To avoid damage up to a given flood depth, dwellings can sometimes be elevated. According to a report of the , if a dwelling is built on a basement, a crawl space, or on open foundations, the first step for elevating it consists generally of separating it from its foundations and raising it on hydraulic jacks while it is held by a temporary support. Then, the existing foundations can be extended or new ones can be built. If the dwelling has slab-on-grade foundations, they are lifted together and new foundations are constructed below the slab

It is also possible to leave the slab on the ground and to construct a new floor after lifting the house.

We analysed the efficacy, cost, and cost-efficiency of elevating dwellings by 50, 100, and 250 cm. We assumed that only single-storey houses can be elevated because other types of dwellings are too large.

Measures can be taken to prevent the water from entering into a dwelling if the flood depth stays below 1 m and the flood lasts less than 48 h. For higher flood depths or longer flood durations, the pressure on the vertical elements of the building structure can cause severe damage.

For a dwelling without a basement, the following measures must be taken together and dimensioned consistently to prevent the water from entering up to a chosen threshold: installing flood barriers, repairing faulty seals in the external walls, waterproofing the external walls, treating cracks in the external walls, installing removable covers on small openings that are below the chosen threshold, installing anti-backflow valves, ensuring that the electrical cable sleeves are watertight, and buying a pumping device

For dwellings with a basement, additional measures must be taken. We only analysed the measures which aim to prevent the water from entering into a dwelling without a basement.

We define “component adaptation” as a measure which consists of using building materials that are resistant to water or cheap to repair or replace. We studied component adaptations which pertain to the ceilings, the walls, the floors, and the openings. We detail in Table  the component adaptations that pertain to each of these building components.

We used a computer tool called floodam and developed in R language to assess the efficacy of the strategies aimed at reducing the vulnerability of existing dwellings. floodam requires, as input, a numerical model of a building and produces a function which associates a level of damage with several combinations of flood depth and duration. We call this type of output a damage function. floodam has been developed to produce national damage functions that are recommended by the French state in its cost–benefit analysis methodology .

The efficacy and the cost of the strategies were assessed for several numerical models of dwellings. The cost of each strategy depends on the geometry of the dwellings and, in some cases, on whether the strategy is taken for an intact or damaged building and for the building materials. The efficacy is assessed by comparing the damage functions of the dwellings obtained with and without a strategy. It depends on the flood depth and duration, on the geometry of the dwellings, and on the building materials.

In this section, we describe the numerical models used as inputs, we provide an overview of how floodam works, we present the advantages of floodam to assess the strategies compared to other flood damage models, and we describe how we modelled the strategies. We then explain how we assessed the cost, the efficacy, and the cost-efficiency of each strategy.

The numerical model of a dwelling is made up of an XML file and a CSV file. The former indicates the ground floor height above ground level, the layout and size of the rooms, and the construction materials used for the building. Figure  provides a graphical interpretation of the XML file of a numerical model. The CSV file indicates the pieces of furniture that are present in each room and their height above the floor. For instance, Table  shows the information that pertains to a bedroom contained in the CSV file of a numerical model.

Originally, we had three numerical models which represent real dwellings: we visited an apartment to establish its plan and make an inventory of its furniture, the European Center for Flood Risk Prevention did the same with a single-storey house and gave us access to their data, and we used the architectural plan of a double-storey house. Since we had no real data regarding the furniture of the double-storey house, we first allocated the pieces of furniture of the single-storey house to the corresponding rooms of the double-storey house. Then, we added pieces of furniture to one room which had no correspondence in the single-storey house (a home office) and to two dressing rooms. More specifically, we assumed that the home office contained a desk, a chair, and a cupboard in chipboard and that the dressing rooms each contained a variety of small items.

In order to be able to generalise our results, we developed several versions of these three numerical models of dwellings by modifying the combination of the building components listed in Table . These components were chosen because they are widely used in French buildings, according to a report of the agency of building quality . Each version contains only one variant of each component listed in Table . Since there are 1728 possible combinations of all components, we developed the same number of versions for each of the three original numerical models of dwellings. In total, we thus have 5184 numerical models of dwellings.

We define the damage suffered by a good as the expected cost of the actions that must be performed after a flood in order to bring it back to its pre-flood state. Using this definition, floodam relies on two main assumptions: (1) the damage suffered by a building is the sum of the levels of damage suffered by its components, and (2) the levels of damage of the components are independent.

floodam relies on a database of 431 elementary damage functions. An elementary damage function associates a level of damage suffered by an elementary component with several combinations of flood depth and duration. A partition wall in plaster and a washing machine are examples of elementary components. The elementary damage functions come from interviews with insurance and construction experts. We used elementary damage functions with data points at immersion depth values from 0 to 500 cm (included) in 10 cm increments and immersion duration values from 0 to 144 h (included)

Some floods can last more than 144 h. The experts interviewed did not provide data regarding the vulnerability of the elementary components to such floods. Thus, we did not study the efficacy of precautionary measures for floods longer than 144 h.

