Assessing and mapping human well-being for sustainable development amid flood hazards: Poyang Lake Region of China

Applied Geography 63 (2015) 66e76

Contents lists available at ScienceDirect

Applied Geography journal homepage: www.elsevier.com/locate/apgeog

Assessing and mapping human well-being for sustainable development amid flood hazards: Poyang Lake Region of China Qing Tian a, *, Daniel G. Brown b, Shuming Bao c, Shuhua Qi d a

Department of Computational Social Science, George Mason University, 4400 University Drive, Fairfax, VA 22030, USA School of Natural Resources and Environment, University of Michigan, 440 Church St., Ann Arbor, MI 48109, USA c China Data Center, University of Michigan, 1007 E Huron St., Ann Arbor, MI 48104, USA d Key Laboratory of Poyang Lake Wetland and Watershed Research, Ministry of Education, Jiangxi Normal University, Nanchang 330022, China b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 8 January 2015 Received in revised form 16 June 2015 Accepted 16 June 2015 Available online xxx

Less developed places that are affected by climatic impacts face great challenges to future development. Place-based assessments that look at both the development level and climatic impacts on development are important for understanding the current state of human well-being and generating insights into how to facilitate sustainable development in the future. We carry out an assessment of human well-being in the Poyang Lake Region of China (PLR), using GIS, remote sensing, and socio-economic data. We measure human well-being in three aspects of (i) development level, (ii) exposure of development to flooding, and (iii) sensitivity of development to flooding. Following the United Nations Development Programme's human development index, we examine development through measures of life expectancy, literacy, and income. We first use a digital elevation model and GIS data on levees to map flood hazard in PLR. Based on the flood hazard map, we then derive quantitative measures of exposure and sensitivity of the development in a town to flooding. Our assessment indicates that development in PLR overall is highly exposed and sensitive to flooding. There are significant variations in different aspects of human wellbeing among the 298 towns in the region. These variations suggest different sustainable development pathways and policy interventions for different places. We discuss the potential usefulness of our approach for other similar places. © 2015 Elsevier Ltd. All rights reserved.

Keywords: Human well-being Sustainable development GIS Quantitative assessment Less developed rural areas Climatic hazards

1. Introduction Less developed places that are affected by climatic hazards face great challenges to future development. While improving development levels in the developing world has proved to be difficult in general (Collier, 2007; UNDP, 1990e2014; World Bank, 2002), extreme climatic events impose an additional constraint on development in such places (Adger, Huq, Brown, Conway, & Hulme, 2003, 2006; Kates, 2000; Kates & Dasgupta, 2007; Takeuchi & Aginam, 2011). Meaningful place-based assessments that examine development and climatic impacts on development in an integrative manner are important for understanding the current state of human well-being and can suggest where and how to make adjustments or improvements in the future, facilitating sustainable

* Corresponding author. Research Hall, Room 374, 4400 University Drive, MS 6B2, Fairfax, VA 22030, USA. E-mail address: [email protected] (Q. Tian). http://dx.doi.org/10.1016/j.apgeog.2015.06.007 0143-6228/© 2015 Elsevier Ltd. All rights reserved.

development. A number of assessment frameworks that are relevant to development and climatic impacts have been used in the literature. The United Nations Development Programme (UNDP) has used the Human Development Index (HDI) to assess human development for its annual human development reports since 1990 (UNDP, 1990e2014). HDI combines variables that represent life expectancy, literacy, and income to measure human development in a country. Though the specific indicators and methods that UNDP uses to calculate HDI have been evolving over time, life expectancy, literacy, and income remain to be the three key dimensions of human development in HDI. These three dimensions of HDI have also been applied to examining human development at other levels than countries (Beavon & Cooke, 2003; Harttgen & Klasen, 2012; Kumar, 1991; Song & Ma, 2004; Thapa, 1995). In natural hazard research, context-based environmental hazard mapping is often performed to assess the impacts of exposure to climatic hazards € schl, 2004; Büchele et al., 2006; Meyer, (Apel, Thieken, Merz, & Blo Scheuer, & Haase, 2009). Such assessments place an emphasis on

Q. Tian et al. / Applied Geography 63 (2015) 66e76

the biophysical aspect of hazard. The UNDP's Human Development Index and environmental hazard mapping each describe some aspects of a system in a place, and each alone is not sufficient for studying development in less developed places that are exposed to extreme climatic events. Due to increasing adverse impacts of climate change (IPCC, 2001, 2007 and 2014), vulnerability assessments have gained much attention in the literature. Social vulnerability assessments often use a composite index that combines socio-economic variables, such as socio-economic status, access to resources, age and gender, the degree of urbanization, occupations, infrastructure, education and social capital, with or without a specific environmental context, to measure vulnerability of people (Cutter, Boruff, & Shirley, 2003; Dwyer, Zoppou, Nielsen, Day, & Roberts, 2004; Rygel, O'sullivan, & Yarnal, 2006; Vincent, 2004). They are useful for identifying vulnerable groups for policy intervention but do not tell much in what ways they are vulnerable, especially when it is not fully understood how each of these variables affects people's vulnerability (Agrawal & Chhatre, 2011). And these variables are likely to have different effects in different places/contexts. The IPCC vulnerability framework (Fussel & Klein, 2006; Houghton et al., 2001; McCarthy, Canzianni, Leary, Dokken, & White, 2001) is a more integrative approach to addressing vulnerability to climate change/variability. It uses exposure, sensitivity, and adaptive capacity to describe vulnerability of a social or natural system to climate change/variability. While capturing the biophysical aspect of vulnerability, it also tries to address how a system may adapt to climate change/variability. Vulnerability assessments in general are centered on climate change/variability and aim to generate insights into how to cope with or adapt to extreme climatic events. Although development is often examined, it is mostly treated as an explanatory variable for vulnerability or a means for climatic impact mitigation. In less developed places that are under influences of extreme climatic events, there are usually multiple stressors and some other more pressing issues than climate that affect the livelihoods of people (o'Brien et al., 2004; Tschakert, 2007; Casale, Drimie, Quinlan, & Ziervogel, 2010). While it is important to look at how development affects vulnerability and capability of people to cope with climatic impacts, it would be more fruitful to examine overall development and view climate as one of the constraints on development to understand human well-being in such places (Tian, 2011). Indeed, researchers, particularly those who have work experience in less developed countries, have long recognized that the central issue of social vulnerability is development, and the well-being of people is the real concern (Eakin, 2005; Ribot, Najam, & Watson, 1996; Ribot, 2009; Wilbanks & Kates, 2010). We need to exam vulnerability in the broad context of sustainable development (Turner et al., 2003). There has been a long tradition of geographic research on natural hazards (Montz & Tobin, 2011). Geographic approaches to natural hazard research have evolved from focusing on understanding the geophysical environment to integrative studies that examine both social and geophysical environments (Burton, Kates, & White, 1978 and 1993; White, 1945). Advanced technological tools, such as GIS and spatial analysis, have also facilitated natural hazard research and proved to be useful for quantifying vulnerability, resilience, and adaptive capacity which have emerged as core  pezconcepts in climate variability/change research (Belmonte, Lo García, & Soriano-García, 2011; Ho & Umitsu, 2011; Frazier, Thompson, Dezzani, & Butsick, 2013; Malcomb, Weaver, & Krakowka, 2014; Santos, Tavares, & Emidio, 2014; Silva, Matyas, & Cunguara, 2015; Varis, Kummu, Lehr, & Shen, 2014). In this study we integrate social and geophysical perspectives to assess human well-being in less developed places that are affected

