A General Overview of the Risk-Reduction Strategies for Floods and Droughts

: Water is a limited resource but essential to sustaining life and supporting economic development. Only 2.5% of all the water on Earth is freshwater and can be used to meet basic human needs such as a clean supply of water for drinking, cooking, and bathing. Water scarcity is the result of an imbalance between supply and demand. E ﬃ cient water resource management is deﬁnitely of interest to research and is a practical topic. At the same time, water-related disasters such as ﬂoods and droughts cause the loss of life and property. Disasters increase the di ﬃ culty of e ﬀ ective water resource management. An increase in climate extremes can also increase the risk of ﬂoods and droughts. This overview covers 150 peer-reviewed journal publications from the last twenty years focusing on risk-reduction strategies for ﬂoods and droughts. First, a deﬁnition and classiﬁcation of ﬂood and drought was introduced. Second, studies and techniques associated with risk reduction were grouped into three themes and discussed: prediction and warning; monitoring; and impact assessment, response, and management. As a result, many studies were solely focused on, and achieve excellence in, their own themes. Special attention was needed to ﬁnd in these studies what can convert the adverse impacts such as ﬂood water to positive outcomes such as drought relief. cooperation is necessary to achieve sustainability and to adapt to climate change. Finally, advanced techniques such as artiﬁcial intelligence (AI) and the internet of things (IoT) were foreseen to have a tremendous impact on future disaster risk reduction.


Introduction
The United Nations (UN) proposed 17 sustainable development goals (SDGs) in 2015, which promote economic, environmental, and social advancements to achieve a better and more sustainable future for all. The SDGs are interconnected, and none of them is less important than the others. It is a call for action by all countries who recognize that the achievement of the goals by the target date of 2030 must go hand in hand with strategies to solve and address different social and economic issues, while tackling climate change and working to preserve our living environment [1]. Water resources are essential and significantly interconnected to the content of the SDGs. For example, Goal 6 demonstrates that clean, accessible water for all is an essential part of the world if humans want to live sustainably. However, water scarcity is still an issue. Only 2.5% of all the water on Earth is freshwater, and the remainder is saltwater. Of this freshwater, two-thirds is snow and ice, and one-third is below ground. Therefore, only 0.3% of all the freshwater on Earth is available as surface water for use. The UN mentioned that more than 2 billion people are living with the risk of reduced access to freshwater resources, and at least one in four people is likely to be affected by chronic or recurring shortages of freshwater by 2050. Water scarcity is the result of an imbalance

