Using the topographic wetness index (TWI) to identify urban flood-prone areas.

This article summarizes the paper “Efficient Hazard Assessment for Pluvial Floods in Urban Environments: A Benchmarking Case Study for the City of Berlin, Germany” published in water. The python script used in this paper is here.

The Topographic Wetness Index (TWI) is a widely used method to analyze the hydrological behaviour of a landscape. It was introduced by Kirkby in 1975 based on the concept that water flows downhill, and calculates a numerical index that represents how “wet” a particular location is based on the combination of its slope and the upslope contributing area.

It is calculated using a simple mathematical formula that incorporates both the topographic slope and the specific catchment area. TWI is used in fields such as hydrology, geomorphology, ecology, and soil science to identify areas of high or low hydrologic potential, to identify potential erosion and sedimentation problems, and to map wetland and upland vegetation. TWI values can also be used as a predictor of soil moisture, evapotranspiration, and runoff. It’s a widely used technique for stream network extraction, and drainage pattern identification, it gives a good representation of the hydrological behaviour of a landscape. It is calculated using python or GIS as follows:

TWI = ln(a / tan(slope))

Where:

• “a” is the specific catchment area (the total area contributing runoff to a particular location)
• “slope” is the local slope of the terrain, measured in radians

Identifying the urban flood-prone area (TWI method)

The method is based on assuming that a location is defined as flood-prone (FP) if its TWI value exceeds a certain threshold (τ). The estimation of τ employs using the maximum likelihood function and the inundation maps from a hydrodynamic model based on the below equation.

Figure. Flood-prone (FP) and inundated (IN) areas.

By using an inundation map for a small area within the city (called microscale ), the estimated TWI threshold identifies the flood-prone areas for the whole city (called mesoscale).

Figure. The spatial extent for the mesoscale and microscale.

We can then estimate the TWI threshold (τ) value associated with the maximum likelihood as shown in the below figure. The estimated threshold is τ = 5.3. This means, hence, that for a 100 mm storm, we would expect all areas in the city (mesoscale) with a TWI>5.3 to inundate.

Figure. Likelihood function L(τ/W) for a 100 mm precipitation depth

Limitations:

1- The TWI method has the obvious advantage that a map of TWI values can be created with minimal effort from a DEM. The TWI method, however, requires a “calibration” in the sense that a TWI threshold τ has to be estimated above which a location is considered as flood-prone (by using maximum likelihood estimation). This calibration has to be repeated for each precipitation depth of interest. To this end, a reference is required, here are the simulation results of a hydrodynamic model (TELEMAC-2D). While this requirement is certainly undesirable, as it again involves the effort to set up and apply a hydrodynamic model.

2- Using maximum likelihood estimation, we were unable to identify a local likelihood maximum and hence a value for τ for lower but still impact-relevant precipitation depths.

3- It identifies flood-prone areas but it does not estimate the floodwater depth which is important for stormwater management.

References

1. Kirkby, M. Hydrograph Modelling Strategies, 69–90; Peel, R., Chisholm, M., Haggert, P., Eds.; Processes in Physical and Human Geography, Heineman: London, UK, 1975. [Google Scholar]