Identification of flood affected roads using topographical data
- 1. Identification of Flood Afflicted Road Blockages using Satellite Imaging, Topographical Data and Elevation Clustering International Water Conference for Sustainable Development Goals (IWCSDG-2024) Presented by: Dr. Swati Sirsant Assistant Professor Department of Civil Engineering Nirma University
- 2. • Floods are one of the most frequent and devastating natural calamities which have huge impacts on life and property. • One of the major impacts of flood include disruption of essential services such as water, electricity, and transportation. • The closure of roads due to flooding has huge impacts on human safety as well as it leads to economic losses. Several past studies have tried to study the impact of flooding on transportation networks. • Traditional flood inundation mapping methods include use of hydraulic and hydrologic models such as HEC- RAS, HEC-HMS, MIKE, LISFLOOD, simplified conceptual models such as Rapid Flood Spreading Method (RFSM), Height Above Nearest Drainage Network (HAND) etc. Introduction
- 3. Literature Review Model Type Description Authors Advantages Disadvantages Empirical Models Based on empirical equations such as Dicken’s formula, Ryves formula, Inglis formula etc. Smith 1997, O'Connor and Costa, 2004), Bellos, 2012, Jun et al. 2016, Teng et al., 2017 Robust models, give accurate results Require collection of huge data sets Hydrodynamic models These models simulate water movement by solving equations formulated by applying laws of physics Ex: HEC-RAS, LISFLOOD, MIKE Caleffi et al., 2003; Alcrudo, 2004; Prakash et al. 2014; Roberts et al., 2015 Fluid dynamics are taken into account Cannot be applied for larger areas, are computationally demanding Simplified conceptual models Substitute for more sophisticated hydrodynamic Models; such as rapid flood spreading method (RFSM), height above nearest drainage network (HAND) model; Lhomme et al. 2008; Nobre et al. 2011; Teng et al. 2017 Suitable for data- sparse areas and large study areas and is particularly computationally efficient These models diverge in accuracy from hydrodynamic models in locations with complex topography and significant flow momentum Conservation; Not suitable for flash flood, tsunami or bank erosion studies
- 4. Literature Review Authors Objective Data Used Findings Bates and Roo, 2000 Development of a new model for simulating flood inundation using only DEM as input Synthetic Aperture Radar (SAR) dataset The developed model provides good approximation to raster based models Sanders (2007) Utility of several on-line DEMs is examined with a set of steady and unsteady test problems Interferometric synthetic aperture radar (IfSAR); Shuttle Radar Topography Mission (SRTM); LiDAR This study highlights utility in SRTM as a global source of terrain data for flood modeling Saksena and Merwade (2015) Relate the errors arising from DEM properties such as spatial resolution and vertical accuracy to flood inundation maps, and then use this relationship to create improved flood inundation maps from coarser resolution DEMs with low accuracy National Elevation Dataset and Shuttle Radar Topography Mission (SRTM) Application of this approach show that improved results can be obtained from flood modeling by using coarser and less accurate DEMs
- 5. Literature Review Authors Objective Findings Fohringer et al. (2015) Used social media posts as a source of identifying the flood prone areas Developed "PostDistiller", a tool to filter geolocated posts from social media services which include links to photos Rosser et al. (2017) Using geotagged photographs sourced from social media, optical remote sensing and high- resolution terrain mapping, for estimation of flood probability Accurate predictions were obtained having receiver operating curve values 0.95 and 0.93 for model fitting and testing respectively. Li et al. (2017) Presented the integrated use of social media data and artificial neural networks for emergency response applications social media data can be used as a complement to remote sensing data sets for emergency response applications
- 6. Research Gaps Most of the past studies rely on extensive data set, which makes them unsuitable to be used for determination of road inundation expeditiously The objective of the present study is to develop a methodology for determining flood afflicted roads using remote sensing data. The methodology is evaluated on the east zone of Bangalore city, which is one of the highly flood-affected areas of the country. The results in terms of probability of inundation of various roads is presented, divided into four categories: severe, safe, vulnerable, highly vulnerable, and extremely vulnerable Research Objectives
- 7. Study Area Bangalore city is one of the most flood affected areas in India, where flooding takes place even for low to moderate amount of rainfall. Bruhat Bengaluru Mahanagara Palike (BBMP) has identified 209 vulnerable spots across the city, out of which 58 have been categorised as sensitive and 151 as moderately sensitive. East zone has the highest number of vulnerable spots amounting to 38, followed by Dasarahalli, which has 37 spots
- 8. Methodology
- 9. Methodology a) Topographical Data Extraction 1.Downloading Elevation Data • The DEM for the Bangalore east zone was downloaded from https://earthexplorer.usgs.gov/ having a spatial resolution of 1 arc-second, which is equal to 30 meters. • Two raster images were downloaded to cover the study area. 2. Downsampling • The DEM raster contains continuous data, resulting in numerous pixels. • To enhance plotting and visualization, downsampling of the raster is required which reduces the pixel density. • This process was implemented using the Pillow module in Python. • By downsampling the DEM, the processing efficiency is enhanced drastically. 3. Extracting Coordinates and Elevation Values • The co-ordinates and elevation values of all the points were extracted from the Rasterio Python Module of the downscaled DEM and exported as CSV file to be used for further analysis.
- 10. Methodology b) Elevation Clustering • The elevation values were clustered into different groups using K-means clustering. • K-means clustering is an unsupervised machine learning algorithm which groups the unlabelled dataset into different clusters. • The objective of K-means clustering is to divide or classify the entire population into a number of groups such that data points within each group are more compatible to each other and different from the data points in other groups. • K-means clustering was applied to the elevation data extracted from the downscaled DEM. • The value of K is a hyper-parameter and is user defined. • Various methods are available for determining the optimal value of K, such as elbow method, silhouette analysis, gap statistic etc. • However, in the present study, a simple heuristic is applied to determine the number of clusters. • To identify large flood regions, a smaller number of clusters should be chosen (e.g., k=4), i.e if we have less cluster that will hold large depth difference so that will give idea for a larger inundated area. On the other hand, if we choose higher K value then it will have vey less depth difference which will help us to consider a small region.
- 11. Methodology c) Plotting Clusters on OpenStreetMap Layer Utilizing the updated CSV with cluster values, the Python Folium module was employed. OpenStreetMap serves as the base layer, enabling the plotting of points based on location coordinates. Each point was assigned a color corresponding to its cluster. This approach enhances visualization, leveraging the clarity of roads and rivers in the OpenStreetMap layer for analysis. d) Visualization of Results Finally, a web interface was created for visualization of the results. The web interface contains two layers, the elevation clusters and the OpenStreetMap. The identification of different elevation regions followed by the identification of surrounding elements such as rivers, water bodies, roads etc. can be useful to determine the flood prone regions. https://karmathecoder.github.io/web2/
- 12. Results Snap of the interface
- 13. Results Wheeler Road and Assayee Road Junction with elevation clusters
- 14. Results Windson Square with elevation clusters
- 15. Conclusions • The present study employed the topographical data for deriving the elevation clusters. • K-means clustering approach was employed for clustering the elevation values into 4 clusters. • These results have immediate practical applications, empowering decision-makers in urban planning, disaster response, and infrastructure management to strategically allocate resources and enhance overall preparedness. • The present study, however, incorporated only elevation as the attributing factor, for flood identification. • Future studies should focus on incorporating other factors into the model such as slope, land use land cover, rainfall, distance to river etc.