Flood
This literature overview classifies floods into three categories: (1) pluvial flooding, (2) fluvial flooding, and (3) coastal flooding. The categories are similar to those in study [7], but this literature review uses fluvial flooding instead of general flooding to describe river floods. The details are described below.
(1) Pluvial flooding occurs in rural areas, when the rate of precipitation falling on an area exceeds the infiltration rate into the ground, and in urbanized areas, when the floodwater exceeds the capacity of the built storm drain systems. Pluvial flooding is defined as flooding that results from rainfall-generated overland flow and ponding before the runoff enters any watercourse, drainage system or storm drain, or cannot enter it because the network is full to capacity [8]. Some studies [8][9][10][11][12][13] differentiate "surface water flooding" from "pluvial flooding", and the former term is usually adopted to describe the source of flooding being surface water during heavy rainfall in urban areas. More broadly, "surface water flooding" includes pluvial flooding, sewer flooding, flooding from small open-channel and culverted urban watercourses, and overland flows from groundwater springs [8]. The discussion of differences in these two terms is beyond the scope of this review and will not be addressed hereafter. It appears that the frequency of pluvial flooding caused by extreme rainfall events is leading to increased impacts in terms of the threat to life and damage. The frequency is possibly increased further as a result of climate change [9,10]. The term "flash flooding" attracts much attention because of its unpredictability and disaster-causing characteristics [13,14]. Flash flooding may also be associated with high-intensity rainfall and cause flooding within 6 hours [14]. Therefore, the details of "flash flooding" are provided in the section on "fluvial flooding".
(2) Fluvial flooding, similar to pluvial flooding, usually occurs when an excessive amount of rainfall exceeds the capacity of a river. In some areas, such as North America, fluvial flooding can also be caused by heavy snowmelt and ice jams [15,16]. There are two main types of riverine flooding: One is "overbank flooding", which is the most common, and occurs when water rises to overflow the edges of a river. Second, "flash flooding" is characterized by an intense, high-velocity torrent of water that occurs in a river within 6 hours or less. The causes of flash floods can be high-intensity rainfall or the sudden breakage of a dam or levee. Flash floods along with debris are very dangerous and destructive not only because of the force of the water but also because of the short warning and response times in comparison with that of normal floods [17]. Flash floods can also occur in mountainous areas due to steep topographic and geological conditions. These conditions may intensify flash-flood damage, but most damage is still found in flat lowlands where population and property are often concentrated [18]. Studies have shown that fluvial flooding is an essential natural process for river and floodplain ecosystems [19,20]. Maintaining the proper functioning of river ecosystems from floods while minimizing damage is definitely a research topic of interest.
(3) Coastal flooding in low-lying areas is usually caused by wind waves and elevated water levels [21]. They are usually generated by large waves, storm surge, high tides, and mean sea-level anomalies. In some areas, such as deltas and estuaries, precipitation and river flow may also contribute to coastal flooding [21]. Many studies have predicted that with climate change, there will be an increase in the intensity and frequency of tropical cyclones and sea-level rise [22][23][24]. As a result, coastal flooding from storm surges will become more frequent with sea-level rise [24]. Furthermore, densely populated regions affected by coastal flooding from tropical cyclones have experienced a rate of sea-level rise near or greater than the global average [23]. It is expected that the population will continue to grow along coastal areas in the future. The frequency of coastal flooding will increase as a result of accelerated sea-level rise and frequent tropical cyclones [22]. Unfortunately, most coastal populations are still not prepared for an increase in the frequency of extreme coastal flooding or a significant rise in sea level [23].

Drought
A precise definition of drought is not clear because of different hydrometeorological conditions, geographical locations, or even stochastic water demands in different regions around the world [25]. For example, the World Meteorological Organization (WMO) [26] and UN [27] identified drought as a condition of deficiency in precipitation. It can be identified as a condition in which crops fail due to the lack of moisture in the soil [28]. It can also be defined as a condition in terms of low annual daily streamflow [29]. Droughts are typically classified into four categories: (1) meteorological drought, (2) hydrological drought, (3) agricultural drought, and (4) socio-economic drought [25,30,31]. The details are described below. It is noted that groundwater drought is not included in this literature overview because only limited research has been done on the occurrence and propagation of droughts in groundwater. Some studies considered groundwater drought as a new type of drought [25,32]. However, groundwater drought is difficult to quantify due to a lack of comprehensive groundwater observations at regional and global scales [32]. Therefore, groundwater drought may be treated as a new type of drought when more data are available [25]. The details of the four types of drought are described below.
(1) Meteorological drought is defined as precipitation deficits over a region for a period of time. It is considered drought when precipitation is below the average values [33]. The time frame to determine drought can be monthly or seasonal precipitation [34,35]. A previous study has shown that meteorological drought is highly related to sea surface temperature (SST) as well as temperature anomalies. However, the level of deficit required to define a meteorological drought is not yet clear [35].
(2) Hydrological drought is related to a period with surface and subsurface water resource (drought in rivers, lakes, and groundwater) shortfalls on adequate water supplies for established water uses (e.g., sources of drinking water or support for aquatic animals and ecosystems) of a given water resource management system [25]. Therefore, the assessment of hydrological drought is crucial for water resource management. Instead of precipitation data used for meteorological drought, streamflow data have been widely applied for hydrologic drought analysis [36,37]. Hydrological droughts can cover extensive areas and can last for months to years [38]. Major differences between meteorological and hydrological droughts are related to the features of climate and catchment [37].
(3) Agricultural drought refers to a period with declining soil moisture and consequent crop failure without any reference to surface water resources [25,39]. Drought starts with meteorological drought and persistent dry conditions and then induces agricultural, hydrological, and agricultural droughts [35]. A decline in soil moisture depends on several factors, such as soil properties, climate, weather, topography, and land cover [40]. For example, when evaporation is greater than precipitation, a depletion of soil water and crop wilting occur to trigger agricultural droughts. Food production is in jeopardy and agricultural droughts cause socio-economic impacts.
(4) Socio-economic drought is associated with the failure of available water resources to meet water demands, thus associating droughts with the supply of and demand for an economic good [25,31]. Socio-economic drought occurs when the demand for an economic good exceeds supply as a result of a weather-related shortfall in water supply [25]. Therefore, the first three drought categories address ways to measure drought as a physical phenomenon. Among the four types of drought, socio-economic drought has a direct impact on human production and daily life [41].

