ABSTRACT
In this research, the authors aim to evaluate the damage caused by floods in Vinh Phuc Province. The study incorporates the construction of hydro-hydraulic models and the development of damage functions for the main elements at risk. In order to ensure the accuracy of the calculated results, the hydro-hydraulic models and damage curves are verified for their reasonableness. The outcomes of this study give us an overview of flood damage in the research area. It indicates that the potential level of flood damage is a matter of significant concern, with an annual damage estimate of approximately 6.9 million USD. The research also reveals the distribution of damage by sector and by space. According to the research findings, the greatest damage was in Vinh Tuong District, where most of the damage was caused by flooding of agricultural land. In contrast, residential buildings in the area suffered relatively little damage compared to agricultural land. This information is crucial for decision-makers to implement effective and targeted mitigation measures.
HIGHLIGHTS
Developing a hydrological and hydraulic model to flooding in Phan-Ca Lo basin. Mathematical models are effective tools for hazard assessment.
Establishing two specific damage functions for residential area and agricultural land of Vinh Phuc Province.
Assessing flood damage in Vinh Phuc Province through damage–frequency curve and the expected annual damage.
INTRODUCTION
Flood is a natural disaster that frequently causes heavy damage to properties and the environment (McGrath et al. 2019; Zeleňáková et al. 2019). Climate change has led to an increase in the frequency and severity of flooding, resulting in a higher potential for damage (Kefi et al. 2018; Cea & Costabile 2022). Despite its significance, research on flood risk in developing countries still has limitations (Romali & Yusop 2021). Flood damage is often divided into direct/indirect and visible/invisible damage (Smith & Ward 1998; Jonkman et al. 2008; Velasco et al. 2016). Direct damage is primarily caused by the physical impact of the hazard, such as structural damage to buildings. Indirect damage includes the disruption of economic activities and the reduction of investment opportunities. Visible damage can be converted into monetary values, while invisible damage cannot be directly measured. Examples of invisible damage include loss of life and damage to cultural heritage. The assessment of indirect and invisible damage is still a significant challenge (Seifert et al. 2010; Meyer et al. 2013). To simplify and make calculations feasible, many studies often focus on the direct and visible damage.
The commonly accepted approach to flood hazard assessment is simulation via mathematical models (de Moel et al. 2015; Kefi et al. 2018; Tran & Nguyen 2024). In particular, flood simulation is often applied by hydraulic models. The foundation of hydraulic models is the use of physical equations that simulate 1D/2D flow, such as the Saint–Venant equation system. The advantage of these models is that they provide detailed results on depth, velocity, and inundation period (de Moel et al. 2015). The levels of hazard can be presented by hazard maps (Tingsanchali & Karim 2010; Binh et al. 2019; Vojtek et al. 2023). Information about the spatial distribution of the key affected factors enables an intuitive overview of hazards. However, the application of hydraulic models requires a large amount of work, including data collection, model establishment, model calibration and validation. The accuracy of the simulation model is an extremely important issue. However, due to the data limitation, many previous studies often only mentioned lightly flood simulation inherited the flood simulation results or did not mention the issue of model calibration and verification. Budiyono et al. (2015) calculated the damage to Jakarta, Indonesia. The study only focused on damage curves, while the flood maps used in the study were reported by village managers. Zeleňáková et al. (2019) modelled the flood risk in Kružlov village, Slovakia using the Hec Ras model and GIS platform. This research showed us the potential result of the hydraulic model, but the model validation process was skipped. In addition to the calibration and validation of the model, the determination of the input boundary in the hydraulic model is a critical factor in the resulting calculations. To the best of the authors' knowledge, this aspect has not been adequately addressed in previous research. Overcoming these limitations is feasible in this study.
Assessing vulnerability is a challenging process because the vulnerability has multiple dimensions. Several studies assess vulnerability by individual proxy indicators or composite indicators (Stathatou et al. 2016; Mansir et al. 2021). However, the availability and uniformity of data related to vulnerability assessment indicators in different areas are some of the major limitations of the method. Assessing a relationship between the level of damage and flood factors is a globally accepted concept (Lang et al. 2010; Martínez-Gomariz et al. 2020). The factors, affecting damage, can vary depending on the sectors being evaluated. Some previous studies have developed damage curves of sectors with different flood depths (Notaro et al. 2014; Shrestha et al. 2021). Some authors combine flooding depth with other factors such as flow velocity (Lazzarin et al. 2022) or flooding period (Thieken et al. 2005) to establish a correlation with the extent of damage. Most authors developed the damage function based on survey data. Therefore, application is limited to specific areas. Furthermore, very few studies have data to verify the accuracy of the established damage functions (Seifert et al. 2010). This is also the remaining gap in previous research. To overcome this gap, the damage curves need to be verified with independent damage data from the locality to check their reasonableness.
