TY - JOUR A1 - Wutzler, Bianca A1 - Hudson, Paul A1 - Thieken, Annegret T1 - Adaptation strategies of flood-damaged businesses in Germany JF - Frontiers in water N2 - Flood risk management in Germany follows an integrative approach in which both private households and businesses can make an important contribution to reducing flood damage by implementing property-level adaptation measures. While the flood adaptation behavior of private households has already been widely researched, comparatively less attention has been paid to the adaptation strategies of businesses. However, their ability to cope with flood risk plays an important role in the social and economic development of a flood-prone region. Therefore, using quantitative survey data, this study aims to identify different strategies and adaptation drivers of 557 businesses damaged by a riverine flood in 2013 and 104 businesses damaged by pluvial or flash floods between 2014 and 2017. Our results indicate that a low perceived self-efficacy may be an important factor that can reduce the motivation of businesses to adapt to flood risk. Furthermore, property-owners tended to act more proactively than tenants. In addition, high experience with previous flood events and low perceived response costs could strengthen proactive adaptation behavior. These findings should be considered in business-tailored risk communication. KW - risk management KW - climate change adaptation KW - floods KW - disaster risk KW - reduction KW - Germany KW - precaution KW - emergency management Y1 - 2022 U6 - https://doi.org/10.3389/frwa.2022.932061 SN - 2624-9375 VL - 4 PB - Frontiers Media CY - Lausanne ER - TY - JOUR A1 - Wutzler, Bianca A1 - Hudson, Paul A1 - Thieken, Annegret T1 - Adaptation strategies of flood-damaged businesses in Germany JF - Frontiers in Water N2 - Flood risk management in Germany follows an integrative approach in which both private households and businesses can make an important contribution to reducing flood damage by implementing property-level adaptation measures. While the flood adaptation behavior of private households has already been widely researched, comparatively less attention has been paid to the adaptation strategies of businesses. However, their ability to cope with flood risk plays an important role in the social and economic development of a flood-prone region. Therefore, using quantitative survey data, this study aims to identify different strategies and adaptation drivers of 557 businesses damaged by a riverine flood in 2013 and 104 businesses damaged by pluvial or flash floods between 2014 and 2017. Our results indicate that a low perceived self-efficacy may be an important factor that can reduce the motivation of businesses to adapt to flood risk. Furthermore, property-owners tended to act more proactively than tenants. In addition, high experience with previous flood events and low perceived response costs could strengthen proactive adaptation behavior. These findings should be considered in business-tailored risk communication. KW - risk management KW - climate change adaptation KW - floods KW - disaster risk reduction KW - Germany KW - precaution KW - emergency management Y1 - 2022 U6 - https://doi.org/10.3389/frwa.2022.932061 SN - 2624-9375 PB - Frontiers Media SA CY - Lausanne, Schweiz ER - TY - GEN A1 - Wutzler, Bianca A1 - Hudson, Paul A1 - Thieken, Annegret T1 - Adaptation strategies of flood-damaged businesses in Germany T2 - Zweitveröffentlichungen der Universität Potsdam : Mathematisch-Naturwissenschaftliche Reihe N2 - Flood risk management in Germany follows an integrative approach in which both private households and businesses can make an important contribution to reducing flood damage by implementing property-level adaptation measures. While the flood adaptation behavior of private households has already been widely researched, comparatively less attention has been paid to the adaptation strategies of businesses. However, their ability to cope with flood risk plays an important role in the social and economic development of a flood-prone region. Therefore, using quantitative survey data, this study aims to identify different strategies and adaptation drivers of 557 businesses damaged by a riverine flood in 2013 and 104 businesses damaged by pluvial or flash floods between 2014 and 2017. Our results indicate that a low perceived self-efficacy may be an important factor that can reduce the motivation of businesses to adapt to flood risk. Furthermore, property-owners tended to act more proactively than tenants. In addition, high experience with previous flood events and low perceived response costs could strengthen proactive adaptation behavior. These findings should be considered in