Adaptation pathways

To develop effective flood risk reduction strategies that can adapt to changing circumstances, planning approaches that incorporate adaptability and flexibility are essential. To accommodate flexibility into decision-making and account for the evolving nature of flood risk, the application of “adaptation pathways” has been identified as a promising approach. Adaptive pathway planning enables decision-making over time, in response to how the future unfolds. The dynamic adaptive policy pathways (DAPP) method is a method in which adaptive pathways are used to develop plans that are subject to uncertainty (Haasnoot et al., 2013).

Figure 1 shows a schematic overview of a pathway map, a feature of the DAPP approach. It shows the available actions and the different pathways that can be chosen. The methodology is based on the idea that investment choices (or actions) have a finite lifespan and may no longer meet goals if circumstances change, i.e., when a threshold is crossed, known as the adaptation tipping point (ATP) (Kwadijk et al., 2010).

In a flood risk reduction strategy, for example, sea level rise can result in a minimum safety level no longer being met. When an action no longer fulfils the objective, new actions are required to meet the standards again, leading to a variety of alternate pathways. The trade-offs between their costs and benefits will determine which pathways are preferred over others. If there is only one initial decision moment, the strategy is considered static. However, if there are opportunities to make decisions based on available knowledge at different points in the design process, the strategy is flexible or adaptive.

While adaptive planning offers advantages, it also presents challenges and weaknesses compared to static approaches. Some challenges include uncertainty in future developments, potential trade-offs between short-term and long-term objectives, and the need for continuous monitoring and adjustment. In contrast, static approaches provide a more straightforward and predictable framework but may lack the ability to respond effectively to changing circumstances. To discover which strategy is optimal for a project location, adequate and thorough evaluation is necessary.

Economic evaluation

Various methods exist for economically evaluating adaptive strategies, including cost-benefit analysis (CBA), real option analysis (ROA) and robust decision making (RDM). Multiple flood risk evaluation methods have been evaluated and it was found that each method comes with its own strengths and limitations. The premise of adaptive pathway planning is the ability to re-evaluate and reassess taken or possible new actions. Currently, no method is found to evaluate strategies under uncertain conditions and include the premise adaptation pathways.

Research objective

This research focussed on filling these identified gaps and contributing to enhancing the economic evaluation of adaptive pathways. Without an adequate evaluation method, validating positive expectations about adaptive pathway planning remains challenging. The study was divided into three subtopics:

1. Evaluation of current evaluation methods.
2. Investigate the value of time
3. Test new evaluation method.

Creating a new method

Analysis of evaluation methods

Multiple approaches exit to develop and evaluate flood risk reduction strategies. Certain methods focus on the uncertainty, other focus more on incorporating flexibility. For this study, the most used techniques were evaluated to derive its strengths and weaknesses. Based on the found results a new method could be formulated.

Robust decision making (RDM)

RDM aims to create decision strategies that are robust and perform well under diverse future scenarios. It emphasises evaluating various options, including worst-case scenarios, to identify robust strategies. Robust refers to the ability to perform well even under conditions of high uncertainty or ambiguity. Identifying strategies that are robust helps decision-makers minimise the risk of negative outcomes and increase the resilience of their systems. Additionally, RDM aims to increase transparency by taking into account various objectives and criteria in the decision-making process.

RDM has the ability to facilitate so called “deliberation with analysis” (Groves et al., 2019). This term describes the process of decision-making where people or groups carefully and methodically consider options and potential outcomes before reaching a final conclusion. It involves careful consideration of relevant information, weighing of alternative options and evaluation of the potential consequences of each decision. This approach is often used in complex situations where there are multiple factors to be considered and where the stakes are high.

Real option analysis (ROA)

ROA, rooted in financial option theory, captures and values flexibility in decision- making. It introduces the concept of real options, allowing for changes in investments based on new information. ROA categorises options as “on” or “in” a system, offering valuable insights for climate adaptation plans. An example of a real option “on” a system is the option to defer or abandon a project. On the other hand, real options “in” a system are options that are incorporated into the design of the system. For example, making allowance for future expansion of a levee by over designing the foundations.

