Probabilistic GIS-based method for delineation of urban flooding risk hotspots

Identifying urban flooding risk hotspots is one of the first steps in an integrated methodology for urban flood risk assessment and mitigation. This work employs three GIS-based frameworks for identifying urban flooding risk hotspots for residential buildings and urban corridors. This is done by overlaying a map of potentially flood-prone areas [estimated through the topographic wetness index (TWI)], a map of residential areas and urban corridors [extracted from a city-wide assessment of urban morphology types (UMT)], and a geo-spatial census dataset. A maximum likelihood method (MLE) is employed for estimating the threshold used for identifying the flood-prone areas (the TWI threshold) based on the inundation profiles calculated for various return periods within a given spatial window. Furthermore, Bayesian parameter estimation is employed in order to estimate the TWI threshold based on inundation profiles calculated for more than one spatial window. For different statistics of the TWI threshold (e.g. MLE estimate, 16th percentile, 50th percentile), the map of the potentially flood-prone areas is overlaid with the map of urban morphology units, identified as residential and urban corridors, in order to delineate the urban hotspots for both UMT. Moreover, information related to population density is integrated by overlaying geo-spatial census datasets in order to estimate the number of people affected by flooding. Differences in exposure characteristics have been assessed for a range of different residential types. As a demonstration, urban flooding risk hotspots are delineated for different percentiles of the TWI value for the city of Addis Ababa, Ethiopia.

1. It should be noted that the Bayesian parameter estimation procedure presented in this work can employ any additional information regarding the potential flooding extent available for selected zones. Therefore, also the historical flood extent information can be used for estimating the TWI threshold. Strictly speaking, when using accurate hydraulic calculations for the estimation of the TWI threshold, only the information provided about the flooding extent is used. In other words, the flood height and velocity information provided by classical hydraulic routines are not going to be employed for estimating the TWI threshold.

2. In the case of more than one spatial window, the probability distribution for TWI threshold obtained based on the information coming from the first window is updated (in the framework of Bayesian updating) in order to incorporate the information from the next spatial window.

3. Assuming that the flood prone areas can be identified by a single TWI threshold value.

4. Note that A _W (FP) is a function of τ since the flood-prone areas are identified as areas with TWI> τ.

5. Note that this term could have also been estimated based on the information contained within window W. However, it was chosen to use the entire extent of the city as a reference. Therefore, in this case the information provided by the inundation profiles within W is not used. In fact, the term P(FP|τ, W) for simplicity is referred to as P(FP|τ), hereafter.

6. The conditioning on W is left out for brevity and simplicity of formulations.

7. Strictly speaking, the formulation in Eq. 9 should have been conditioned on the "correct identification" of the flood-prone areas for window W (see Eq. 2). However, for the sake of simplicity and tractability of the equations, we have used the symbol W in order to imply in, a concise manner, all the additional information about the inundation profile contained within window W.

8. It should be noted that the urban morphology type for a specific spatial unit represents the predominant morphology type for the unit in question. In other words, a proportion of the area is going to have alternative land cover, different from the predominant type. In this particular case, it is reasonable to presume that the rest of the population lives in the residential areas located in units that are not designated as residential.

9. Assuming that the uncertainty in the TWI threshold is the only source of uncertainty.

This work was supported in part by the European Commission’s seventh framework programme Climate Change and Urban Vulnerability in Africa (CLUVA), FP7-ENV-2010, Grant No. 265137. This support is gratefully acknowledged.

Authors and Affiliations

1. Department of Structures for Engineering and Architecture, University of Naples Federico II, Via Claudio 21, 80125, Naples, Italy

   Fatemeh Jalayer, Raffaele De Risi & Gaetano Manfredi

2. Department of Civil, Architectural and Environmental Engineering, University of Naples Federico II, Naples, Italy

   Francesco De Paola & Maurizio Giugni

3. Analysis and Monitoring of Environmental Risks (AMRA), Scarl, Naples, Italy

   Fatemeh Jalayer, Maurizio Giugni, Gaetano Manfredi, Paolo Gasparini & Maria Elena Topa

4. Ethiopian Institute of Architecture, Building Construction and City Development, Addis Ababa, Ethiopia

   Nebyou Yonas, Kumelachew Yeshitela & Alemu Nebebe

5. Division of Geography and Environmental Management, Manchester Metropolitan University, Manchester, UK

   Gina Cavan

6. School of Environment and Development, The University of Manchester, Manchester, UK

   Sarah Lindley

7. Chair for Strategic Landscape Planning and Management, Technical University of Munich, Munich, Germany

   Andreas Printz & Florian Renner

Corresponding author

Correspondence to Fatemeh Jalayer.

Reprints and permissions