The conventional approach to flood risk assessment encompasses hazard assessment, exposure assessment, vulnerability assessment, and their integration to form a comprehensive risk assessment (Kron 2005; Foudi et al. 2015; Luu et al. 2020). Hazard assessment involves evaluating flood characteristics, such as water depth corresponding certain frequency. The outcome of this step typically includes inundation maps that depict water depth and flooded areas. Exposure assessment involves defining the elements at risk within the flooded area, which can include residential, non-residential, agricultural, and other uses, as well as people. However, each element possesses its own susceptibility to flooding, necessitating an evaluation of their vulnerability in the subsequent step. The final stage in this process involves risk assessment, which integrates the three preceding components. Risk results can be expressed qualitatively (risk classification) or quantitatively (usually converting the level of damage in terms of money).
Damage assessment provides a foundation for establishing damage mitigation solutions. However, accurately assessing the level of damage caused by floods is still limited. The primary objective of this research endeavour is to gain a more comprehensive understanding of the flood damage situation in Vinh Phuc Province. This study encompasses a broader scope beyond merely assessing damage from a particular flood or frequency. Instead, it presents an approach that offers an overview of damage across space and frequency. To achieve this, the study develops integrated hydrological and hydraulic models to simulate flood hazards in the study area. It is imperative to validate these models to ensure the reliability of the calculated results. Leveraging the advantages of mathematical models, the study meticulously determines the flood hazard corresponding to design flood frequencies in detail. Furthermore, the study develops damage functions for some of the main elements at risk. The damage curves elucidate the specificity of the area through field surveys and data collection from localities. All of the aforementioned contents collectively aim to address existing research gaps and provide a more accurate method of damage assessment.
STUDY AREA
Vinh Phuc Province is located in the Red River Delta region of the Northern midlands and mountains, with coordinates from 21°35′15″N to 21°08′55″N (on the Red River in Dai Tu commune, Yen Lac District); from 105°20′25″E to 105°47′15″E. In recent years, the level of flood damage has become increasingly serious. The main cause is the phenomenon of prolonged heavy rain in a large area combined with the draining system not draining in time, plus the typical low-lying terrain of Vinh Yen City (Nguyen 2018). According to reported data, some typical major floods, such as in 2020 and 2022, have caused thousands of billions of Vietnam Dong (VND) in damage in Vinh Phuc Province (Vinh Phuc Province 2020, 2023). Therefore, assessing flood damage becomes urgent in Vinh Phuc Province.
METHODS
In the first step, various types of data on the research area are collected. They include many types of data such as hydrometeorological data, topography, structures affecting flood drainage, land use, and damage information. These data are analysed and processed as input for subsequent assessment steps. Based on topographic data from the 1/10,000 scale map, the study area was divided into nine sub-basins, as shown in Figure 1. The contribution flows to the river Phan – Ca Lo system were estimated from rainfall on these sub-basins. These flows were used as the boundaries for the hydraulic model. The areas of sub-basins are shown in Table 1. Additionally, the researchers gathered river cross-sections to establish hydraulic models.
Phan-Ca Lo river sub-basins, Vinh Phuc Province
ID . | Sub-basin . | F (km2) . |
---|---|---|
1 | Ca Lo Basin | 227.34 |
2 | Phan Basin | 282.89 |
3 | LV01 | 79.16 |
4 | LV02 | 130.02 |
5 | LV03 | 54.29 |
6 | LV04 | 94.74 |
7 | LV05 | 58.56 |
8 | LV06 | 49.50 |
9 | LV07 | 44.99 |
ID . | Sub-basin . | F (km2) . |
---|---|---|
1 | Ca Lo Basin | 227.34 |
2 | Phan Basin | 282.89 |
3 | LV01 | 79.16 |
4 | LV02 | 130.02 |
5 | LV03 | 54.29 |
6 | LV04 | 94.74 |
7 | LV05 | 58.56 |
8 | LV06 | 49.50 |
9 | LV07 | 44.99 |
Summary of information and data collected through questionnaire surveys
Data categories . | Collected information/data . |
---|---|
Residential buildings | Flood depth |
Flood damage (building damaged, asset damaged, etc.) | |
Building area | |
Estimate total value of residential building | |
Agriculture | Flood depth |
% productivity deduction |
Data categories . | Collected information/data . |
---|---|
Residential buildings | Flood depth |
Flood damage (building damaged, asset damaged, etc.) | |
Building area | |
Estimate total value of residential building | |
Agriculture | Flood depth |
% productivity deduction |
Survey location of subjects at risk: (a) residential area and (b) agriculture land.