67

by extreme climatic events. We propose a framework for assessing human well-being to guide sustainable development in such places. The framework combines measures of (1) development level, which includes various aspects of human development, (2) exposure of the human system, which characterizes the nature and degree to which the human system is exposed to climatic variations or extremes and is determined by the natural environment, and (3) sensitivity of human development, which reflects how human development is affected by climatic variations or extremes. We use a GIS approach and combine environmental and socioeconomic data to carry out an assessment of human well-being for 298 towns in the Poyang Lake Region of China (PLR). PLR is a less developed rural area in Jiangxi province. The region has been historically subject to flooding from Poyang Lake, the largest fresh water lake in China. We use a digital elevation model and GIS data on levees to map flood hazard. Based on the flood hazard map, we derive quantitative measures of exposure and sensitivity of the development. Following the UNDP's Human Development Index, we measure development level using variables that represent life expectancy, literacy, and income. After a brief introduction to the study area below, we present our data and methods in detail. We then present the assessment results. Our discussion focuses on (i) possible implications of the assessment results for future development in PLR and (ii) how our approach can be potentially useful for assessing human well-being in similar places. 2. Study area PLR is largely a rural area in Jiangxi Province composed of ten counties and two cities (Nanchang and Jiujiang) with a total area of 20,970 km2. According to the Chinese Census in 2000, the total population in PLR was about 7.7M. PLR is a major agricultural production base in Jiangxi. Rice cultivation is the most traditional agricultural activity and still widely practiced in the region. In 2004, the average farmer per capita income was 2450CNY (1 US dollar was 8.28 CNY in 2004). Rural development in PLR is affected by flooding from Poyang Lake. Situated in a structural depression, Poyang Lake collects water from five major rivers in Jiangxi and drains into the Yangtze River at Hukou (about 700 km downstream of Three Gorges Dam). Its water level varies considerably throughout the year (Min, 1997). From April to June, the water levels of the five rivers are high due to seasonal rainfalls, and the lake level rises as well. From July to September, the Yangtze River is high due to seasonal rains, and the water can flow back to Poyang Lake, making the lake level rise. Historically, the most severe floods occurred when high water levels in the five rivers and the Yangtze River coincided. Global climate change and engineering work, such as Three Gorges Dam, may increase the uncertainty in the future flood regime. Although no severe floods have occurred in the past decade, according to some local scientists who study the hydrology of Poyang Lake (Min & Liu, Pers. Comm.), concerns about flooding remain because the lake responds to long-term climatic and hydrologic cycles and may be in a low level stage. Overall, flooding from Poyang Lake imposes a significant constraint on economic development in the region and especially on the agricultural sector (Chen & Zhao, 2007; Huang & Dai, 2004; Huang, Wang, Jiang, Zhao, & Shi, 2006; Jiang et al., 2008; Shankman & Liang, 2003; Wang, Yan, & Wu, 2006; Ma, 2007; Zhao & Guo, 2001). Throughout history, people in PLR have built levees to mitigate flood impacts. Levees are also used to reclaim land for increased grain production and to accommodate population growth. In PLR levees either enclose an area or connect with coastlines to form polders. The levees in PLR are categorized as crucial levees, major levees, and minor levees based on the amount of enclosed farmland and the area of settlement they contain, forming different types of

68

Q. Tian et al. / Applied Geography 63 (2015) 66e76

polders (Jiang, 2006; Peng, 1999, Fig. 1). Crucial levees enclose more than 66.7 km2 of farmland in addition to large cities. The height of crucial levees (i.e., the elevation of the top of a levee) is 22.37 m on average, ranging from 21.76 m to 22.81 m. Major levees are charged with protecting more than 33.3 km2 of farmland and may contain large towns. The height of major levees is 22.18 m on average, ranging between 19.26 m and 28.18 m. Crucial and major levees are built strong with concrete and are also well maintained by the governments. They are designed to withstand floods with a 100year recurrence interval but are not adequate for this purpose in reality (Qi et al., 2009). Minor levees protect less than 33.3 km2 of farmland and do not contain cities or large towns. The heights of minor levees vary largely from 16.46 m to 28.26 m with an average of 21.99 m. Minor levees are usually not well built or maintained, often by local people. Some of them have an earthen structure. Levees can protect human lives and property from flooding, but extensive levee construction also reduces water storage capacity and changes the flood regime of Poyang Lake. Though land reclamation from the lake has a history of several centuries, many of the current levees around Poyang Lake were built between the 1950s and 1970s after the tremendous flood of 1954 that nearly destroyed all of the levees in PLR (Jiang, 2006; Qi et al., 2009). During this period some new levees were also built to enlarge farmland area, guided by national policies that placed a high priority on grain production. Farmland was consequently reclaimed from the lake area, and the floodplain was enclosed into polders. It is estimated that there were about 565 levees around Poyang Lake in 1998, and the free surface of Poyang Lake was reduced from 8932 km2 to 3850 km2 due to levee construction (Jiang, 2006). To improve the lake's floodwater storage capacity, four polders were designated for floodwater storage by the government of Jiangxi in 1986 (Fig. 1). They are intended to be filled up with floodwater in the largest floods. The flood storage polders are built with concrete and may contain farmland but little urban area. The height of flood storage polders is 21.86 m on average, ranging from 21.26 m to 22.37 m. In addition to increasing the risk of dramatic flood catastrophes, more economic loss is involved when levees fail. The disastrous flood in 1998 resulted in the failure of many important levees, which caused significant damage to the economy. After the 1998 flood event, the government implemented a policy called “Returning Farmland to Lake,” which resulted in the abandonment of many minor polders. The abandoned levees all have an earthen

