Articles | Volume 23, issue 2
https://doi.org/10.5194/nhess-23-809-2023
© Author(s) 2023. This work is distributed under
the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
https://doi.org/10.5194/nhess-23-809-2023
© Author(s) 2023. This work is distributed under
the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Research article
| 
24 Feb 2023
Research article |
| 24 Feb 2023
| 24 Feb 2023
Transferability of data-driven models to predict urban pluvial flood water depth in Berlin, Germany
Department of Hydrology and Climatology, Institute of Environmental Science and Geography, University of Potsdam, Potsdam, Germany
Georgy Ayzel
Department of Hydrology and Climatology, Institute of Environmental Science and Geography, University of Potsdam, Potsdam, Germany
Axel Bronstert
Department of Hydrology and Climatology, Institute of Environmental Science and Geography, University of Potsdam, Potsdam, Germany
Maik Heistermann
Department of Hydrology and Climatology, Institute of Environmental Science and Geography, University of Potsdam, Potsdam, Germany
Related authors
No articles found.
Sophia Dobkowitz, Leon Frederik De Vos, Deva Charan Jarajapu, Sarah Lindenlaub, Guilherme Samprogna Mohor, Omar Seleem, and Axel Bronstert
Hydrol. Earth Syst. Sci., 30, 3041–3060, https://doi.org/10.5194/hess-30-3041-2026, https://doi.org/10.5194/hess-30-3041-2026, 2026
Short summary
Short summary
Surface sealing makes cities vulnerable to flooding caused by heavy rain. Green infrastructure, such as green roofs, can reduce flooding. This modelling study investigates the potential of green infrastructure to reduce runoff, flood water depth and building damage. Bioretention systems turned out to be the most space efficient compared to green roofs and pervious pavement. For larger rain events, more green infrastructure implementation is needed to achieve relevant flood mitigation.
Paul Voit, Felix Fauer, and Maik Heistermann
Nat. Hazards Earth Syst. Sci., 26, 2189–2201, https://doi.org/10.5194/nhess-26-2189-2026, https://doi.org/10.5194/nhess-26-2189-2026, 2026
Short summary
Short summary
Reliable flood estimates are vital for effective flood risk management, yet short observation records often limit accuracy. We present a new approach that adds realistic "what if" flood scenarios from nearby similar catchments. Applied to thousands of small German basins, our method improves flood frequency estimates, reducing uncertainty and supporting more robust planning for extreme events.
Amalie Skålevåg, Lena Katharina Schmidt, Nele Eggers, Jana Tjeda Brettin, Oliver Korup, and Axel Bronstert
Hydrol. Earth Syst. Sci., 30, 2717–2739, https://doi.org/10.5194/hess-30-2717-2026, https://doi.org/10.5194/hess-30-2717-2026, 2026
Short summary
Short summary
As glaciers retreat in mountain regions, heavy rainstorms increasingly control how much soil and sediment rivers carry downstream. We analysed rainfall and sediment data over 21 years in the Austrian Alps and found heavy precipitation events becoming more important for sediment movement, although total annual suspended sediment transport is declining as glaciers shrink. This shift may increase flood hazards, affecting ecosystems and water quality downstream.
Patricio Yeste, Lieke A. Melsen, João Paulo L. F. Brêda, Nicolás Tacoronte, Andrea Saltelli, Giulia Vannucci, Roberta Siciliano, and Axel Bronstert
EGUsphere, https://doi.org/10.5194/egusphere-2026-1787, https://doi.org/10.5194/egusphere-2026-1787, 2026
This preprint is open for discussion and under review for Hydrology and Earth System Sciences (HESS).
Short summary
Short summary
Understanding which factors most influence streamflow is key for accurate hydrologic modelling. This study shows that machine learning methods, like random forests and neural networks, can efficiently identify the most important model inputs, producing results similar to traditional sensitivity analysis but with far less computation. This approach helps scientists explore complex models faster and more reliably, improving insights into how catchments respond to changing conditions.