Survey location of subjects at risk: (a) residential area and (b) agriculture land.
In the third step, a survey is performed to gather damage information. The accuracy of the damage results depends greatly on the quality of the collected damage data (Shrestha et al. 2021). Therefore, in this study, damage statistics were gathered through survey questionnaires. With limited time and budget, the research only focused on agricultural land and residential areas, which were the main damaged sectors based on past statistics. This study had 20 agricultural land and 15 residential area samples. The survey locations were distributed throughout the study area to ensure representativeness, as shown in Figure 3. The information was collected corresponding to the most recent major flood in 2022. There was the historical flood that occurred recently in the basin, so the collected information from local people ensures accuracy.
Due to a shortage of investigation samples, the damage curves had a large uncertainty, which can affect the reliability of the results. To address this issue, the research collected independent damage information for all sectors in Vinh Phuc Province caused by a recent flood event. This information was the damaged statistical information was summarized by the local government.
At the end of this process, the study conducted damage assessments corresponding to design flood frequencies to construct damage-probability curves (DPC) and expected annual damage (EAD). DPC represents the relationship between the level of damage associated with a specific flood frequency, while EAD represents the average level of damage over a long period (Foudi et al. 2015). The study also created damage maps corresponding to design frequencies, which allow for prioritization based on the location. This information will be useful for decision-makers in terms of socio-economic development and mitigation measures.
RESULTS
Validate hydrological and hydraulic models
The calibration and validation result for the hydrological model
Station . | Characteristic . | Calibration . | Validation . |
---|---|---|---|
Qmax (m3/s) . | Qmax (m3/s) . | ||
Ngoc Thanh | Calculate | 134 | 131 |
Observer | 122 | 118 | |
Peak error | 12.0 | 12.7 | |
NASH (%) | 91.6 | 91.8 |
Station . | Characteristic . | Calibration . | Validation . |
---|---|---|---|
Qmax (m3/s) . | Qmax (m3/s) . | ||
Ngoc Thanh | Calculate | 134 | 131 |
Observer | 122 | 118 | |
Peak error | 12.0 | 12.7 | |
NASH (%) | 91.6 | 91.8 |
Calculated and actually measured discharge flow path at Ngoc Thanh Station during flood: (a) July 1971; (b) October 1978.
Calculated and actually measured discharge flow path at Ngoc Thanh Station during flood: (a) July 1971; (b) October 1978.
Calculated and measured water level profile at Manh Tan Station for flood: (a) November 2008; (b) June 2022.
Calculated and measured water level profile at Manh Tan Station for flood: (a) November 2008; (b) June 2022.
The set of parameters is reliable and accurate in simulating runoff for the study area. This is demonstrated by the fact that the simulation results match the observed data with a Nash-Sutcliffe Efficiency Index (NASH coefficient) above 90% at both the calibrated and the calibrated steps. The calculated flood peak value is also only slightly higher than the actual measured value, providing a safety margin. The values of the parameters are shown in Table 4. In this set of parameters, the parameters CQOF, and CK1,2 are the most important. The parameter CQOF is the overland flow runoff coefficient. Therefore, it greatly affects the total flow volume. This parameter is very important because it determines the amount of excess water to form runoff and the amount of water infiltration. In basins with flat terrain, composed of coarse sand, the CQOF value is relatively small. In basins where soil permeability is poor such as clay and boulders, its value will be very huge. Meanwhile, the CK1,2 is the time constants for routing overland flow. They are very important parameters, affecting the maximum discharge and the shape of the hydrograph. Based on the set of parameters of the Ngoc Thanh basin, the study conducted micro-adjustments to suit the conditions of each sub-basin to calculate the boundaries for hydraulic models.