business-tailored risk communication. T3 - Zweitveröffentlichungen der Universität Potsdam : Mathematisch-Naturwissenschaftliche Reihe - 1304 KW - risk management KW - climate change adaptation KW - floods KW - disaster risk reduction KW - Germany KW - precaution KW - emergency management Y1 - 2023 U6 - http://nbn-resolving.de/urn/resolver.pl?urn:nbn:de:kobv:517-opus4-577350 SN - 1866-8372 IS - 1304 ER - TY - GEN A1 - Sultana, Zakia A1 - Sieg, Tobias A1 - Kellermann, Patric A1 - Müller, Meike A1 - Kreibich, Heidi T1 - Assessment of business interruption of flood-affected companies using random forests T2 - Postprints der Universität Potsdam : Mathematisch-Naturwissenschaftliche Reihe N2 - Losses due to floods have dramatically increased over the past decades, and losses of companies, comprising direct and indirect losses, have a large share of the total economic losses. Thus, there is an urgent need to gain more quantitative knowledge about flood losses, particularly losses caused by business interruption, in order to mitigate the economic loss of companies. However, business interruption caused by floods is rarely assessed because of a lack of sufficiently detailed data. A survey was undertaken to explore processes influencing business interruption, which collected information on 557 companies affected by the severe flood in June 2013 in Germany. Based on this data set, the study aims to assess the business interruption of directly affected companies by means of a Random Forests model. Variables that influence the duration and costs of business interruption were identified by the variable importance measures of Random Forests. Additionally, Random Forest-based models were developed and tested for their capacity to estimate business interruption duration and associated costs. The water level was found to be the most important variable influencing the duration of business interruption. Other important variables, relating to the estimation of business interruption duration, are the warning time, perceived danger of flood recurrence and inundation duration. In contrast, the amount of business interruption costs is strongly influenced by the size of the company, as assessed by the number of employees, emergency measures undertaken by the company and the fraction of customers within a 50 km radius. These results provide useful information and methods for companies to mitigate their losses from business interruption. However, the heterogeneity of companies is relatively high, and sector-specific analyses were not possible due to the small sample size. Therefore, further sector-specific analyses on the basis of more flood loss data of companies are recommended. T3 - Zweitveröffentlichungen der Universität Potsdam : Mathematisch-Naturwissenschaftliche Reihe - 939 KW - business interruption KW - floods KW - Random Forests KW - companies KW - variable importance Y1 - 2020 U6 - http://nbn-resolving.de/urn/resolver.pl?urn:nbn:de:kobv:517-opus4-459778 SN - 1866-8372 IS - 939 ER - TY - JOUR A1 - Sultana, Zakia A1 - Sieg, Tobias A1 - Kellermann, Patric A1 - Müller, Meike A1 - Kreibich, Heidi T1 - Assessment of business interruption of flood-affected companies using random forests JF - Water N2 - Losses due to floods have dramatically increased over the past decades, and losses of companies, comprising direct and indirect losses, have a large share of the total economic losses. Thus, there is an urgent need to gain more quantitative knowledge about flood losses, particularly losses caused by business interruption, in order to mitigate the economic loss of companies. However, business interruption caused by floods is rarely assessed because of a lack of sufficiently detailed data. A survey was undertaken to explore processes influencing business interruption, which collected information on 557 companies affected by the severe flood in June 2013 in Germany. Based on this data set, the study aims to assess the business interruption of directly affected companies by means of a Random Forests model. Variables that influence the duration and costs of business interruption were identified by the variable importance measures of Random Forests. Additionally, Random Forest-based models were developed and tested for their capacity to estimate business interruption duration and associated costs. The water level was found to be the most important variable influencing the duration of business interruption. Other important variables, relating to the estimation of business interruption duration, are the warning time, perceived danger of flood recurrence and inundation duration. In contrast, the amount of business interruption costs is strongly influenced by the size of the company, as assessed by the number of