In this blog, we address five factors that drive the flooding risk of real assets in Europe. This blog builds further on our previous blog, titled ‘Predicting asset flooding using machine learning and random forest classification’ which can be accessed using the button below.
In short, this blog series discusses a project conducted by Ian Ronk under supervision from Sander Van Splunter (UvA) and KR&A to predict the risk of flooding over a few return periods using local features such as the surrounding artificial imperviousness. This project resulted in a classifier that can predict whether a location will flood within a 20-year return period with an accuracy of 92%. This blog discusses a part of the literary review that was conducted as part of this project. In this blog, we will take a look at the top five biggest contributing intrinsic features to the predictability of the model, which are as follows:
- Soil Type
- Artificial Imperviousness
- Distance to River
- Extreme Precipitation
- GDP per capita
1. Soil Type
Every soil type has a different natural imperviousness: the amount of what that can pass through. For example, dry sand has a very low imperviousness, while wet clay is almost not possible to let any water through. The drainage of an area is an important aspect of flooding as good drainage results in the water not reaching further places, as it is absorbed by the ground. If the water, however, can travel further, it is able to reach more distant locations, seen from the river. Each soil type has a different natural imperviousness and is therefore considered as a local feature in predicting fluvial flooding.
2. Artificial Imperviousness
Artificial imperviousness is the percentage of permeation of a location as a result of artificial construction. This construction for example includes roofing: if a roof is constructed, the water cannot use this area to siphon into the ground. The same is true for tarmac roads. Unlike the soil type, where the water is usually able to be drained by the ground, although up to a certain degree, this artificial imperviousness makes the ground impenetrable and is therefore an even bigger contributor to the impermeability of an area. Artificial imperviousness has become a bigger and bigger problem, with urbanisation and the consequent expansion of cities. Studies have shown that the increase in the frequency of urban flooding can largely be attributed to this artificial imperviousness.
3. Distance to river
A logically reasoned feature is the distance to the river: if a location is far removed from a river, it will never experience fluvial flooding. On the other hand, if an asset is next to a river, it has a higher chance of flooding. A local feature was constructed by using a river map to calculate the distance to the nearest river.
4. Extreme precipitation
The most important contributor to fluvial flooding is precipitation. A river floods when there is an excess of water that flows through the river at a certain point in time. If there is a period of extreme precipitation, the river cannot channel the water away and as a result, the river overflows. An example of such an occurrence is the flash-flood that happened in London in July 2021 as a result of extreme summer precipitation. Data is available on the most precipitation in one day and in five consecutive days, which were both used as features to predict fluvial flooding.
5. GDP per capita
A more out-of-the-box feature is the GDP per capita in a region. The logic behind this feature is that more developed regions will have flood counter measurements in place to both protect expensive infrastructure and have the capital to protect themselves against the river. The amount of counter measurements is important for flood prediction. While there is no method to quantify this aspect, the GDP per capita, in addition to the quality of the infrastructure and government long-term vision are relatively good substitutes.
Investors can get insights into the risk of their assets, or any other asset in Europe, with KR&A’s Flash Due Diligence. Flooding risk may seem like a negligible element in your investment decisions, but once flooding occurs, the damages are profound and costly. Therefore, assessing the flooding risk of an asset is essential for every data-driven investor.
Improving portfolios
As an asset manager of a REIT or a private real estate fund you want to know where the flooding risk is concentrated. KR&A’s European Data Platform allows REITs and other management teams to rank their assets, in order to be able to weed out the high-risk assets, in a stealth exercise, through our off-market information.
As an asset manager of a REIT or a private real estate fund you want to know where the flooding risk is concentrated. Our data platform allows REITs and other management teams to rank their assets, in order to be able to weed out the high-risk assets, in a stealth exercise, through our off-market information.
Above a screenshot of the European Data Portal in which data can be aggregated. It ranks the top five most flooding-sensitive sub-portfolios of five European and UK retail REITs: Hammerson, Mercialys, Carmila, IGD SIIQ and Lar Espana. We classify retail property as high-street when it is below 5,000 sqm, as convenience retail between 5,000 and 50,000 sqm and Reg Mall when above 50,000 sqm. Subportfolios are ranked by High flood risk(%) One of the three Spanish Regional Malls of Lar comes up as a high flood risk asset, hence 33.3% of the Lar Espana sub-portfolio of Spanish Regional Malls.
