@phdthesis{Seleem2023,
  author    = {Seleem, Omar},
  title     = {Towards urban pluvial flood mapping using data-driven models},
  doi       = {10.25932/publishup-59813},
  url       = {http://nbn-resolving.de/urn:nbn:de:kobv:517-opus4-598137},
  school      = {Universit{\"a}t Potsdam},
  pages     = {xv, 95},
  year      = {2023},
  abstract  = {Casualties and damages from urban pluvial flooding are increasing. Triggered by short, localized, and intensive rainfall events, urban pluvial floods can occur anywhere, even in areas without a history of flooding. Urban pluvial floods have relatively small temporal and spatial scales. Although cumulative losses from urban pluvial floods are comparable, most flood risk management and mitigation strategies focus on fluvial and coastal flooding. Numerical-physical-hydrodynamic models are considered the best tool to represent the complex nature of urban pluvial floods; however, they are computationally expensive and time-consuming. These sophisticated models make large-scale analysis and operational forecasting prohibitive. Therefore, it is crucial to evaluate and benchmark the performance of other alternative methods. The findings of this cumulative thesis are represented in three research articles. The first study evaluates two topographic-based methods to map urban pluvial flooding, fill-spill-merge (FSM) and topographic wetness index (TWI), by comparing them against a sophisticated hydrodynamic model. The FSM method identifies flood-prone areas within topographic depressions while the TWI method employs maximum likelihood estimation to calibrate a TWI threshold (τ) based on inundation maps from the 2D hydrodynamic model. The results point out that the FSM method outperforms the TWI method. The study highlights then the advantage and limitations of both methods. Data-driven models provide a promising alternative to computationally expensive hydrodynamic models. However, the literature lacks benchmarking studies to evaluate the different models' performance, advantages and limitations. Model transferability in space is a crucial problem. Most studies focus on river flooding, likely due to the relative availability of flow and rain gauge records for training and validation. Furthermore, they consider these models as black boxes. The second study uses a flood inventory for the city of Berlin and 11 predictive features which potentially indicate an increased pluvial flooding hazard to map urban pluvial flood susceptibility using a convolutional neural network (CNN), an artificial neural network (ANN) and the benchmarking machine learning models random forest (RF) and support vector machine (SVM). I investigate the influence of spatial resolution on the implemented models, the models' transferability in space and the importance of the predictive features. The results show that all models perform well and the RF models are superior to the other models within and outside the training domain. The models developed using fine spatial resolution (2 and 5 m) could better identify flood-prone areas. Finally, the results point out that aspect is the most important predictive feature for the CNN models, and altitude is for the other models. While flood susceptibility maps identify flood-prone areas, they do not represent flood variables such as velocity and depth which are necessary for effective flood risk management. To address this, the third study investigates data-driven models' transferability to predict urban pluvial floodwater depth and the models' ability to enhance their predictions using transfer learning techniques. It compares the performance of RF (the best-performing model in the previous study) and CNN models using 12 predictive features and output from a hydrodynamic model. The findings in the third study suggest that while CNN models tend to generalise and smooth the target function on the training dataset, RF models suffer from overfitting. Hence, RF models are superior for predictions inside the training domains but fail outside them while CNN models could control the relative loss in performance outside the training domains. Finally, the CNN models benefit more from transfer learning techniques than RF models, boosting their performance outside training domains. In conclusion, this thesis has evaluated both topographic-based methods and data-driven models to map urban pluvial flooding. However, further studies are crucial to have methods that completely overcome the limitation of 2D hydrodynamic models.},
  language  = {en}
}
@phdthesis{Marc2016,
  author    = {Marc, Odin},
  title     = {Earthquake-induced landsliding},
  url       = {http://nbn-resolving.de/urn:nbn:de:kobv:517-opus4-96808},
  school      = {Universit{\"a}t Potsdam},
  pages     = {xvi, 171},
  year      = {2016},
