Assessing the inundation risk resulting from extreme water levels under sea-level rise: a case study of Rongcheng, China

ABSTRACT Driven by global climate change, sea-level rise would exacerbate the hazard of extreme water level as a disaster-inducing factor. Based on Representative Concentration Pathway (RCP) 2.6, 4.5, and 8.5, this study explored the inundation risk of extreme water levels under climate change and Rongcheng was a case study. Pearson Type III (P-III) distribution was used for refitting recurrence periods of extreme water level. Expected losses exposed to extreme water levels were assessed through inundated area and depth per-unit loss values and vulnerability curves of land-use types. Results indicated that sea-level rise significantly shortened recurrence period in 2050 and 2100, which suggested a higher frequency of extreme water level in future. A large increase in expected direct losses would reach an average of 60% with a 0.82-m sea-level rise (under RCP 8.5) in 2100. Moreover, affected population and gross domestic product would grow 4.95% to 13.87% and 3.66% to 10.95% in 2050, respectively, while the increment in 2100 would be twice. Residential land and farmland were demonstrated as at greater inundation risk because of higher exposure and losses. Consequently, the intensifying hazard and the increase in possible losses suggested that sea-level rise would exacerbate future inundation risk in coastal region.


Introduction
Extreme events occur frequently as a result of climate change, such as the super storm Sandy (Trenberth et al. 2015). Extreme water levels always result in coastal inundation when a storm surge is concurrent with an astronomical high tide (e.g. Pugh 2004;Quinn et al. 2014). Recent research indicates that sea-level rise with global mean rates of 1.6-1.9 mm yr ¡1 over the past 100 years (Holgate 2007;Church and White 2011;Ray and Douglas 2011), is strongly aggravating coastal inundation (Winsemius et al. 2016). Global mean sea level is expected to rise more than 1 m by the end of this century (Levermann et al. 2013;Dutton et al. 2015), even if global warming could be controlled within 2 C. Thus, the inundation risk coupled with continuous sea-level rise should be paid attention to.
Extreme water level is one of the major threats to the people and assets in the coastal region. Compositions that affect extreme water level mainly include storm surge, tide, wave, crustal subsidence as well as sea-level rise. The contribution of storm surges to extreme water CONTACT Shaohong Wu wush@igsnrr.ac.cn levels has been concerned to-date (e.g. Sindhu and Unnikrishnan 2012). Exceedance probabilities of current extreme water levels, which were induced by tropical and extra-tropical cyclones were estimated by Haigh et al. (2014aHaigh et al. ( , 2014b. Responses of extreme inundation to wave and subsidence were highly varied on temporal and spatial scales (Sheng et al. 2010;Yang et al. 2014). For a long time span, sea-level rise resulted from undisputed global warming and is an indispensable part of extreme water level ). Due to sea-level rise, coastal flooding induced by extreme water levels would become more serious and 85% of deltas experienced severe flooding in global delta (Syvitski et al. 2009). Feng and Tsimplis (2014) showed that extreme water level around China's coastline increased 2.0-14.1 mm yr ¡1 between 1954 and 2012. Based on an ensemble of projection to global inundation risk, some researchers argued that the frequency of flooding in Southeast Asia would likely increase substantially (e.g. Hirabayashi et al. 2013). Previous studies indicated that an extreme event with current century level would become 'decade' (with 10% probability of occurrence annually) or more frequent event in 2050 (Tebaldi et al. 2012). The proportion of global urban land exposed to the highfrequency floods would increase to 40% by 2030 from a 30% level in 2000 (Guneralp et al. 2015). Conservative projections suggested that over a half of surface areas in global delta would be inundated as a result of sea-level rise by 2100 (Syvitski et al. 2009). Impacts of coastal inundation on socio-economy are considered. With the socio-economic development, large assets and aggregated population exposed to inundation risk would increase in future (Mokrech et al. 2012;Strauss et al. 2012;Alfieri et al. 2015). For example, urbanization of China was rapid in the world and many low-lying coastal cities were confronted with high probabilities of flooding (Nicholls and Cazenave 2010). More than 30% of China's coast was evaluated as high vulnerability according to the study of Yin et al. (2012), and the magnitude of population exposed to flooding risk was also great (Neumann et al. 2015). A number of China's cities including Guangzhou, Shenzhen, and Tianjin were in the top 20 among global cities on account of huge average annual losses while expected losses would increase due to the rising of sea level (Hallegatte et al. 2013).
Risk assessment of inundation integrated sea-level rise is very vital for coastal disaster mitigation and adaptation. In this study, the inundation suffered from high water levels was analysed by a combination of storm surges, astronomical high tides and changes in sea level. Comprehensive analyses of inundation risk were presented including two periods of 2050 and 2100, and three Representative Concentration Pathways (RCPs) of 2.6, 4.5, and 8.5 mentioned in Intergovernmental Panel on Climate Change Fifth Assessment Report (IPCC 2013). Combined with the hazard of extreme water levels and the vulnerability of hazard-affected bodies, effects of sea-level rise on the risk of coastal inundation were explored. Using Rongcheng as a case study, the main objectives of this study are to: (1) evaluate the recurrence period variation of extreme water level; (2) investigate the change of inundation areas and expected direct economic loss; and (3) analyse the effect of inundation on population and Gross Domestic Product (GDP).