Paul Voit, Marie Hundhausen, Larisa Seregina, Hendrik Feldmann, Bodo Ahrens, and Maik Heistermann
EGUsphere, https://doi.org/10.5194/egusphere-2026-1229, https://doi.org/10.5194/egusphere-2026-1229, 2026
This preprint is open for discussion and under review for Natural Hazards and Earth System Sciences (NHESS).
Short summary
Short summary
Extreme rainfall is expected to increase with a warming climate. However, it remains unclear how this will impact future floods. This study utilizes an ensemble of 6 new datasets, containing the future projected rainfall, and compares the simulated flash floods of the past (1971–2000) and future (2071–2100) under the RCP8.5 scenario for the German part of the Danube catchment. Our results indicate an increase of 100-yr flood return levels of 30 % across all subcatchments under this scenario.
Peter M. Grosse, Elodie Marret, Lena Scheiffele, Katya Dimitrova Petrova, Till Francke, Daniel Altdorff, Maik Heistermann, Merlin Schiel, Carsten Neumann, Daniel Scheffler, Mehdi Saberioon, Matthias Kunz, Jonas Marach, Marcel Reginatto, Miroslav Zboril, Anna Balenzano, Daniel Rasche, Christine Stumpp, Benjamin Trost, and Sascha E. Oswald
Earth Syst. Sci. Data, 18, 1703–1727, https://doi.org/10.5194/essd-18-1703-2026, https://doi.org/10.5194/essd-18-1703-2026, 2026
Short summary
Short summary
This data paper describes a comprehensive collection of soil moisture and related data from an extensive cosmic-ray neutron sensing (CRNS) network at an agricultural research site in north-east Germany. The data set comprises not only soil moisture observations at different spatio-temporal scales, but also a wealth of accompanying data that provide the context to interpret soil moisture dynamics within a broader hydrological and environmental framework.
Maik Heistermann, Daniel Altdorff, Till Francke, Martin Schrön, Peter M. Grosse, Arvid Markert, Albrecht Bauriegel, Peter Biró, Sabine Attinger, Frank Beyrich, Peter Dietrich, Rebekka Eichstädt, Jakob Terschlüsen, Ariane Walz, Steffen Zacharias, and Sascha E. Oswald
Nat. Hazards Earth Syst. Sci., 26, 465–486, https://doi.org/10.5194/nhess-26-465-2026, https://doi.org/10.5194/nhess-26-465-2026, 2026
Short summary
Short summary
The German federal state of Brandenburg is particularly prone to soil moisture droughts. To support the management of related risks, we introduce a novel soil moisture monitoring network based on cosmic-ray neutron sensing technology. The resulting observations can be particularly helpful to evaluate and improve soil moisture products from modelling or remote sensing.
Gerd Bürger and Maik Heistermann
EGUsphere, https://doi.org/10.5194/egusphere-2025-3584, https://doi.org/10.5194/egusphere-2025-3584, 2025
Short summary
Short summary
For four regions of Germany we project future extreme rainfall events from GCM projections. The fields are linked to the local events by classifying observed fields. Classifiers are conventional procedures along with modern schemes. Positive trends are identified for two antagonists model fields, causing large uncertainties of the local trends. Uncertainty is analyzed using the factors of event severity, emissions, GCMs, classifiers, and region. Positive trends outweigh the negative ones.
Xiaoxiang Guan, Viet Dung Nguyen, Paul Voit, Bruno Merz, Maik Heistermann, and Sergiy Vorogushyn
Nat. Hazards Earth Syst. Sci., 25, 3075–3086, https://doi.org/10.5194/nhess-25-3075-2025, https://doi.org/10.5194/nhess-25-3075-2025, 2025
Short summary
Short summary
We evaluated a multi-site stochastic regional weather generator (nsRWG) for its ability to capture the cross-scale extremity of heavy-precipitation events (HPEs) in Germany. We generated 100 realizations of 72 years of daily synthetic precipitation data. The performance was assessed using WEI and xWEI indices, which measure event extremity across spatiotemporal scales. The results show that nsRWG simulates the extremity patterns of HPEs well, although it overestimates short-duration small-extent events.
