@phdthesis{SamprognaMohor2022,
  author    = {Samprogna Mohor, Guilherme},
  title     = {Exploring the transferability of flood loss models across flood types},
  doi       = {10.25932/publishup-55714},
  url       = {http://nbn-resolving.de/urn:nbn:de:kobv:517-opus4-557141},
  school      = {Universit{\"a}t Potsdam},
  pages     = {XXIV, 182},
  year      = {2022},
  abstract  = {The estimation of financial losses is an integral part of flood risk assessment. The application of existing flood loss models on locations or events different from the ones used to train the models has led to low performance, showing that characteristics of the flood damaging process have not been sufficiently well represented yet. To improve flood loss model transferability, I explore various model structures aiming at incorporating different (inland water) flood types and pathways. That is based on a large survey dataset of approximately 6000 flood-affected households which addresses several aspects of the flood event, not only the hazard characteristics but also information on the affected building, socioeconomic factors, the household's preparedness level, early warning, and impacts. Moreover, the dataset reports the coincidence of different flood pathways. Whilst flood types are a classification of flood events reflecting their generating process (e.g. fluvial, pluvial), flood pathways represent the route the water takes to reach the receptors (e.g. buildings). In this work, the following flood pathways are considered: levee breaches, river floods, surface water floods, and groundwater floods. The coincidence of several hazard processes at the same time and place characterises a compound event. In fact, many flood events develop through several pathways, such as the ones addressed in the survey dataset used. Earlier loss models, although developed with one or multiple predictor variables, commonly use loss data from a single flood event which is attributed to a single flood type, disregarding specific flood pathways or the coincidence of multiple pathways. This gap is addressed by this thesis through the following research questions: 1. In which aspects do flood pathways of the same (compound inland) flood event differ? 2. How much do factors which contribute to the overall flood loss in a building differ in various settings, specifically across different flood pathways? 3. How well can Bayesian loss models learn from different settings? 4. Do compound, that is, coinciding flood pathways result in higher losses than a single pathway, and what does the outcome imply for future loss modelling? Statistical analysis has found that households affected by different flood pathways also show, in general, differing characteristics of the affected building, preparedness, and early warning, besides the hazard characteristics. Forecasting and early warning capabilities and the preparedness of the population are dominated by the general flood type, but characteristics of the hazard at the object-level, the impacts, and the recovery are more related to specific flood pathways, indicating that risk communication and loss models could benefit from the inclusion of flood-pathway-specific information. For the development of the loss model, several potentially relevant predictors are analysed: water depth, duration, velocity, contamination, early warning lead time, perceived knowledge about self-protection, warning information, warning source, gap between warning and action, emergency measures, implementation of property-level precautionary measures (PLPMs), perceived efficacy of PLPMs, previous flood experience, awareness of flood risk, ownership, building type, number of flats, building quality, building value, house/flat area, building area, cellar, age, household size, number of children, number of elderly residents, income class, socioeconomic status, and insurance against floods. After a variable selection, descriptors of the hazard, building, and preparedness were deemed significant, namely: water depth, contamination, duration, velocity, building area, building quality, cellar, PLPMs, perceived efficacy of PLPMs, emergency measures, insurance, and previous flood experience. The inclusion of the indicators of preparedness is relevant, as they are rarely involved in loss datasets and in loss modelling, although previous studies have shown their potential in reducing losses. In addition, the linear model fit indicates that the explanatory factors are, in several cases, differently relevant across flood pathways. Next, Bayesian multilevel models were trained, which intrinsically incorporate uncertainties and allow for partial pooling (i.e. different groups of data, such as households affected by different flood pathways, can learn from each other), increasing the statistical power of the model. A new variable selection was performed for this new model approach, reducing the number of predictors from twelve to seven variables but keeping factors of the hazard, building, and preparedness, namely: water depth, contamination, duration, building area, PLPMs, insurance, and previous flood experience. The new model was trained not only across flood pathways but also across regions of Germany, divided according to general socioeconomic factors and insurance policies, and across flood events. The distinction across regions and flood events did not improve loss modelling and led to a large overlap of regression coefficients, with no clear trend or pattern. The distinction of flood pathways showed credibly distinct regression coefficients, leading to a better understanding of flood loss modelling and indicating one potential reason why model transferability has been challenging. Finally, new model structures were trained to include the possibility of compound inland floods (i.e. when multiple flood pathways coincide on the same affected asset). The dataset does not allow for verifying in which sequence the flood pathway waves occurred and predictor variables reflect only their mixed or combined outcome. Thus, two Bayesian models were trained: 1. a multi-membership model, a structure which learns the regression coefficients for multiple flood pathways at the same time, and 2. a multilevel model wherein the combination of coinciding flood pathways makes individual categories. The multi-membership model resulted in credibly different coefficients across flood pathways but did not improve model performance in comparison to the model assuming only a single dominant flood pathway. The model with combined categories signals an increase in impacts after compound floods, but due to the uncertainty in model coefficients and estimates, it is not possible to ascertain such an increase as credible. That is, with the current level of uncertainty in differentiating the flood pathways, the loss estimates are not credibly distinct from individual flood pathways. To overcome the challenges faced, non-linear or mixed models could be explored in the future. Interactions, moderation, and mediation effects, as well as non-linear effects, should also be further studied. Loss data collection should regularly include preparedness indicators, and either data collection or hydraulic modelling should focus on the distinction of coinciding flood pathways, which could inform loss models and further improve estimates. Flood pathways show distinct (financial) impacts, and their inclusion in loss modelling proves relevant, for it helps in clarifying the different contribution of influencing factors to the final loss, improving understanding of the damaging process, and indicating future lines of research.},
