Flood management is multifaceted, affected by the interplay of several climatologic, hydrologic, socioeconomic and environmental factors, involving various stakeholders, competing alternatives and different tradeoffs [1
]. While the focus of conventional flood management was more on economy and society, sustainable flood management (SFM) intends to incorporate the economic, social and environmental risk of flooding in concert. To undertake such a complex task, multi-criteria decision making (MCDM) techniques has been used in flood management by providing decision makers with a systematic framework to deal with such a complex problems. MCDM approaches are capable of structuring these complex problems into a quantifiable transparent format.
The two general applications of MCDM methods in flood management are the prioritization of flood management alternatives and the mapping of flood hazard, vulnerability and risk. The latter is the focus of this study, with emphasis on flood hazard. The MCDM-based flood hazard mapping has been frequently studied [3
]. However, there has been a limited effort in transforming the environmental impacts of flooding into the MCDM-based flood hazard analysis. Understanding flood influences in terms of sustainability is essential for risk reduction [9
With the recent shift from conventional flood mitigation toward SFM [13
], more attention is given to the environmental impacts of flooding. In the past, the focus of flood management lay more on economic and technical aspects, while social and environmental criteria did not receive remarkable attention [15
]. On the other hand, SFM requires all the aforementioned factors to be considered [19
]. SFM can be defined as a framework that efficiently minimizes flood influences together with stakeholders while taking into consideration economy, society and environment both in short-term and long-term perspectives [14
]. Decision-making by using the conventional frameworks may cause conflicts between economic/social and environmental interests or bring unacceptable environmental standards particularly in a long-term perspective [24
]. Kundzewicz [22
] stated that a set of suitable selection criteria is needed for the sustainability appraisal of flood management systems. Evers et al. [25
] stated the need for a transparent and organized framework to undertake SFM. However, such frameworks have been rarely elaborated and development of well-established and coherent frameworks remains a research niche.
Bearing in mind the inherent multidisciplinarity of the SFM and the capabilities of MCDM approach in handling multiple factors, a MCDM-based SFM is essential. Despite the advantages, there have been rare applications of MCDM techniques in the SFM studies. To our knowledge, the only study was conducted by Edjossan-Sossou et al. [24
], in which a MCDM-based framework was proposed to undertake SFM. Overlooking the environmental impacts of flooding might lead to incomplete representation of the effectiveness of flood management alternatives and subsequent selection of suboptimal strategies. As a result, millions of dollars might be wasted and the ecosystem might be threatened. Thus, a MCDM-based SFM framework needs to be urgently established.
This study develops a MCDM-based framework for sustainability-based flood hazard mapping. Supported with a hydrologic model for rainfall–runoff transformation and a two-dimensional (2D) hydraulic model for flood simulation, the framework employs a GIS-based weighted summation method (WSM) to perform flood hazard mapping. Three categories of economic, social, and environmental flood hazard are used as the hazard components in order to undertake the SFM framework. Utilizing the proposed framework, the aggregated flood hazard is provided in each geographic location. The analysis is performed to analyze the severity of the flooding and does not consider the vulnerability of the receptors (e.g., buildings and people). The framework is demonstrated for the fluvial overbank flooding in the Swannanoa River watershed in the state of North Carolina, U.S. The proposed approach advances the current practice of flood hazard mapping and enhances the sustainability of the risk-reduction strategies considering the environmental impacts of flooding in concert.
2. Study Area
The study area is the Swannanoa River watershed in the state of North Carolina, U.S. The watershed, which is a part of the larger French Broad River basin, is located in the western North Carolina Mountains, from Asheville to Montreat. It is selected due to its proximity to the southeastern coast of the U.S. that exposes it to the potential path of flood-causing hurricanes and tropical storms. With the presence of steep slopes (greater than 45%), the watershed can be considered hilly. Although the watershed is predominantly rural, there are developed areas such as City of Asheville. Fluvial overbank flooding is prevalent in the watershed and there are no tidal effects. Figure 1
shows the study area including the proposed computational domain and cities as well as the U.S. states. The average topographic slope is 20.9% within the selected computational domain, implying a hilly case is analyzed here. The area has experienced several harmful floods in the past, including the events in 1916, 1928, 1940, 1964, 1977 and 2004. The most severe flooding occurred in 2004 during hurricanes Francis and Ivan, and caused $54 million damages to the structures, 11 fatalities as well as disruption to the communities in the watershed [26
]. While there are warning systems in the watershed, no certified flood control reservoir or levee is implemented. For this study, the 33.3 km Swannanoa River reach is selected, which is bounded by an area of 173.1 km2
, upstream of the confluence of the Swannanoa River and French Broad River, including some parts of Asheville.
The findings through the presented framework are subject to the limitations of the H&H models. In the hydrologic model, the nearest gaging approach was used to spatially distribute the rainfall depth to the subwatersheds. This approach impacts the timing of the derived design hydrograph and subsequently the time-variant flood parameters (velocity and duration). Therefore, the estimated social and environmental hazards, which depend on flood velocity and duration, are impacted. The impact is expected to be less on economic hazard (which was determined upon flood depth). The hydraulic model assumes a uniform Manning’s roughness for the entire simulation boundary. Since this parameter represents the resistance to flow, an uncertainty is introduced to the populated flood parameters. Flood2D-GPU is also less accurate in hilly areas due to bed discontinuity. A first-order upwind numerical scheme is likewise employed by the hydraulic model to solve the Saint Venant equations, which is incapable of capturing shocks and may yield flow discontinuities. This is unlikely to have a major impact on a rainfall-driven flooding, which was analyzed here, but can have a significant influence in dam break flooding.