structure and are further classified into two types: “partial return” and “complete return” (Jiang, 2006, Fig. 1). In the partial return polders, villagers were resettled to higher ground, but farmland could still be cultivated. The height of partial return levees is 21.57 m on average, ranging from 20.26 m to 22.76 m. In the complete return polders, villagers were resettled and farmland was restored to wetland. The heights of partial return levees range between 19.26 m and 21.76 m with a mean of 20.44 m. Flood hazard in PLR varies spatially and is closely associated with levee type. Floods occurred in polders often because levees were unable to withstand floodwater and failed before floodwater reached the top. For example, about 240 levees including some important levees in PLR were damaged in 1998 (Jiang, 2006). Levee type reflects how well levees are constructed and maintained, which directly affects the chance and frequency that polders are flooded. In the event of a flood, the government also differentiates resources and efforts in preventing levee failure according to potential economic loss due to levee collapse (Jiang, 2006). As a consequence, flood hazard is relatively low within crucial polders, slightly higher within major polders and high within minor polders. Because flood storage polders are designed to be breached only when extreme floods occur, flood hazard within flood storage polders is relatively low. In fact, flood storage polders have never been used for this purpose, and even in the most severe flood of 1998, these polders were not breached (Jiang et al., 2008). But flood hazard within returned polders is high because they have low quality and are designed to be breached more frequently. PLR also possesses ecological importance. The coastal zone and wetlands around Poyang Lake serve as important habitats for more than 332 different species of birds, of which 13 are internationally protected, and the Siberian Crane is critically endangered. 3. Data and methods We used a digital elevation model, GIS data on levees and historical data on lake levels to map flood-hazard zones. The floodhazard zones serve as the basis for measuring exposure of the towns to flooding. We then combined land-use data interpreted from remote sensing images and population distribution map with the flood-hazard zones to derive measures of sensitivity to flooding. Assessments were performed and mapped using individual variables and then combined to produce a map of overall wellbeing. 3.1. Mapping flood hazard

Fig. 1. Polders created by different types of levees.

Environmental flood hazard is often described using the flooding frequency over a particular period (for example, 50 years or 100 years), which reflects the empirical probability that the location has been flooded in the past (Dunne & Leopold, 1978). Historical records of flood events are often used to generate flood frequencies. Such data, however, are usually aggregated at high levels of administrative units or based on point samples collected with insufficient frequency to provide detailed spatial variability of hazard over a large area. A continuous spatial surface of flooding frequency can be derived if maps of flood inundation over a multiple-year period are available. Satellite-based remote sensing images have been used to map the extent of flood inundation and provide an effective way to create maps of flood inundation over large areas (Andreoli, Yesou, Li, & Desnos, 2007; Bhavsar, 1984; Deutsch, Ruggles, Guss, & Yost, 1973; Rango & Solomonson, 1974; Wang, Chen, Ouyang, 2002). However, satellite images are not yet available over a 50-year period, and optical sensors are mainly used to observe post-flood inundation extent because they cannot penetrate clouds, which nearly always accompany flood events. An

Q. Tian et al. / Applied Geography 63 (2015) 66e76

alternative approach is to model floods based on digital elevation models (DEMs). Because DEMs characterize the topographic basin within which flood events occur, they provide important information on flood hazard. DEMs have been used in various ways to aid flood mapping and modeling often in combination with other data, such as hydrological and hydraulic models and satellite observations of inundation extent (Correia, Rego, Saraiva, & Ramos, 1998; Liu & De Smedt, 2005; Qi et al., 2009). We combined a 30 m-resolution digital elevation model (DEM), GIS data on levee location, levee height and construction quality, and annual high lake levels from 1951 to 2001 to map flood-hazard zones in PLR. Two elements are essential in determining flood hazard in PLR: elevation and levee construction. As a floodplain of Poyang Lake, the terrain in PLR is in general flat close to the lake and gradually rises further away from the lake. It is estimated that about 57 percent of the flood prone area (defined as the area below 20.75 m) in PLR is protected by levees, and the remainder is mainly permanent and seasonal water surface (Jiang, 2006). In the areas within polders, the height and quality of levees are the important factors that affect geographic variations of flood hazard, and it is within the same polder that elevations further modify variations of flood hazard. A map of levees around Poyang Lake was created through interpretation of Landsat TM/ETM þ imagery, with additional information from published sources and field surveys (Jiang, 2006). The levee GIS data was used to adjust the DEM in order to characterize the terrain as modified by levee construction. Based on adjusted elevations, historical high lake levels recorded at Hukou were used to produce a flooding frequency map. Flood-hazard zones were then identified according to flooding probability. Levees change the natural terrain as if lifting the protected areas to a new height. But because levees can fail, and floods often occur due to levee failure, this virtual height does not provide the same level of protection as natural elevation. We borrowed the concept of discount rate from economics to discount the virtual height created by a levee and reflect its probability of failure. The modified elevation of a place behind a levee was computed as:

E0 ¼ E þ ðH  EÞ*R

69

Fig. 2. Elevations in PLR, modified to represent the effects of levees. The area in white is above 30 m.

summarized to generate the flooding frequency over a 50-year period. The flood-frequency map was then classified to create floodhazard zones (Fig. 3) using the definitions described in Table 1. These flood-hazard zones allow us to evaluate the spatial variability of flood hazard in PLR and serve as the basis for calculating exposure and sensitivity of human development to flooding. By classifying into zones, we retained the information on various locations' relative flood hazard, while simplifying quantitative information about the flood frequencies.

(1)

Where E and E0 are the DEM-based and modified elevations of a pixel respectively, H is the levee height, and R is the discount rate, which is an (inverse) indicator of a levee's failure probability. As we discussed in the section on study area, levee type is an important determinant of flood hazard in PLR and indicates how often a levee is expected to fail or be breached. We used a discount rate of 98% for crucial levees, 95% for major levees, 98% for flood storages, 90% for minor levees, and 80% for return levees respectively. These discount rates reflect the likelihood of levee failure. For example, the 98% discount rate for crucial levees can be interpreted as indicating that these levees fail once every 50 years, or that they will stand strong against floods less severe than those that occur once every 50 years. We took into account local scientists' opinions in determining the discount rates. After the DEM was adjusted (Fig. 2), historical data on lake levels from 1951 to 2001 were used to generate a flooding frequency map. During this period, the historical high-water level reached 22.59 m in 1998 at Hukou, and the lowest high-water level of 15.84 m occurred in 1972 (Jiang, 2006; Qi et al., 2009). On average, the highwater level was 19.11 m. Nine major floods occurred in 1973, 1977, 1980, 1983, 1992, 1995, 1996, 1998, and 1999 when the high-water level exceeded 20.89 m. If the adjusted elevation of a place was lower than the high-water level of a year, it was counted as having flooded once. The total number of years in which the adjusted elevation of a place was lower than the high-water level was

Fig. 3. Mapped flood-risk zones in PLR.

