focus groups described the 2003 flood as ‘‘very sudden,’’ occurring afterheavy rains described as ‘‘mua den’’ (black rain), and that advanced warning of a day or soallowed households to have some time to prepare Around 70% the households in KienXuong reported listening to TV carefully prior to flood events and during heavy rains,while 49% reported using radios to listen for flood forecasts 52% of households alsoreported getting direct information about floods from either a village leader or via arelative/friend who had spoken to a village leader In terms of rapid preparation, sandbagswere filled to pile along dikes and in front of houses, and bamboo stakes cut to help securetrees, houses and other structures Livestock were brought to pens or evacuated to higherground, and some animals simply ran away themselves and had to be caught by ownersafter the floods had receded Household goods were lifted to second stories or attics, ormoved to friends’ or relatives’ houses on higher ground

When the flood hit, most households recalled floodwaters of at least knee-high in yardsduring the worst of the flood, and some houses had waters up to the level of beds on thefirst floor In these cases, family members evacuated to relatives’ or neighbors’ houses.75% of households surveyed reported that they perceived their houses were relatively safeduring flood events, and that they could stay at home safely to wait out flooding Localgovernment officials interviewed disputed these risk perceptions somewhat, stating that inthe future, faced with a similar flood as in 2003, they would need to be more proactiveabout evacuating people and not letting them stay in their own houses, in order to reducerisks of health effects or injury 58% households in the sample did report that they hadknowledge of safe places to evacuate to if they were needed (such as buildings of thePeople’s committee or schools, which are often located on the higher land and which serve

as shelters) In the case of evacuations, both boats and motorbikes could be used,depending on the level of floodwaters (see Fig.3)

4.3 Impact from flood event

The 2003 flood had significant impacts on households Immediate impacts includeddamage to homes and livelihoods, and access to transportation 64% of surveyed house-holds got some floodwater in their yards (the average depth of yard floodwaters was 4 m),while 20% of the interviewed households got some water inside their house (but less than

1 m); only 1% of the interviewed households had more than 1 m of water in their house.The duration of the water in yards and houses was from 5–10 days for most households.There was no correlation in the sample between depth of floodwater and poverty status ofthe household, indicating that poor families did not seem more physically exposed to floodsthan others

The household survey asked sampled households about their major income sourcesbefore and after 2003 flood event, and 59% of interviewed households in Thai Binh agreedthat the event had caused ‘‘very serious’’ damage, while the rest of households said thedamage was average or minimal Given an average household income of around $1337USD before the flood, 8% of households estimated their total property and income damageabove 20 million VND (around $1130 USD), while 23% estimated from 10 to 20 million($570–1130 USD) The rest reported damage under 10 million VND ($570 US) 52% ofhouseholds estimated needing from 1–6 months to recover from the floods, 23% needed

6–12 months, and 26% needed over a year, with some even reporting taking up to 3 years

to recover losses

When asked to estimate the amount of damage from the flood to different incomesectors, households provided varying estimates (Table4) 95% of the total interviewedhouseholds reported a decrease in income from agriculture from the flood, explained by thetiming of the flood event, as the 2003 flood hit in early September when paddy rice wasstill in the field, causing the near-complete destruction of the summer–autumn rice crop.Vegetable crops (which were often planted outside of protective dikes and thus exposed tofloodwaters) were also heavily damaged Livestock was a differentially affected sector,with 52% of households reporting a loss of income after the flood from livestock, while the

Fig 3 Example of concrete boats available to use during floods in Kien Xuong

Table 4 Households reporting income changes before and after flood event Source Household survey, 2009

Source of income Baseline income (in 2002)

rest reported little income decrease Aquaculture was also affected for some households(42%), reflecting the fact that while only around one-quarter of households had a fish pond,other families without fishponds usually had fishing income from river fishing, whichdecreased after floods, as people were either too busy or wary of catching fish that mighthave been in polluted floodwaters Salary and wage income showed the least impact fromflooding, with 80% of households reporting that it was the same after the flood event, oreven increased (2% of households).