Both options “in” as options “on" a system are valuable for climate adaptation plans. So, in contrast with the traditional planning approach in which only one-off investment options are recognised, the real option’s concept is able to take management flexibility and volatility into account by enabling changes to an investment, in case new information becomes available in the future (Buurman and Babovic, 2016).

In a traditional CBA, uncertainty is included by expected values depending on probability distributions. The downside of this approach is that it connects a “now or never” quality to the decision moment. This quality is only suitable in case there is no flexibility. However, when the possibility exists to modify the decision, a traditional CBA tend to undervalue. With ROA, at every moment, the option to invest or not to invest is evaluated. The value of options of taking measures later or now is valued.

The ability to choose a different course of action or to decide to postpone an action until more information is available, results in the opportunity to limit the negative effects of making a poor choice while also maximising the positive effects of the newly available information. This aspect is the main premise of the adaptive planning concept. However, the downside of ROA is that it is complex to perform as many uncertainties need to be quantified, integrated and discretised in scenarios, as showed by Kind et al. (2018).

Extended cost-benefit analysis (CBA)

The study of de Ruig (2020) and Haasnoot et al. (2020) extended the traditional CBA framework to evaluate adaptation pathways. They both extended the time horizon of the traditional CBA and included evaluation of sequential measures. De Ruig’s method (2020) incorporates both the temporal and spatial dimensions of climate change impacts and evaluates a range of adaptation measures and their timing to identify the most cost-effective and efficient pathway. Similarly, Haasnoot et al. (2020) extended the traditional CBA framework by incorporating multiple scenarios and an extended time horizon to evaluate sequences of investments or adaptation options. The most effective pathway is determined by the climate and socio-economic scenario that is considered. Transfer costs are included that quantify the path-dependency of options.

Both methods build on the traditional CBA framework and provide a more thorough analysis by taking into account a range of factors and considering the long-term effects of climate change. The recommendation that followed from this research is to incorporate the flexibility in the economic assessment that would enable alterations in the type and/or height of subsequent measures for conditions different from assumed.

Reduced uncertainty: the value of time

A flood risk reduction strategy is subject to various uncertainties which makes it difficult to design the optimal strategy. The performance of the strategy could be increased in case there is a possibility to base decisions on new and more accurate information. The “updated” knowledge which could lead to reduced uncertainty is one of the possible drivers of the added value of adaptive pathway planning.

Figure 2 illustrates this principle of “updated knowledge” and is also referred to as the value of time. The schematisation depicts two pathway maps. Map 1 illustrates the scenario in which no updated knowledge is used. In Map 2, new input variables, based on knowledge retrieved over the past time span, are included at the decision node and transfer station, marked in orange. To describe it more simply, do we learn over time about relevant future conditions? However, although time passes, not all input variables will experience a reduced uncertainty. In this article, the analysis of the three most dominant factors are included: the sea level rise (SLR), economic growth and damages.

Sea level rise projections

For this research, a study was conducted to find a trend in the uncertainty ranges of the past SLR projections. Over the past 200 years, observations of sea levels have mainly been based on tide gauge measurements. Technological advancements, such as satellite altimetry in 1992 and high precision gravity measurements in 2002, have enhanced our knowledge of global SLR and our understanding of the magnitude and relative contributions of the different processes causing sea level change.

Various factors contribute to projection uncertainty, including glacier and ice sheet mass loss, thermal expansion and changes in non-glacial water storage. There is not a universally accepted best estimation technique or a single agreed-upon probability distribution. To refrain from choosing a particular projection, decision makers often use the IPCC projections, which are an ensemble mean or consensus estimate.