Till Francke and Maik Heistermann
Nat. Hazards Earth Syst. Sci., 25, 2783–2802, https://doi.org/10.5194/nhess-25-2783-2025, https://doi.org/10.5194/nhess-25-2783-2025, 2025
Short summary
Short summary
Brandenburg is among the driest federal states in Germany. The low groundwater recharge (GWR) is fundamental to both water supply and the support of natural ecosystems. In this study, we show that the decline of observed discharge and groundwater tables since 1980 can be explained by climate change in combination with an increasing leaf area index. Still, simulated GWR rates remain highly uncertain due to the uncertainty in precipitation trends.
Till Francke, Cosimo Brogi, Alby Duarte Rocha, Michael Förster, Maik Heistermann, Markus Köhli, Daniel Rasche, Marvin Reich, Paul Schattan, Lena Scheiffele, and Martin Schrön
Geosci. Model Dev., 18, 819–842, https://doi.org/10.5194/gmd-18-819-2025, https://doi.org/10.5194/gmd-18-819-2025, 2025
Short summary
Short summary
Multiple methods for measuring soil moisture beyond the point scale exist. Their validation is generally hindered by not knowing the truth. We propose a virtual framework in which this truth is fully known and the sensor observations for cosmic ray neutron sensing, remote sensing, and hydrogravimetry are simulated. This allows for the rigorous testing of these virtual sensors to understand their effectiveness and limitations.
Georgy Ayzel and Maik Heistermann
Nat. Hazards Earth Syst. Sci., 25, 41–47, https://doi.org/10.5194/nhess-25-41-2025, https://doi.org/10.5194/nhess-25-41-2025, 2025
Short summary
Short summary
Forecasting rainfall over the next hour is an essential feature of early warning systems. Deep learning (DL) has emerged as a powerful alternative to conventional nowcasting technologies, but it still struggles to adequately predict impact-relevant heavy rainfall. We think that DL could do much better if the training tasks were defined more specifically and that such specification presents an opportunity to better align the output of nowcasting models with actual user requirements.
Patricio Yeste, Matilde García-Valdecasas Ojeda, Sonia R. Gámiz-Fortis, Yolanda Castro-Díez, Axel Bronstert, and María Jesús Esteban-Parra
Hydrol. Earth Syst. Sci., 28, 5331–5352, https://doi.org/10.5194/hess-28-5331-2024, https://doi.org/10.5194/hess-28-5331-2024, 2024
Short summary
Short summary
Integrating streamflow and evaporation data can help improve the physical realism of hydrologic models. We investigate the capabilities of the Variable Infiltration Capacity (VIC) to reproduce both hydrologic variables for 189 headwater located in Spain. Results from sensitivity analyses indicate that adding two vegetation parameters is enough to improve the representation of evaporation and that the performance of VIC exceeded that of the largest modelling effort currently available in Spain.
Paul Voit and Maik Heistermann
Nat. Hazards Earth Syst. Sci., 24, 4609–4615, https://doi.org/10.5194/nhess-24-4609-2024, https://doi.org/10.5194/nhess-24-4609-2024, 2024
Short summary
Short summary
Floods have caused significant damage in the past. To prepare for such events, we rely on historical data but face issues due to rare rainfall events, lack of data and climate change. Counterfactuals, or
what ifscenarios, simulate historical rainfall in different locations to estimate flood levels. Our new study refines this by deriving more-plausible local scenarios, using the June 2024 Bavaria flood as a case study. This method could improve preparedness for future floods.
Amalie Skålevåg, Oliver Korup, and Axel Bronstert
Hydrol. Earth Syst. Sci., 28, 4771–4796, https://doi.org/10.5194/hess-28-4771-2024, https://doi.org/10.5194/hess-28-4771-2024, 2024
Short summary
Short summary
We present a cluster-based approach for inferring sediment discharge event types from suspended sediment concentration and streamflow. Applying it to a glacierised catchment, we find event magnitude and shape complexity to be the key characteristics separating event types, while hysteresis is less important. The four event types are attributed to compound rainfall–melt extremes, high snowmelt and glacier melt, freeze–thaw-modulated snow-melt and precipitation, and late-season glacier melt.