70

Q. Tian et al. / Applied Geography 63 (2015) 66e76 Table 1 Definitions of flood-risk zones. Flood-risk zone

Flooding frequency over 50 years (F)

Interpretation

Very low risk Low risk Medium risk High risk Very high risk

F¼0 F¼1 1 < F <¼ 5 5 < F <¼ 10 F > 10

Never flooded Flooded once every 50 years Flooded more than once every 50 years Flooded more than once every 10 years Flooded more than once every 5 years

3.2. Land-cover and population-density data A land-cover layer was interpreted from a pair of Landsat 7 ETM þ images on December 10, 1999 and July 5, 2000 (Jiang et al., 2008, Fig. 4). Two images for Landsat path 121/row 40 were used to cover land-use differences in the winter and the summer. The image in 2000 was acquired from the China Remote-Sensing Satellite Ground Station of the Chinese Academy of Sciences and orthorectified using the DEM. The image in 1999 was acquired from and orthorectified by Earth Satellite Corporation. The two images were then geometrically registered to each other. With reference to field land-use data that were collected for 131 locations around Poyang Lake, the images were initially classified into seven land-cover categories: Paddy Rice, Upland Crops, Forest, Wetland/Water, Fishpond, Urban, and Bare land. Paddy Rice and Upland Crops, which were often mixed with vegetables or orchards, were then combined into a single Farmland category. For further details of image interpretation, please refer to our paper on land cover and land use change in PLR (Jiang et al., 2008). Because Landsat TM/ETM path 121/row 40 did not cover the entire study area, we were able to collect land-cover data for only 270 of the 298 towns in PLR. Based on our analysis on land cover and land use change (Jiang et al., 2008), farmland as the main land-cover type in PLR occupied 41 to 45 percent of the total area in 1987, 1993, 1999, and 2004. There was little change in proportion of farmland especially from 1993 to 2004, during which farmland remained about 45 percent of the total area. In addition, the proportions of farmland within different types of polders were relatively stable from 1987 to 2004. About 63% of farmland below 21 m was protected by crucial and major polders. The China Data Center at the University of Michigan provided

Fig. 4. Land use land cover map in PLR, interpreted from Landsat 7 ETM þ images (path 121/row 40).

this study a population-density map at one sq km grid level and a geographic data layer that approximates town boundaries in PLR. The population-density map was derived from population data at the town level from the 2000 census with references to other GIS data that included residential areas, roads, rivers, lakes, elevations, administrative boundaries of counties and districts, and administrative areas of towns at 1:250,000 scale. The boundaries of towns were first created based on town administrative locations using Thiessen polygon method within ArcGIS. The boundaries were then adjusted using an iterative procedure implemented as a separate computer program, which took into account the impacts of roads, elevations, water bodies etc. on human dwellings and aimed at matching the areal sizes of all the towns. 3.3. Measuring the well-being of towns We used several variables to represent the three aspects of wellbeing at the town level (Table 2). For the assessment, we reclassified the five flood-hazard zones defined in Table 1 into three zones of high, medium, and low flood hazard. The high flood-hazard zone now included areas of high and very high flood hazard, the low flood-hazard zone included areas of low and very low flood hazard, and the medium flood-hazard zone remained the same. Exposure was represented by the percentage of land in the high flood-hazard zone because this measure reflects the biophysical environment of a town with respect to flood hazards and cannot easily be changed by the town. The percentage of people and the percentage of farmland in the high flood-hazard zone were used to represent the sensitivity of human development to flooding because these measures reflect the outcome of interactions between human activities and the biophysical environment with respect to flood hazards, as well as how human development can be affected by flooding. Unlike the measure of exposure (percentage of land in the high flood-hazard zone), these measures of sensitivity are changeable. We considered human lives and economic activities in examining the impact of climatic hazards. Because the major economic activities in PLR are agricultural, the spatial pattern of farmland is important. We used ArcGIS to calculate these measures of exposure and sensitivity. The percentage of the population living in the high flood-hazard zone was calculated by overlaying the population-density surface with the flood-hazard zone map and summarizing across towns. Similarly, the percentage of farmland was calculated by overlaying the farmland surface on the floodhazard zone map. We used three variables to represent human development with regard to income, literacy, and life expectancy. They were the closest match to UNDP's human development measures from among the variables available in the 2000 census data (provided by the China Data Center at the University of Michigan) at the town level in PLR. Because income was not reported in the census, the number of households per thousand that spent at least 50,000CNY in building or purchasing a house was used to capture economic development. The percentage of the population with at least high school education and the number of deaths per thousand infants under one year old were used to capture broader social aspects of

Q. Tian et al. / Applied Geography 63 (2015) 66e76

71

Table 2 Summary of variables used in the regional assessment. Well-being

Variables

Measurement scheme

Exposure

Percentage of land in high flood-risk zone

Sensitivity

Percentage of people living in high flood-risk zone

Degree of exposure: Ist Quartile: 1 2nd Quartile: 2 3rd Quartile: 3 4th Quartile: 4 Degree of sensitivity of human lives: Ist Quartile: 1 2nd Quartile: 2 3rd Quartile: 3 4th Quartile: 4 Degree of sensitivity of farmland: Ist Quartile: 1 2nd Quartile: 2 3rd Quartile: 3 4th Quartile: 4 Rank the values from low to high for all towns, and then group every 30 towns into one category. The first category scores 1, The second category scores 2, … Rank the values from low to high for all towns, and then group every 30 towns into one category. The first category scores 1, The second category scores 2, … Rank the values from high to low for all towns, and then group every 30 towns into one category. The first category scores 1, The second category scores 2, …

Percentage of farmland in high flood-risk zone

Development Number of households spending 50,000yuan (or more) in building or Level purchasing house per thousand households

Percentage of people with a high school diploma (or above)

Number of deaths per thousand infants under one year old

Overall sensitivity: the HIGHER of the two

Overall development level: the SUM of the three

Note: For those towns with missing land-use data, we used the sensitivity of population to represent the overall sensitivity of development to flooding.

development. The infant mortality rate is related to health, and reducing infant mortality rate has been specified as a major Millennium Development Goal (MDG, 2008). Using quartile assignments on each of the three aspects of wellbeing provides a good understanding of PLR towns' relative development, exposure, and sensitivity levels. These assignments also reduce the amount of information to make the assessment easily accessible to policy makers and remedy the problem created by the lack of a direct measure of income.

4. Results and discussion 4.1. Assessment results and implications for future development in PLR About one third of the land and one fifth of the farmland in PLR are at risk of being flooded more than once every ten years (Table 3). Approximately one quarter of the population lives in a location at risk of being flooded more than once every ten years (Table 3). These numbers suggest that development in PLR overall is highly exposed and sensitive to flooding. Variables representing exposure and sensitivity exhibit similar spatial patterns, with both exposure and sensitivity appearing higher closer to the lake (Fig. 5; Fig. 6). The percentage of land in the high flood-hazard zone is negatively associated with the distance to Poyang Lake (with a correlation coefficient of 0.47). The percentage of population in the high flood-hazard zone and the percentage of farmland in the high flood-hazard zone are both significantly correlated with the percentage of land in the high

Table 3 Land, population and farmland in each flood-risk zone in PLR. Flood-risk zone