The flood hazard mapping framework presented in this study can greatly help floodplain managers and insurance companies in better understanding the overall flood hazard for sustainability. With flood hazard mapping being a key element of FEMA’s National Flood Insurance Program (NFIP), the presented approach can assist in development of flood insurance rate maps (FIRMs) and planning for proper mitigation actions. A more sustainable hazard mapping approach advances the current practice and enhances the sustainability of the risk-reduction strategies proposed by the federal and local authorities. We illustrated how the integration of economic, social and environmental components promotes sustainability for flood management.
Development of a sustainability-based framework provides a more holistic understanding of the negative influences of flooding. This research focused on a single case, which allows making certain conclusions. Additional case studies using this framework will highlight the impact of watershed characteristics on the findings of this analysis. As mentioned in the Methodology section, the study watershed is mostly hilly and the flooding is often deep, flashy and short-lived. In a flat watershed, the economic and social hazard (depth and velocity were the indicators) are likely to be lower (for a given hydrograph and Manning’s roughness). Such a conclusion cannot be made for environmental hazard since it depends on all the three flood parameters (depth, velocity and duration). Furthermore, as described in Section 2
, the Swannanoa watershed is predominantly rural. A more urbanized area with a greater Manning’s value, implying deeper, slower and longer flows (for a given hydrograph and ground slope). As a result, a greater hazard in terms of economy is most likely by applying the presented framework. However, such a conclusion cannot be made for other areas in terms of social and environmental hazards since they depend on a combination of flood depth, velocity and duration (in the presented flood hazard mapping framework). Further general conclusions about the overall flood hazard in other areas, which is directly related to consequence estimation, can be drawn only through additional research and case study applications.
The hazard components in the presented framework were considered upon direct impacts of flooding. To draw more comprehensive conclusions, indirect impacts of flooding such as delays and reduced connectivity in transportation network [61
] should be taken into account. This might be done by revisiting the indicators and flood parameters (e.g., arrival time, rate of rise, time of flooding (day/night and summer/winter) and extent) that were used for each hazard component. Even in terms of direct impacts of flooding, all the influences were not taken into account. For instance, the economic hazard focused only on structural damages, while hazard to roads was neglected. Adding such additional indicators (i.e., damage to roads) would require adding other flood parameters into the economic hazard such as velocity [62
]. In environmental hazard, other indicators such as the impact on uptake rate of soil can be incorporated alongside potential erosion and pollution.
In addition to the flood hazard components, hazard estimation can be performed using alternate methods for the indicators that were considered. For instance, the commonly-used practice to estimate structural damages in the U.S. is based on the flood depth [39
], while European practice uses velocity in concert [37
]. Alternatively, other studies proposed other parameters such as duration by Thieken et al. [47
] and warning time by Graham [36
]. Using other techniques can affect the number of cells in each hazard class and subsequently the aggregated flood hazard. Therefore, the presented results should not be taken as definitive flood hazard in the watershed. Instead, the study suggests an integrated framework, which provides a more sustainable approach to floodplain managers. The framework serves the decision makers with a more informative and holistic perspective of flood hazard.
The presented framework is to determine the severity of the flooding by considering only the flood parameters (and their combinations). The vulnerability of the receptors (e.g., buildings and people) and the exposure was not taken into account. Therefore, the results should not be mistaken as flood risk, which is a combination of hazard and vulnerability [63
]. To estimate the flood risk, the information about the receptors must be overlaid with the flood hazard results of this study.
As mentioned earlier, a rainfall-driven flooding was studied in this paper. The presented sustainability-based framework is also proposed for this type of flooding. That said, the applicability of this framework to other flood types such as flash flooding and snowmelt requires additional research and should be cautiously applied to those conditions.
Since the focus of this study was to present a sustainability-based flood hazard mapping framework, a deterministic modeling approach was employed without accounting for the uncertainties. Uncertainty may arise from input variables, observed data, the choice of goodness-of-fit measures in calibration/validation, model structure and parameters [64
], nonstationarity events such as changes in climate and LULC [67
]. None of these were analyzed in this study. Therefore, it is recommended that these types of uncertainties be incorporated into the analysis and later be communicated to the managers and decision makers as they can significantly affect planning and decision-making [70
A MCDM-based framework was presented for sustainability-based flood hazard mapping in the Swannanoa River watershed, located in the state of North Carolina, U.S. The framework used a hydrologic model for rainfall–runoff transformation, a 2D unsteady hydraulic model for flood simulation and a GIS-based weighted summation method for flood hazard mapping. Three hazard components, economic, social, and environmental hazards, were used as the components to undertake the sustainability-based flood hazard mapping. Supported with a survey results, these components were combined on a cell-by-cell basis and an aggregated flood hazard map was generated, which showed the overall hazard in each geographic location. A sensitivity analysis was also performed to evaluate the sensitivity of the populated sustainability-based flood hazard map with respect to the three components weights.
While the commonly-used practice of flood risk analysis focuses on economic and social impacts of flooding, this research presented a framework to account for the environmental influences concurrently. A better understanding of the flooding impacts can be used to develop new strategies for protecting flood-prone areas. The presented framework promotes the sustainability of floodplain management decisions and guides future policy and planning decisions. The case study results serve as an example to floodplain managers and decision makers working on complex floodplain systems but our proposed approach can be applied to other regions due to its versatility.