Flood hazard potential reveals global floodplain settlement patterns

Flooding is one of the most common natural hazards, causing disastrous impacts worldwide. Stress-testing the global human-Earth system to understand the sensitivity of floodplains and population exposure to a range of plausible conditions is one strategy to identify where future changes to flooding or exposure might be most critical. This study presents a global analysis of the sensitivity of inundated areas and population exposure to varying flood event magnitudes globally for 1.2 million river reaches. Here we show that topography and drainage areas correlate with flood sensitivities as well as with societal behaviour. We find clear settlement patterns in which floodplains most sensitive to frequent, low magnitude events, reveal evenly distributed exposure across hazard zones, suggesting that people have adapted to this risk. In contrast, floodplains most sensitive to extreme magnitude events have a tendency for populations to be most densely settled in these rarely flooded zones, being in significant danger from potentially increasing hazard magnitudes given climate change.

In South and Central America, there are 97 million people settled on floodplains, which accounts for 15% of the total population across the region. Partially confined, confined, and laterally unconfined floodplains are found on 81%, 14% and 5% of all river reaches. Central America has the largest proportion of the population exposure found on confined floodplains (24%). Mexico has the largest population living on floodplains in this region (28 million people) and of this, 22% are living on confined floodplains.
In Europe, there are 98 million people settled on floodplains. Partially confined, confined and laterally unconfined floodplains are found on 83%, 13% and 5% of reaches across the region, however, 13% of population exposure is found on laterally unconfined floodplains, and only 3% is found on confined floodplains. Germany, Italy, and France have the highest total population living on floodplains, with 16 million, 14 million and 12 million people exposed, respectively.
In North America, 16% of the total population live on floodplains (59 million people), and of this, 55 million are in the United States (17% of the population are exposed to flooding). In the US, partially confined, confined and laterally unconfined floodplains are found on 79%, 17% and 4% of all river reaches, however, 11% of the population exposure is found on laterally unconfined floodplains, and only 5% is found on confined floodplains.

Fig. S1.
Correlation between physical attributes and river reach sensitivity to flooding. Physical properties have been calculated for each river reach, and their influence on the sensitivity of the river reach (br) to changing flood magnitude is assessed. Reaches are put into 500 bins (2000 points per bin) each representing 0.2% of the total data. Local topography is the dominant control. Spearman Rank Correlation Coefficients are given for the unbinned data. The error bars display the standard error of the mean for each bin. Flood extents grow most rapidly during rare, extreme magnitude events due to steep slopes next to the river channel constraining how flood extents can grow laterally during frequent, low magnitude events. Partially confined are the dominant type of floodplain across the globe and are found on transitional streams, e.g., valley bottoms. Flood extents grow most rapidly during frequent, low magnitude events, which is due to the wide and flat terrain next to the river channel that meets a break in slope which can constrain the growth of inundation. Laterally unconfined floodplains are very wide and flat and are found on low slope alluvial streams. They are unbounded which allows for exponential growth in inundation area. Flood extents grow most rapidly during rare, extreme magnitude events.  Figure 3 & S4, where we have used a log scale when plotting the data. This is due to the values that br and bpop take. When b is less than 1, values range between 0 and 1, however, when b is greater than 1, theoretically values can range between 1 and infinity. where population in absolute terms is great in frequently flooded areas of the floodplain) have high standards of structural protection. Continent specific plots are included in Fig. S5 for ease of visualisation.

Fig. S5.
Interaction between flood hazard and population exposure sensitivity by continent. Country level mean flood extent and population exposure sensitivity parameters. The regions presented in Figure 3 (main text) and Figure S4 have been separated in order to highlight the regional differences. Point shapes represent the floodplain category for each country (circles for partially confined, triangles for laterally unconfined, squares for confined), the sizes have been scaled based on the number of people living on the different floodplain types in each country, and the colour refers to the region/continent that the country is in.   Exposure density of the floodplain categories. River reaches have been separated into their sub-region and the floodplain category. Exposure density has been calculated as the number of people who are exposed per kilometre of river reach in each of the floodplain categories. This highlights the high density of exposure that is found on floodplains in Asia.

Fig. S9.
Diagram of quantifying the sensitivity of flooded extents. 1) Upstream inflow points and downstream boundary conditions are used to define reaches along the global river network. Flooded extents are extracted between these upstream and downstream points along a diagonal transect using the reach length as an effective search radius. This is done for each return period. 2) extracted flooded areas are normalised and plotted against the normalised return period. A power law is fitted to these points. The exponent parameter of this power law is used to describe the shape of the growth curve of the flooded extents.