Area of Land

Population

Area of farmland

Low risk Medium risk High risk Total

63.3% 7.4% 29.3% 19,874 km2

68.2% 8.6% 23.2% 7,955,966 persons

66.5% 14.0% 19.5% 7849 km2

flood-hazard zone (with a correlation coefficient of 0.97 and 0.86, respectively). The relative level of sensitivity is identical to the relative level of exposure for most towns (247 towns), and only 17 towns have sensitivity one level lower than exposure (Table 4). These facts suggest that the sensitivity of development to flooding is affected or confined by exposure to some degree. Variables representing development level do not appear to have a spatial pattern similar to that of exposure (Fig. 4; Fig. 5), and overall development level is not statistically associated with exposure (with a correlation coefficient of 0.02). There are some towns in which exposure and development influence the population in opposite directions. Fourteen towns have exposure and development levels both in the lowest quartile, and fifteen towns had exposure and development levels both in the highest quartile (Table 4). Variations in development level among towns are more related to location relative to cities and degree of urbanization. Though the correlation between development level and distance to county capital are not statistically strong among all towns (with a correlation coefficient of 0.31), the mean distance to the county capital for towns in the highest quartile of development level is significantly smaller than that of towns in other quartiles (p < 0.001). Development level and the percentage of rural population are negatively correlated (with a correlation coefficient of 0.61). The percentage of rural population alone explains 57% of the variation in development level among all towns. Also note that the three measures of development in income, health, and education are not closely related to each other. The housing variable and the education variable are correlated to some degree (with a correlation coefficient of 0.56). But the housing variable is not correlated with the health variable (with a correlation coefficient of 0.17), suggesting that a higher level of economic achievement does not guarantee improved health. Therefore to increase overall human well-being, only focusing on economic growth is not sufficient. The government needs to look at and promote both social and economic development. While the assessment in development level, exposure and sensitivity provides a comprehensive view of the state of overall development in PLR in the context of flood hazards, more

72

Q. Tian et al. / Applied Geography 63 (2015) 66e76

Fig. 5. Mapped patterns of each variable. The categories used for the variables are defined in Table 2. Green or yellow dots indicate county capitals. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 6. Classification of towns according to the three aspects of well-being. The categories for exposure, sensitivity and development level are defined in Table 2.

importantly, it reveals large variations in different aspects of wellbeing among 298 towns (Table 4; Fig. 5; Fig. 6). These variations suggest possibly different future development pathways in different places and the need for different policy interventions to improve well-being. Several types of towns deserve particular attention. Towns with an extremely high degree of exposure and sensitivity and low level of development could be candidates of the government's wetland restoration program. Many of them are near the lake (Fig. 6), and this makes them appropriate candidates. “Returning Farmland to Lake” is a first step that the government has taken towards a more ecologically-sound means of flood mitigation. This assessment provides some useful information for the government to move further in this direction. Towns with an extremely high degree of exposure and sensitivity and low level of development could also be candidate sites for natural reserves. The current natural reserves around Poyang Lake for wildlife protection are not sufficient to provide wintering habitat for the cranes, and the variety and extent of protected wetland habitats needs to be expanded (Bird Life International, 2000; Kanai et al., 2002). For this purpose, additional information on local-scale variations in lake hydrology and wetland habitats need to be combined with the measures here to prioritize preserves based on both habitat quality

Q. Tian et al. / Applied Geography 63 (2015) 66e76

73

Table 4 Descriptive statistics of well-being of towns.

Min Median Mean Max SD

Pct. farmland in the Pct. land in the high Pct. people in the flood-risk zone (%) high flood-risk zone high flood-risk zone (%) (%)

Number of households per thousand spending 50,000CNY (þ) in housing (‰)

Pct. people with a high Number of deaths per school (þ) diploma (%) thousand infants under one year (‰)

0.0 13.7 25.6 99.8 29.0

0.0 16.0 45.7 459.2 92.6

2.5 6.3 11.9 76.7 13.1

0.0 13.6 25.2 99.8 28.5

0.0 10.4 20.6 99.9 24.7

and human well-being. For highly exposed towns whose populations are also extremely sensitive, policies that induce people to migrate out may be necessary in the long run. Extremely high levels of exposure alone can reduce human well-being to such a low level that outmigration is perhaps the best solution, particularly when human life is threatened. Given long-established livelihoods tied to a particular place, it can be very difficult for households to move to a new place. Therefore, assisting farmer households in finding new livelihoods in cities or elsewhere (with a particular focus on future generations through education) should be a key aspect of migration efforts. Without such efforts, migration will be a failed strategy. In PLR, some farmers who were resettled under the Returning Farmland to Lake Policy returned to farm the land that was restored to wetland because they were unable to acquire new livelihoods (Jiang, 2006). We do need to respect local people's right to choose, though. Twelve towns in PLR had more than 90% of both their land and people in the high flood-hazard zone, of which five had more than 95% of both land and people in the high flood-hazard zone. Notice that many of those highly exposed towns whose populations are also extremely sensitive are located near the lake (Fig. 5), and they could also be considered as potential sites for wetland restoration. For those towns whose farmland is highly sensitive to flooding, agricultural practices that can reduce flood damage are important for achieving sustainable development, especially because agricultural production still significantly contributes to economic output in PLR. New land-use practices that aim to reduce flood damage and increase land profitability have been developed by agricultural scientists in PLR (Wang, Colby, Mulcahy, 2002; Yu, 2002; Yuan, Xiao, & Liu, 2002a; Yuan, Xiao, Liu, & Liu, 2002b, 2007). These practices include new rice breeds with different growth cycles or rotation patterns that can avoid severe floods or spatial patterns planned based on elevation, such as growing more flood tolerable crops in low-lying areas. They have not been widely adopted, in part because government agencies have limited human and financial resources for promoting them. The information generated by this assessment can help government agencies target dissemination efforts to places that need them the most. Thirty towns in PLR had more than 50% of their farmland in the high floodhazard zone, and only three of them had less than 50% of their land in the high flood-hazard zone. Thirteen towns had more than 80% of both their land and farmland in the high flood-hazard zone. Towns that are not highly exposed to flooding but have low levels of development may need to examine social systems to look for ways to increase their development level in the future. Most of them are far from the lake (Fig. 6). Fourteen towns in PLR had both exposure and development levels in the lowest quartile. They are the “hotspots” for the regional government to conduct further investigation into the causes of low development in order to identify possible solutions. Town governments there need to have a clear understanding of the various aspects of local development and improve the social-economic-political processes to create a