For a given numerical model, floodam computes the height above the ground of all elementary components and the size or quantity of the elementary components of the building. Then, for each combination of depth and duration of immersion, it (1) computes the depth of immersion of each elementary component, (2) retrieves the corresponding levels of damage by measurement unit for the duration of immersion considered, (3) multiplies each level of damage by the size or quantity of the corresponding elementary component, and (4) sums the obtained values to obtain the level of damage at the scale of the dwelling. The damage function of the numerical model considered is made up of the levels of damage computed for each combination of immersion depth and duration.

floodam is used to estimate the damage due to floods that do not cause failure of walls and that do not involve salt water.

A more detailed description of floodam can be found in .

Numerous empirical and synthetic flood loss models exist. floodam belongs to the latter category. Empirical models are based on observed flood loss data, whereas synthetic models rely on a description of flood damage mechanisms .

Some empirical flood damage models include precautionary measures as explanatory variables (see for example ) and can be used to estimate their mean efficacy . These models account for the mean effect on flood damage of all the measures observed in the case studies used to produce them. Thus, they cannot be used to estimate the efficacy of specific precautionary measures. Moreover, the influence of the flood parameters and building characteristics on the efficacy of precautionary measures cannot be deduced from these models.

The damage mechanisms are more explicit in synthetic models (see for example ). They can be altered to depict the effect of specific precautionary measures.

To our knowledge, floodam is the synthetic model based on the most detailed database of elementary damage functions. This characteristic enabled us to examine the influence of a wide variety of building materials on the efficacy of specific precautionary measures.

As shown in Table , the cost of a given strategy does not depend on the immersion depth and duration. It always varies with the type of dwelling and the characteristic of the strategy (the height of elevation, the threshold, or the adaptation context). In the specific case of the component adaptation strategy, it also depends on the original variant used for the building components.

The cost of elevating a single-storey house lies between EUR 66 000 and EUR 109 000. Thus, it is always higher that the highest value of efficacy for this strategy. It increases with the elevation and is always the highest for dwellings that have slab-on-grade foundations.

The cost of dry proofing a dwelling ranges from EUR 6000 to EUR 10 000. For a given threshold, the cost of dry proofing is always maximum for the single-storey house and minimum for the double-storey house. This is due to the fact that the single-storey house has the greatest perimeter (54 m) and the double-storey house the smallest (40 m). The area on which some measures must be applied increases with the threshold. Thus, the cost of dry proofing is greater when the threshold is 100 cm than when it is 50 cm.

Regarding the adaptation of all building components, if it takes place on a damaged building (repair context), it is often less expensive to adapt the dwelling than to install the original variants of the components again. More precisely, it is the case for 91 % of the 5184 numerical models of dwellings. On the contrary, if the dwelling is intact when the adaptation takes place (prevention context), the mean cost of adapting all components is approximately 4 times higher than the cost of dry proofing. In the prevention context, the minimum costs relate to versions of the dwellings for which the original variants of the components are highly similar to the recommended ones. Note that we did not compute numerical models of dwellings that were made up of all the recommended variants of the components. The cost and efficacy of adapting all building components of such dwellings would be null. The highest adaptation costs relate to versions of the dwellings for which almost all components must be adapted.

Figure  shows the range of efficacy obtained for each strategy. The efficacy of a strategy depends on the type and components of the dwelling, on the immersion depth and duration, and sometimes on some characteristics of the strategy.

More specifically, the efficacy of elevating a dwelling increases with the value of elevation and with the value of the components of the dwelling. For this strategy, the highest efficacy is obtained for numerical models of dwellings which contain a lot of wooden components (joist boards, parquet, and opening frames and shutters in wood). The highest efficacy is observed for an immersion depth equal to the elevation value.

As for dry proofing, its efficacy increases with the threshold, the value of the components of the dwelling, and the floor area. The highest maximum efficacies relate to the single-storey house and the lowest to the double-storey house because the former has the largest floor area (133.5 m2) and the latter the smallest (98 m2). For a given numerical model of a dwelling, the efficacy of dry proofing increases up to an immersion depth equal to the threshold and an immersion duration of 48 h. In keeping with the assumptions used when modelling dry proofing, the efficacy is equal to zero for higher immersion depths and longer immersion durations.