Disaster Classification
Definition Reference
[ [36][37][38] Agricultural Declining soil moisture and consequent crop failure are without any reference to surface water resources. [35,39,40] Socio-economic The demand for an economic good exceeds supply of a weather-related shortfall in water supply. [41]

Recent Research on Disaster Reduction
Measures taken to prevent or mitigate the impacts of floods and droughts can be identified as structural and nonstructural measures. Common structural measures, such as retention ponds, dams, river improvement, urban drainage systems, and levees or dikes, are used to store floodwater, decrease peak stage, and retain floodwater to mitigate the impact of floods [42][43][44]. Nonstructural measures include early flood warning, hydrological taxes, flood emergency planning and response, and environmental education [44,45]. Structural measures are engineering constructions that aim to reduce or remove the volume of floods. Some studies argue that structural measures are fundamental to solving most flood problems, but in addition to being costly, they do not by themselves effectively and sustainably resolve the most complex problems [45,46]. Nonstructural measures are applied to prevent and reduce damage through actions, legislation, standards, and programs [45,47]. One study showed that flood impacts are more efficiently reduced with structural strategies compared to nonstructural strategies [48]. Nevertheless, the benefits based on the avoided damages from these measures must be compared to their costs. Therefore, a cost-benefit analysis must be carried out for further evaluation [48]. Measures to mitigate drought impacts are classified into three categories: (1) water-supply measures, (2) water-demand measures, and (3) impact minimization measures [49]. The first two measures tackle water-scarcity issues, and the third measure minimizes the social, economic, and environmental impacts of drought on human society. Structural measures can increase new water storage, such as the construction of new reservoirs, and improve existing water system efficiency, such as leak detection. Nonstructural measures are the development of drought warning systems to identify early warning signals of drought and the taking of response measures and policy-related actions that can be taken to reduce the impact on society. For example, Texas in 2011 suffered from drought and the government used an insurance program to pay out $10.8 billion for agriculture compensation [50]. Other actions, such as the reduction of water usage for industrial sectors and incentives for low flow fixtures, can also be taken according to the drought stage and existing conditions. This overview focuses on nonstructural measures that are more proactive and easier to apply and understand from the public point of view. Meanwhile, nonstructural measures are also relatively cost effective, require a short time to implement, and supplement structural measures well. Many published papers were reviewed following the major themes and subtopics in the years from 2000 to 2020. A typical disaster-reduction cycle consists of prevention, preparedness, and emergency response as well as recovery and rehabilitation. In terms of research themes corresponding to the cycle, this review identified three major themes: prediction and warning; monitoring; and impact assessment, response, and management ( Figure 1). The three categories are interconnected. In other words, the initiated item can be any of them. For example, flood disaster mitigation can start with the establishment of a monitoring network and then provide prediction and warning information to decision makers for taking necessary response measures. A systematic overview of these three categories is provided here, and Table 2 shows the number of published papers according to different themes from 2000 to 2020. The details of each theme are discussed below.