A study by Garner et al. (2018) compared 70 SLR projections, published over a time span of 35 years. It revealed a slow reduction in the range of projections from 1983 till 2000, however the uncertainty bands remained relatively great. The projections between 1983 and 1989 contain the greatest range from all other periods across the 35 years. The projections included many assumptions, which resulted in higher uncertainty factors, and future research should overcome these shortcomings.

The third assessment report (TAR) of the IPCC in 2001 and the studies after assumed no major contributions to SLR due to loss of founded ice from the West Atlantic ice sheet could be expected before 2100. Which resulted in a reduction in the uncertainty bands of the projections in the years between 2000 and 2007. In 2007 this trend reversed and greater ranges were projected, experts argue that this reflects the uncertainty about the maximum contribution of the Greenland and Antarctic ice sheets.

It is sensible to say that the increase of understanding has also resulted in an increased of understanding of the elements that are not understood yet. The latest IPCC report emphasises the critical next decade for gaining insights into future projections and their implications. This underscores the importance of ongoing research in the coming years. Figure 3 shows the evolution of SLR projections divided by the publications of the IPPC assessment reports. Initially, a reducing uncertainty is clearly visible, which changes with the fourth assessment report (AR4) in 2007.

Economic damages

For the economic evaluation in this study, only direct physical damages are included in the risk assessment. This procedure consists of three elements: 1) determination of flood characteristics; 2) assembling data on land use and maximum damage amounts; and 3) application of stage-damage functions (Jonkman et al., 2008). These three elements were evaluated in order to determine whether they are subject to reduced uncertainty when time passes.

The first element involves determination of flood characteristics, which are prone to model uncertainties and will therefore not benefit from more time. For the second element, the uncertainty in captured in the future developments of the land (use). The land use of the project area can change over time due to, for example, development projects. Besides, in case of land use change or not, the value of the land can also change over time. These are local uncertainties, so no generic conclusions about a reduction over time can be drawn. The third element concerns the damage curves. There are methods to increase the accuracy but is not assumable to expect a reduction over time achieved by more data.

Socio-economic growth in flood risk context

Socio-economic growth, describing the future value evolution of a project area, is influenced by various factors. Economic growth indicators like gross domestic product, coupled with development plans, impact this growth rate. Six key drivers of economic growth, including natural resources, infrastructure, population, human capital, technology and law, contribute to uncertainties. The growth rate's uncertainty is dependent on the initial GDP per capita and the current level of development in the project area. The specific characteristics of a project area determine whether the uncertainty range changes over time.

Impact on the framework

The performed study into the strengths and limitations of existing methods and the premise of adaptive pathways lead to a focus on incorporating flexibility and evaluation metrics of a new framework.

Flexibility

To enhance flexibility and therefore create adaptability in a strategy, it's crucial to consider two key factors: sea level rise (SLR) rate and the timescale of adaptation measures. The timescale involves the functional lifetime of a measure and the time needed for its completion. Adaptive pathways planning emphasises flexibility and shorter timescales enhance the ability to make informed, low-regret decisions. The framework evaluates flexibility by alternating between measures with different functional lifetimes and required timescales. The envisioned lifetime of a measure depends on the protection level, which is influenced by which SLR scenario is considered. Secondly, shorter lead time (time needed from planning till completion) contribute to increased strategy flexibility. The considered SLR scenario determine the expected timing of the adaptation tipping point (ATP).

While low SLR measures have more predictable ATP timing, extreme scenarios introduce uncertainty, emphasising the need for adaptable strategies. Creating decision moments in the strategy is essential to achieve different levels of adaptability and find the right balance between flexibility and costs. When smaller measures, for shorter expected lifetimes with short lead times, more flexibility is implemented in the strategy. However, more flexibility is not directly better and the right balance should be found.

