Prof Willem Vervoort

Abstract
Floods have a devastating effect on human lives and local economies, and therefore the traditional government risk management approach has been to reduce the occurrence probability of floods. As such, the preferred method of flood control is through river regulation and building of dams. While this has been successful in reducing the occurrence of minor to intermediate floods, it comes at significant costs, socio-economic disruption and environmental impact. In addition, even the most regulated rivers are not risk-free in terms of flooding. Given these observations, alternative options in flood management such as “living with floods” are worth considering. With projected changes in the climate, such alternatives become even more attractive.

Introduction
Floods have a devastating effect on the livelihoods of people in many countries (e.g. Yin and Li, 2001) and the actual severity of the floods is possibly increasing in time (Birkland et al., 2003; Varis, 2005). Flooding can cause substantial economic and human losses (Kundzewicz, 1999; Varis, 2005), create a major disruption of daily lives of the downstream population (Tran and Shaw, 2007), destroy infrastructure (Morss et al., 2005), and can lead to diseases and other health issues once the flood peak has passed (Ivers and Ryan, 2006; Tran et al., 2008). In the United States, floods are considered to be the costliest natural hazards (Birkland et al., 2003). Due to public concerns about safety and economic losses in developed countries, the traditional response to flooding in a river basin has been to regulate the river through dams or levees (Birkland et al., 2003). Such construction of dams and levees is based on the view that nature can be effectively controlled by humans (van Ogtrop et al., 2005). In addition, river regulation created advantages for river navigation and can regulate the water supply for human consumption and industry.

Floods are the result of runoff generated by heavy rainfall events in which either the rainfall rate exceeds the infiltration capacity, or total rainfall exceeds the storage capacity of the catchment.  The key determinants for the generation of floods are therefore 1) the intensity and the amount (or recurrence) of the rainfall event, and 2) catchment characteristics related to landuse and soil properties. In Australia, there is evidence that the rainfall amount per storm is increasing due to changes in the climate (Alexander et al., 2007; Gallant et al., 2007). Globally, the total amount of rainfall is also expected to increase due to climate change. Models predict a 3% increase while a 7% increase was observed based on satellite derived rainfall data over the last 20 years (Lambert et al., 2008; Previdi and Liepert, 2008).

Land use changes impact the storage capacity of the soil through changes in root depths, litter layers and evapotranspiration, and can also change the infiltration capacity, particular as a result of urbanization and related increases in impervious surfaces in a catchment (Dietz and Clausen, 2008; Kundzewicz et al., 2005; Wheater and Evans, 2009). Furthermore, in many countries, land use has been changing rapidly over the last decades with increased urbanization and clearing of forest and range land areas (Dietz and Clausen, 2008; McAlpine et al., 2007; Yin and Li, 2001). As a result, this has led to major changes in the runoff characteristics, which could explain the observed increase in the number and severity of floods (Birkland et al., 2003).

Most of the past research on the management of floods has taken place in Europe and the United States, resulting in an emphasis on humid systems. River regulation in these countries is extensive (Nilsson et al., 2005) as limiting economic losses through flood prevention is seen as important. Most humid river systems are characterized by regular seasonality in the flow and relatively low coefficients of variation (McMahon et al., 1992), which makes river flow relatively predictable and regulation easier. In contrast, semi-arid systems are characterized by irregular flow and high coefficients of variation in flow. Transmission losses can be high making predictions of flood wave propagation difficult. Some rivers in India, Indochina and China are also characterized by very high sediment loads, which means there geomorphology is dynamic (Ludwig and Probst, 1998). In many areas of the tropics, river regulation is further made problematic due to the occurrence of cyclonic weather systems or very strong seasonality and large flows such as in monsoonal systems. This creates additional problems in terms of managing the natural flows in the river using engineering tools. In addition, lack of human and financial capital in some countries in the tropical development zone, makes it difficult to undertake large engineering projects to manage floods (van Ogtrop et al., 2005).