Flood
Prediction and warning are proactive approaches to disaster management. The most common information used for the prediction and warning for floods are flood levels and peak discharges. Along with an emergency response strategy, this approach can be implemented before and during the events to evacuate residents or to prevent the failure of structural measures, such as levees or reservoirs. The impact of disasters can be decreased due to precautionary measures. Studies have shown that the development of an early warning system is the most cost-effective measure because it saves lives and property [51,52]. There are various approaches to simulate flooding when rainfall information is given, and the most common approaches are divided into three groups: data-drivenbased methods, physics-based models, and other alternatives [53]. The latter considers mass conservation and moment equations to better describe flow behaviors and provide high spatial Prevention Preparedness Emergency response

Recovery and rehabilitation
Predictions and warnings Monitoring Impact assesment, response, and management

Flood
Prediction and warning are proactive approaches to disaster management. The most common information used for the prediction and warning for floods are flood levels and peak discharges. Along with an emergency response strategy, this approach can be implemented before and during the events to evacuate residents or to prevent the failure of structural measures, such as levees or reservoirs. The impact of disasters can be decreased due to precautionary measures. Studies have shown that the development of an early warning system is the most cost-effective measure because it saves lives and property [51,52]. There are various approaches to simulate flooding when rainfall information is given, and the most common approaches are divided into three groups: data-driven-based methods, physics-based models, and other alternatives [53]. The latter considers mass conservation and moment equations to better describe flow behaviors and provide high spatial details in terms of flood forecasts. For example, one-dimensional modeling in the hydrodynamic model usually assumes that flow discharge and water level along the flow direction are parameters of interest. The most well-known model is the HEC-RAS model, which was developed and is continuously maintained by the US Army Corps of Engineers [54]. It is a commercial free software, and many studies have applied it for various purposes, not only limited to flood modeling but including sediment transport simulation and dam break, etc. (e.g., [55][56][57]). The newer version of HEC-RAS even extends its capabilities to two-dimensional inundation modeling [58]. Regarding two-dimensional models, Saint-Venant equations are the most applicable assumptions in two-dimensional flood modeling. Most well-known models, such as HEC-RAS, CCHE2D, MIKE21, and SOBEK, are designed to provide more details regarding fluid motion, such as flow velocities in the longitudinal and transverse directions (e.g., [59][60][61]). Some studies considered lumped and distributed hydrological models, such as SWAT, AnnAGNPS, WEPP, or APEX, to provide peak flow at watershed scale [62][63][64]. The usefulness of lumped hydrological models for forecasting may be limited by their coarse resolution, the need for long-term historical data for calibration, and an inapplicability to poorly gauged catchments [14]. When the data is limited or there are no gauging stations at the study area, for example precipitation data is missing or there is not enough, numerical models can be applied to generate future data of flow used in models such as HEC-RAS for flood forecasting [65][66][67][68]. However, research on data scarcity or ungauged basins is another fast-growing topic but beyond the scope of this overview. The topic itself deserves another systematic review. Computational cost and data requirements are significant in solving the detailed governing equations. During emergency response, efficiency and numerical stability become very important for further applications in flood prediction and warning. As a result, a variety of alternatives have been developed to improve the computing efficiency. Despite active research in the field, rapid and accurate flood modeling at high spatiotemporal resolutions remains a significant challenge in hydrologic and hydraulic studies. The data-driven-based model, also referred to as the black-box model, is one of the models to attract the most attention. The model itself may not include the abovementioned governing equations but instead involves mathematical equations from the analysis of time series data. The analysis can then be used to provide flood predictions and warnings e.g., [69,70]. The quantity and quality of data have a significant impact on the performance. Its performance also deteriorates when the forecast lead time is extended [71]. Other than the abovementioned categories, simple physics-(or conceptual models), threshold-, or index-based models have also recently been discussed by many studies (e.g., [18,[72][73][74]). Both simple physics-and index-based models can provide rapid flood hazard assessments in different spatial and temporal solutions. Therefore, the requirements of emergencies in terms of efficiency and reliability are easily met. However, due to their simplicity in assumptions, they should not be considered as an alternative to the complex hydrometeorological models but can be immediate tools for purely nontechnical decision makers in the case of early warnings and flash floods [75]. As a conclusion, the trend of prediction and modeling from traditional physics-based flood models moves to simple physics-based flood models to meet the needs of operational purpose. Because of rapid development in data science, the artificial intelligent or data-driven models are getting most researchers' attention if observation data is provided. Among all models, Adaptive Neuro Fuzzy Inference System (ANFIS), Support Vector Machine (SVM), and Artificial Neural Network (ANN) are most commonly used [69][70][71].