Study area
Rongcheng is located at the tip of the Shandong Peninsula, China, surrounded by the Yellow Sea on three sides with a coastline length of 500 km (Figure 1(a)). This city has a low elevation and a flat topography as well as an area of more than 1500 km 2 . It experiences a monsoonal climate with an average annual rainfall of 757 mm and a temperature of 11.7 C (data from http://data.cma.cn/). Rongcheng is a typical coastal city in China, which is confronted with extreme marine disaster. The study area has caused great loss of $1 billion because of frequent storm surges according to historical records. Besides, it is densely populated, rapidly developed, and highly vulnerable with less defences. As shown in Figure 1(b), liveable land-use types distribute along coastal areas, such as residential land. Its population of 0.67 million people and GDP of $12.31 billion make it become one of the top hundred counties in China. Substantial additional capital investment is expected in this region where the Shandong Peninsula National High-tech Zone was approved as a part of the National Independent Innovation Demonstration Zone by China's State Council in 2016 (http://www.gov.cn/). The future social economy will present rapid development and the risk situation is urgent. Hence, it is bound to focus on disaster risk and comprehensive prevention in the regional development. Our work is a necessary foundation, especially under the uptrend of sea-level rise under global climate change.

Data sources
Data sources are summarized in Table 1 including hydrological, geographical, and statistical data.
(1) Hydrological data comprise storm surges, astronomical high tides, and sea-level rise inputs to construct extreme water levels in Section 3.2.1. Storm surges and astronomical high tides data are from observations at tidal gauge stations (Chengshantou, Shidao, Yantai, and Qianliyan), which have been continuously monitoring for approximately 50 years. Taking the accuracy and applicability of the method into consideration, astronomical high tide for each station is calculated using the T-Tide method (Pawlowicz et al. 2002), which is predicted by mainly harmonic constants (refer Zijl et al. 2013) over 19 calendar years. Storm surge series are extracted from observed time-series data by subtracting the periodically tidal component for each tide-gauge station. Given the National Third Assessment Report on Climate Change in China (2015), the future regional sea-level rise height in the Yellow Sea around Shandong Peninsula is slightly larger than global mean sea-level rise. Thereby, the heights of global mean sea-level rise based on RCP 2.6, 4.5, and 8.5 for 2050 and 2100 (IPCC 2013) are used to provide a minimum degree of future risk prevention and mitigation.
(2) Geographical data include high-precision digital elevation map, land-use map, and spatial distribution inputs on GDP and population. High-precision digital elevation map of 1:10,000 is obtained from the Bureau of Land Management and is created into10-m £ 10-m resolution using ArcGIS. Moreover, 30-m £ 30-m land-use map is acquired by interpreting Landsat TM remote sensing