Paul Voit and Maik Heistermann
Nat. Hazards Earth Syst. Sci., 24, 2147–2164, https://doi.org/10.5194/nhess-24-2147-2024, https://doi.org/10.5194/nhess-24-2147-2024, 2024
Short summary
Short summary
To identify flash flood potential in Germany, we shifted the most extreme rainfall events from the last 22 years systematically across Germany and simulated the consequent runoff reaction. Our results show that almost all areas in Germany have not seen the worst-case scenario of flood peaks within the last 22 years. With a slight spatial change of historical rainfall events, flood peaks of a factor of 2 or more would be achieved for most areas. The results can aid disaster risk management.
Maik Heistermann, Till Francke, Martin Schrön, and Sascha E. Oswald
Hydrol. Earth Syst. Sci., 28, 989–1000, https://doi.org/10.5194/hess-28-989-2024, https://doi.org/10.5194/hess-28-989-2024, 2024
Short summary
Short summary
Cosmic-ray neutron sensing (CRNS) is a non-invasive technique used to obtain estimates of soil water content (SWC) at a horizontal footprint of around 150 m and a vertical penetration depth of up to 30 cm. However, typical CRNS applications require the local calibration of a function which converts neutron counts to SWC. As an alternative, we propose a generalized function as a way to avoid the use of local reference measurements of SWC and hence a major source of uncertainty.
Lena Katharina Schmidt, Till Francke, Peter Martin Grosse, and Axel Bronstert
Hydrol. Earth Syst. Sci., 28, 139–161, https://doi.org/10.5194/hess-28-139-2024, https://doi.org/10.5194/hess-28-139-2024, 2024
Short summary
Short summary
How suspended sediment export from glacierized high-alpine areas responds to future climate change is hardly assessable as many interacting processes are involved, and appropriate physical models are lacking. We present the first study, to our knowledge, exploring machine learning to project sediment export until 2100 in two high-alpine catchments. We find that uncertainties due to methodological limitations are small until 2070. Negative trends imply that peak sediment may have already passed.
Gerd Bürger and Maik Heistermann
Nat. Hazards Earth Syst. Sci., 23, 3065–3077, https://doi.org/10.5194/nhess-23-3065-2023, https://doi.org/10.5194/nhess-23-3065-2023, 2023
Short summary
Short summary
Our subject is a new catalogue of radar-based heavy rainfall events (CatRaRE) over Germany and how it relates to the concurrent atmospheric circulation. We classify reanalyzed daily atmospheric fields of convective indices according to CatRaRE, using conventional statistical and more recent machine learning algorithms, and apply them to present and future atmospheres. Increasing trends are projected for CatRaRE-type probabilities, from reanalyzed as well as from simulated atmospheric fields.
Maik Heistermann, Till Francke, Lena Scheiffele, Katya Dimitrova Petrova, Christian Budach, Martin Schrön, Benjamin Trost, Daniel Rasche, Andreas Güntner, Veronika Döpper, Michael Förster, Markus Köhli, Lisa Angermann, Nikolaos Antonoglou, Manuela Zude-Sasse, and Sascha E. Oswald
Earth Syst. Sci. Data, 15, 3243–3262, https://doi.org/10.5194/essd-15-3243-2023, https://doi.org/10.5194/essd-15-3243-2023, 2023
Short summary
Short summary
Cosmic-ray neutron sensing (CRNS) allows for the non-invasive estimation of root-zone soil water content (SWC). The signal observed by a single CRNS sensor is influenced by the SWC in a radius of around 150 m (the footprint). Here, we have put together a cluster of eight CRNS sensors with overlapping footprints at an agricultural research site in north-east Germany. That way, we hope to represent spatial SWC heterogeneity instead of retrieving just one average SWC estimate from a single sensor.
Lena Katharina Schmidt, Till Francke, Peter Martin Grosse, Christoph Mayer, and Axel Bronstert
Hydrol. Earth Syst. Sci., 27, 1841–1863, https://doi.org/10.5194/hess-27-1841-2023, https://doi.org/10.5194/hess-27-1841-2023, 2023
Short summary
Short summary
We present a suitable method to reconstruct sediment export from decadal records of hydroclimatic predictors (discharge, precipitation, temperature) and shorter suspended sediment measurements. This lets us fill the knowledge gap on how sediment export from glacierized high-alpine areas has responded to climate change. We find positive trends in sediment export from the two investigated nested catchments with step-like increases around 1981 which are linked to crucial changes in glacier melt.