  language  = {en}
}
@phdthesis{Bryant2024,
  author    = {Bryant, Seth},
  title     = {Aggregation and disaggregation in flood risk models},
  doi       = {10.25932/publishup-65095},
  url       = {http://nbn-resolving.de/urn:nbn:de:kobv:517-opus4-650952},
  school      = {Universit{\"a}t Potsdam},
  pages     = {ix, 116},
  year      = {2024},
  abstract  = {Floods continue to be the leading cause of economic damages and fatalities among natural disasters worldwide. As future climate and exposure changes are projected to intensify these damages, the need for more accurate and scalable flood risk models is rising. Over the past decade, macro-scale flood risk models have evolved from initial proof-of-concepts to indispensable tools for decision-making at global-, nationaland, increasingly, the local-level. This progress has been propelled by the advent of high-performance computing and the availability of global, space-based datasets. However, despite such advancements, these models are rarely validated and consistently fall short of the accuracy achieved by high-resolution local models. While capabilities have improved, significant gaps persist in understanding the behaviours of such macro-scale models, particularly their tendency to overestimate risk. This dissertation aims to address such gaps by examining the scale transfers inherent in the construction and application of coarse macroscale models. To achieve this, four studies are presented that, collectively, address exposure, hazard, and vulnerability components of risk affected by upscaling or downscaling. The first study focuses on a type of downscaling where coarse flood hazard inundation grids are enhanced to a finer resolution. While such inundation downscaling has been employed in numerous global model chains, ours is the first study to focus specifically on this component, providing an evaluation of the state of the art and a novel algorithm. Findings demonstrate that our novel algorithm is eight times faster than existing methods, offers a slight improvement in accuracy, and generates more physically coherent flood maps in hydraulically challenging regions. When applied to a case study, the algorithm generated a 4m resolution inundation map from 30m hydrodynamic model outputs in 33 s, a 60-fold improvement in runtime with a 25\% increase in RMSE compared with direct hydrodynamic modelling. All evaluated downscaling algorithms yielded better accuracy than the coarse hydrodynamic model when compared to observations, demonstrating similar limits of coarse hydrodynamic models reported by others. The substitution of downscaling into flood risk model chains, in place of high-resolution modelling, can drastically improve the lead time of impactbased forecasts and the efficiency of hazard map production. With downscaling, local regions could obtain high resolution local inundation maps by post-processing a global model without the need for expensive modelling or expertise. The second study focuses on hazard aggregation and its implications for exposure, investigating implicit aggregations commonly used to intersect hazard grids with coarse exposure models. This research introduces a novel spatial classification framework to understand the effects of rescaling flood hazard grids to a coarser resolution. The study derives closed-form analytical solutions for the location and direction of bias from flood grid aggregation, showing that bias will always be present in regions near the edge of inundation. For example, inundation area will be positively biased when water depth grids are aggregated, while volume will be negatively biased when water elevation grids are aggregated. Extending the analysis to effects of hazard aggregation on building exposure, this study shows that exposure in regions at the edge of inundation are an order of magnitude more sensitive to aggregation errors than hazard alone. Among the two aggregation routines considered, averaging water surface elevation grids better preserved flood depths at buildings than averaging of water depth grids. The study provides the first mathematical proof and generalizeable treatment of flood hazard grid aggregation, demonstrating important mechanisms to help flood risk modellers understand and control model behaviour. The final two studies focus on the aggregation of vulnerability models or flood damage functions, investigating the practice of applying per-asset functions to aggregate exposure models. Both studies extend Jensen's inequality, a well-known 1906 mathematical proof, to demonstrate how the aggregation of flood damage functions leads to bias. Applying Jensen's proof in this new context, results show that typically concave flood damage functions will introduce a positive bias (overestimation) when aggregated. This behaviour was further investigated with a simulation experiment including 2 million buildings in Germany, four global flood hazard simulations and three aggregation scenarios. The results show that positive aggregation bias is not distributed evenly in space, meaning some regions identified as "hot spots of risk" in assessments may in fact just be hot spots of aggregation bias. This study provides the first application of Jensen's inequality to explain the overestimates reported elsewhere and advice for modellers to minimize such artifacts. In total, this dissertation investigates the complex ways aggregation and disaggregation influence the behaviour of risk models, focusing on the scale-transfers underpinning macro-scale flood risk assessments. Extending a key finding of the flood hazard literature to the broader context of flood risk, this dissertation concludes that all else equal, coarse models overestimate risk. This dissertation goes beyond previous studies by providing mathematical proofs for how and where such bias emerges in aggregation routines, offering a mechanistic explanation for coarse model overestimates. It shows that this bias is spatially heterogeneous, necessitating a deep understanding of how rescaling may bias models to effectively reduce or communicate uncertainties. Further, the dissertation offers specific recommendations to help modellers minimize scale transfers in problematic regions. In conclusion, I argue that such aggregation errors are epistemic, stemming from choices in model structure, and therefore hold greater potential and impetus for study and mitigation. This deeper understanding of uncertainties is essential for improving macro-scale flood risk models and their effectiveness in equitable, holistic, and sustainable flood management.},
  language  = {en}
}