**1. Introduction**

Like most developing countries, Taiwan's coast has been alternatively used for settlement, agriculture, trade, industry, and recreation without careful and thorough planning in the development stage since 70s. The continuous expansion and diversity of urbanization together with the accumulation of deleterious effects on the coastal zone has intensified natural disasters in certain areas and their consequences for coastal residence. The conflicts between coastal exploitation and restriction continued because the Coastal Zone Management Act had not yet become a statutory law. This made it difficult to draw up strategies for Integrated Coastal Zone Management (ICZM), as there was no legal ground for the planning of land use in coastal areas. The Act came into effect in February 2015. It regulates the classification of coastal areas as first- or second-level coastal protection areas; and specifies which authorities are responsible for drawing up the respective coastal protection plans. Huang et al. (2016) (see also Chien et al., 2016) showed the zoning principles of the coastal protection areas which were based on the severity level of coastal hazards, however, neglecting the vulnerability [1,2]. The possible risks that a coastal area faces are not clearly identified. Thus, the most important step in facilitating coastal management is to generate risk maps in order to develop robust adaptation strategies and measures for different levels of protection areas respectively.

The concept of hazard risk analysis proposed by the United Nations Disaster Relief Organization (UNDRO, 1980) [3] involves a comprehensive examination of the relationship between hazard and vulnerability (risk = hazard × vulnerability). The hazard potential factor refers to the variability of hazards. In general, an increase in hazard intensity and frequency causes more serious damage and loss. There are five kinds of hazards for coastal areas in Taiwan, namely, storm surge, coastal erosion, flooding, ground subsidence, and tsunami (Chien et al., 2012) [4]. While the first four hazards are most frequent, the last one is relatively rare. The Coastal Zone Management Act therefore listed the first four hazards as the 'primary concerns' of coastal risks.

Coastal vulnerability can be defined as a measure of the degree to which natural hazards can affect coastal residents (McCarthy et al., 2001; van der Veen and Logtmeije, 2005; Parkinson and McCue, 2011) [5–7]. Possible losses increase when vulnerability increases (Cutter, 1996) [8]. Different variables have been used to evaluate coastal vulnerability according to the research orientation and perspective, for example the effect of sea-level rise (Khouakhi et al., 2013; Özyurt and Ergin, 2010), coastal erosion (Fitton et al., 2016; Merlotto, et al., 2016; Tarragoni et al., 2014), and sustainable development (Schernewski et al., 2014) [9–14]. Generally speaking, the variables can be categorized into geophysical, social, and socio-environmental contexts (Zanetti et al., 2016) [15]. The geophysical vulnerability focused on the relationships between physical features and coastal hazards. Hammar-Klose and Thieler (2001) used the six physical variables proposed by Gornitz et al. (1994) and Shaw et al. (1998) to assess the vulnerability of coasts in the United States [16–18]. Social vulnerability identifies the characteristics of coastal communities that enable them to respond to and recover from hazards (Cutter et al., 2003) [19]. In socio-environmental vulnerability the combined effects of both social and environmental vulnerabilities were taken into consideration (Wang et al., 2014) [20]. The United Nations Office for Disaster Risk Reduction (UNISDR, 2004) assessed vulnerability by grading physical, environmental, social and economic variables [21]. As social and economic conditions were taken into consideration, the potential threats of coastal hazards to natural surroundings and coastal residents can be clearly evaluated.

The combinations of the potential hazards and coastal vulnerability can be used to generate risk maps. Chien et al. (2013) used the risk maps to assess existing hazard prevention and coastal management measures in Taiwan [22]. Wang et al. (2014) conducted a comprehensive risk assessment strategy based on the risk matrix approach (RMA), which consisted of a probability phase and a severity phase [20]. Note that a relationship between the hazard map and the coastal protection criteria has not been clearly identified in any of these studies. Carrasco et al. (2012) and Ward et al., (2014) pointed out that flood hazard maps based on return periods could be useful in evaluating the physical damage to infrastructure, economy, and ecological resources of a region [23,24]. The risk maps with classified grades can be helpful in understanding the possible risks that a coastal area faces and determining the criteria for coastal protection.

Traditionally, coastal hazard management has concentrated on providing protection against floods through coastal defenses. However, severer marine conditions including sea level rise and the increasing intensity of storms due to climate change appear to be unavoidable and will inevitably threaten the residents of coastal areas (Van Vuren et al., 2004) [25]. To strengthen coastal defenses unlimitedly against climate change effects may lead to significant ecological losses and high cost (Penning-Rowsell et al., 2014) [26]. Nowadays, the coastal hazard management has moved away from engineering-dominated approaches to formulate adaptation strategies (Dinh et al., 2012; ESCAP/UNISDR, 2012; Luo et al., 2015; Salik et al., 2015) [27–30]. Non-engineering measures can be incorporated, for those hazards beyond the design criteria of coastal defenses. The non-engineering measures should be encouraged since, now the "Coast Zone Management Act" has come into effect, the government will have a legal basis to take action. Regulations to limit or even ban further exploitation of hazard-prone coastal areas can thus now be enacted and climate change adaptation strategies enforced.