Drought
Short-term or immediate measures are of great benefit to reduce flood damage. Drought, unlike floods, is usually a long-term effect. The slow progression and inception of drought can be advantageous. Decision makers can foresee these events in advance and therefore take corresponding responses [76]. Therefore, long-term but not short-term advance prediction and warning are most important for early preparedness to reduce the impact of drought. As addressed in the previous section, drought can be classified into four categories. Among these, meteorological drought can precede and lead to the three other categories [77]. Because of this prerequisite and the fact that drought involves a long-term variation in climate, prediction and warning are usually focused on meteorological drought. The two most common approaches to provide early drought information are dynamic modeling and statistical modeling. The latter approach includes regression analysis and stochastic, probabilistic, and artificial intelligence-based modeling [78,79]. Dynamic modeling can provide real-time results. Given accurate real-time environmental monitoring information, such as rainfall, streamflow, and temperature, and remote sensing data, such as broader earth weather or the water content of the soil, drought monitoring and prediction systems can be established [80][81][82]. The results from dynamic modeling can provide predictions and warnings for meteorological drought. Along with hydrological modeling and socio-economic methods, three other types of drought can also be evaluated [83]. Dynamic modeling usually involves large-scale observations in larger areas, and its complexity, such as the model structure itself and the requirement of computational resources, makes dynamic drought modeling difficult to adopt by individuals. Monitoring information such as remote sensing data serves as input, and its accuracy, therefore, has a significant impact on the performance of dynamic modeling. In contrast, statistical modeling is relatively simpler than dynamic modeling. Many drought indices have been derived for assessing the effect of a drought and defining drought parameters, such as the intensity severity and spatial extent [79]. Among them, as rainfall is the primary source for water resources, the standardized precipitation index (SPI) is the index most adopted by different countries [84]. Many studies have applied the SPI as a drought-quantifying factor to predict drought conditions at various time scales (up to 24 months lead time) with reasonable accuracy [69,[85][86][87]. Different statistical models have their advantages and disadvantages. For example, the regression model is the simplest and most direct model. Its linearity assumption may limit its application for longer lead time forecasting. Probabilistic models show good performance but need larger computational resources in comparison with that of other statistical models [78]. Similar to flood forecasting, drought forecasting traditionally was done based on dynamic modeling or observation-based indexing. The trend of recent research is moving to the application of artificial intelligence [79,84,88,89]. As with most applied modeling in flood forecasting, ANN, ANFIS, and SVM are also the most mentioned models for drought forecasting [78,90,91]. The SVM model seems to outperform other artificial intelligence models, especially for long-term forecasting [86,90]. This ca be explained by SVMs seeking to minimize the generalization error while other models seek to minimize the training error. Its advantage is to avoid overfitting and local minima [78]. Regardless of what data-driven models are used to provide predictions or warnings, long-term and accurate monitoring information is important to serve as inputs and has significant impacts on forecast performance.

