IMPROVING FLOOD PREDICTION ASSIMILATING UNCERTAIN CROWDSOURCED DATA INTO HYDROLOGIC AND HYDRAULIC MODELS IMPROVING FLOOD PREDICTION ASSIMILATING UNCERTAIN CROWDSOURCED DATA INTO HYDROLOGICAL AND HYDRAULIC MODELS DISSERTATION Submitted in fulfilment of the requirements of the Board for Doctorates of Delft University of Technology and of the Academic Board of the UNESCO-IHE Institute for Water Education for the Degree of DOCTOR to be defended in public on Monday, 28 November 2016, at 10:00 hours In Delft, the Netherlands by Maurizio MAZZOLENI Master of Science in Environmental Engineering University of Brescia, Brescia born in Brescia, Italy This dissertation has been approved by the promotor: Prof. dr. D.P. Solomatine and copromotor: Dr. L. Alfonso Composition of the doctoral committee: Chairman Rector Magnificus TU Delft Vice-Chairman Rector UNESCO-IHE Prof.dr. D.P. Solomatine, UNESCO-IHE/TU Delft, promotor Dr. L. Alfonso, UNESCO-IHE, copromotor Independent members: Prof. dr. ir. A.W. Heemink, TU Delft Prof. dr. ir. A. Weerts, Wageningen University Dr. ir. H. Madsen, University of Copenhagen, Denmark Prof. dr. ir J.A. Roelvink, UNESCO-IHE/TU Delft Prof. dr. N.C. van de Giesen, TU Delft, reserve This research was conducted under the auspices of the Graduate School for Socio- Economic and Natural Sciences of the Environment (SENSE) CRC Press/Balkema is an imprint of the Taylor & Francis Group, an informal business © 2016, Maurizio Mazzoleni Although all care is taken to ensure the integrity and quality of this publication and information herein, no responsibility is assumed by the publishers, the author nor UNESCO-IHE for any damage to the property or persons as a result of the operation or use of this publication and/or the information contained herein. A pdf version of this work will be made available as Open Access via http://repository.tudelft.nl/ihe This version is licensed under the Creative Commons Attribution-Non Commercial 4.0 International License, http://creativecommons.org/licenses/by-nc/4.0/ Published by: CRC Press/Balkema PO Box 11320, 2301 EH Leiden, The Netherlands [email protected] www.crcpress.com – www.taylorandfrancis.com ISBN 978-1-138-03590-4 (Taylor & Francis Group) To my parents, Mirella and Luciano To Jessica v A CKNOLEDGMENTS Once Nelson Mandela said “It always seems impossible until it's done”. Indeed, during my PhD research I had this feeling many times, but it is because of the help of my supervisors, colleagues, family and friends that I have finally made it. First of all, I would like to express my sincere gratitude to my promotor Prof. Dimitri Solomatine and co-promotor Leonardo Alfonso. Dear Dimitri, in these four years you provided me with creative and innovative ideas, critical thought and advice that inspired me during my PhD research. Thanks for always believing in me, and involving me in numerous research activities and projects. Leo, thank you for always being available for a nice word or suggestion, for sharing your ideas and for the insightful discussions on innovative research topics. This PhD research project was part of the WeSenseIt Project which was funded by the Seventh Framework Programme of the European Union. I really enjoyed being part of such an important project and I would like to thank all the partner for their support. In particular, I appreciate the dedication of Martina, Daniele and Michele of the Alto Adriatico Water Authority for sharing flood data together with their hydrological and hydraulic models with us. I am grateful to Jolanda, Anique and Jos for all the administrative support, Paula Derkse for the suggestion for printing this thesis, Benno for helping me with the samenvatting and Joanne for the proofread of the thesis. Thanks to all the Hydroinformatics staff members, Ioana, Andreja, Schalk Jan, Biswa, and Gerald for involving me in the teaching activities of our group. I am very thankful to Prof. Dong-Jun Seo for the opportunity to spend four months as a visiting scholar at the University of Texas at Arlington where I also met Dr. Seong-Jin Noh. Dear D.J. and Seong-Jin, thank you very much for your hospitality and your precious scientific advice on data assimilation, I'm sure in the future there will be opportunities to strengthen the collaboration created between UTA and UNESCO-IHE. Special thanks go to Behzad Z., Hamideh R., Hamideh H., Ray and Bahzad N. for making me feel at home in Arlington. Luigia, thank you very much for always being there when I needed it, for your sincere comments on the readability of our papers and for the funny moments during the coffee breaks. Thank you Giuliano for being not only a supervisor but also a friend. It is also because of you if I am now here. I wish to you, Luigia and Tommaso all the best. i There are no words to convey my feelings to all my friends in Delft. Thanks Yared for the 3 years sharing the workspace, for the Saturday football in TU, for all the fun during conferences and many other moments. Juan, project buddy, thanks for all the discussions about data assimilation, for the support during the PhD and for helping to finish all that vodka in