Coastal risk maps are important for both decision-makers and engineers. They contain essential information for the former to make policies, for the latter to assess the design criteria of the coastal defenses, and are also required to implement the Coastal Zone Management Act. In this study, the coastal risk map was drawn to assess the current coastal areas whether coastal areas should be managed by mitigation and/or adaptation. We believe that this could be helpful in achieving integrated coastal management in Taiwan.

#### **2. Background**

#### *2.1. Coastal Protection Areas in Taiwan*

To promote the sustainable development of coastal areas, coastal zone planning under the Act consists of designating conservation and protection areas with the aim of establishing a spatial development strategy. Conservation areas afford legal protection to natural resources, whereas protection areas are aimed at preventing coastal hazards and protecting the lives and assets of residents. Coastal areas, consisting of shore land areas and offshore areas, were demarcated and declared by the Construction and Planning Agency of the Ministry of the Interior (CPAMI). Shore areas were defined as extending from the mean high tide line to the nearest provincial highway, major coastal road, or ridgeline. Relevant assessments were restricted to the areas defined in this manner, that is, conservation or protection areas cannot be demarcated beyond these coastal areas. The two areas were demarcated and announced in the "Integrated Coastal Zone Management Plan" (CPAMI, 2017) [31]. This study focuses on the management of coastal protection areas.

Coastal protection areas were designated as first- and second-grade according to the severities of aforementioned four hazards. Table 1 lists the severity of the hazards and zoning principles. Two types of inundations, storm surge and flood, are considered to be hazardous to coastal areas. The surge hazard is defined as the coastal area inundated by sea water. It should be noted that most of the sea walls in Taiwan are higher than the 100-year surge water levels. Most of the flooding by seawaters is due to run-ups and overtopping of waves, where the duration is very short and flooded areas are relatively small. The "Integrated Coastal Zone Management Plan" thus treats the potential of surge hazard as the elevation difference between the heights of shore land and storm surge, ignoring the function of seawalls. Flooding is due to overland flows. Coastal erosion takes erosion rates and potential effects into consideration. Ground subsidence has an added effect to the other three hazards. However, it is considered as to pose no immediate threats to coastal residences alone. Thus, areas having ground subsidence were demarcated into protection areas when overlaid with areas having one or more of the other three hazards.


**Table 1.** Grading of coastal protection areas.

After determination of the potential severity of coastal hazards, the "Coastal Protection Sectors" can be delimited following the zoning principles. First-grade coastal protection sectors are those with high potential coastal hazard. Sectors with severe ground subsidence and having areas with one other potential coastal hazard are also categorized as the first grade coastal protection sectors. The second-grade coastal protection sectors are those with one or more medium potential compound

hazards. Furthermore, coastal sectors with similar natural hazards and protection requirements were incorporated in the same grade and their extents were zoned according to the appropriate administrative boundaries or landmarks.

Figure 1 presents the distributions of four kinds of hazard prone areas. The total coastline of the first- and second-level protection areas was 478.3 km and 181 km, respectively. Most of the first-level protection areas were distributed in the southwestern region of Taiwan, in the counties of Changhua, Yunlin, Kaohsiung, and Pingtung. All these areas suffer from severe ground subsidence induced by the excessive pumping of groundwater for aquaculture. We selected Yunlin, one of these counties, for further discussion.

**Figure 1.** Distribution of hazard prone areas, and first- and second-level coastal protection areas.

#### *2.2. Coastal Protection Strategies*

In the past, coastal protection measures were formulated based on the Seawall Management Regulations, which stipulate that the measures can only be implemented within the extent of seawall areas. The zoning of seawall areas was greatly restricted because of inflexible coastal protection measures due to peripheral social and economic developments. The need to protect coastal areas from the impact of tides and waves when other auxiliary protection measures were lacking led to the construction of hard engineering structures (e.g., seawalls) with relatively strict design criteria and resulting in the construction of rather massive structures. However, extreme climate events have become more frequent and severe (Webster and Holland, 2005; Landsea et al., 2006; Elsner et al., 2008) [32–34]. For example, in recent decades, the annual minimum typhoon pressure of typhoons that have made landfall in Taiwan has gradually decreased, while the maximum wind speed has increased (Lan et al., 2013) [35]. The conventional mode of using a single protection strategy for coastal areas has thus become outdated. To depend on conventional protection modes, current protection structures must be reinforced to respond to the unpredictable trends of environmental changes. Nevertheless, the use of a single protection measure is limited with regard to economics, environmental impact, and protection effectiveness. Global practice for coastal protection remedies has gradually demarcated setback lines for coastal areas with high hazard risks. In other words, there is no longer a complete reliance on protection defense when facing unpredictable natural hazards. Instead, the conventional conception of zero disaster is discarded, hazards are allowed to occur to an acceptable extent, and attempts are made to reduce hazard-induced damage through risk management. It is suggested that both engineering and non-engineering measures should be conducted to facilitate the goals of ICZM.