As a result of these difficulties, alternative strategies for managing floods have been suggested over the last two decades (van Ogtrop et al., 2005; Werritty, 2006). There are a range of reasons why the views on flood management have been changing. This paper intends to first review traditional flood management, its advantages and its impact on the river system. Secondly this paper will discuss alternative management of floods and rivers in particular in relation to uncertainty about climate variability now and into the future. Given the amount of work already focusing on humid systems, this paper will emphasize semi-arid systems and monsoonal systems, such as in Australia and parts of India.

A review of flood management approaches
Dams, the engineering approach: management of floods using statistics
Managing floods basically deals with managing risk (Krzysztofowicz, 2001; van Ogtrop et al., 2005). Risk is defined as being the combination of damage and occurrence: 

          (1)

Traditionally most governments have aimed to reduce the occurrence of floods to manage risk and therefore much research has concentrated on the area of statistical hydrology and engineering. Control structures such as dams and levees have been popular as they give a sense of security (Krzysztofowicz, 2001). Apart from reducing the flood occurrence, dams and levees could serve more than one purpose such as improved navigation and water security (Birkland et al., 2003; Morss et al., 2005). The statistical flood management approach focuses on calculating the recurrence of floods of a certain height and basing the policy decisions about flood management on a “calculated risk”.

The recurrence of downstream flood events is based on the statistical analysis of past data (Lave and Balvanyos, 1998). Basically the cumulative probability distribution is calculated and the flow volumes are plotted against the inverse of the frequency, i.e. 1/frequency equals the return period (Figure 1). In this way, the design of a structure, or zoning for flooding can be based on a desired return frequency. For dam construction the return frequency needs to be very high, i.e. a 1 in 1000 year flood was often a standard (Figure 1). This standard related to the maximum flood that a dam would need to be able to withstand without breaking rather than the maximum flood it should be able to store. It related to the design of the spillway. More recently, the concept of probable maximum precipitation (PMP), and the related probable maximum flood (PMF) have been introduced to further improve dam safety (Pessoa and Cluckie, 1990). This is based on the idea that the observed data are often insufficient and therefore the data need to be extended to include the most severe reasonably possible flood (Pessoa and Cluckie, 1990).

There are different ways to estimate PMF (Lave and Balvanyos, 1998), but in the end, other factors such as construction costs will also influence the decision maker. The PMF does not really have a frequency associated with it and depending on the length of the record used, different values can be found. For example, in many cases peak flows cannot be measured due to limitations on the gauging station and this many of the estimates of high flow have high uncertainty (Pessoa and Cluckie, 1990) and further uncertainty is introduced through regionalization (the combination of data within a region to derive the flood curve). Given that the PMF is a relatively recent development, many existing structures needed to be upgraded, which involved significant costs. Thus the choice of PMF method can become an economic decision rather than a scientific one. Again, the PMF design guidelines relate to the design of the spillway and the amount of water the dam would need to be able to spill to safeguard the dam against failure.

In semi-arid areas flooding is often unpredictable and not regular and data series are often very short. The high variation and therefore unpredictability is particularly visible in the presented frequency curve for the Lower Balonne in South East Queensland (Figure 1). The curve is very steep, even though the return frequency is plotted on a log scale. In addition, much of the climate in Australia appears to be influenced by very long climate cycles, possibly as long as 50 years, as can be seen from the lack of regularity in the Lower Balonne timeseries. Good dam design would therefore require very long data series, which can be problematic in many areas of the world including India and Australia.

In essence, dams are only designed to “hold” smaller floods, because, for a dam to be fully effective for flood control, it needs to include a massive over capacity to cope with the one very large flood. Given that a dam is a major infrastructure investment, most dams are not only designed for flood management but are multi-purpose. This means they combine flood management with power generation, water supply for irrigation or drinking water, and recreational purposes. However these other purposes counteract the effective flood prevention role. A water manager focusing on flood prevention would want a dam to be as empty as possible to store the maximum flood, while a manager focusing on irrigation water storage or power generation would like the dam to be as full as possible. As a result the risk of early spilling is increased resulting in floods downstream.