Monitoring
Monitoring data that are ground or remote sensing based, such as rainfall, weather, crop conditions, soil moisture, stream stage, and flow, serve as water availability inputs in early warning systems for prediction [18,[92][93][94]. The temporal length and spatial coverage of the data can also help the understanding of extreme events in detail [17,34]. The reliability and credibility of data determine the performance of prediction [95]. The spatial coverage of information such as rainfall observation data is always an issue because gauging stations are never dense enough to characterize events [88]. Other than being inputs to the models, information from ground-based gauges is also used as ground truth to calibrate and validate the model performance. Ground-based monitoring systems can only offer limited coverage in terms of spatial extent. Remote sensing-based observations such as satellite and radar can complement data collected by ground systems. Satellites can observe soil and vegetation moisture or provide images to estimate the SPI to understand drought conditions [96][97][98]. A technique called data assimilation integrates remote sensing-based data and improves the performance of large-scale models for extreme forecasting, such as weather prediction models (WRFs) [99][100][101]. Along with artificial intelligence, satellite observations can then provide precipitation estimations [102]. Recent studies have applied satellite images for flood detection or flood mapping [103,104]. This approach is particularly helpful for those areas or countries without adequate ground-based observations [105,106]. The improvement in radar observation techniques in dual-polarization radar instead of single-polarization radar provides more accurate rainfall estimation [107]. Quality assurance (QA) and quality control (QC) for the data are essential for subsequent applications. The common approach is to adaptively correct remote sensing-based data with ground-based observations [108,109]. Recent studies have used data-driven or data-mining techniques to identify possible data bias or extremes and further correct them simultaneously to meet the needs of disaster response [110][111][112]. Finally, because of rapid progress in the internet of things (IoT), sensors and associated components are becoming cheaper than ever before. The deployment of large-scale and economically efficient sensory networks is expected, and the quantity and quality of the monitoring can be increased and improved accordingly [113,114]. New techniques, such as the application of commercial microwave links for rainfall observations, the analysis of pictures from cameras, or the crowdsourcing of personal weather stations, are rapidly developing to extend the horizon of monitoring for sustainability [115][116][117][118]. Other than the monitoring task itself, gauging stations according to their locations (e.g., upstream of a watershed) can be used to for flood forecasts at a downstream area [119,120]. The early warning systems based on upstream gauging stations can provide predictions within a watershed but only with a limited lead time. If the time of concentration of the watershed is short (i.e., less than a few hours), then the time window of disaster response is also narrow. Efficient numerical models such as HEC-RAS, rapid flood evaluation models, or data-driven models are commonly applied to meet the requirement of a short response time [119,121]. In conclusion, much research has put a lot of effort into remote sensing monitoring technologies such as Synthetic Aperture Radar (SAR) flood detection [122]. Its 24-hour, all-weather monitoring capability can improve the accuracy of inundation area detection. In addition, it is hard to build a dense monitoring network due to high cost of monitoring devices. It is expected that much research will pay attention to low-cost devices and use them to provide accurate and high-frequency monitoring data.

Flood
Impact assessment, in terms of different disaster types, is carried out on the basis of the persistence of stressed conditions, population, and critical infrastructure and, finally, its effect on social and economic losses. The response means the measures during and after events to mitigate the impacts of disasters. Management includes planning for water management through policymaking, better water and crop management, and increased public awareness and education. Impact assessment is a tool to support decision making. It not only provides flood risk assessment but also directly translates the risk into the expected socio-economic impacts [123]. Along with flood or drought warning and prediction models, the integration of different information and tools, such as geographic information systems (GISs), satellite images, consensus data, the locations of critical structures, and socio-economic impact assessment tools, is necessary [123,124]. Impact assessment can be classified according to its evaluation time into pre-or during-disaster assessment. Pre-disaster impact assessments consider climate change [22,125] and planning-or preparedness-related applications [126,127]. During-disaster impact assessment is usually directly associated with immediate response [128][129][130]. A comprehensive and well-established monitoring network is necessary for during-disaster or so-called near real-time impact assessment. Given real-time observations along with detailed socio-economic investigation data, the efficiency of predictions is another important issue. The computer models or evaluation approaches must adopt the input information (such as real-time observations) and provide assessments in a timely manner to meet the need of emergency response. As mentioned before, drought usually develops slowly and has a prolonged impact. Floods usually have a clear and sudden start and end, but drought does not. However, drought can be unexpectedly ended by extreme precipitation [131]. Therefore, long-term instead of short-term impact assessment (here, short-term means a few hours to a few days) is usually more important for drought. The assessment scale varies from a city to a national scale according to the problems. Responses are consistent with being pre-disaster or during disasters accordingly. In terms of flood reduction, Hegger et al. [132] distinguish five types of flood management strategies: (1) flood defense through structural measures; (2) risk prevention by considering proactive spatial planning to avoid high-risk areas; (3) risk mitigation by the flood-adapted design and use of buildings; (4) the preparedness for response, e.g., using flood warnings and evacuation plans; and (5) recovery considering risk transfer mechanisms such as flood insurance to compensate for flood losses. These five management strategies cover structural (e.g., dams and dikes) and nonstructural measures (e.g., land use planning and insurance programs).