The average value of flood losses was estimated at US$447/HH in damages, or aroundone-third the average income for households the previous year Because insurance forcrops or livestock, as well as for houses, is virtually nonexistent in Vietnam (Hansson et al

2008), these losses had to be borne entirely by households themselves For the 42% ofhouseholds that had reported aquaculture damages, the income losses were almost total:92% of the value of the previous years’ fish income was reported as losses or damaged bythe 2003 flood (Table5) This was attributed to the fact that the floods often swept awayfish, losing not only the ability to sell the fish, but also wiping out the investment that hadbeen made in fish fry Salary and business income was the least affected by the floods, withonly 1% damage reported

4.4 Damage from flooding by income and age classes

To understand the differential impact of flood damage on income classes, we stratified oursurveyed households into three income terciles (lowest one-third, middle one-third, andrichest one-third) Damage to cropping, livestock, and aquaculture was significant acrossincome terciles (Kruskal–Wallis test), although impacts need to be weighed both in terms

of absolute (total) damages incurred and relative impacts (the damage as a percentage ofincome) While poorer households had the most relative damage from floods, in that theylost higher percentages of their total income, their overall absolute damage was lowercompared to richer households (Table6) Richer households had higher amounts ofabsolute damage from the flooding events, as they were more likely to have invested largesums in risky ventures like new aquaculture ponds or large flocks of ducks that were badlyhit by the flooding (significant at 0.00) Yet for poorer households, even the lower amounts

of overall damage were significant as they took up a large percentage of total income, andrelative agricultural damage was highest for the poorest households (sig at 002) Thereported time needed for recovery from the flood event varied by household from only

1 week to over 3 years (15% of households) Interestingly, middle-class householdsreported needing less time to recover than either poor or rich households did

To judge impacts based on household’ own assessments of the impacts from floods,rather than by income classes, we asked our respondents themselves to state whether they

Table 5 Value of flood losses across survey sample Source Household survey, 2009

business income Total value of aggregate income reported for 2002

across all households surveyed (USD)

48,761 67,852 18,627 101,689 Reported losses in 2003 flood -27,307 -17,642 -17,084 -1017

were ‘‘highly affected’’ by the flood of 2003, ‘‘moderately affected,’’ or ‘‘little to no affect’’(Table7) These self-assessments not surprisingly were highly correlated with total incomedamage, as those noting they were ‘‘highly affected’’ had nearly three times the income andproperty damage as the ‘‘moderately affected’’ group (sig at 0.00) We assessed how each

of these groups may have differed from each other using Kruskal–Wallis tests, whichshowed that the age of the household head and the educational level were correlated withhigher self-reported impacts Age showed a clear correlation, with less impacted house-holds more likely to be older This is likely due to the fact older households were morelong settled and more likely to have land inside dikes in protected areas and thus lessexposed to floodwaters, as opposed to younger families who were more likely to havemarginal lands they had settled more recently that were not protected within the dikes.Educational levels however did not show a clear relationship to impact, and were likely

Table 6 Relationship of economic class to flood vulnerability Source Household survey, 2009

(n = 51)

Middle 1/3 (n = 50)

Richest 1/3 (n = 50)

Sig.

Total income in 2007 5,341157 15,718,220 38,881,560 000**

Total property and HH asset damage 896,471 566,400 3,151,600 488 Agricultural damage (absolute) 2,070,549 1,875,540 4,753,060 033** Relative agricultural damage (% of income from

this source lost by flood)

Relative livestock damage (% of income from

this source lost by flood)

Aquaculture/fishing damage 792,255 579,000 3,292,000 0.02** Total cost of damages 4,383,798 4,694,280 15,811,460 000** Time to recover from floods 476 days 303 days 515 days 0.04**

** indicates significance of the results (p \ 05)

Table 7 Comparison of different levels of impacted households Source Household survey, 2009 Variable Highly affected HH

(n = 89)

Moderately affected

HH (n = 55)

Low affected HH (n = 7)

Sig.

Total average income in