Risk assessment
Risk assessment is based on the theory of disaster risk as shown in Equation (1) (UN/ISDR 2007): In this study, R is the inundation risk resulting from extreme water levels, H represents the hazard of extreme water levels, and V represents the vulnerability of hazard-affected bodies. Risk assessment of inundation follows three steps. First, the recurrent periods of extreme water levels are calculated using storm surge data, astronomical high tides, and sea-level rise heights using Pearson Type III (P-III) distribution. Here, changes of recurrence period (or probability) reflect the hazard of extreme water levels. Second, the inundated area and depth are identified by the flood model using the data of extreme water levels and a Digital Elevation Model (DEM). Third, the expected direct losses (EDL) are assessed by combining extreme water levels with different recurrence periods; meanwhile, the magnitudes of affected population/GDP are also estimated. Process of risk assessment is presented in Figure 2.
In this equation, R future and R current represent the future and current inundation risk of extreme water level, respectively; DR expresses the risk variation between future and current stage under sealevel rise. 3.2.1. Refitting cumulative probability distribution of extreme water levels Extreme water level is a compound event, which is generally caused by storm surge and astronomical high tide. Under global climate change, sea-level rise contributes to extreme water levels. Therefore, future extreme water level is a combination of current extreme water level (CEWL) with projected sea-level rise, which is defined as scenario extreme water level (SEWL). The cumulative probability distribution curves of CEWLs and SEWLs are refitted using the P-III distribution as shown in Equation (3). The details of this method are shown in our published article of Wu et al. (2016).
In this expression, f x ð Þ denotes the cumulative probability distribution of storm surge; a, b, and a 0 are the shape, scale, and location parameters, respectively; x is the annual maximum values for storm surge (namely, annual extreme storm surge); and p is the probability of occurrence.
When storm surge encounters astronomical high tide, extreme event is very likely to occur. Combining the two factors, the expression of extreme water level is simplified as Equation (4): where CEWL denotes current extreme water level, ST is storm surge and AHT is astronomical high tide. Above CEWL, sea-level rise in RCP scenarios is considered. Thus, SEWL is simplified as a superposition of CEWL and sea-level rise in the absence of their interaction. SEWL is expressed as Equation (5): where SEWL denotes scenario extreme water level and SLR is the projected height of sea-level rise. Thus, f CEWL ð Þ and f SEWL ð Þ separately denote the cumulative probability distribution of CEWL and SEWL.
where T stands for the recurrence period of extreme water level and the T-year recurrence level means that an event of extreme water level has a 1/T probability of occurrence in any given year (Cooley et al. 2007). T current and T future denote the current and future recurrent period of extreme water levels; DT represents the recurrent period change between T current and T future , which reflects the probability of extreme events. Because of the uncertain impacts of sea-level rise on storm surges, the future probabilities of storm surge are assumed to be unchanged (e.g. Hunter 2012; Kopp et al. 2013). In order to reduce uncertainties and meet the interpolation conditions, extreme water levels from Chengshantou, Shidao, Yantai, and Qianliyan stations are used to interpolate into layers. The locations of Yantai and Qianliyan stations are marked in Figure 1. As shown in Figure 3, probability distribution curves of extreme water levels under RCP 2.6, 4.5, and 8.5 are refitted by Equations (3-5) with recorded data of Chengshantou (a), Shidao (b), Yantai (c), and Qianliyan (d).

Identification of inundation
Inundated area is extracted by the flood model (the four nearest neighbours' algorithm, FNNA) based on the high-resolution DEM (grid size: 10 m £ 10 m), and extreme water level layers (grid size: 10 m £ 10 m) which are interpolated by the tool of inverse-distance-weighted (IDW) in Arc-GIS. Flooding criteria of FNNA are that extreme water level in each layer cell must be greater than or equal to the elevation in the DEM, while inundated cells must be individually connected to the sea (Xu et al. 2016). The impacts of urban landscapes and other buildings on flooding process are not considered. In this section, inundated area and depth are computed.

