Article first published on China Water Risk, August 2020
Flooding poses severe threat to human societies, especially in low-lying and coastal regions. World Resources Institute (2020) estimates the number of people at risk of coastal and river flooding will double from 72 million in 2020 to 147 million in 2030. IPCC (2012) states the projected change in precipitation and temperature will alter the probability of flooding.
Although there is insufficient evidence linking anthropogenic warming and increased flooding, compounding events including sea level rise (SLR) and precipitation-induced high river discharge is likely to increase the risk of flooding.
Traditional flood risk assessment
Traditional flood risk assessment stem from the belief that risks can be accurately predicted through models.
Traditional flood risk assessments believe risks can be accurately predicted through modelling
After establishing the confidence interval, a line can be fitted in the model to find out the likelihood of very high flows. Using probability distribution, damage can be calculated at each level, then generalise an extreme value. Economic damage is assumed to be the only aspect of risk and can be well characterised by the damage function. This provides an economic case for risk reduction, where economic decisions can be informed through quantified estimates.
In addition, the assessment method is structured based on stationarity, which implies natural systems fluctuating in unchanged variability. Flow statistics are regarded as stationary in models, such that the properties of any variable with time-invariant density function can be estimated from instrument records. Probability density functions are used to assess risks to water supplies, floodplains and infrastructure development.
The modelled data can generate substantial economic benefits…
…Insurance companies are using it
The utilisation of modelled data can generate substantial economic benefits of certain sectors. Back in 1980s, insurance premiums are determined based on previous claims. Category 5 Hurricane Andrew created massive destruction in the States, causing USD15.5 billion of insured losses and subsequent insolvency of 11 insurance companies (Lamb, 1995). Since then, insurance companies began adopting modelling to estimate levels of premiums and reinsurance. They use catastrophe model with different versions of flood risk modelling. This can improve the accuracy of setting premium levels in protection of companies’ interest.
In the context of floods, perceiving risks as predictable leads to adaptation approaches that aim for optimality, which maximises net benefits. For instance, dam construction is considered the panacea to flood management historically, which generate associated benefits of clean hydroelectric power provision and irrigation water storage. Significant benefits outweigh the cost of construction. With cost-benefit analysis, dam is therefore the optimal approach to manage flooding.
In African countries, dam construction is still the dominant flood management and power generation strategy with growing number being built. Globally, over USD5.7 trillion has been spent on water infrastructure investment, but the benefits are expected to accrue in the long term to offset the cost.
Cascading uncertainty posed by climate change
Increasing evidence illustrate climate change as a problem of ‘strict uncertainty’, which has unknown probabilities of future states (Knight, 1921). Uncertainty cascades downwards from state of society to adaptation responses as illustrated by Wilby and Dessai (2010).
Increasing evidence illustrate climate change as a problem of ‘strict uncertainty’…
…which cascades downwards from state of society to adaptations
How the future society evolves determine greenhouse gas emission pathways, which feed into climate model, regional scenario, impact model, ultimately translating to local impacts and adaptation responses. With complexities in climate system and modelling, there are more permutations and increasing envelop of socioeconomic uncertainty when adaptation responses are planned.
Climate-ready assessment and adaptation
Revisiting the traditional flood risk assessment model, the assumptions are critiqued based on the deeply uncertain nature of climate change.
Firstly, stationarity is compromised under climate change and human influence, as Milly et al. (2008) claims ‘Stationarity is Dead’. There are long-term human disturbances in river basins and floodplains. Water supply, water quality and water flow are all affected by infrastructure, land use change and channel modification.
At the same time, exogenous factors of changing natural climate, internal variability under ocean and ice sheet dynamics all hold the stationarity assumption false. Non quantifiable and intangible damages brought by flooding are omitted in the model, such as loss of social networks and cultural heritage. Another problem is the data fed into climate models have both natural uncertainty and also human subjectivity. It is affected by the preferences of scientists who build the models, as well as policymakers involved in generating scenarios.
The flawed assumptions & subjectivity of the traditional model do not diminish the value of risk assessment…
…it cautions us to only use it as a guide, not a dictating tool
The flawed assumptions and subjectivity do not completely diminish the value of risk assessment in economic and policy decision making. Yet, it cautions us to only use the assessment as a guide instead of dictating tool. Quantified numbers related to risk and damages only offer an indication of an uncertain trend. Politicians should avoid making use of the exact numbers generated from the exercise. It is important to keep a complex suite of tools to avoid political manipulation (Stirling, 2010).
Proportionate adaptation aims for robustness across a range of climate scenarios (Hall et al., 2012). Proportionate adaptation identifies tolerable risk level in the societal context and stay within the threshold with low-regret strategies. Such measures perform acceptably well across a range of scenarios and avoid costly maladaptation. This can be achieved using a scenario-neutral, policy-first approach (below figure, Wilby & Dessai, 2010).
Proportionate adaptation identifies tolerable risk level in the societal context…
…and to stay within the threshold with low-regret strategies
The government first comes up with a series of flood adaptation measures, then test their sensitivities to different climate scenarios and other parameters like technical feasibility. This can reduce the reliance on trusting modelled data on flood risks. A common approach is adaptation pathways.
When planning for the Thames Estuary in 2100, the UK government did not begin with a detailed flood risk assessment. Rather, a range of adaptation options were proposed, such as improving existing defense system and building new barrage. The office then monitors sea level rise, which will trigger different adaptation options from the list following the pathways (below figure).
The UK govt has come up with a series of adaptation options that it can implement against sea level rise
When designing a robust policy measure, we must first recognize the differences in perception and acceptability of risks across different contexts. As Mary Douglas (1983) proposes the social construction of risk, in which social norms and cultures deviate how risks arise and are interpreted. It is important for governments to identify the ‘tolerable’ level in their countries to tailor a policy suite of adaptation measures (below figure).
An example of misjudgment is the World Food Crisis in 2007 triggered by droughts, the rise of grain prices was not properly managed and eventually led to social unrest in many developing countries. While the principle of robust decision-making is useful under deep uncertainty, it requires a deep level of understanding in risk acceptability of local governments.
Conclusion
Modern risk assessments take into account the societal perception of risks and embrace uncertainties entailed by climate change. All risk assessments involve a degree of subjectivity with judgements and assumption from scientists and policymakers.
Flood risk assessment should be part of broader flood management
Flood risk assessment should be treated as a part of broader decision-making process in flood management. Exact figures generated from the exercise should not be taken at face value, rather be utilised to develop structural understanding of wicked problems. Proportionate adaptation is proposed as a robust decision-making strategy that minimizes losses caused by uncertain outcomes under climate change.
References
Natalie Chung 鍾芯豫 