Extracting inundation patterns from flood watermarks with remote sensing SfM technique to enhance urban flood simulation: The case of Ayutthaya, Thailand Vorawit Meesuk a,b,d, ⁎, Zoran Vojinovic a,c , Arthur E. Mynett a,d a UNESCO-IHE Institute for Water Education, Westvest 7, 2611AX Delft, The Netherlands b Hydro and Agro Informatics Institute, eight floors, Bangkok Thai Tower 108, Rangnam Rd., Phayathai, Ratchathewi, Bangkok 10400, Thailand c University of Exeter, Exeter EX4 4QF, UK d Delft University of Technology, Faculty of Civil Engineering and Geosciences, Stevinweg 1, 2628, CN, Delft, The Netherlands abstract article info Article history: Received 24 October 2016 Received in revised form 2 February 2017 Accepted 9 March 2017 Available online xxxx Flood watermarks stipulate peak water depths from a flood event, indicating a magnitude of inundation that took place. Such information is invaluable for instantiation and validation of urban flood models. However, collecting and processing such data from land surveys can be costly and time-consuming. New remote sensing and data processing technologies offer improved opportunities to address these issues. The present paper deals with the new structure from motion (SfM) technology and its application in extracting flood watermarks. For this purpose, the first of its kind, side-view SfM surveys with two mobile units were utilised. Survey works were carried out in the vicinity of Ayutthaya heritage area (Thailand) and data obtained were used for setting up numerical models and simulations of the 2011 flood event. The work undertaken demonstrates the significant capability of SfM technology for extraction of flood watermarks. With such technology, it was possible to indicate façades, low- level structures, and susceptible openings, which in turn have improved schematizations of two-dimensional (2D) flood models. The resulting model simulations were found to be more accurate (i.e., more close to the mea- surements of flood watermarks) than those obtained from models with conventional top-view light detection and ranging (LiDAR) data. © 2017 Published by Elsevier Ltd. 1. Introduction Effective flood risk management in urban areas has become a grow- ing priority for city managers and disaster risk agencies. In view of fac- tors such as migration of people to urban areas, unplanned development, changing climate, and increasing operational and mainte- nance costs of urban water systems, this task is a challenge for all those involved in planning and managing urban water systems (Vojinović & Van Teeffelen, 2007; Sathish, Arya, & Vojinović, 2013; Sanchez, Medina, Vojinović, & Price, 2014; Singh, Arya, Taxak, & Vojinović, 2016). New remote sensing and data processing technologies offer im- proved opportunities to address these factors. One way of building resilience to floods and flood-related disasters is by investing in data collection and flood modelling activities. Corre- spondingly, city managers are increasingly engaging in collection, ar- chiving, and analysis of data for their urban areas, especially through facilities offered by advanced Geographic Information Systems (GIS) and remote sensing. As shown in many researches (Mynett & Vojinović, 2009; Vojinović & Abbott, 2012; Vojinović, 2015; Vojinović et al., 2016), GIS maps of areas at risk represent valuable information and communication facilities in their own right. Flood maps, which are typically based on numerical model results, can delineate flood- plains, zone areas for development and flood protection measures (Barreto, Vojinović, Price, & Solomatine, 2006; Vojinović, Solomatine, & Price, 2006a; Barreto, Vojinović, Price, & Solomatine, 2008; Vojinović, Sanchez, & Barreto, 2008; Barreto, Vojinović, Price, & Solomatine, 2010; Alves et al., 2016). Accuracies of flood models, and corresponding flood maps, crucially depends on nature of physical con- ditions and availability and quality of data. Nowadays, urban topogra- phy, drainage network or river channel layouts and even detailed geometry of urban surface can be readily surveyed using current tech- nology. However, there are numerous issues and uncertainties associat- ed with data collections, model calibrations, and modelling approaches (Abbott, Tumwesigye, & Vojinović, 2006). If flood flows are confined to well-defined conduits or river chan- nels, a robust 1D model can usually be sufficient to produce results safe for decision-making. As soon as model domain becomes more com- plex modelling task becomes a much greater challenge. Presence of Computers, Environment and Urban Systems 64 (2017) 239–253 ⁎ Corresponding author at: WSE, UNESCO-IHE Institute for Water Education, PO Box 3015, 2601AX Delft, The Netherlands. E-mail address: v.meeuk@unesco-ihe.org (V. Meesuk). http://dx.doi.org/10.1016/j.compenvurbsys.2017.03.004 0198-9715/© 2017 Published by Elsevier Ltd. Contents lists available at ScienceDirect Computers, Environment and Urban Systems journal homepage: www.elsevier.com/locate/ceus