0.0 17.0 31.7 352.2 56.8

favorable environment for their citizens. Towns with degrees of sensitivity higher than exposure may need to examine their development patterns carefully to further reduce sensitivity. Thirty four towns in PLR belong to this category. Twelve towns in PLR had development, exposure, and sensitivity levels all in the highest quartile. For their development to be sustainable, strengthening engineering work (i.e. levees) may be necessary to reduce exposure and sensitivity, in addition to making appropriate adjustments to development. This assessment has several limitations due to the paucity of data at the town level in PLR. Direct measures of income would better capture the economic aspect of human development. An assessment using more recent data would generate a better understanding of the current situation. Our discussions on future development of several types of towns are also limited because we do not have detailed information about those towns, and further investigation is needed to understand their situations and provide more specific recommendations for their future development. The accuracy of above assessment can be affected by errors resulting from the classification of remote sensing images. The producer's accuracy and user's accuracy for farmland were 82% and 94% respectively (Jiang et al., 2008). Because major errors associated with farmland classification were due to confusion of farmland with forest on the images, the assessment accuracy for those towns that have significant forest coverage may suffer more than other towns. Therefore, when applying these assessment results, we need to pay special attention to those towns that have large forest areas. 4.2. Potential usefulness of the approach to other places Our approach can be applied to assessing human well-being in other less developed places that are under adverse climatic influences. The concepts of exposure and sensitivity as defined here allow us to understand the nature and impacts of extreme climatic events on development. They offer insights into whether development in a place is likely to be sustainable regarding climate variations/extremes: as long as a place is exposed to climatic variations/extremes and the development is sensitive to some degree, its development will be affected when hazardous climate events manifest and therefore unlikely to be sustainable. Exposure also serves as a reference point to sensitivity and reveals if people are doing things that exacerbate or ameliorate their risks. Together with levels of development in various aspects, they provide a comprehensive view of human well-being in less developed places that are exposed to extreme climatic events and suggest where people should make adjustments or what people may do to achieve sustainability in the context of climate variations/extremes (Table 5). Our assessment of human well-being in three aspects of development level, exposure, and sensitivity also offers some insights into adaptation to climatic change/variability and helps illuminate the concept of adaptive capacity, which is an important part of the IPCC vulnerability framework but difficult to assess due

74

Q. Tian et al. / Applied Geography 63 (2015) 66e76

Table 5 System states and possible implications. Development Exposure Sensitivity Possible implication High High High High Low Low Low Low

Low Low High High Low Low High High

Low High Low High Low High Low High

Desired state. Not doing right things e need to locate the sensitive part of development and make appropriate adjustments. Good e doing things that mitigate risks. Serious problem e may need to seek both engineering works and “soft” means to reduce sensitivity. Key issue is development, but need to make sure not to do things that exacerbate risk. Key issue is development, but need to reduce sensitivity at the same time. Key issue is development, but need to pay close attention to sensitivity and may need engineering works to keep sensitivity low. Might consider migration away as an ultimate solution.

to the dynamic nature of the concept. In the literature, adaptive capacity assessments mostly focus on the resources that facilitate coping/mitigating climatic impacts rather than the ability of dealing with change and often use indicators that reflect various aspects of development (http://www.sciencedirect.com/science/ article/pii/S0959378011000203Vincent, 2007; Engle, 2011). Despite some efforts that look at the role of institutions in adaptive capacity building (Agrawal, 2008; Eakin & Lemos, 2006; Engle & Lemos, 2010; Gupta et al., 2010; Wilby & Vaughan, 2011), research is much needed to illustrate the process of adaption so to generate useful insights for adaptive capacity building in local places (Adger et al., 2003; Hill, 2013; Lemos, Boyd, Tompkins, Osbahr, & Liverman, 2007). To adapt to environmental change appropriately and in a timely fashion, people must be willing to make change to development patterns, anticipate and take hazard into consideration when making future development plans in addition to having the resources that are necessary to make change. Thinking in terms of development level, exposure, and sensitivity forces people to consider hazard and consequently reduce sensitivity. Deliberate adaptation is an iterative process in which people assess the current situation and make adjustments accordingly. Meaningful assessments can guide people making appropriate adjustments, and assessing human well-being is the first step toward adaptive capacity building. In our framework, not only do development level, exposure and sensitivity point to where and how to make future adjustments, sensitivity as defined here in a way also provides a measure of the outcome of human adaptation: if over time people make development less sensitive to climate variations/extremes, they are adaptive and adapt in the right direction. Our assessment demonstrates that a GIS approach can capture the spatial variations in both biophysical hazard and socioeconomic development and generate useful insights for sustainable development. Such assessments can help policy makers to understand the overall state of development in the region as well as variations in development level, exposure and sensitivity across places. They can provide scientific support for policy makers in designing policies accordingly to target different problems in different places or identifying hot spots for further investigation. Using the format of maps to represent and visualize assessment results also makes the assessments easily accessible to policy makers.

5. Conclusions and future work We used a GIS approach and combined measures of development level, exposure and sensitivity to assess human well-being amid flood hazards in PLR. The assessment suggests that development in PLR overall is highly exposed and sensitive to flooding hazard with approximately one fifth of the farmland and one quarter of the population in a location at risk of being flooded more than once every ten years. We find that the sensitivity of development to flooding at the town level is closely related to (and

perhaps bound by) exposure, with both higher closer to the lake. The development level, however, is more closely associated with degree of a town's urbanization, and higher development levels are found in towns closer to county capitals. There are significant variations regarding the three aspects of well-being among the 298 towns in the region. These variations indicate different sustainable development pathways and policy interventions for different places. Future work will further investigate those “hotspot” towns identified by this regional assessment to provide additional insight into their future development. Future work will also analyze the social, institutional, and environmental factors at multiple levels and their interactions to understand the fundamental processes underlying human well-being, which will generate important insight into how to shape development in a less developed place toward a sustainable growth path. Additionally, using data from multiple time periods to examine changes will help us evaluate if the system has been developing and adapting in the right direction. Acknowledgments The first author appreciates the NASA Earth and Space Science Fellowship, Graham Environmental Sustainability Institute Graduate Fellowship and the Rackham Graduate School One-Term Dissertation Fellowship that provided financial support for this research. References Adger, W. N., Huq, S., Brown, K., Conway, D., & Hulme, M. (2003). Adaptation to climate change in the developing world. Progress in Development Studies, 3, 179e195. Adger, N. W., Paavola, J., Huq, S., & Mace, M. J. (Eds.). (2006). Fairness in adaptation to climate change. Cambridge, MA: MIT Press. Agrawal, A. (2008). The role of local institutions in adaptation to climate change. In The social dimensions of climate change. Washington, DC: The World Bank. Agrawal, A., & Chhatre, A. (2011). Against mono-consequentialism: multiple outcomes and their drivers in socialeecological systems. Global Environmental Change, 21(1), 1e3. Andreoli, R., Yesou, H., Li, J., & Desnos, Y. (2007). Synergy of low and medium resolution ENVISAT ASAR and optical data for lake watershed monitoring: Case study of Poyang Lake (Jiangxi, P.R. China). Paper Presented at ENVISAT Symposium 2007, Montreaux, Switzerland, April 2007. €schl, G. (2004). Flood risk assessment and Apel, H., Thieken, A. H., Merz, B., & Blo associated uncertainty. Natural Hazards and Earth Systems Science, 4, 295e308. Beavon, D., & Cooke, M. (2003). An application of the united nations human development Index to registered Indians in Canada, 1996. Aboriginal Conditions: Research as a Foundation for Public Policy, 201e221. pez-García, M. J., & Soriano-García, J. (2011). Mapping Belmonte, A. M. C., Lo temporally-variable exposure to flooding in small Mediterranean basins using land-use indicators. Applied Geography, 31(1), 136e145. Bhavsar, P. (1984). Review of remote sensing applications in hydrology and water sources management in India. Advances in Space Research, 4(11), 193e200. Bird Life International. (2000). Threatened birds of the world. Cambridge, UK: Bird Life Intl. Büchele, B., Kreibich, H., Kron, A., Thieken, A., Ihringer, J., Oberle, P., et al. (2006). Flood-risk mapping: contributions towards an enhanced assessment of extreme events and associated risks. Natural Hazards and Earth System Science, 6(4), 485e503. Burton, I., Kates, R. W., & White, G. F. (1978). The environment as Hazard. New York:

Q. Tian et al. / Applied Geography 63 (2015) 66e76 Oxford Univ. Press. Burton, I., Kates, R. W., & White, G. F. (1993). The environment as Hazard. New York: Oxford Univ. Press. Casale, M., Drimie, S., Quinlan, T., & Ziervogel, G. (2010). Understanding vulnerability in Southern Africa: comparative findings using a multiple-stressor approach in South Africa and Malawi. Regional Environmental Change, 10(2), 157e168. Chen, W., & Zhao, X. (Eds.). (2007). Land-use patterns and construction of security patterns in the Poyang lake region, China. Beijing, China: China Agriculture Press (in Chinese). Collier, P. (2007). The bottom billion: Why the poorest countries are failing and what can be done about it. Oxford: Oxford University Press. Correia, F. N., Rego, F. C., Saraiva, M. D. S., & Ramos, I. (1998). Coupling GIS with hydrologic and hydraulic flood modeling. Water Resources Management, 12, 229e249. Cutter, S. L., Boruff, B. J., & Shirley, W. L. (2003). Social vulnerability to environmental hazards. Social Science Quarterly, 84, 242e261. Deutsch, M., Ruggles, F., Guss, P., & Yost, E. (1973). Mapping the 1973 Mississippi floods from the earth resource technology satellites. In , 17. Proceedings of the international Symposium on remote sensing and water resource Management (pp. 39e55). Burlington, Canada: American Water Resource Association. Dwyer, A., Zoppou, C., Nielsen, O., Day, S., & Roberts, S. (2004). Quantifying social vulnerability: a methodology for identifying those at risk to natural hazards. Geoscience Australia. Record 14 Available at: http://www.ga.gov.au. Eakin, H. (2005). Institutional change, climate risk, and rural vulnerability: cases from central Mexico. World Development, 33(11), 1923e1938. Eakin, H., & Lemos, M. C. (2006). Adaptation and the state: Latin America and the challenge of capacity-building under globalization. Glob Environ Chang, 16(1), 7e18. http://dx.doi.org/10.1016/j.gloenvcha.2005.10.004. Engle, N. L. (2011). Adaptive capacity and its assessment. Global Environmental Change, 21(2), 647e656. Engle, N. L., & Lemos, M. C. (2010). Unpacking governance: building adaptive capacity to climate change of river basins in Brazil. Glob Environ Chang, 20(1), 4e13. http://dx.doi.org/10.1016/j.gloenvcha.2009.07.001. Frazier, T. G., Thompson, C. M., Dezzani, R. J., & Butsick, D. (2013). Spatial and temporal quantification of resilience at the community scale. Applied Geography, 42, 95e107. Fussel, H. M., & Klein, R. J. T. (2006). Climate change vulnerability assessments: an evolution of conceptual thinking. Climatic Change, 75,301e75,329. Gupta, J., Termeer, C., Klostermann, J., Meijerink, S., Van den Brink, M., Jong, P., et al. (2010). The adaptive capacity wheel: a method to assess the inherent characteristics of institutions to enable the adaptive capacity of society. Environmental Science Policy, 13(6), 459e471. http://dx.doi.org/10.1016/j.envsci.2010.05.006. Harttgen, K., & Klasen, S. (2012). A household-based human development index. World Development, 40(5), 878e899. Hill, M. (2013). The assessment of adaptive capacity. In Climate change and water governance (pp. 53e71). Netherlands: Springer. Houghton, J. T., Ding, Y. D. J. G., Griggs, D. J., Noguer, M., van der Linden, P. J., Dai, X., et al. (2001). Climate change 2001: The scientific basis. Cambridge: Cambridge Univ. Press. Ho, L. T., & Umitsu, M. (2011). Micro-landform classification and flood hazard assessment of the Thu Bon alluvial plain, central Vietnam via an integrated method utilizing remotely sensed data. Applied Geography, 31(3), 1082e1093. Huang, J., & Dai, S. (2004). Agricultural structure adjustment and industrial development in the Poyang Lake Region. Journal of Nanchang University, 28(2), 34e139 (in Chinese). Huang, G., Wang, S., Jiang, H., Zhao, X., & Shi, Q. (2006). Research on the development of plant industry in Poyang Lake and its surrounding economic zone. Journal of Jiangxi Agricultural University, 5(4), 20e28 (in Chinese). IPCC. (2001). Assessment report. Available at: https://www.ipcc.ch/publications_ and_data/publications_and_data_reports.shtml. IPCC (2007). IPCC (2014). Jiang, L. (2006). Flood risk and land use change in the wetland restoration area around poyang Lake, China. PhD Dissertation. Beijing, China: Institute of Geographic Science and Natural Resources Research, Chinese Academy of Sciences (in Chinese). Jiang, L., Bergen, K. M., Brown, D. G., Zhao, T., Tian, Q., & Qi, S. (2008). Land-cover change and vulnerability to flooding near Poyang Lake, Jiangxi Province, China. Photogrammetric Engineering and Remote Sensing, 74(6), 775e786. Kanai, Y., Ueta, M., Germogenov, N., Nagendran, M., Mita, N., & Higuchi, H. (2002). Migration routes and important resting areas of Siberian Cranes. Biology and Conservation, 106, 339e346. Kates, R. W. (2000). Cautionary tales: adaptation and the global poor. Climatic Change, 45(1), 5e17. Kates, R. W., & Dasgupta, P. (2007). African poverty: a grand challenge for sustainability science. Proceedings of the National Academy of Sciences, 104(43), 16747e16750. Kumar, A. S. (1991). UNDP's human development index: a computation for Indian states. Economic and Political Weekly, 2343e2345. Lemos, M. C., Boyd, E., Tompkins, E. L., Osbahr, H., & Liverman, D. (2007). Developing adaptation and adapting development. Ecology and Society, 12, 26. Leopold, L. B., & Dunne, T. (1978). Water in environmental planning. New York: W. H. Freeman and Company. Liu, Y. B., & De Smedt, F. (2005). Flood modeling for complex terrain using GIS and