The set of parameter of the Ngoc Thanh hydrological model
ID . | Parameter . | Value . | Unit . | Range . | Note . |
---|---|---|---|---|---|
1 | Umax | mm | 10–20 | Maximum water content in surface storage | |
2 | Lmax | mm | 100–300 | Maximum water content in root zone storage | |
3 | CQOF | – | 0.1–1 | Overland flow runoff coefficient | |
4 | CQIF | – | 200–1,000 | Time constant for interflow | |
5 | CK1,2 | hour | 3–48 | Time constants for routing overland flow | |
6 | TOF | – | 0–0.99 | Root zone threshold value for overland flow | |
7 | TIF | – | 0–0.99 | Root zone threshold value for inter flow | |
8 | TG | – | 0–0.99 | Root zone threshold value for ground water recharge | |
9 | CKBF | hour | 1,000–4,000 | Time constant for routing baseflow |
ID . | Parameter . | Value . | Unit . | Range . | Note . |
---|---|---|---|---|---|
1 | Umax | mm | 10–20 | Maximum water content in surface storage | |
2 | Lmax | mm | 100–300 | Maximum water content in root zone storage | |
3 | CQOF | – | 0.1–1 | Overland flow runoff coefficient | |
4 | CQIF | – | 200–1,000 | Time constant for interflow | |
5 | CK1,2 | hour | 3–48 | Time constants for routing overland flow | |
6 | TOF | – | 0–0.99 | Root zone threshold value for overland flow | |
7 | TIF | – | 0–0.99 | Root zone threshold value for inter flow | |
8 | TG | – | 0–0.99 | Root zone threshold value for ground water recharge | |
9 | CKBF | hour | 1,000–4,000 | Time constant for routing baseflow |
The hydraulic model also was validated in this study. For hydraulic models, the main parameter of the model is the roughness coefficient. A trial-and-error process is used to find the appropriate roughness coefficient. Roughness coefficient values vary according to land cover. The reference value of the roughness coefficient through technical documents (Chow 1959; Brunner 2020). The study compared the calculated water level data from the model with observed data from the Manh Tan Station during two heavy flood events. The calibration process used heavy rains from 1 to 10 November 2008 and the flood from 23 May to 8 June 2022, was used to validate the model. The calculation results show that the error of maximum water level between calculation and measurement is, respectively, 0.20 and 0.21 m (Table 5), and the calculated and observed flood peaks appear at the same time and the hydrographs have the same shape (Figure 5). The study also compared the model calculation results with the survey flood marks of the 2022 historic flood at Ca Lo bridge. The results show that the water level between calculation and actual measurement is 8.72 and 8.84 m, respectively. From the results, it is clear that the model is also reliable in simulating flowing steps.
Flood damage curves
The study in question also compared the outcomes it found with damage functions from other research conducted in areas with comparable conditions, such as the average damage function for the Asian region developed by the Joint Research Centre (JRC) (Huizinga et al. 2017). Upon comparing the two damage curves perceived, in terms of visible trends, the study's damage function was lower than the JRC 2017 damage curve for both residential buildings and agricultural land. As the flooding depth increased, this difference became more pronounced. However, analysis showed that this difference was approximately 10%.
The comparison between statistical and calculated damage for residential building (a) and agriculture (b).
The comparison between statistical and calculated damage for residential building (a) and agriculture (b).