Figure 1 Example frequency curve for monthly river flow for the lower Balonne river at St. George 1921 – 200. The dashed line indicates the 1000 year return frequency.

The key point is that, despite engineering advances and careful construction, dams neither provide full flood prevention (Krzysztofowicz, 2001; Kundzewicz, 1999; Morss et al., 2005) nor are they fail proof (Lave and Balvanyos, 1998) as their design is essentially based on a statistical approximation of possible floods and are never designed to hold all floods. A clear example of this was the recent floods in January 2011 around Brisbane in Queensland, Australia. In this case the major dam (Wivenhoe dam), which was designed to prevent flooding since the previous flood in 1974 was forced to spill rapidly, causing downstream flooding due to the size of the occurring climatic event. In addition, any estimate includes uncertainty. This uncertainty will most probably increase in the future given climate change effects on rainfall and runoff. Overall dams are thus a costly investment in infrastructure which is not risk free. Further investment will always be needed to manage flood damage, the other component of equation 1.
The fact that dams for flood management are never risk free is probably the reason that most dams have been designed with a multi-use purpose, with decreased downstream flooding as only an “additional benefit”. As an example the newly constructed Three Gorges Dam in China lists improved navigation, flood management and hydropower as the three main purposes of dam construction (Wu et al., 2003).

Environmental impacts of river regulation

Figure 2 Example of the effect of a dam on the low flow frequency in a river. Lachlan river at Cowra (NSW) before (1893-1935) and after construction of Wylangala dam (1972 -2007). After McMahon and Finlayson (2003)

The ecological impact of dams on rivers has been extensively documented (e.g. Johnson et al., 1995; Kingsford, 2000; Puckridge et al., 2000). Globally, about half of all major rivers are impacted by dams, this is called “fragmentation of flow” (Nilsson et al., 2005). In the continental U.S. only 42 large rivers (longer than 200 km) are unimpaired (Graf, 1999; Poff and Hart, 2002). The main impact of river regulation is through the change in flow patterns in the river (Graf, 1999; Magilligan and Nislow, 2005). Low flows are particularly affectded (McMahon and Finlayson, 2003) (Figure 2), but other aspects of flows are also impacted. Ecosystems rely on three main aspects of flow: flow regime (the long term nature of flow), flow history (the sequence of low flow and high flow events) and flow pulse (the height of floods) (Sheldon et al., 2000). Regulation of rivers due to dams changes all three these aspects resulting in major changes in ecology of the riverine system (Magilligan and Nislow, 2005; Puckridge et al., 2000). Changes in the flow patterns can result in the separation of the main channel from the floodplain resulting in reduced recruitment in riparian species, changes in downstream food webs and aquatic productivity (Poff and Hart, 2002).

In winter rainfall dominated areas in Australia and the United States, changes in the timing of the water use from the dam and in the river can lead to flow inversion. This is due to the release of water during summer to match crop demand in irrigation and the collection of high flows in the dam in winter  (Magilligan and Nislow, 2005; Walker, 1985). Similar problems would occur in monsoon driven systems, where most flow would be normally be expected during the monsoon season and ecosystems will have adapted to such seasonal trends.

In semi-arid environments, most rivers are losing, that is the groundwater tables are well below the water level in the river. Transmission losses from rivers during flooding are the most important contributions of recharge to the local fresh groundwater (Barbier, 2003; Williams et al., 1989). The loss of flooding downstream and change in flows downstream will therefore greatly impact downstream groundwater resources (Barbier, 2003).

The release of water from a dam generally occurs from a so-called “off take” which takes water from the bottom of the impoundment. Water bodies deeper than 8 m, which are not regularly disturbed, develop a strong temperature and dissolved oxygen gradient (Håkanson et al., 2004).  Due to these gradients, releases from dams tend to much cooler (up to 10 oC) and more anoxic than the receiving or original river water and this effect can travel downstream over distances of 100 km (Poff and Hart, 2002; Walker, 1985). Such major changes in temperature and dissolved oxygen can severely affect sensitive ecological processes such as fish spawning and aquatic productivity.