#### **3. Methodology**

Coastal areas have different characteristics and degrees of exploitation; therefore, a single set of protection design criteria cannot satisfy the aim of sustainable coastal development. Environmental characteristics should be taken into consideration when formulating design criteria for the coastal defense of different regions. The process should be based on hazard risk classes to provide references for developing corresponding design criteria and formulating hazard protection and management measures.

As stated earlier, coastal protection areas in Taiwan were designated according to the criteria set out in the Coastal Zone Management Act, which did not include tsunami hazards. Wave gauging stations around Taiwan also indicate that there has been no tsunami that has caused any casualties in the last century (Central Weather Bureau, CWB; Chen and Chen, 2011; Kontar et al., 2014) [36–38]. Nevertheless, Taiwan is located in the Circum-Pacific seismic zone, and the threat of tsunamis is not negligible. Particularly, the potential threat of tsunamis originating from the Manila Trench, the East Luzon trench, and the Ryukyu trench is of concern (Lin et al., 2015; Wu and Huang, 2009; Wu et al., 2015) [39–41]. Therefore, the tsunami hazards are included in consideration.

The present study aims to establish methods for assessing the design criteria of coastal defense and land use management in various coastal areas. These methods were mainly based on a set of systematic assessment principles, from which relevant indicators were selected for further management.

Based on the risk management policy proposed by the Executive Yuan, Taiwan, the hazard risk can be defined in terms of the product of potential hazards and vulnerability. Coastal hazards were classified into the five coastal hazard types discussed above. While the hazard index can be quantified based on hazard severity, the coastal vulnerability index (CVI) must be developed relative to specific combinations of different objectives, processes, and spatial and temporal scales (Özyurt and Ergin, 2010) [10]. Furthermore, both natural and anthropogenic factors should be considered. The objective of a protection area is to minimize the impacts of coastal hazards on residents, which means that the focus is on socio-environmental concerns; in contrast, natural factors are incorporated into the hazard indices. Chien et al. (2012) suggested that vulnerability in protection areas should refer to the

possibility of life-threatening events or property loss induced by potential hazard factors in a given hazard-prone area [42]. We therefore chose socio-environmental indicators for the assessment of CVI.

First, we decided on the spatial units used to estimate vulnerability in our analysis. Although adopting large-scale units may allow easy and rapid operations and high data accessibility, the resulting failure to reflect local or regional characteristics may lead to their underrepresentation during analysis of the results. To ascertain coastal characteristics accurately, we adopted townships/villages as the analytical and statistical spatial unit, and used currently accessible data in this study. According to the spatial overlay, coastal areas in Taiwan were comprised of 110 townships, which were further subdivided into 898 villages.

Second, we selected indicators for grading before conducting risk analyses and assessments. From a statistical perspective, adopting more indicators generates results that are more representative of the characteristics of analyzed targets. However, in practice, the information required for indicators frequently fails to satisfy analytical requirements for spatial units and accuracy, and relevant survey data may even be completely lacking. This study proposed the following principles for selecting indicators:


Accordingly, the indicators chosen to assess vulnerability in this study were as follows: population density, annual comprehensive income, and land use.

Regarding indicator weights, expert consensus (e.g., the Analytic Hierarchy Process or the Delphic Hierarchy Process) has generally been relied upon in previous studies, albeit still modified by the personal approaches of the experts involved and the number of survey samples (Ward, 2014) [24]. For this reason, this study still focused on establishing a methodology and assessing its feasibility. Hazard and vulnerability factors were given equal weights in the calculation.

Table 2 graded the scores of each hazard factors and CVI for further risk analysis. Each factor is independent. Hazard potential was defined as the ratio of the hazard-prone area to the shore land area; the greater of these two indicators defined the score of the hazard factor.


**Table 2.** Classification of hazard factors and CVI.

<sup>1</sup> Hazard potential is defined as ratio of the hazard-prone area to shore land area.

Hazard-prone areas were demarcated following the criteria for "High potential" (Table 1). Areas prone to flooding due to storm surges and floods were estimated through numerical simulations. Coastal erosion and ground subsidence areas were demarcated based on survey data sourced from the Water Resource Agency of Taiwan. Furthermore, potential tsunami threats for coasts around Taiwan were taken from the results of the National Science and Technology Center for Disaster Reduction (NCDR, 2015) [43]. The detailed procedure can be found in the following case study.

Vulnerability was scored on a scale of 1 to 5, with 1 indicating the least vulnerability, and 5 indicating greatest vulnerability. The population density and comprehensive income of the 898 villages within coastal areas in Taiwan were divided into five classes, by ranking them in 20% increments in ascending order (Figure 2). Five classes were also used to score land use, with vulnerability referring to the impact on human life and property (Table 2). The level of vulnerability estimated in the risk matrix is the average score of the three indicators.

**Figure 2.** Population density and comprehensive income within coastal areas.

The hazard and vulnerability factors were multiplied in a 6 × 5 risk matrix, generating six risk classes ranging from A to F that denoted high, high-intermediate, moderate, low-intermediate, low, and minimal protection levels, respectively. These risk classes were subsequently used to determine the appropriate design criteria. The assessment procedure and framework for this method are presented in Figure 3.

**Figure 3.** Procedure and framework for coastal risk assessment.