Katharina Lengfeld, Paul Voit, Frank Kaspar, and Maik Heistermann
Nat. Hazards Earth Syst. Sci., 23, 1227–1232, https://doi.org/10.5194/nhess-23-1227-2023, https://doi.org/10.5194/nhess-23-1227-2023, 2023
Short summary
Short summary
Estimating the severity of a rainfall event based on the damage caused is easy but highly depends on the affected region. A less biased measure for the extremeness of an event is its rarity combined with its spatial extent. In this brief communication, we investigate the sensitivity of such measures to the underlying dataset and highlight the importance of considering multiple spatial and temporal scales using the devastating rainfall event in July 2021 in central Europe as an example.
Alberto Caldas-Alvarez, Markus Augenstein, Georgy Ayzel, Klemens Barfus, Ribu Cherian, Lisa Dillenardt, Felix Fauer, Hendrik Feldmann, Maik Heistermann, Alexia Karwat, Frank Kaspar, Heidi Kreibich, Etor Emanuel Lucio-Eceiza, Edmund P. Meredith, Susanna Mohr, Deborah Niermann, Stephan Pfahl, Florian Ruff, Henning W. Rust, Lukas Schoppa, Thomas Schwitalla, Stella Steidl, Annegret H. Thieken, Jordis S. Tradowsky, Volker Wulfmeyer, and Johannes Quaas
Nat. Hazards Earth Syst. Sci., 22, 3701–3724, https://doi.org/10.5194/nhess-22-3701-2022, https://doi.org/10.5194/nhess-22-3701-2022, 2022
Short summary
Short summary
In a warming climate, extreme precipitation events are becoming more frequent. To advance our knowledge on such phenomena, we present a multidisciplinary analysis of a selected case study that took place on 29 June 2017 in the Berlin metropolitan area. Our analysis provides evidence of the extremeness of the case from the atmospheric and the impacts perspectives as well as new insights on the physical mechanisms of the event at the meteorological and climate scales.
Paul Voit and Maik Heistermann
Nat. Hazards Earth Syst. Sci., 22, 2791–2805, https://doi.org/10.5194/nhess-22-2791-2022, https://doi.org/10.5194/nhess-22-2791-2022, 2022
Short summary
Short summary
To better understand how the frequency and intensity of heavy precipitation events (HPEs) will change with changing climate and to adapt disaster risk management accordingly, we have to quantify the extremeness of HPEs in a reliable way. We introduce the xWEI (cross-scale WEI) and show that this index can reveal important characteristics of HPEs that would otherwise remain hidden. We conclude that the xWEI could be a valuable instrument in both disaster risk management and research.
Lena Katharina Schmidt, Till Francke, Erwin Rottler, Theresa Blume, Johannes Schöber, and Axel Bronstert
Earth Surf. Dynam., 10, 653–669, https://doi.org/10.5194/esurf-10-653-2022, https://doi.org/10.5194/esurf-10-653-2022, 2022
Short summary
Short summary
Climate change fundamentally alters glaciated high-alpine areas, but it is unclear how this affects riverine sediment transport. As a first step, we aimed to identify the most important processes and source areas in three nested catchments in the Ötztal, Austria, in the past 15 years. We found that areas above 2500 m were crucial and that summer rainstorms were less influential than glacier melt. These findings provide a baseline for studies on future changes in high-alpine sediment dynamics.
Maik Heistermann, Heye Bogena, Till Francke, Andreas Güntner, Jannis Jakobi, Daniel Rasche, Martin Schrön, Veronika Döpper, Benjamin Fersch, Jannis Groh, Amol Patil, Thomas Pütz, Marvin Reich, Steffen Zacharias, Carmen Zengerle, and Sascha Oswald
Earth Syst. Sci. Data, 14, 2501–2519, https://doi.org/10.5194/essd-14-2501-2022, https://doi.org/10.5194/essd-14-2501-2022, 2022
Short summary
Short summary
This paper presents a dense network of cosmic-ray neutron sensing (CRNS) to measure spatio-temporal soil moisture patterns during a 2-month campaign in the Wüstebach headwater catchment in Germany. Stationary, mobile, and airborne CRNS technology monitored the root-zone water dynamics as well as spatial heterogeneity in the 0.4 km2 area. The 15 CRNS stations were supported by a hydrogravimeter, biomass sampling, and a wireless soil sensor network to facilitate holistic hydrological analysis.