Floodplains of many rivers act as filters for nutrients and reducing flooding concentrates nutrients in the river, which are subsequently being deposited in the ocean. In the case of the Mississippi river this has led to toxic algae blooms in the Gulf of Mexico (Sparks, 1995). In Australia, agriculture and river regulation are having similar impacts on the Great Barrier Reef (McCulloch et al., 2003).

Further and more long-term changes will be geomorphologically. Dams not only affect flow volumes and velocities but also act as a catcher for all sediment in the river, as the water slows down and sediment can settle from the water column within the dam. Sediment is important for the maintenance of fertility in natural and agricultural flood plain systems (Mingzhou et al., 2007; Ogden et al., 2007). Conversely, sediment can have negative impacts on the ecology due to anthropogenic contamination with heavy metals or chemicals (Costa et al., 2006; Lecce et al., 2008; Pease et al., 2007). However, sediment is also important for the river to maintain its geomorphological structure. Hence decreases in flow and reduced delivery of sediment load can change the overall channel and floodplain structure (Grubaugh and Anderson, 1989; Ligon et al., 1995) and particularly in semi-arid areas these changes can be rapid (Petts and Gurnell, 2005).

Dam releases and the concentration of flow in the river channel not only lead to a disconnection, but could also increase the risk of flooding. Flooding of floodplains decreases the velocity of the flood wave and decreases the flood peak through attenuation. Concentrating more of the flow in the river channel will therefore increase the risk of flooding downstream (Sparks, 1995). This is further exacerbated by the fact that clay landscapes of the floodplains which are not regularly flooded will subside, thus creating an even greater potential for flooding (Sparks, 1995).

Social and economic impacts of river regulation
People have a difficult relationship with floods. In western countries such as Europe and the United States and Australia, floods are generally treated as damaging and a risk (Kundzewicz et al., 2005). However despite this, many people perceive the risk of actually being affected by a flood as small (Kundzewicz et al., 2005; McPherson and Saarinen, 1977) even if they live on the floodplain (Krzysztofowicz, 2001; McPherson and Saarinen, 1977), or other risks are seen as more pressing (López-Marrero and Yarnal, 2010). Flood mitigation through dam building and river regulation can create further complacency due the misinterpretation of the risk by the population (Krzysztofowicz, 2001; McPherson and Saarinen, 1977).

In contrast, in many other countries, floods are seen as life giving and important sources of moisture for agriculture (Adams, 1999; Adams, 1985). The disruption of flows by a large dam thus has a similar impact on the agricultural productivity downstream as on the riparian ecology (Adams, 1999), which means dam building includes a socio economic disturbance of a similar magnitude (Adams, 1999; Barbier, 2003; Lerer and Scudder, 1999; Varis and Lahtela, 2002). This is particularly the case if 1) local downstream farmers use so-called “recession farming” and thus grow corps on the residual moisture after the flood (Adams, 1999; Adams, 1985; Barbier, 2003), or 2) the downstream farmers rely on the floods to replenish local groundwater tables (Barbier, 2003). In addition, changes in the river ecology can have major impacts on the opportunities of fishermen downstream from dams (Adams, 1999; Adams, 1985; Varis and Lahtela, 2002).

Dams are often built for more than one purpose, flood mitigation being only one of them (Poff and Hart, 2002). This means that releases are also related to such other purposes, either irrigation water supplies or hydropower generation. As a result dam managers have a tendency to store water in the dam at large volumes for future use and therefore releases are sometimes wrongly timed (Adams, 1999) or insufficient for flood plain agriculture (Adams, 1985).

Finally, while floods are seen as a health risk (Ivers and Ryan, 2006), dams can also pose a health risk (Lerer and Scudder, 1999), partly through the loss of access to water for the poorer communities and partly through an increase in vector borne diseases related to the storage of water. As a sad additional detail, even with all the dam construction occurring in the world, basic sanitation and water needs of many communities are still not being met (Gleick, 2003). 

