In this study, we examine the interplay of flood risk and regional characteristics in a temporal and spatial (urban and rural) context by analyzing flood property losses in Korea. The findings show that 1) some variables are consistently important for property damage from floods over 15 years, 2) effects on property damage from floods change over time, not only in importance but also in the direction of some variables, and 3) geographic locations , which are characterized by The degree of urbanization of the area has a different pattern of flood risk in terms of land use and socio-economic factors. This study provided evidence of trends and changes in flood risk over the past 15 years by examining important factors associated with property losses from floods in a temporal and spatial (urban and rural) approach.
Introduction
In short, the objectives of this study were 1) to identify if there are trends and changes in flood risk over the past 15 years; 2) to examine whether geographical localities characterized by the degree of urbanization of an area have a different pattern of flood risk; and 3) investigate which factors are significantly associated with property losses from flooding in the temporal and city classification approach (urban and rural). First, through a literature review, we examine the theoretical concept of flood risk and its relationship with urbanization. Property damage from flooding was analyzed with 6 component variables using multiple linear regression: urbanization, hazard, topographic, land use, capacity, and socio-economic factors.
Literature Review
The determinants of flood damage
As a result, the occurrence pattern of flood damage between highly developed area and low developed area becomes different. Kunkel, Andsager and Easterling (1999) note that flood property damage may be increased in high income households due to the increase in their economic values. Although the relationship between income and disaster damage is still complex and not fully understood, it is clear that an effect of natural hazard may differ with socioeconomic status.
Researches on the flood risk in Korea
Furthermore, these two studies were difficult to apply to the national model because they focused on a particular local area for analysis. On the other hand, Choi and Seo (2013), Shim and Kim (2012) and Yoo et al (2013) try to analyze the reasons for property damage in disasters through land use and topographic features. Shim and Kim (2012) present the effect of various land use characteristics on property damage caused by natural disasters in Seoul, Incheon and Gyuonggi regions.
Research Design
Scope and procedure
We assume that the significant variable would be different between urban and rural areas. This assumption begins with the recognition of 1) the different mechanism of flooding between urban and rural areas, and 2) the close connection between independent variables - each variable has its own unique meaning and specificity, but they share a spatial characteristic that is at the same time inherent them.
Measurement
To estimate the effect of urbanization on property flood damage, we measured two indicators, the degree of urbanization of an area and the population density of an urbanized area. The degree of urbanization of an area was measured by dividing the urbanized area by the total area based on data from the National Statistics Office. River area-1 (density of main stream area), River area-2 (density of branch and small river area), floodplain area and steep slopes.
We calculated the ratio of the area of the main stream and a sum of the area of the tributaries and the small river with an area of the region. We also used a GIS to measure the floodplain area and land slope area. Finally we measured the slope-soil slope variable to assess the effect of soil erosion and landslides.
We have extracted a 60-100° slope area from the soil data and DEM file, and then calculated the soil slope degree in the area of the region. We measured six land-use variables expected to influence the extent of flood damage: Impervious area in the floodplain, Agricultural area in the floodplain area, Impervious area on sloping land, Agricultural land area in steep slope, Bare Forest Land-1 (no forest), and Bare Forest Land-2 (covered with road, stream, rock, etc.). We calculated the impervious area and agricultural land in a flood plain or sloping land using a GIS land cover layer derived from the classification of Korean Satellite-2 (Arirang-2) images with 5-meter spatial resolution from the Department of Environment in .
Finally, we measured the ratio of persons with disabilities to the total population based on the data inventoried in the KOSIS databases.
Analytical Methods
We measured four socio-economic predictors that have been shown to influence the level of flood damage: the ratio of old buildings, the ratio of vacant houses and the ratio of the number of disabled people to the total population. First, we measured old buildings by dividing the sum of buildings older than 30 years old by the total number of residential buildings in a region. We used the data from the online search engine Korea Statistical and Geographical Information Service.
It makes sense to consider old buildings more vulnerable to flooding because dilapidated buildings are more likely to develop cracks and leaks. We also measured this by dividing the number of empty houses by the total number of residential buildings in the region. Vacant houses, like variable old buildings, can easily be damaged by floods because they are usually not well managed.
Cutter (1996) suggested that the hazard potential is related to two factors; geographic context and social structure, and those two factors interact to produce place vulnerability. From this we can investigate whether socially vulnerable populations are susceptible to flood events in Korea.