Till Francke, Maik Heistermann, Markus Köhli, Christian Budach, Martin Schrön, and Sascha E. Oswald
Geosci. Instrum. Method. Data Syst., 11, 75–92, https://doi.org/10.5194/gi-11-75-2022, https://doi.org/10.5194/gi-11-75-2022, 2022
Short summary
Short summary
Cosmic-ray neutron sensing (CRNS) is a non-invasive tool for measuring hydrogen pools like soil moisture, snow, or vegetation. This study presents a directional shielding approach, aiming to measure in specific directions only. The results show that non-directional neutron transport blurs the signal of the targeted direction. For typical instruments, this does not allow acceptable precision at a daily time resolution. However, the mere statistical distinction of two rates is feasible.
Maik Heistermann, Till Francke, Martin Schrön, and Sascha E. Oswald
Hydrol. Earth Syst. Sci., 25, 4807–4824, https://doi.org/10.5194/hess-25-4807-2021, https://doi.org/10.5194/hess-25-4807-2021, 2021
Short summary
Short summary
Cosmic-ray neutron sensing (CRNS) is a powerful technique for retrieving representative estimates of soil moisture in footprints extending over hectometres in the horizontal and decimetres in the vertical. This study, however, demonstrates the potential of CRNS to obtain spatio-temporal patterns of soil moisture beyond isolated footprints. To that end, we analyse data from a unique observational campaign that featured a dense network of more than 20 neutron detectors in an area of just 1 km2.
Cited articles
Adiba, A., Hajji, H., and Maatouk, M.: Transfer learning and U-Net for buildings segmentation, in: Proceedings of the New Challenges in Data Sciences: Acts of the Second Conference of the Moroccan Classification Society, Kenitra Morocco, 28–29 March 2019, Association for Computing Machinery, New York, NY, United States, 1–6, ISBN 978-1-4503-6129-3, 2019. a
ATKIS: Digitale Geländemodelle – ATKIS DGM, Senatsverwaltung für Stadtentwicklung, Bauen und Wohnen, http://www.stadtentwicklung.berlin.de/geoinformation/landesvermessung/atkis/de/dgm.shtml (last access: 22 February 2022), 2020. a
Berghäuser, L., Schoppa, L., Ulrich, J., Dillenardt, L., Jurado, O. E.,
Passow, C., Mohor, G. S., Seleem, O., Petrow, T., and Thieken, A. H.:
Starkregen in Berlin: Meteorologische Ereignisrekonstruktion und
Betroffenenbefragung, task force report, University of Potsdam, 44 pp., https://doi.org/10.25932/publishup-50056, 2021. a, b, c
Biau, G. and Scornet, E.: A random forest guided tour, Test, 25, 197–227,
2016. a
Breiman, L.: Random forests, Mach. Learn., 45, 5–32, 2001. a
Chen, W., Li, Y., Xue, W., Shahabi, H., Li, S., Hong, H., Wang, X., Bian, H., Zhang, S., Pradhan, B., and Ahmad, B. B.: Modeling flood susceptibility using data-driven approaches of naïve bayes tree, alternating decision tree, and random forest methods, Sci. Total Environ., 701, 134979, https://doi.org/10.1016/j.scitotenv.2019.134979, 2020. a
Chollet, F. et al.: Keras, https://keras.io (last access: 22 February 2023), 2015. a
Costabile, P., Costanzo, C., and Macchione, F.: Performances and limitations of the diffusive approximation of the 2-d shallow water equations for flood simulation in urban and rural areas, Appl. Numer. Math., 116, 141–156, 2017. a
Galland, J.-C., Goutal, N., and Hervouet, J.-M.: TELEMAC: A new numerical model for solving shallow water equations, Adv. Water Resour., 14,
138–148, 1991. a
Grant, T. D. and Wischik, D. J.: Finding Patterns as the Path from Input to