Table 1 Summary of downstream impacts of dam construction for flood mitigation

Alternative approaches of flood management
As alternative approach to minimizing the occurrence of flooding, we can also minimize damage in equation (1). The concept of “living with floods” has therefore been gaining ground (Kundzewicz, 1999; van Ogtrop et al., 2005). Here flood management focuses on co-existing with floods and adapting society and land development to flood levels. This concept is currently guiding flood management in the Netherlands (van Ogtrop et al., 2005). The aim is to reduce the risk of flood damage rather than reducing the flood occurrence, such as through using dams. In addition, living with floods focuses on public awareness of floods and minimizing environmental degradation (van Ogtrop et al., 2005).

Resilience is a concept which has mainly been used in economical and ecological context (Walker et al., 2004). A resilient system is a system that is able to absorb shocks without changing state. In contrast a resistant system is able to withstand shocks up to certain magnitude after which the system changes state. This concept can also be applied to flood management (van Ogtrop et al., 2005), where a system of dams and levees can be defined as a resistant system, while a system which copes with regular flooding can be seen as a resilient system. The difference between the two is again through the focus on the different elements of equation (1).

Suggestions for resilient flood management systems that minimize damage have ranged from evacuating susceptible low lying areas (Lave and Balvanyos, 1998; Varis, 2005), to improved flood forecasting and upstream catchment management (Varis, 2005). But this could be further expanded with innovative ways of living on floodplains (Kundzewicz, 1999; Tran and Shaw, 2007; van Ogtrop et al., 2005). Sustainable flood management can therefore be defined in terms of three actions: 1) modify susceptibility to flood damage 2) modify flood waters 3) modify impact of flood (Kundzewicz, 1999). Smaller dams and levees might still be needed to protect crucial infrastructure (Kundzewicz, 1999).

But systems can also go backwards. In a study in Vietnam it was found that the traditional system was more resilient than the current system due to changes in the socioeconomics of the region (Tran et al., 2008). In particular, social cohesion and bonding was very important in terms of reducing the impact of flooding on the local community (Tran et al., 2008). Deforestation in the upper catchment due to export demands and a decline in traditional systems of environmental management resulted in an increase of both flood risk and flood damage (Tran and Shaw, 2007).

For Australian semi-arid catchments (inland rather than coastal) damage is generally not a major concern as the population densities are low. In fact, floods are generally welcomed as life giving. Problems only occur around urban centres where economic losses tend to be higher, such as recently in Brisbane. A further example of problems related to human encroachment on the river is related to the recent floods around Rockhampton, Queensland. Here most of the damage was related to an urban area known as “the swamp”. This neighbourhood was locally known as the swamp because it was built on a low lying area adjacent to the river and prone to flooding.

In contrast, some of the irrigated systems in south-east Queensland and northern New South Wales in Australia are dependent on flood water to supplement the uncertain rainfall, but this has, similarly to the construction of dams a major impact on the flood frequency and magnitude (Kingsford, 2000). A system of adaptive flood management that protect small urban centers, but allows widespread flooding elsewhere could easily be implemented. However this would require changes to planning regulations.

Future flood management under increased climate variability
The predicted changes to global climate (IPCC, 2007) will throw up a range of new challenges for flood management. Resilient alternatives in flood management will therefore have to include the predicted effects of climate change. Future climate change effects are predicted to increase global rainfall with the main increases probably occurring in the mid latitude areas (Dore, 2005; Huntington, 2006; Kundzewicz et al., 2005; Lambert et al., 2008; Previdi and Liepert, 2008). More semi-arid areas, such as Australia will probably see increases in the time between rainfall events, while the amounts per event could also slightly rise (CSIRO, 2007; Pitman and Perkins, 2008). In terms of flood management, it means that if the storage capacity in the dams is assumed to stay constant, this will lead to increased overflows from the dams. From a statistical design point of view, the PMF will shift up. This will require further (costly) upgrading of existing structures to reduce the risk of dam failure, as this risk would increase, a trend which might be already evident in the current data (Lave and Balvanyos, 1998).