Results
Descriptive statistics
The results of correlation analysis
Meanwhile, rainfall variables are negatively correlated with property damage and are not significant between 2010 and 2014. Third, river area-1 (main stream) and river area-2 (branches and small rivers) are statistically significant in all periods but in different directions. Flooded area is negatively associated with property damage flooding, contrary to our expectations, and the steepness slope is not significant in all periods.
This result seems to reflect the effects of our flood mitigation policy, because our flood mitigation measures are mostly focused on preventing river floods. Fourth, regarding land use variables, all variables have significance except deforested forest land-1 (unwooded) from 2000-2009. Meanwhile, during the 2010-2014 period, separated land use variables measured from the steep slope-land variable and bare forest land-1 were not statistically significant.
Fifth, all capacity-related variables – degree of financial independence, volume of backwater basin, and rate of supply to the sewer system – are negatively related to property damage. The degree of financial independence and the degree of supply of the sewage system show a strong correlation, but the coefficient has also decreased over the last five years. Finally, old buildings and empty houses are highly significant with a positive direction in all periods.
On the other hand, the invalid variable shows a negative correlation with property damage during the period 2000-2004.
Flood risk assessment using multiple regression analysis
On the other hand, topographic variables are statistically significant only in the rural group, pointing in different directions. The backwater volume is not statistically associated with property loss in either group, and sewer system inflow rate is negatively significant only in the urban group. But the disabled variable negatively affects property damage in the rural group, unlike the result of the urban group.
And the flood frequency and steep slope-land variable still positively predict the probability of flood damage to property. Second, capacity-related variables are not associated with flood property damage in the final model. In the urban group, five significant variables explain 48.7% (adjusted 43.7%) of the variance in the flood property damage model.
And the impervious area, farmland, vacant houses, and disability variable significantly predict property damage from flooding in the rural group. Among the components, risk-related variables and socio-economic variables are strongly associated with flood property damage in the full model. But rainfall-2 significantly reduces the possibility of property losses only in the urban group (b=-.632, p<0.05).
In the urban group, the probability of property loss due to flooding decreases as the percentage of steep slopes increases. In terms of land use variables, agricultural land in the floodplain is 0.183 times more likely to reduce property damage from flooding in urban areas, and agricultural land in areas with steep slopes is 0.259 times more likely to reduce property losses in rural areas. As with the overall model, capacity-related variables in the regional model are not significant in the period 2010-2014.
Discussion
Floodplain area negatively predicts the probability of property damage due to flooding, while area with steep slopes is positively associated with flood property damage. For example, for an area damaged by land reclamation and fire, precautions should be taken before the rainy season begins. For example, the ratio of impervious area to floodplain significantly increases the probability of flood damage in an urban area during 2000–2004.
Furthermore, the ratio of farmland in the floodplain area positively predicts the odds of flood damage over 10 years in the urban area, but is negatively associated over the past five years. This result suggests that decision-makers in urban areas should consider the land-use characteristics of the floodplain area in disaster prevention planning. Fifth, our results show that empty houses significantly increase flood damage in Korea for 15 years.
Even if the number of empty houses is increasing in rural areas as the population ages, it is difficult for the government to control this because most of the empty houses are privately owned. Due to the fact that its actual price is not assessed, the result of the flood risk assessment may differ if the damage to the building is aggregated based on its real value. Thus, additional flood risk assessment using specific property damage types appears essential to examine regional flood risk characteristics in further studies.
More information on nonstructural mitigation techniques that affect the degree of flood damage will help identify the relationship between capacity-related variables and property flood damage.
Conclusion
Detection of changes in streams and floods as a result of climate change and changes in land use and drainage. Assessing the impact of the factors of urban characteristics on damages caused by natural disasters using panel model. 2013) Recent change in the frequency–magnitude of summer extreme precipitation events over the Republic of Korea.
An analysis of the time trend of direct economic damage due to natural disasters and their determinants: the case of Gangwon-do. Urban renewal strategy for climate change adaptation: using green infrastructure in flood mitigation. The Determinants and Economic Impact of Damage from Natural Disasters (Doctoral dissertation), Seoul National University of Environmental Planning, Seoul, Korea. in Korean with English summary) Liu, J., Hertel, T.
A study on the research for the method to strengthen the disaster reduction activities and increase the investments. An analysis of the interrelationship between land use characteristics and damage caused by natural hazards.