Inundation risk assessment
EDL are calculated using the data of inundated area and depth mentioned in 3.2.2., vulnerability curves, and per-unit loss values for each land-use type. The land-use map of 30-m resolution is resampled to be 10-m cells using the raster processing tool in ArcGIS in order to match inundated cells. The assessment model for EDL is where EDL stands for the expected direct losses of extreme floods; i denotes land-use type including residential land, farmland, woodland, grassland, and unused land; j denotes a flooded cell under land-use type i; A denotes inundated area; h stands for flood depth; r stands for loss rate (vulnerability curves); and V stands for the per-unit loss value ($/m 2 ). The amounts of affected population and GDP are estimated by the grid distribution data of population and GDP (a resolution of 1 km, published in 2010), which exposed to the region of inundation. Until the inundated depth of more than 10 cm, population and GDP will be affected. When calculating the magnitude of affected population and GDP, the zones of less than 10 cm in depth are excluded. Besides, land-use cover change and socio-economic development are not considered (e.g. Hallegatte

Recurrence periods variation of extreme water levels due to sea-level rise
Recurrence periods of extreme water levels would be shortened due to climate change through refitting SEWLs combined CEWLs with sea-level rise (Figure 4). Under each RCP scenario in 2050, the recurrence period of 50-to 1000-year would be shortened rapidly. Results demonstrate that the CEWL of 100-year recurrence period would be once in 8-31 years (RCP 2.6), 7-26 years (RCP 4.5), and 5-21 years (RCP 8.5), respectively. In 2100, it is more distinct that CEWLs would occur more frequently and even a small probability event would become normal, especially under RCP 8.5. The worst-case situation is that the CEWL of 1000-year recurrence period would occur once in three years and that of 100-year is likely to become normal, which are likely to be annual at the end of this century. As a result, the shortening of the extreme water levels would significantly increase inundation risk of coastal areas.

Inundated area and depth
Under the extreme situation, the areas flooded by CEWLs and SEWLs are shown in Figure 5. At the present stage, inundated areas range from 156.60 km 2 to 168.8 km 2 when Rongcheng encounters extreme water levels. An expanding trend in inundated area is inevitable in consideration of sealevel rise. In this analysis, the least increase in inundated area would be seen under RCP 2.6 while the most would be seen under RCP 8.5, and it would be enlarged more significantly with rising sea level in 2100 compared to 2050. Under the extreme scenario of RCP 8.5, results predict that the scope of flooding area ranges from 168.35 km 2 to 186.46 km 2 in 2050, and that it may be 187.72 km 2 to 199.18 km 2 by 2100. According to this projection, the largest inundated area nearly accounts for 13% of the whole city by the end of the century. At high level for each RCP scenario in 2100, the increase of inundated area is likely to be 14.21% to 19.54% given a 100-year return level.
Land-use types of residential land, farmland, woodland, and grassland are involved in the inundated area while water bodies and unused land could be ignored. Thus, through classifying inundated areas, the inundated land-use types are shown in Figure 6 (taking the RCP 8.5 as an example). Results show that residential land and farmland are more exposed to inundation than woodland and grassland. Indeed, at 50-to 1000-year of extreme inundation, 42.63 km 2 to 46.77 km 2 of  residential land and 34.15 km 2 to 39.97 km 2 of farmland would be affected in Rongcheng, respectively. Given a high level of RCP 8.5, inundated areas of residential land and farmland would increase to 47.61 km 2 and 41.13 km 2 in 2050 and increase to 52.88 km 2 and 51.47 km 2 in 2100. More seriously, the sum of residential land and farmland exposed to flooding would rise to around 50 km 2 in 2050 and 56 km 2 in 2100, respectively.
The inundated depth of the affected area also varies due to the height change of extreme water levels under sea-level rise. Spatial distribution of inundated depth in RCP 8.5 (high level) is shown in Figure 7 and area statistics of different depth are shown in Table 2. In 2050, depth is generally less than 3 m when Rongcheng encounters the extreme water level once in 50 years. In the case of 100-and 200-year, the submerged areas where the depth is 2-3 m and greater than 3 m are obviously enlarged. At 500-year return level, the depth of the central, southwestern, and southeastern coastal areas of Rongcheng is more than 3 m. However, more than 10% of inundated region are flooded more than 3 m at 1000-year. At the same recurrence period of extreme water level, total submerged areas and depth are significantly increased in 2100 than that in 2050. In 2100, the submerged depth at 50-year approximates that of 2050 at 1000-year. The areas with a submerged depth of greater than 3 m are mainly expanded.