Climate change and land use change might go hand in hand. Changes in rainfall patterns and temperature would affect vegetation survival and cropping patterns. Increased pressure on forest resources and limited arable land decreases the amount of forest cover in the upper catchments and could increase populations in flood prone areas. It is not clear which might go faster, climate change or land use change, particular in areas with high population pressures, such as South East Asia and India. This is a smaller concern in the less densely populated areas in Australia, where in fact major reforestation is needed, due to increased salinity risks and past land clearing (McAlpine et al., 2007; Pannell and Ewing, 2006).

For Australia, there are some further interesting considerations. Currently large storage dams are used to manage floods and irrigation waters in many of the rivers in the Murray Darling Basin. However, due to the large variability of the climate in Australia, surface water resources are often uncertain and evaporation losses from irrigation storage basins and dams can be high. Reliable groundwater resources would be a preferred option, but there are limitations in pumping capacity and sustainable yield of good quality groundwater. Future climate predictions for Australia indicate an increase in the variability of rainfall affecting both the recurrence of floods and drought periods. Increased recharge into groundwater through increased opportunity of flooding would allow increased use of groundwater for irrigation, i.e. similar to the objectives of rainwater harvesting in India. While the overall amount of water available for irrigation might decrease, the reliability would increase. In terms of dam management, this would either require removal of dams or an increase of so-called “translucent flows” (inflows which are immediately released). Clearly, construction of new dams is not a good choice.

In many countries, flood plain areas are crowded by population due to the high fertility of the areas, or preferences for living “on the water”. For example, during the recent floods in January 2001 in Brisbane in Queensland, the hardest hit suburbs were quite wealthy where people had paid premiums for living close to the river. Moreover in areas with less developed infrastructure, or less developed economies, where people choose to live in flood plain areas because of farming opportunities, this creates additional problems in terms of avoiding flood damage.

As an alternative to high levels of government investment, public participation could be used in finding resilient flood management solutions. Public participation in flood management has two advantages. The first is that solutions can be found which are flexible and low cost. The second is that through participation there is an increased awareness of the flood risk which leads to better preparedness and a decrease in the loss of lives in future floods. As an example, in the earlier mentioned study in Vietnam, it was noted that limited public participation meant that the linkage between environmental management (i.e. land use and land degradation) and flood hazard were not clear (Tran and Shaw, 2007). In addition, social cohesion and bonding were very important in disaster management, such as flooding (Tran et al., 2008). In areas of rapid economic growth, public participation is additionally important to increase awareness of the risk of building and development in flood prone areas (Tran et al., 2008).

In terms of resilience, climate change can deliver some of the shocks which might test the system. A system based on flood management using a dam would have a higher risk of failure and thus a smaller resistance and precariousness  than a system based on a “living with floods” concept, or any other system which includes high levels of public participation and a focus on damage minimization (Gersonius et al., 2010; Walker et al., 2004). Under future climatic change it will be even more difficult to minimize the occurrence of floods than to minimize damage from floods.

Summary and Conclusions
In summary, the review in this paper indicates that there are many issues related to management of floods using dam construction (Table 1).  Given the projected changes in climate, flood management using dams is not a real viable alternative for the future, because related costs and downstream impacts are significant. Given the increased pressures on government monies and the range of other priorities, alternatives should be considered. An assumption has been that the proposed dam would primarily be constructed for flood mitigation and not for other purposes (such as hydropower, town water supply and irrigation). However, in this paper, it is also indicted that some of these purposes could also be met with other means (using groundwater). Even if there are multiple uses for a proposed dam, arguments for and against construction can be given, but the analysis is more complex.

Flood management using dams is costly, greatly disrupts the environment and is not fail safe. The methods for assessing dam safety are often based on extrapolation of data and include large levels of uncertainty. This is particularly exacerbated in areas of low data density or high variability such as India and Australia. Alternative flood management strategies, such as “living with floods” have opportunities to increase public awareness of flood danger and have less environmental impact. In the light of future changes in climate, increased population and increased pressures on fertile flood plain soils, alternative flood management strategies become even more attractive. In addition, in Australia, increased flooding would allow increased recharge into valuable groundwater resources.