Krakow. Kun, sometimes it was more difficult to explain my jokes to you than data assimilation to my mother, thanks for all those nice memories. Thanks Angi, my friiiend, for all the fonny moments and your patience with me in the gym. Thanks Maribel, I could not ask for a better roommate, even if we had some bread-related issues. Thank you to all of you guys, Benno, Elena, Paolo, Jojo, Juliette, Aki, Pato, Miguel, Alex, Alessio, Alessandro, Neiler, Vero, Mark, Fer, Peter, Alida, Pablo, Claudia, Sara, Arlex, ShanShan, Erika, Andres, Juan Pablo, Paola, Natalia, Micah, Mario, Quan, Mohaned, Stefan, Quintilia, Dibesh, Joanne, Imra, Hadi, Feroz, Marianne, Laura, Victor, and many more for being here all this time. I would like to express my immense gratitude to Vale and Matteo for their continuous support, Ale and Livio for unforgettable moments in the mountains and inspiring conversations and Davide and Michela for their Sunday pizza dinners and wedding talks. I would like to thank also Cristian, Manuela, Ippolita, Gianni, Marco, Ciro, Manuel and Laura for always finding time for me when I was back in Italy even if very busy. Last but not least, I would like to give a special thanks to someone who has been constantly with me for these last four years, in the rainy and sunny days, in the sad and happy moments and, many times, overnight. We had difficult moments but at the end, somehow, it always “ran” smoothly. Thank you my dear Laptop. Un ringraziamento speciale va ai miei genitori, Mirella e Luciano. Grazie per la vostra fiducia, paziena e per avermi sempre sostenuto incondizionatamente in questi ultimi quattro anni (e non solo) anche quando le mie scelte mi hanno portato lontano da casa per molto tempo. Un pezzo di questo dottorato é anche vostro. Grazie di cuore ai miei zii Rossella, Battista e Gianni per essere sempre stati li quando c’era bisogno e per aver aiutato la mamma negli ultimi due anni. Grazie Daniela per aver supportato ma soprattutto sopportato quel vecchio brontolone in questi ultimi anni. I do not think I would ever get to finish my PhD without you, my wife, my lover, my best friend, Jessica. In these last four years you saw the best and worst parts of me and you were always there to cheer me up. My journey would not have been possible without your continuous and unlimited support and help. Thanks for your love, encouraging me in all of my pursuits and inspiring me to follow my dreams. Te amo ii S UMMARY Monitoring stations have been used for decades to measure hydrological variables, and mathematical water models used to predict floods can be enhanced by the incorporation of these observations, i.e. by data assimilation. The assimilation of remotely sensed water level observations in hydrological and hydraulic modelling has become more attractive due to their availability and spatially distributed nature. In recent years, continued technological advances have stimulated a spread of low- cost sensors that has triggered crowdsourcing as a way to obtain observations of hydrological variables in a more distributed way than the classic static physical sensor networks. The main advantage of using this type of sensors is that they can be used not only by technicians, as with observations from traditional physical sensors, but also by regular citizens. However, there are also drawbacks of using these observations, e.g. their relatively limited reliability, varying accuracy in time and space, and their irregular and non a-priori defined availability. For this reason, crowdsourced observations have not been widely integrated in hydrological and/or hydraulic models for flood forecasting applications. Instead, they have generally been used to validate model results against observations, in post-event analyses. Model updating is a strategy that aims at improving models using observations. A particular case of model updating is data assimilation, which often uses measured data such as streamflow, soil moisture, etc. coming from static physical stations. However, only a few studies have considered the integration of crowdsourced observations into water-related models. The main objective of this research is to investigate the benefits of assimilating crowdsourced observations coming from a distributed network of heterogeneous physical and social (static and dynamic) sensors within hydrological and hydraulic models, in order to improve flood forecasting. Standard data assimilation approaches, such as Kalman filtering, ensemble Kalman filtering, nudging, etc. are applied to the three different case studies to assimilate crowdsourced observations of variable accuracy and random life-span. The results of this study demonstrate that crowdsourced observations can significantly improve flood prediction if they are properly integrated in hydrological and hydraulic models. In particular, this research proved that assimilation of streamflow observations from static physical sensors provides improvements in model performance, the magnitude of which depends on the observation locations and model structure. In case of the Brue catchment, the best model improvement is achieved by assimilating iii