**4. Discussion and Conclusions**

This study has collected empirical data on the cost of flood adaptation measures, using case studies and project-based information from various sources in the literature. The focus is on construction and maintenance costs for six categories of flood adaptation measures. The amount and quality of the data varies considerably, though recent research on specific flood-management issues has advanced the empirical basis of cost data. Data quality and uncertainty have been addressed by using cost intervals with upper and lower limits and an indication of the data quality of non-peer reviewed reports.

One of the issues with finding reliable costs is what cost categories are included in the aggregate cost estimate [13]. The studies used in this report are often unclear as to what cost components have been addressed in the final estimate, or have different components. This shows that cost estimates should be handled with care and with certain error margins; however, in most cases the estimates are probably on the conservative side, as some cost components have not been valued [14].

The issue of uncertainty in cost estimates also plays a role in the stage of the project at which the cost estimate was made: The error margins of cost estimates in the design and planning phases are obviously much higher than those of cost figures derived after project implementation. Wright and Pierce [33], for example, estimate 20% contingencies in a design for the dry flood-proofing of pumping stations. Addressing unexpected costs in the design phase holds especially for complex engineering projects such as the development of storm surge barriers, where the contingencies in the design phase are about 50% [106]. These projects suffer from rapid cost increases during the design phase, as requirements may change (e.g., a request for a higher protection standard), unforeseen complications in geographical conditions may arise (e.g., geological stability of the underground), or weather conditions which cause delays may occur.

Uncertainties also pertain to estimating the cost of operation and maintenance [13]. Such costs depend, for example, on the frequency of inspections and maintenance requirements that change over time due to the aging of the structure [26]. In the extensive review by Bayraktarov et al. [19], only a few of the underlying studies differentiate between cost components (e.g., between capital and operating costs) or provide information on other cost factors, such as planning, land acquisition, financing, monitoring, and repair/replacement. Some of the studies in this review estimate the annual operation and maintenance costs as a percentage of the construction costs but do not provide the data that underpins these estimates. Though yearly maintenance costs are low at first sight compared to construction costs, they add up quickly in an economic analysis, as they are valued over the lifetime of the measure.

Additional temporal aspects play a role in the interpretation and use of cost estimates [107]. While in existing studies, cost estimates are addressed as one-time investment numbers, such investments are often phased out over time. Examples are cost-benefit analyses for the planning of levee-reinforcement programs [106,108]. Furthermore, the lifetime of the proposed measures plays an important role in economic analysis. The expected lifetime of the larger investments (levees, storm surge barriers, sewer systems) are usually >50 years, whereas the lifetime is usually shorter (20–30 years) for cheaper measures such as flood-proofing buildings [26]. However, the required lifetime of an investment differs per country and even per project: The assumed lifetime for measures in a flood-management study in New York City is 50 years [29], whereas it is 100 years for measures in a comparable study in The Netherlands [109].

In addition to empirical data, studies exist that use modeling techniques to fit empirical cost data against explanatory variables. For example, a study by Mauer et al. [110] applies a model that calculates the length and size distribution of the sewer pipes in an urban area on the basis of rainfall intensity, housing densities, and area. Future research could expand such approaches. This also addresses the concerns of some researchers [13] that applying unit cost estimates assumes a fixed linear relationship between, for example, dike cost and some variable such as "meter height raised". Such costs may increase non-linearly with increasing dike heights, and non-linear models are needed to describe such relations—especially in the face of future climate change. Research by Lenk et al. [13], however, shows, on the basis of empirical data in The Netherlands and Canada, that a unit cost expression is adequate to express the cost of flood protection per height raised or per unit length.

Research on the economic evaluation of the cost and benefits for nature-based solutions (for example, in urban drainage) compared to hard-engineered drainage measures is at an early stage. This is due to a few factors: (1) Nature-based solutions, for example, in urban drainage, require irrigation and possible replanting until the vegetation is fully established, and these costs are still difficult to accurately determine [23]; (2) developers and city planners may be concerned that natural drainage options may decrease the area suitable for economic production; and (3) it is difficult to find data on the hydrologic benefits of the measures as reflected, for example, in a design standard. This is more straightforward in flood protection projects, where the design standard refers to a return period of maximum water level that should be incorporated in the design of an embankment. Though nature-based solutions are basically designed to do the same (i.e., lowering water levels, absorb wave energy, store water, etc.), more research and modeling is needed on the "hydrologic and hydraulic return" of a dollar investment in nature-based solutions.

Future research may address these issues, and expand the research with estimating both the cost and benefits of flood adaptation measures, and assess the benefit cost ratios (BCRs) of such measures. With such numbers, a comparative study can be conducted.

**Funding:** This work was conducted under VICI Grant nr. 453-13-006, from the Dutch Science Foundation.

**Acknowledgments:** I thank Sanne van Amsterdam for the detailed preparatory work and Lars de Ruig and Max Tesselaar for reading earlier versions of this manuscript.

**Conflicts of Interest:** The author declares no conflict of interest.