Expected direct losses of inundation
Inundation damage does not only depend on inundated area and depth, but also is related to the loss rate (vulnerability curves) and value of land-use types described in Table 1. As displayed in Figure 8, the future EDL would grow compared to the loss of current extreme inundation. Exposed to CEWLs of 50-to 1000-year, the magnitude of EDL is up to $0.53 billion and $0.69 billion. EDL are enhanced by more than 20% in the RCP that the rise of sea level exceeds 0.3 m while the increase rates expand to beyond 40% with a 0.5-m sea-level rise. In 2050, estimated loss would be between $0.6 billion and $0.84 billion under RCP 2.6, which is less than RCP 4.5 and 8.5. Analysis shows that EDL would be more aggravated by the end of the century. In 2100, the minimum estimation of expected damage would appear at the low level of RCP 2.6 with a range of $0.63 billion to $0.81 billion. Moreover, the maximum range of expected damage would be from $0.88 billion to $1.08 billion at the high level of RCP 8.5. It is worth noting that the loss increase rates of different recurrence periods would reach an average of 60% at the high level of RCP 8.5 with a 0.82-m sea-level rise. However, the largest increase in direct damage would be up to 29% in 2050 and 67% in 2100. Given the high level of RCP 8.5, EDL of main land-use types are shown in Table 3. Results indicated that the possible loss of residential land is particularly prominent followed by farmland. Exposed to the extreme water levels of 50-to 1000-year, the current direct loss of residential land is about $ 437-572 million. Compared to the current losses, the EDL of residential land, farmland, woodland, and grassland would averagely increase by 134.78, 18.97, 6.53, and 2.51 million dollars in 2050 and increase by 306.62, 42.07, 15.70, and 5.48 million dollars in 2100, respectively. However, the maximum increase in submerged losses is as high as 29% in 2050 and 67% in 2100.

Population and GDP affected by extreme water levels
With the rapid socio-economic development, population and GDP distribute along the coastal region. Thus, a proportion of population and GDP are possible to be affected by extreme water levels and that exposed to flooding would be larger as a result of sea-level rise.
Affected population under RCP scenarios of 2.6, 4.5, and 8.5 are demonstrated in Figure 9(a). The amount of affected population who may suffer from the CEWLs of 50-to 1000-year is between 70,000 and 79,000. In the next decades, the increase of affected population is great and the maximum increase is close to 20,000 in 2050 and 30,000 in 2100. Considering RCP 4.5, more people (5.57% to 12.36%) would be confronted with inundation risk in 2050, while the affected population would increase by 9.52% to 23.53% in 2100.
Similarly, sea-level rise also leads to an increase in the exposure of GDP and the scope of affected GDP is presented in Figure 9(b). At present, the possible GDP of Rongcheng would be around $1.80 billion at risk of extreme floods. With the increase in inundated area under sea-level rise, a total of affected GDP rises. By 2100, additional GDP affected by flooding disaster would range from $1.82 billion to $2.23 billion. At the scenario of high level of RCP 8.5, affected GDP would have an increase of approximately 20% at the end of the century.