Drought
Drought, as mentioned before, is a slow-developing process; therefore, the general public usually does not feel the response in comparison to mitigation measures for floods. Short-term measures would include providing early-warning information, increased emphasis on water conservation (demand reduction), increased water supplies through other backup resources such as groundwater, and water reutilization and recycling. A short-term strategy like water recycling can immediately reduce the need of a water supply and the installation of a water recycling system can be done in a timely and cost-effective manner. However, a comprehensive plan and long-term measures are key to solving the drought issue fundamentally. Long-term measures include the construction of reservoirs, joint operation of water supplies (interconnecting water supplies between neighboring communities or nations), and drought preparedness planning to build greater institutional capacity and awareness building and education [133]. A comprehensive plan includes long-term measures to meet the long-term water demand along with short-term measures to provide a buffer to respond to the uncertainty of the future. Among all the measures, demand reduction is always the most common and efficient measure to take. It is considered that the water consumption of agriculture is always the highest. Therefore, many studies have focused on agricultural drought responses [134][135][136]. Table 2. Research themes and associated studies related to flood and drought mitigation.

Other Associated Studies
It has been confirmed that the state has the primary responsibility to protect people and property from hazards. However, it has also been indicated that strengthening community-level capacities to reduce disaster risk is especially needed, considering that appropriate disaster-reduction measures enable communities and individuals to significantly reduce their vulnerabilities to hazards [137]. The traditional paradigm for emergency response is a "top-down" approach: an executive decision maker or other top person makes the decisions for emergency response. The community-based approach, a so called "bottom-up" approach, is the opposite approach, in which everything is initiated locally and designed to respond to a disaster immediately. Recent studies have shown the efficiency of community-based emergency plans [138,139]. The first priority of disaster response is to protect people. However, the interference of economic activities and the associated loss receives attention. Post-disaster impact, such as interruption to the supply chain of industries, is also significant as floods can occur in high-tech industry parks. To maintain a resilient business or a sustainable society, the concept of a business continuity plan (BCP) has been generally applied and studied [140,141]. In conclusion, many response and management strategies rely on historical records or past experiences. However, there are more and more unprecedented extreme events, which mean no one thought they would happen. The paradigm of disaster management must shift from prevention to management or mitigation in order to achieve the goal of sustainability. The concept of BCP is a very good practice for business sectors to adapt to climate extremes and return to regular trade in a timely manner [142]. Tables 3 and 4 show examples of risk reduction for floods and droughts, respectively. The tables provide a systematic strategy for the risk reduction of floods and droughts. Furthermore, an integration of structural and nonstructural measures is considered. The monitoring is an important role for the risk reduction of floods and drought. There is no big difference in terms of monitoring for both floods and droughts. Establishing a comprehensive monitoring network is essential for detailed observation. It provides the condition of status quo and serves as a reference for future forecasts. Prediction and warning can be done by using the monitoring information as input. At this stage, numerical models, a nonstructural measure, are applied to provide results, such as river stage forecasts for floods and SPI for droughts. Government and public entities can evaluate possible impacts, such as flood affected area and drought duration, using the abovementioned forecasts. The socio-economic impact is obtained, and decision makers can take precautionary measures as a response. The "response" is the biggest difference between floods and droughts. Due to the characteristic of drought, it is almost impossible to prevent drought from happening. Therefore, the main idea in response to droughts is to decrease the socio-economic impact. Structural measures for drought are mainly used to increase water resources. When a drought occurs, decision makers can only take nonstructural measures, such as lower water pressure, water supply reduction, and compensation, against the drought loss. Differently from droughts, the first idea in the response to floods is prevention and the second is mitigation. Therefore, structural measures such as levees and emergency bypasses are built to prevent flooding. However, people realize that the disaster cannot be prevented due to climate change and unprecedented extreme weather events. Nonstructural measures are applied to mitigate the impact of flooding and adopt to climate change. A combination of structural and nonstructural measures is the way forward to maximize the disaster loss reduction. While structural measures are mostly initiated by governments, the nonstructural measures mentioned above are more effectively supported and engaged in by the general public.