Evaluation metrics

From the performed study into the existing evaluation methods, it was evident that incorporation of uncertainty is crucial. Utilising simulation techniques like Monte Carlo analysis, where uncertain variables are expressed through probability density distributions, enhances the analysis. To provide a more comprehensive evaluation and increase understanding, three additional metrics next to the commonly used net present value (NPV) and benefit cost ratio (BCR) have been identified. The search for more metrics came forward after the assessment of existing evaluation methods.

Equivalent annual costs (EAC)

EAC introduces a valuable indicator, particularly when dealing with strategies of different lifetimes. EAC facilitates the fair comparison of cost-effectiveness for assets with unequal lifespans, ensuring a more nuanced financial assessment. Important to note that as the NPV is used, a positive EAC value implies benefits. In this formulation, the higher the EAC, the better. The following formulas are involved in this concept:

with A = annuity factor, r = discount rate and t = numer of years of entire lifespan.

Coefficient of variation (CV)

The spread of results in flood risk reduction strategies is influenced by stochastic variables. While variance is a commonly used metric to describe spread, the coefficient of variation (CV) provides a unique perspective. Unlike variance, CV offers a relative measure of variability, showcasing the standard deviation as a percentage of the mean. This dimensionless metric allows for meaningful comparisons across datasets with varying scales or units. This ensures decision-makers can effectively assess and compare the variability of outcomes, irrespective of differences in scales or units.

In which:
X = variance; X = each value in the data set; X = Mean of all values in the data set; and X = number of value in the data set.

In which: X = standard deviation and X = mean.

Probability of loss (PoL)

While metrics such as (NPV) offer insights into economic performance, relying solely on the highest mean NPV can be misleading. The probability of loss (PoL) can turn out to be crucial in evaluating the potential financial impact of a flood risk reduction investment. It represents the probability that the NPV will be negative. By considering the PoL, decision-makers can gain a more nuanced understanding of strategy performance, avoiding a distorted view based solely on mean NPV. As the example in Figure 4 shows, when only considering the highest mean NPV, Strategy A would be preferred. However, the high chance of a negative outcome would have been missed.

Formula for probability of loss:

In summary, these evaluation metrics collectively provide a robust framework for assessing flood risk reduction strategies, considering uncertainty, variability and potential financial impacts.

Conceptual case study

The analysis of existing methods showed that to capture the value of adaptive pathways, it is important to consider uncertainty, include sufficient and relevant evaluation metrics, and acknowledge the value of time. Uncertainty needs to be incorporated in the evaluation of a strategy, either through a scenario- based approach or a sampling technique. Multiple evaluation metrics should be used to obtain a deeper understanding of the performance of a strategy which supports decision-making. The value of time lies in the ability to reassess and reevaluate with newly obtained data that becomes available over time, which connects with the premise of adaptive planning. Overall, by considering these focus points, a new framework for evaluating flood risk strategies was established.

Incorporation of value of time

Earlier was explained that, multiple stochastic variables play a role in a flood risk reduction strategy and if and how extra time would impact its uncertainty. For this conceptual case study, it was assumed that only the amount of SLR is influenced by the value of time. As discussed, it was concluded that no evidence was found that the uncertainty of SLR will reduce over time. However, the coming years will tell whether we are heading for an extreme or moderate climate change scenario. Therefore, to include this prospect, in this research it is assumed that the absolute uncertainty band would remain stable with respect to the time horizon.

For clarity, this assumption is schematised in Figure 5. In this example, the horizon of the first measure is 65 years (2020-2085) and expected horizon for the second measure is also 65 years (2085-2150). Since the projected time is equal, so should be the absolute variance. In other words, when an ATP is reached in 2085, a narrower uncertainty band can be used to find the optimal measure for the next action. This extra knowledge is used to optimise the decision of the new investment at the decision moment and leverages the newly retrieved data, and therefore the value of time. The moment of the ATP is a stochastic variable as it is influenced by the rate of SLR. The pink dot expresses the ATP for a single simulation.

Results

The framework was tested on a conceptual coastal case study. Seven different strategies were formulated that differed in the level of flexibility.