Output, in: On the path to AI, Springer, 41–48, https://doi.org/10.1007/978-3-030-43582-0_4, 2020. a
Grinsztajn, L., Oyallon, E., and Varoquaux, G.: Why do tree-based models still outperform deep learning on tabular data?, arXiv [preprint], https://doi.org/10.48550/arXiv.2207.08815, 18 July 2022. a
Guo, Z., Leitão, J. P., Simões, N. E., and Moosavi, V.: Data-driven
flood emulation: Speeding up urban flood predictions by deep convolutional
neural networks, J. Flood Risk Manag., 14, e12684, https://doi.org/10.1111/jfr3.12684, 2021. a, b, c
Hofmann, J. and Schüttrumpf, H.: floodGAN: Using Deep Adversarial Learning to Predict Pluvial Flooding in Real Time, Water, 13, 2255, https://doi.org/10.3390/w13162255, 2021. a
Ioffe, S. and Szegedy, C.: Batch normalization: Accelerating deep network
training by reducing internal covariate shift, in: International conference
on machine learning, Lille, France, 6–11 July 2015, PMLR, 448–456, arXiv [preprint], https://doi.org/10.48550/arXiv.1502.03167, 11 February 2015. a
Kingma, D. P. and Ba, J.: Adam: A method for stochastic optimization, arXiv [preprint], https://doi.org/10.48550/arXiv.1412.6980, 22 December 2014. a
Kirkby, M.: Hydrograph modeling strategies, in: Process in physical and human geography, Heinemann Educational, 69–90, ISBN 978-0435356255, 1975. a
Lee, S., Kim, J.-C., Jung, H.-S., Lee, M. J., and Lee, S.: Spatial prediction of flood susceptibility using random-forest and boosted-tree models in Seoul metropolitan city, Korea, Geomatics, Natural Hazards and Risk, 8, 1185–1203,
2017. a
Lei, X., Chen, W., Panahi, M., Falah, F., Rahmati, O., Uuemaa, E., Kalantari, Z., Ferreira, C. S. S., Rezaie, F., Tiefenbacher, J. P., and Lee S.: Urban flood modeling using deep-learning approaches in Seoul, South Korea, J. Hydrol., 601, 126684, https://doi.org/10.1016/j.jhydrol.2021.126684, 2021. a
Löwe, R., Böhm, J., Jensen, D. G., Leandro, J., and Rasmussen, S. H.:
U-FLOOD–topographic deep learning for predicting urban pluvial flood water
depth, J. Hydrol., 603, 126898, https://doi.org/10.1016/j.jhydrol.2021.126898, 2021. a, b, c, d, e, f, g, h, i, j, k, l, m, n, o, p, q, r, s, t, u, v, w, x, y
Maas, A. L., Hannun, A. Y., and Ng, A. Y.: Rectifier nonlinearities improve neural network acoustic models, in: Proceedings of the 30th International Conference on Machine Learning, Atlanta, Georgia, USA, 16–21 June 2013, p. 3, Citeseer, https://ai.stanford.edu/~amaas/papers/relu_hybrid_icml2013_final.pdf (last access: 22 February 2023), 2013. a
Miikkulainen, R., Liang, J., Meyerson, E., Rawal, A., Fink, D., Francon, O., Raju, B., Shahrzad, H., Navruzyan, A., Duffy, N., and Hodjat B.: Evolving deep neural networks, in: Artificial intelligence in the age of neural networks and brain computing, Elsevier, 293–312, https://doi.org/10.1016/B978-0-12-815480-9.00015-3, 2019. a
Nash, J. E. and Sutcliffe, J. V.: River flow forecasting through conceptual
models part I – A discussion of principles, J. Hydrol., 10,
282–290, 1970. a
Peel, M. C., Finlayson, B. L., and McMahon, T. A.: Updated world map of the Köppen-Geiger climate classification, Hydrol. Earth Syst. Sci., 11, 1633–1644, https://doi.org/10.5194/hess-11-1633-2007, 2007. a
Prechelt, L.: Early stopping – but when?, in: Neural Networks: Tricks of the
trade, Springer, 55–69, https://doi.org/10.1007/978-3-642-35289-8_5, 1998. a
Probst, P., Wright, M. N., and Boulesteix, A.-L.: Hyperparameters and tuning