  abstract  = {Earthquakes deform Earth's surface, building long-lasting topographic features and contributing to landscape and mountain formation. However, seismic waves produced by earthquakes may also destabilize hillslopes, leading to large amounts of soil and bedrock moving downslope. Moreover, static deformation and shaking are suspected to damage the surface bedrock and therefore alter its future properties, affecting hydrological and erosional dynamics. Thus, earthquakes participate both in mountain building and stimulate directly or indirectly their erosion. Moreover, the impact of earthquakes on hillslopes has important implications for the amount of sediment and organic matter delivered to rivers, and ultimately to oceans, during episodic catastrophic seismic crises, the magnitude of life and property losses associated with landsliding, the perturbation and recovery of landscape properties after shaking, and the long term topographic evolution of mountain belts. Several of these aspects have been addressed recently through individual case studies but additional data compilation as well as theoretical or numerical modelling are required to tackle these issues in a more systematic and rigorous manner. This dissertation combines data compilation of earthquake characteristics, landslide mapping, and seismological data interpretation with physically-based modeling in order to address how earthquakes impact on erosional processes and landscape evolution. Over short time scales (10-100 s) and intermediate length scales (10 km), I have attempted to improve our understanding and ability to predict the amount of landslide debris triggered by seismic shaking in epicentral areas. Over long time scales (1-100 ky) and across a mountain belt (100 km) I have modeled the competition between erosional unloading and building of topography associated with earthquakes. Finally, over intermediate time scales (1-10 y) and at the hillslope scale (0.1-1 km) I have collected geomorphological and seismological data that highlight persistent effects of earthquakes on landscape properties and behaviour. First, I compiled a database on earthquakes that produced significant landsliding, including an estimate of the total landslide volume and area, and earthquake characteristics such as seismic moment and source depth. A key issue is the accurate conversion of landslide maps into volume estimates. Therefore I also estimated how amalgamation - when mapping errors lead to the bundling of multiple landslide into a single polygon - affects volume estimates from various earthquake-induced landslide inventories and developed an algorithm to automatically detect this artifact. The database was used to test a physically-based prediction of the total landslide area and volume caused by earthquakes, based on seismological scaling relationships and a statistical description of the landscape properties. The model outperforms empirical fits in accuracy, with 25 out of 40 cases well predicted, and allows interpretation of many outliers in physical terms. Apart from seismological complexities neglected by the model I found that exceptional rock strength properties or antecedent conditions may explain most outliers. Second, I assessed the geomorphic effects of large earthquakes on landscape dynamics by surveying the temporal evolution of precipitation-normalized landslide rate. I found strongly elevated landslide rates following earthquakes that progressively recover over 1 to 4 years, indicating that regolith strength drops and recovers. The relaxation is clearly non-linear for at least one case, and does not seem to correlate with coseismic landslide reactivation, water table level increase or tree root-system recovery. I suggested that shallow bedrock is damaged by the earthquake and then heals on annual timescales. Such variations in ground strength must be translated into shallow subsurface seismic velocities that are increasingly surveyed with ambient seismic noise correlations. With seismic noise autocorrelation I computed the seismic velocity in the epicentral areas of three earthquakes where I constrained a change in landslide rate. We found similar recovery dynamics and timescales, suggesting that seismic noise correlation techniques could be further developed to meaningfully assess ground strength variations for landscape dynamics. These two measurements are also in good agreement with the temporal dynamics of post-seismic surface displacement measured by GPS. This correlation suggests that the surface healing mechanism may be driven by tectonic deformation, and that the surface regolith and fractured bedrock may behave as a granular media that slowly compacts as it is sheared or vibrated. Last, I compared our model of earthquake-induced landsliding with a standard formulation of surface deformation caused by earthquakes to understand which parameters govern the competition between the building and destruction of topography caused by earthquakes. In contrast with previous studies I found that very large (Mw>8) earthquakes always increase the average topography, whereas only intermediate (Mw ~ 7) earthquakes in steep landscapes may reduce topography. Moreover, I illustrated how the net effect of earthquakes varies with depth or landscape steepness implying a complex and ambivalent role through the life of a mountain belt. Further I showed that faults producing a Gutenberg-Richter distribution of earthquake sizes, will limit topography over a larger range of fault sizes than faults producing repeated earthquakes with a characteristic size.},
  language  = {en}
}