Damage-probability curve and EAD
The calibration and validation result for the hydraulic model
Station . | Characteristic . | Calibration . | Validation . |
---|---|---|---|
Zmax (m) . | Zmax (m) . | ||
Manh Tan | Calculate | 8.01 | 7.89 |
Observer | 7.81 | 7.68 | |
Peak error | 0.20 | 0.21 | |
NASH (%) | 97 | 92 |
Station . | Characteristic . | Calibration . | Validation . |
---|---|---|---|
Zmax (m) . | Zmax (m) . | ||
Manh Tan | Calculate | 8.01 | 7.89 |
Observer | 7.81 | 7.68 | |
Peak error | 0.20 | 0.21 | |
NASH (%) | 97 | 92 |
Statistics on damage level by administrative unit (USD)
District . | Residential area . | Agricultural land . | Total . |
---|---|---|---|
Binh Xuyen | 26,658 | 3,497,250 | 3,523,868 |
Phuc Yen | 24,386 | 1,606,041 | 1,630,427 |
Tam Duong | 7,507 | 285,087 | 292,635 |
Vinh Tuong | 157,881 | 4,884,708 | 5,042,590 |
Vinh Yen | 27,389 | 890,158 | 917,546 |
Yen Lac | 84,317 | 3,850,627 | 3,934,904 |
District . | Residential area . | Agricultural land . | Total . |
---|---|---|---|
Binh Xuyen | 26,658 | 3,497,250 | 3,523,868 |
Phuc Yen | 24,386 | 1,606,041 | 1,630,427 |
Tam Duong | 7,507 | 285,087 | 292,635 |
Vinh Tuong | 157,881 | 4,884,708 | 5,042,590 |
Vinh Yen | 27,389 | 890,158 | 917,546 |
Yen Lac | 84,317 | 3,850,627 | 3,934,904 |
DISCUSSION
The current investigation has validated the effectiveness of a more comprehensive approach to damage assessment. The findings of this study indicate three key aspects. Firstly, the mathematical models are effective tools for hazard assessment. Secondly, the study has established two distinct damage functions, specifically for residential areas and agricultural lands in research area. Thirdly, the researchers have employed the damage–frequency curve and the EAD to assess the extent of flood damage in Vinh Phuc Province. The results of this study are more accurate than those of earlier investigations. Previous research has often skipped the model validate process, while the development of damage curves was typically without independent verification. These factors, along with the uncertainty associated with them, can significantly impact the accuracy of calculations. Therefore, the validation of both models and damage curves represents a significant advancement in this area of study.
For research area that regularly suffers from flooding, accuracy damage assessment is very necessary. This is the foundation for localities to prepare annual reserve budgets for natural disaster prevention. Moreover, when a damage mitigation solution is proposed, a new level of damage will be computed to compare with the current situation. The effectiveness of the solution can be quantified in monetary terms, making it convenient to compare costs and benefits of mitigation solutions. This provides a good starting point for discussion and further research.
Although mathematical models require a large amount of input data as well as high-quality personnel to conduct model set-up. The mathematical model shows many outstanding advantages. Firstly, the mathematical model gives detailed results in space and time. This is very important for assessing hazards and elements at risk. In this study, we only focus on flooding depth. However, factors, such as velocity or inundation period, can be determined. These can be researched more detailed in future studies. Another advantage of using mathematical models is their flexibility. The mathematical model gives quick results so hazard assessment after each flood or with different frequencies can be done quickly. If mitigation solutions are applied, the effectiveness of the solution can also be quickly determined quantitatively. The accuracy of the damage curve is of great concern. The damage information is often compiled after each flood event. This information can be collected from localities to verify to improve the appropriateness of the calculation results. Due to limited time and data, the study only focuses on two main sectors in Vinh Phuc Province, and the number of samples collected is limited and has not met the requirements of sampling technique. One other limitation of the study is that it did not take into account land-use changes over time. The study assumes that flooding occurs while agricultural land is under cultivation, so if flooding occurs after the harvest, the level of damage may not be as high as calculated.
In the future, the study will continue to investigate, collect more samples, and include other sectors such as roads and infrastructure to provide a more comprehensive view of flood damage in Vinh Phuc Province. Additionally, the detailed information in commune level needs to be done. This would be more helpful information in minimizing damage in the study area.
CONCLUSIONS
The current study presents a comprehensive examination of the damage inflicted by flooding on residential buildings and agricultural land in the Phan-Ca Lo river basin of Vinh Phuc Province. Employing a simulation technique that integrates mathematical models with field investigations, the study relies on these key factors to conduct its research. The research result shows that mathematical models are effective tools for hazard assessment. The mathematical models give fast, detailed results and are adaptable to many scenarios. The study also assesses the vulnerability of elements at risk using damage functions. The accuracy of the damage functions is validated with independent local data sources. This reliability of the results is crucial for the study's findings.
The method described in this study will facilitate a model-based study to assess the damage that can be employed as a management tool for river basin management. In this study, the damage–frequency curve and the EAD due to flooding in the study area were calculated. The EAD in this area is 6.9 million USD, with the majority of the damage being attributed to agricultural land. Nevertheless, our research provides a detailed spatial and sector distribution of damage in this area. This information is particularly noteworthy and should receive attention from policymakers and the local people.
DATA AVAILABILITY STATEMENT
All relevant data are included in the paper or its Supplementary Information.
CONFLICT OF INTEREST
The authors declare there is no conflict.