By varying in envisioned lifetimes of measures, and measures with different lead times, different levels of adaptability can be accomplished. Furthermore, from real option analysis, the concept of formulating and creating extra flexibility in terms of options was retrieved. This was formulated in one of the strategies, in which a flexibility premium was included. It entails higher costs upfront to achieve lower costs for possible future investments.

In Figure 6, the results of two different strategies are plotted. In Figure 6A, a static strategy was followed in which a levee was built for an envisioned lifetime of 100 years, based on a low SLR scenario. In Figure 6B, a more adaptive strategy was followed in which initially dry-flood proofing was applied and when the ATP was reached, a levee was built. For the adaptive strategies, the second action was optimised based on the retrieved knowledge over time. When the strategies were tested for a situation in which the knowledge was not used, it was found that these strategies showed a reduced performance of 6% in terms of NPV. This indicates that when the value of time was not acknowledged, the adaptive strategies were undervalued. Subsequently, the evaluation method was able to show the difference in performance in case of a fully deterministic situation (using no uncertainties) and in case the more SLR scenarios were used. Finally, the performance for the strategies that included a flexibility premium increased when the probability on a high SLR scenario increased.

Discussion

SLR knowledge

In the framework, the assumption was made that the absolute variance of SLR projections remains stable over time. However, recent findings from the IPCC Synthesis Report (March 2023) emphasise the urgency to curb temperature rise. This raises questions about the stability of SLR projections. The currently used assumption is arguable to be conservative and might underestimate the potential impact of future insights. However, this assumption can be easily altered in the evaluation procedures.

Economic appreciation

The CBA used a deterministic discount rate to value future cash flows. The discount rate, considered as a political decision, holds ethical and political implications. Ethically, it shapes the distribution of costs and benefits across generations, with higher rates potentially favouring the present. Politically, divergent stakeholder views on discount rates reflect varied interests and priorities. An analysis with a 2% discount rate instead of 4%, highlights the significant impact on the performance of adaptive strategies, emphasising the need for thoughtful consideration in aligning discount rates with societal values and long-term goals.

Real case study

While the framework demonstrated functionality in a conceptual case study and sensitivity tests, its true validation lies in application to a real case study. Currently, a relatively simple project area with straight forward measures were evaluated.

Besides, the framework used only direct damages in the CBA. Questions can be raised about the framework's completeness in capturing non-economic factors and potential limitations when applied to more complex case studies.

Conclusions

In this research it was investigated how to incorporate the value of adaptive pathway planning. It was found that an adequate evaluation method should include uncertainties, use diverse evaluation metrics and finally acknowledge the value of time

The study examined variables in adaptive planning with potential reduced uncertainty over time. Notably, there's no scientific basis for expecting a reduction in uncertainty regarding future sea level rise (SLR) projections. The next two decades are crucial for gaining clarity on potential scenarios, assuming the stability of SLR uncertainty. Economic growth rate and related factors show no conclusive evidence of decreasing uncertainty over time and are highly influenced by local conditions.

The framework incorporates uncertainty by using a Monte Carlo to include all stochastic variables. While certain metrics like the PoL and CV did not provide meaningful insights for the conceptual case study, the value of time was confirmed. Strategies adapting to observed sea level rise performed approximately 6% better. Additionally, the framework succeeded to show the difference in performances between static and adaptive strategies when conditions were altered. Secondly, within the adaptive strategies, the evaluation method also showed the difference in performance for the strategies with a flexibility premium, inspired on the ROA concept.

The developed framework is able to capture the impact of uncertainty and the value of time successfully. Although these features did not yield a significant impact on the results, the framework provides a proper foundation for further studies. When the framework is refined and validated through real case studies, it has the potential to serve as a valuable tool for decision-making. It enables the evaluation of adaptive pathways’ performance and supports the justification for either a static and robust strategy or a more flexible adaptive strategy.