Discussion
Based on the theory of disaster risk, the extreme risk of inundation was assessed by integrating hazard with vulnerability. In this study, the future extreme risk under sea-level rise was highlighted. Results showed that recurrence periods would be likely reduced by more than 70% by 2050 and the decrease would exceed 80% by 2100 given high scenario of RCP. Similarly, Nicholls (2002) reported that a 0.2-m rise of sea level would significantly shorten recurrence periods, such as a 10-year would be converted into a 6-month. Indeed, as the shrinkage of recurrence periods, low-lying coastal areas would have a higher probability of flooding damage over the next few decades. Sea-level rise would enhance the inundation disaster. Results demonstrated that the potential inundated area of Rongcheng would be extended by 3% to 11% in 2050 (0.17-0.38 m) and by 5% to 20% in 2100 (0.26-0.82 m) due to sea-level rise under RCPs scenario. In contrast, sea-level rise increased the inundated area in Bangladesh by 15% with a 0.3-m rise (Karim and Mimura 2008). In terms of the analysis of land-use types, this study showed that residential land and farmland were more vulnerable to sea-level rise owing to a large potential inundated area and direct damage. According to projected SEWLs under RCPs scenario, residential land was at great risk, where EDL would be up to $0.6 billion in 2050 and even exceed $1.00 billion by 2100. Attributed to rising sea levels, it was estimated that average annual flood losses of Tianjin would be as high as $2.3 billion in 2050 (Hallegatte et al. 2013) while Shanghai would be submerged by 46% in 2100 with its seawalls and levees . Other researchers have highlighted that continuous sea-level rise would induce inundation disaster in many coastal cities including San Francisco (e.g. Gaines 2016). At the end of the century, 0.2% to 4.6% of the global population would be at risk of flooding while expected annual GDP losses would be 0.3% to 9.3% without adaptation (Hinkel et al. 2014). Based on this study, it was worth noting that rising sea levels would lead to a large number of people and property that would be faced with flooding risk, especially the fast growth of China's coastal cities (McGranahan et al. 2007;Smith 2011).
Given the recurrence of periods of shortening of extreme water levels, property and assets exposed to inundation would be more likely in future. For instance, results showed that under RCP 8.5, an extreme water level of 1000-year which might cause direct damage of $0.7 billion would occur about once every 50 years in 2050 and even to be normal in 2100. Under these circumstances, many people and industries at extreme risk from floods would have no choice but to retreat from coastal regions. However, studies indicated that most coastal inhabitants were unprepared for an increasing risk of extreme events, especially in developing countries (Woodruff et al. 2013).
This study shows that sea-level rise would significantly exacerbate the inundation risk. However, some uncertainties remain. First, because of spatial heterogeneity, regional sea-level rise and its local influence factors, such as crustal movement should be focused on the future work. Second, the dynamic interaction of tides, storm surge, and SLR are very critical to coastal risk management. In future work, coastal risk induced by climate change would be estimated and explained in the perspective of mechanism. Hydrodynamic models coupling the three interactions and other influencing factors, such as water flow, could further quantify the future coastal risk under different global warming and climate scenarios. Third, the combination of climate and weather extremes including storm surges, astronomical tides, rainfall, and sea-level rise need to be focused on, which would also generate extreme conditions or would amplify extreme events (Leonard et al. 2014). Because of the monsoonal climate of the coastal zone in China, inundation risk assessment in consideration of rainfall is particularly important (Bart van den et al. 2015;Wahl et al. 2015). At last, land-use change, population and economic development, which are feedbacks of human activities, are the driving forces of future inundation risk (Stevens et al. 2015) so that this result may be a low margin of extreme risk. Consequently, the deeper exploration aiming at these uncertainties would be undertaken.

Conclusions
This study assessed the inundation risk resulting from extreme water levels, which were combined with sea-level rise projected under different RCP scenarios in 2050 and 2100. Results demonstrated that continuous sea-level rise would augment the inundation risk by increasing expected losses and potential effect as well as shortening recurrence periods of extreme water levels. (1) Inundation risk would be increased as the increment of inundated area, direct damage, and affected population and GDP.
(2) The analysis presented here showed that sea-level rise would principally threaten the landuse types for human survival, especially residential land and farmland. (3) Projections indicated that inundation risk would continue to increase up to 2100 and would be the most serious under the RCP 8.5 scenario. (4) Sea-level rise would make low-lying coastal regions more possible to be exposed to inundation because of the recurrence periods shortening of extreme water levels. In this case, expected losses and damage caused by inundation would have a big probability under sea-level rise. In summary, these results reveal that the inundation risk would be significantly increased by sea-level rise under climate change. Effective mitigation and adaptation programs are needed to deal with the increasing coastal risk.