Conclusions
The occurrence frequency of floods and droughts is increasing due to climate change [143,144]. This overview provides definitions of floods and droughts and discusses the related research for disaster reduction. However, the research outcomes regarding disaster reduction are too numerous to be covered. There is always the possibility that relevant contributions or other advanced research studies have not been covered in the overview. Moreover, the overview only includes articles published in English. Therefore, articles published in other languages are not included. In conclusion, an integrated view of the research area is provided, including three research themes corresponding to the disaster-reduction cycle. Because of rapid advances in technology, the application of artificial intelligence and the internet of things in disaster reduction are also discussed in this paper. The abovementioned components and their interrelationships provide a cohesive overview of the literature on the special issue of "sustainable water resource management for disaster risk reduction". The findings of this review further indicate that further research is required to enhance the environmental and social sustainability of disaster risk reduction. The results of this review show areas lacking research or needing further development, especially through multidisciplinary research and cooperation. The detailed conclusions are provided as follows:

Implications for Flood and Drought Researchers
The results of this review help to identify the current status and studies of floods and drought in the literature. This work provides a foundation for further examination of the topic and theory development by pointing out needs for future research from various disciplines. An important contribution to this work is the definition of flood and drought classification. The core contribution is classified into three research themes (prediction and warning, monitoring, and impact assessment and response) based on the disaster-reduction cycle: preparedness, response, recovery, and prevention. This contribution is important because it helps researchers position their research better in the flood-and drought-related disaster-reduction domain. Furthermore, this work provides thoughts for future research of interest, particularly highlighting the need for sustainable development and urbanization aspects.

Future Research of Interest
This systematic review provides an overview of three research themes related to floods and drought and the associated disaster reduction. All the research works achieve significant contributions to each theme. However, multidisciplinary research and cooperation across fields are necessary to accommodate the urgent need for disaster reduction. For example, an improvement in drought and flood monitoring can be achieved by considering hybrid ground-based observation and remote sensing techniques [145]. The integration of physics-based and artificial intelligence-based models is also another research topic worth investigating. Finally, only a few studies have focused on converting a negative impact (e.g., flooding) into a positive impact (e.g., extra water resource). Disasters such as floods or drought not only occur frequently but also occur alternatively. This occurrence creates an enormous challenge for scientists and engineers to build up a resilient living environment adopted for future climate change scenarios. Rapid urbanization and population growth in countries such as China, India, and others has changed the environment and worsened the impact of disasters. Therefore, integrated urban flood/drought policymaking with sustainable urbanization policymaking to best contribute to minimizing flood and drought risks in cities is necessary [146]. Some sustainable technologies or concepts, such as low-impact development (LID), green roofs, or sponge cities, may be considered to achieve sustainable goals [147,148].
Special attention is initially paid to those research outcomes that convert negative impacts into positive impacts. For example, rapid urbanization and population growth are considered as a double-edged sword [146]. The progress exposes the population to a higher risk of disaster, but the resulting socioeconomic development creates a better living environment. However, it is a good opportunity for a city to build up the capacity to alleviate flood and drought risks by increasing investment in sustainable urban structure, developing risk warning and management systems, and reducing disaster risk through land-use planning. Another example-that of shifting drought water storage from reservoirs to underground aquifers-has made substantially more storage capacity available for winter flood management in California [149]. This requires a comprehensive and integrated reoperation of many flood and water supply system elements [15]. As a conclusion, sustainable disaster-reduction strategies should represent the ability or potential of a system to respond successfully to climate variability and change, including adjustments in both behavior and in resources and technologies responding to both floods and droughts. However, existing studies or measures still remain within in this category and there is a lack of multidisciplinary cooperation to meet the needs of future challenges.