75

remote sensed information. Water Resources Management, 19(5), 605e624. Ma, D. (2007). Analysis of socio-economical vulnerability to flood in the poyang Lake region. PhD Dissertation. Beijing, China: Institute of Geographic Science and Natural Resources Research, CAS (in Chinese). Malcomb, D. W., Weaver, E. A., & Krakowka, A. R. (2014). Vulnerability modeling for sub-Saharan Africa: an operationalized approach in Malawi. Applied Geography, 48, 17e30. McCarthy, J. J., Canzianni, O. F., Leary, N. A., Dokken, D. J., & White, K. S. (Eds.). (2001). Climate change 2001: Impacts, adaptation and vulnerability. Cambridge: Cambridge Univ. Press. Meyer, V., Scheuer, S., & Haase, D. (2009). A multicriteria approach for flood risk mapping exemplified at the Mulde river, Germany. Natural Hazards, 48(1), 17e39. Millennium Development Goals (MDG) (2008). Available at http://siteresources. worldbank.org/DATASTATISTICS/Resources/MDGsOfficialList2008.pdf. Min, Q. (1997). The flooding trends of Poyang Lake and strategies for flood protection and damage reduction. Disaster Reduction in China, 7(2), 33e37 (in Chinese). Montz, B. E., & Tobin, G. A. (2011). Natural hazards: an evolving tradition in applied geography. Applied Geography, 31(1), 1e4. o’Brien, K., Leichenko, R., Kelkar, U., Venema, H., Aandahl, G., Tompkins, H., et al. (2004). Mapping vulnerability to multiple stressors: climate change and globalization in India. Global Environmental Change, 14(4), 303e313. Peng, Z. (Ed.). (1999). Levee atlas of Jiangxi province. Jiangxi Provincial Government Levee Department Publication (in Chinese). Qi, S., Brown, D. G., Tian, Q., Jiang, L., Zhao, T., & Bergen, K. M. (2009). Inundation extent and flood frequency mapping using LANDSAT imagery and digital elevation models. GIScience & Remote Sensing, 46(1), 101e127. Rango, A., & Solomonson, V. V. (1974). Regional flood mapping from space. Water Resource Research, 10(3), 473e484. Ribot, J. C. (2009). Vulnerability does not just fall from the sky: toward multi-scale pro-poor climate policy. In R. Mearns, & A. Norton (Eds.), Social dimensions of climate change: Equity and vulnerability in a warming world. Washington, DC: The World Bank. Ribot, J. C., Najam, A., & Watson, G. (1996). Climate variation, vulnerability and sustainable development in the semi-arid tropics. In J. C. Ribot, A. R. Magalhaes, & S. S. Panagides (Eds.), Climate variability, climate change and social vulnerability in the semi-arid tropics (pp. 13e51). Cambridge: Cambridge Univ. Press. Rygel, L., O'sullivan, D., & Yarnal, B. (2006). A method for constructing a social vulnerability index: an application to hurricane storm surges in a developed country. Mitigation and Adaptation Strategies for Global Change, 11(3), 741e764. Santos, A., Tavares, A. O., & Emidio, A. (2014). Comparative tsunami vulnerability assessment of an urban area: an analysis of Setúbal city, Portugal. Applied Geography, 55, 19e29. Shankman, D., & Liang, Q. (2003). Landscape changes and increasing flood frequency in China's Poyang Lake Region. Professional Geographer, 55(4), 434e445. Silva, J. A., Matyas, A. J., & Cunguara, B. (2015). Regional inequality and polarization in the context of concurrent extreme weather and economic shocks. Applied Geography, 61, 105e116. Song, H., & Ma, Y. (2004). Measuring rural-urban disparity in China by human development index method. Economic Research Journal, 11, 4el5. Takeuchi, K., & Aginam, O. (2011). Sustainability challenges and opportunities in Africa. Sustainability Science, 6(1), 3e5. Thapa, S. (1995). The human development index: a portrait of the 75 districts in Nepal. Asia-Pacific Population Journal, 10(2), 3e14. Tian, Q. (2011). From vulnerability to sustainability: Rural development in the Poyang Lake region of China amid institutional changes and flood hazards. PhD Dissertation. School of Natural Resources and Environment, the University of Michigan. Available online at: http://deepblue.lib.umich.edu/handle/2027.42/89699. Tschakert, P. (2007). Views from the vulnerable: understanding climatic and other stressors in the Sahel. Global Environmental Change, 17(3), 381e396. Turner, B. L., Kasperson, R. E., Matson, P. A., McCarthy, J. J., Corell, R. W., Christensen, L., et al. (2003). A framework for vulnerability analysis in sustainability science. Proceedings of the national academy of sciences, 100(14), 8074e8079. UNDP. (1990-2014). Human development report. Available at: http://hdr.undp.org/ en/global-reports. Varis, O., Kummu, M., Lehr, C., & Shen, D. (2014). China's stressed waters: societal and environmental vulnerability in China's internal and transboundary river systems. Applied Geography, 53, 105e116. Vincent, K. (2004). Creating an index of social vulnerability to climate change for Africa. Working Paper Series at Tyndall Centre for Climate Change Research. Available at: http://www.tyndall.ac.uk. Vincent, K. (2007). Uncertainty in adaptive capacity and the importance of scale. Global Environmental Change-Human and Policy Dimensions, 17(1), 12e24. Wang, D., Chen, W., & Ouyang, Y. (2002). A study on techniques of growing highyield early rice in the flood-prone Yangzi River basin. Journal of Jiangxi Agricultural University, 24(4), 467e470 (in Chinese). Wang, Y., Colby, J. D., & Mulcahy, K. A. (2002). An efficient method for mapping flood extent in a coastal floodplain using Landsat TM and DEM data. Int. J. Remote Sensing, 23(18), 3681e3696. Wang, X., Yan, B., & Wu, G. (Eds.). (2006). The program of Mountain-River-lake. Beijing, China: Science Press (in Chinese). White, G. F. (1945). Human adjustment to floods: A geographical approach to the flood problem in the United States. Research Paper No. 29. Chicago: University of

76

Q. Tian et al. / Applied Geography 63 (2015) 66e76

Chicago Press. Wilbanks, T. J., & Kates, R. W. (2010). Beyond adapting to climate change: embedding adaptation in responses to multiple threats and stresses. Annals of the American Association of Geographers, 100(4), 719e728. Wilby, R. L., & Vaughan, K. (2011). Hallmarks of organisations that are adapting to climate change. Water Environ J, 25, 271e281. World Bank. (2002). Global economic prospects and the developing countries 2002: Making trade work for the world's poor. Washington, DC: World Bank. Yu, Y. (2002). The application of several high-return production models for flooddamage-reduction agriculture in Nanchang County. Jiangxi Agricultural Science and Technology, 3, 7e8 (in Chinese). Yuan, Z., Xiao, Y., & Liu, G. (2002a). Research on selection of flood-damage-

reduction short growth early rice breeds and their adaptability. Jiangxi Agricultural Science and Technology, 6, 2e4 (in Chinese). Yuan, Z., Xiao, Y., Liu, G., & Liu, R. (2002b). Research on flood- damage- reduction rice rotation methods and their evaluation. Jiangxi Agricultural Science and Technology, 5, 37e39 (in Chinese). Yuan, Z., Xiao, Y., Liu, G., & Liu, R. (2007). Research and practice on flooding risk reduction agriculture in the Poyang Lake Region. Journal of Jiangxi Agricultural University, 19(7), 112e114 (in Chinese). Zhao, X., & Guo, X. (2001). The analysis of sustainable development and use of agricultural land resources in the Poyang Lake region. Economic Geography, 21(4), 483e495 (in Chinese).