strategies for random forest, WIRES Data Min. Knowl., 9, e1301, https://doi.org/10.1002/widm.1301, 2019. a
Rahmati, O., Pourghasemi, H. R., and Zeinivand, H.: Flood susceptibility mapping using frequency ratio and weights-of-evidence models in the Golastan
Province, Iran, Geocarto Int., 31, 42–70, 2016. a
Ronneberger, O., Fischer, P., and Brox, T.: U-net: Convolutional networks for biomedical image segmentation, in: 18th International Conference, Munich, Germany, 5–9 October 2015, Proceedings, Part III, Springer, 234–241, https://doi.org/10.1007/978-3-319-24574-4_28, 2015. a, b
Santurkar, S., Tsipras, D., Ilyas, A., and Madry, A.: How does batch normalization help optimization?, arXiv [preprint], https://doi.org/10.48550/arXiv.1805.11604, 29 May 2018. a
Seleem, O.: Transferability of data-driven models to predict urban pluvial flood water depth in Berlin, Germany, Zenodo [data set], https://doi.org/10.5281/zenodo.7516408, 2023a. a
Seleem, O.: omarseleem92/Urban_flooding: Python script and handouts, Zenodo [code], https://doi.org/10.5281/zenodo.7661174, 2023b. a
Skougaard Kaspersen, P., Høegh Ravn, N., Arnbjerg-Nielsen, K., Madsen, H., and Drews, M.: Comparison of the impacts of urban development and climate change on exposing European cities to pluvial flooding, Hydrol. Earth Syst. Sci., 21, 4131–4147, https://doi.org/10.5194/hess-21-4131-2017, 2017. a
Srivastava, R. K., Greff, K., and Schmidhuber, J.: Training very deep networks, arXiv [preprint], https://doi.org/10.48550/arXiv.1507.06228, 22 July 2015. a
Tan, C., Sun, F., Kong, T., Zhang, W., Yang, C., and Liu, C.: A survey on deep transfer learning, in: International conference on artificial neural networks, 5–7 October 2018, Rhodes, Greece, Springer, 270–279, https://doi.org/10.1007/978-3-030-01424-7_27, 2018. a, b
van Dijk, E., van der Meulen, J., Kluck, J., and Straatman, J.: Comparing modelling techniques for analysing urban pluvial flooding, Water Sci. Technol., 69, 305–311, 2014. a
Weiss, A.: Topographic position and landforms analysis, in: Poster
presentation, ESRI user conference, San Diego, CA, USA, 9–13 July 2001, poster presnetation, 2001. a
Zahura, F. T., Goodall, J. L., Sadler, J. M., Shen, Y., Morsy, M. M., and Behl, M.: Training machine learning surrogate models from a high-fidelity physics-based model: Application for real-time street-scale flood prediction
in an urban coastal community, Water Resour. Res., 56, e2019WR027038, https://doi.org/10.1029/2019WR027038, 2020. a, b
Zhao, G., Pang, B., Xu, Z., Peng, D., and Zuo, D.: Urban flood susceptibility assessment based on convolutional neural networks, J. Hydrol., 590,
125235, https://doi.org/10.1016/j.jhydrol.2020.125235, 2020. a, b, c
Zhao, G., Pang, B., Xu, Z., Cui, L., Wang, J., Zuo, D., and Peng, D.: Improving urban flood susceptibility mapping using transfer learning, J. Hydrol., 602, 126777, https://doi.org/10.1016/j.jhydrol.2021.126777, 2021. a, b, c
Short summary
Data-driven models are becoming more of a surrogate that overcomes the limitations of the computationally expensive 2D hydrodynamic models to map urban flood hazards. However, the model's ability to generalize outside the training domain is still a major challenge. We evaluate the performance of random forest and convolutional neural networks to predict urban floodwater depth and investigate their transferability outside the training domain.
Data-driven models are becoming more of a surrogate that overcomes the limitations of the...
Altmetrics
Final-revised paper
Preprint