Refine
Has Fulltext
- yes (2)
Document Type
- Doctoral Thesis (2) (remove)
Language
- English (2)
Is part of the Bibliography
- yes (2)
Keywords
- flood risk (2) (remove)
Institute
- Institut für Geowissenschaften (2) (remove)
Technological progress allows for producing ever more complex predictive models on the basis of increasingly big datasets. For risk management of natural hazards, a multitude of models is needed as basis for decision-making, e.g. in the evaluation of observational data, for the prediction of hazard scenarios, or for statistical estimates of expected damage. The question arises, how modern modelling approaches like machine learning or data-mining can be meaningfully deployed in this thematic field. In addition, with respect to data availability and accessibility, the trend is towards open data. Topic of this thesis is therefore to investigate the possibilities and limitations of machine learning and open geospatial data in the field of flood risk modelling in the broad sense. As this overarching topic is broad in scope, individual relevant aspects are identified and inspected in detail.
A prominent data source in the flood context is satellite-based mapping of inundated areas, for example made openly available by the Copernicus service of the European Union. Great expectations are directed towards these products in scientific literature, both for acute support of relief forces during emergency response action, and for modelling via hydrodynamic models or for damage estimation. Therefore, a focus of this work was set on evaluating these flood masks. From the observation that the quality of these products is insufficient in forested and built-up areas, a procedure for subsequent improvement via machine learning was developed. This procedure is based on a classification algorithm that only requires training data from a particular class to be predicted, in this specific case data of flooded areas, but not of the negative class (dry areas). The application for hurricane Harvey in Houston shows the high potential of this method, which depends on the quality of the initial flood mask.
Next, it is investigated how much the predicted statistical risk from a process-based model chain is dependent on implemented physical process details. Thereby it is demonstrated what a risk study based on established models can deliver. Even for fluvial flooding, such model chains are already quite complex, though, and are hardly available for compound or cascading events comprising torrential rainfall, flash floods, and other processes. In the fourth chapter of this thesis it is therefore tested whether machine learning based on comprehensive damage data can offer a more direct path towards damage modelling, that avoids explicit conception of such a model chain. For that purpose, a state-collected dataset of damaged buildings from the severe El Niño event 2017 in Peru is used. In this context, the possibilities of data-mining for extracting process knowledge are explored as well. It can be shown that various openly available geodata sources contain useful information for flood hazard and damage modelling for complex events, e.g. satellite-based rainfall measurements, topographic and hydrographic information, mapped settlement areas, as well as indicators from spectral data. Further, insights on damaging processes are discovered, which mainly are in line with prior expectations. The maximum intensity of rainfall, for example, acts stronger in cities and steep canyons, while the sum of rain was found more informative in low-lying river catchments and forested areas. Rural areas of Peru exhibited higher vulnerability in the presented study compared to urban areas. However, the general limitations of the methods and the dependence on specific datasets and algorithms also become obvious.
In the overarching discussion, the different methods – process-based modelling, predictive machine learning, and data-mining – are evaluated with respect to the overall research questions. In the case of hazard observation it seems that a focus on novel algorithms makes sense for future research. In the subtopic of hazard modelling, especially for river floods, the improvement of physical models and the integration of process-based and statistical procedures is suggested. For damage modelling the large and representative datasets necessary for the broad application of machine learning are still lacking. Therefore, the improvement of the data basis in the field of damage is currently regarded as more important than the selection of algorithms.
In the past, floods were basically managed by flood control mechanisms. The focus was set on the reduction of flood hazard. The potential consequences were of minor interest. Nowadays river flooding is increasingly seen from the risk perspective, including possible consequences. Moreover, the large-scale picture of flood risk became increasingly important for disaster management planning, national risk developments and the (re-) insurance industry. Therefore, it is widely accepted that risk-orientated flood management ap-proaches at the basin-scale are needed. However, large-scale flood risk assessment methods for areas of several 10,000 km² are still in early stages. Traditional flood risk assessments are performed reach wise, assuming constant probabilities for the entire reach or basin. This might be helpful on a local basis, but where large-scale patterns are important this approach is of limited use. Assuming a T-year flood (e.g. 100 years) for the entire river network is unrealistic and would lead to an overestimation of flood risk at the large scale. Due to the lack of damage data, additionally, the probability of peak discharge or rainfall is usually used as proxy for damage probability to derive flood risk. With a continuous and long term simulation of the entire flood risk chain, the spatial variability of probabilities could be consider and flood risk could be directly derived from damage data in a consistent way.
The objective of this study is the development and application of a full flood risk chain, appropriate for the large scale and based on long term and continuous simulation. The novel approach of ‘derived flood risk based on continuous simulations’ is introduced, where the synthetic discharge time series is used as input into flood impact models and flood risk is directly derived from the resulting synthetic damage time series.
The bottleneck at this scale is the hydrodynamic simu-lation. To find suitable hydrodynamic approaches for the large-scale a benchmark study with simplified 2D hydrodynamic models was performed. A raster-based approach with inertia formulation and a relatively high resolution of 100 m in combination with a fast 1D channel routing model was chosen.
To investigate the suitability of the continuous simulation of a full flood risk chain for the large scale, all model parts were integrated into a new framework, the Regional Flood Model (RFM). RFM consists of the hydrological model SWIM, a 1D hydrodynamic river network model, a 2D raster based inundation model and the flood loss model FELMOps+r. Subsequently, the model chain was applied to the Elbe catchment, one of the largest catchments in Germany. For the proof-of-concept, a continuous simulation was per-formed for the period of 1990-2003. Results were evaluated / validated as far as possible with available observed data in this period. Although each model part introduced its own uncertainties, results and runtime were generally found to be adequate for the purpose of continuous simulation at the large catchment scale.
Finally, RFM was applied to a meso-scale catchment in the east of Germany to firstly perform a flood risk assessment with the novel approach of ‘derived flood risk assessment based on continuous simulations’. Therefore, RFM was driven by long term synthetic meteorological input data generated by a weather generator. Thereby, a virtual time series of climate data of 100 x 100 years was generated and served as input to RFM providing subsequent 100 x 100 years of spatially consistent river discharge series, inundation patterns and damage values. On this basis, flood risk curves and expected annual damage could be derived directly from damage data, providing a large-scale picture of flood risk. In contrast to traditional flood risk analysis, where homogenous return periods are assumed for the entire basin, the presented approach provides a coherent large-scale picture of flood risk. The spatial variability of occurrence probability is respected. Additionally, data and methods are consistent. Catchment and floodplain processes are repre-sented in a holistic way. Antecedent catchment conditions are implicitly taken into account, as well as physical processes like storage effects, flood attenuation or channel–floodplain interactions and related damage influencing effects. Finally, the simulation of a virtual period of 100 x 100 years and consequently large data set on flood loss events enabled the calculation of flood risk directly from damage distributions. Problems associated with the transfer of probabilities in rainfall or peak runoff to probabilities in damage, as often used in traditional approaches, are bypassed.
RFM and the ‘derived flood risk approach based on continuous simulations’ has the potential to provide flood risk statements for national planning, re-insurance aspects or other questions where spatially consistent, large-scale assessments are required.