Adapting all components can sometimes generate the same or a higher level of damage than keeping all the original components. This is the case for the 576 versions of each type of dwelling, which originally have masonry internal walls and tiles or textiles as coatings of floors. There is indeed a probability greater than zero that the adapted walls in plaster must be replaced for immersion depths greater than or equal to 30 cm, no matter the immersion duration. On the contrary, the probability that masonry walls must be replaced for immersion durations lower than 72 h is equal to zero. This type of wall needs only to be repaired in such cases. As a consequence, for some immersion depths, the elementary damage function of an internal wall in plaster is above the one of a masonry internal wall for immersion durations lower than 72 h. Negative levels of efficacy are observed only for dwellings which originally have tiles or textile as coatings of floors because the high efficacy of replacing parquet floors by sealed tiling floors compensates the negative efficacy of replacing masonry walls by walls in plaster. The efficacy of adapting all components is always positive for immersion durations higher than or equal to 72 h. The variance in the efficacy increases with the quantity of building components to adapt.

Table  shows the distribution of the efficacy of each strategy. The efficacy of elevating a dwelling lies between EUR 0 and EUR 65 000. Dry proofing a dwelling leads to a reduction of damage ranging between EUR 0 and EUR 36 000, and the efficacy of adapting all the components of the building is between EUR -14 000 and EUR 54 000. These results highlight the high variability of the efficacy of each strategy.

Figure  indicates the maximum cost-efficiency of each strategy, depending on the type of dwelling and on some characteristics of the strategy (the elevation, the threshold, or the adaptation context) and for several combinations of the time horizon and return period.

For the analysed values of the time horizon, it is never cost-efficient to elevate a single-storey house which is only exposed to floods with a return period higher than 30 years. If the dwelling has slab-on-grade foundations, the minimum return period for which it could be cost-efficient to elevate a dwelling is 20 years approximately.

Similarly, according to our results, adapting all building components is never cost-efficient for intact dwellings that are not exposed to floods that have a return period of less than 20 years. However, when component adaptations take place on a damaged dwelling, our results do not indicate ranges of the time horizon and return period for which this strategy is never cost-efficient.

Regarding dry proofing, the results are similar for the three types of dwellings and are not affected by the threshold. They suggest that dry proofing is never cost-efficient for dwellings that are not exposed to floods with a return period of 100 years or less.

Except in the case of adapting all the building components of a damaged building, we observe that the maximum return period for which a strategy is cost-efficient increases with the time horizon. For instance, for a dwelling that must be entirely dry proofed again after 20 years and that is not exposed to floods with a return period of less than 60 years, this strategy will never be cost-efficient.

The value of the building components by square metre has a positive effect on the efficacy of dry proofing and elevating a dwelling, while it does not affect the cost of these strategies. Hence, the more expensive the components of a dwelling, the more relevant it can be to elevate or dry proof it. By contrast, both the efficacy and the cost of dry proofing and elevation increase with the flood depth below which damage must be avoided and with the size of the dwelling. Consequently, these parameters do not affect the maximum cost-efficiency of dry proofing and elevation. According to our results, these strategies are unlikely to be cost-efficient for dwellings only exposed to floods with a return period higher than 100 and 30 years, respectively.

The efficacy of adapting the building components strongly depends on their original materials. It can even be negative for floods that last less than 72 h if the internal walls are originally in masonry. The cost of adapting the building components is influenced by the adaptation context and by the original materials. If the adaptation takes place on an already-damaged building, it is most of the time less expensive to adapt it than to reinstall the original variants of the components. However, it costs approximately EUR 40 000 on average to adapt an intact building. Since the cost and efficacy both increase with the quantity of components to adapt, the maximum cost-efficiency does not depend on the size of the dwelling. Adapting an intact dwelling that is not exposed to floods with a return period lower than 20 years is unlikely to be cost-efficient. In a repair context, we could not identify exposure levels for which it is never cost-efficient to adapt all building components.

Note that the cost-efficiency of all strategies increases with their lifespan and with the level of exposure in terms of frequency of floods.

Our results seem reliable, since they are mostly in line with previous studies. They suggest that elevating or adapting the building components of intact dwellings that are not exposed to frequent floods (i.e. with a return period lower than 30 years) should not be fostered by policymakers who wish to limit material damage due to floods. However, after a flood, it could be efficient to take advantage of the reconstruction phase to adapt the building components, since it is often less expensive to install the recommended components than to rebuild dwellings as they were before the event. Given that the post-disaster recovery often occurs in a climate of urgency, it could be useful to design in advance policy tools to help people adapt their dwelling during this phase.

Moreover, policymakers should not promote the installation of dry-proofing measures in dwellings that are not exposed to floods with a return period lower than 100 years.

Besides the level of exposure, the building components should be taken into account to assess the vulnerability of the dwellings and thus the relevance of implementing precautionary measures that sometimes generate higher costs than benefits.

Decisions regarding the allocation of public funds to communicate about the strategies considered here or to subsidise their installation cannot be based solely on the present study because of several limits relating either to its perimeter or to the method on which it relies. Several limits due to the method are linked to the assumptions and data used in floodam.