employees, emergency measures undertaken by the company and the fraction of customers within a 50 km radius. These results provide useful information and methods for companies to mitigate their losses from business interruption. However, the heterogeneity of companies is relatively high, and sector-specific analyses were not possible due to the small sample size. Therefore, further sector-specific analyses on the basis of more flood loss data of companies are recommended. KW - business interruption KW - floods KW - Random Forests KW - companies KW - variable importance Y1 - 2018 U6 - https://doi.org/10.3390/w10081049 SN - 2073-4441 VL - 10 IS - 8 PB - MDPI CY - Basel ER - TY - JOUR A1 - Schmidt, Lennart A1 - Hesse, Falk A1 - Attinger, Sabine A1 - Kumar, Rohini T1 - Challenges in applying machine learning models for hydrological inference BT - a case study for flooding events across Germany JF - Water resources research N2 - Machine learning (ML) algorithms are being increasingly used in Earth and Environmental modeling studies owing to the ever-increasing availability of diverse data sets and computational resources as well as advancement in ML algorithms. Despite advances in their predictive accuracy, the usefulness of ML algorithms for inference remains elusive. In this study, we employ two popular ML algorithms, artificial neural networks and random forest, to analyze a large data set of flood events across Germany with the goals to analyze their predictive accuracy and their usability to provide insights to hydrologic system functioning. The results of the ML algorithms are contrasted against a parametric approach based on multiple linear regression. For analysis, we employ a model-agnostic framework named Permuted Feature Importance to derive the influence of models' predictors. This allows us to compare the results of different algorithms for the first time in the context of hydrology. Our main findings are that (1) the ML models achieve higher prediction accuracy than linear regression, (2) the results reflect basic hydrological principles, but (3) further inference is hindered by the heterogeneity of results across algorithms. Thus, we conclude that the problem of equifinality as known from classical hydrological modeling also exists for ML and severely hampers its potential for inference. To account for the observed problems, we propose that when employing ML for inference, this should be made by using multiple algorithms and multiple methods, of which the latter should be embedded in a cross-validation routine. KW - machine learning KW - inference KW - floods Y1 - 2020 U6 - https://doi.org/10.1029/2019WR025924 SN - 0043-1397 SN - 1944-7973 VL - 56 IS - 5 PB - American Geophysical Union CY - Washington ER - TY - GEN A1 - Schmidt, Lennart A1 - Heße, Falk A1 - Attinger, Sabine A1 - Kumar, Rohini T1 - Challenges in applying machine learning models for hydrological inference: a case study for flooding events across Germany T2 - Postprints der Universität Potsdam : Mathematisch-Naturwissenschaftliche Reihe N2 - Machine learning (ML) algorithms are being increasingly used in Earth and Environmental modeling studies owing to the ever-increasing availability of diverse data sets and computational resources as well as advancement in ML algorithms. Despite advances in their predictive accuracy, the usefulness of ML algorithms for inference remains elusive. In this study, we employ two popular ML algorithms, artificial neural networks and random forest, to analyze a large data set of flood events across Germany with the goals to analyze their predictive accuracy and their usability to provide insights to hydrologic system functioning. The results of the ML algorithms are contrasted against a parametric approach based on multiple linear regression. For analysis, we employ a model-agnostic framework named Permuted Feature Importance to derive the influence of models' predictors. This allows us to compare the results of different algorithms for the first time in the context of hydrology. Our main findings are that (1) the ML models achieve higher prediction accuracy than linear regression, (2) the results reflect basic hydrological principles, but (3) further inference is hindered by the heterogeneity of results across algorithms. Thus, we conclude that the problem of equifinality as known from classical hydrological modeling also exists for ML and severely hampers its potential for inference. To account for the observed problems, we propose that when employing ML for inference, this should be made by using multiple algorithms and multiple methods, of which the latter should be embedded in a cross-validation routine. T3 - Zweitveröffentlichungen der Universität Potsdam : Mathematisch-Naturwissenschaftliche Reihe - 1193 KW - machine learning KW - inference KW - floods Y1 - 2019 U6 - http://nbn-resolving.de/urn/resolver.pl?urn:nbn:de:kobv:517-opus4-523843 SN - 1866-8372 IS - 5 ER - TY - JOUR A1 - Schmidt, Lennart A1 - Heße, Falk A1 - Attinger, Sabine A1 - Kumar, Rohini T1 - Challenges in applying machine learning models for hydrological inference: a case study for flooding events across Germany JF - Water Resources Research N2 - Machine learning (ML) algorithms are being increasingly used in Earth and Environmental modeling studies owing to the ever-increasing availability of diverse data sets and computational resources as well as advancement in ML algorithms. Despite advances in their predictive accuracy, the usefulness of ML algorithms for inference remains elusive. In this study, we employ two popular ML algorithms, artificial neural networks and random forest, to analyze a large data set of flood events across Germany with the goals to analyze their predictive accuracy and their usability to provide insights to hydrologic system functioning. The results of the ML algorithms are contrasted against a parametric approach based on multiple linear regression. For analysis, we employ a model-agnostic framework named Permuted Feature Importance to derive the influence of models' predictors. This allows us to compare the results of different algorithms for the first time in the context of hydrology. Our main findings are that (1) the ML models achieve higher prediction accuracy than linear regression, (2) the results reflect basic hydrological principles, but (3) further inference is hindered by the heterogeneity of results across algorithms. Thus, we conclude that the problem of equifinality as known from classical hydrological modeling also exists for ML and severely hampers its potential for inference. To account for the observed problems, we propose that when employing ML for inference, this should be made by using multiple algorithms and multiple methods, of which the latter should be embedded in a cross-validation routine. KW - machine learning KW - inference KW - floods Y1 - 2019 VL - 56 IS - 5 PB - John Wiley & Sons, Inc. CY - New Jersey ER - TY - THES A1 - Laudan, Jonas T1 - Changing susceptibility of flood-prone residents in Germany T1 - Die Änderung der Anfälligkeit von Hochwassergefährdeten Anwohnern in Deutschland BT - mental coping and mitigation behaviour in the context of different flood types BT - Mentale Bewältigung und Schadensminderungsverhalten im Zusammenhang mit Verschiedenen Hochwassertypen N2 - Floods are among the most costly natural hazards that affect Europe and Germany, demanding a continuous adaptation of flood risk management. While social and economic development in recent years altered the flood risk patterns mainly with regard to an increase in flood exposure, different flood events are further expected to increase in frequency and severity in certain European regions due to climate change. As a result of recent major flood events in Germany, the German flood risk management shifted to more integrated approaches that include private precaution and preparation to reduce the damage on exposed assets. Yet, detailed insights into the preparedness decisions of flood-prone households remain scarce, especially in connection to mental impacts and individual coping strategies after being affected by different flood types. This thesis aims to gain insights into flash floods as a costly hazard in certain German regions and compares the damage driving factors to the damage driving factors of river floods. Furthermore, psychological impacts as well as the effects on coping and mitigation behaviour of flood-affected households are assessed. In this context, psychological models such as the Protection Motivation Theory (PMT) and methods such as regressions and Bayesian statistics are used to evaluate influencing factors on the mental coping after an event and to identify psychological variables that are connected to intended private flood mitigation. The database consists of surveys that were conducted among affected households after major river floods in 2013 and flash floods in 2016. The main conclusions that can be drawn from this thesis reveal that the damage patterns and damage driving factors of strong flash floods differ significantly from those of river floods due to a rapid flow origination process, higher flow velocities and flow forces. However, the effects on mental coping of people that have been affected by flood events appear to be weakly influenced by different flood types, but yet show a coherence to the event severity, where often thinking of the respective event is pronounced and also connected to a higher mitigation motivation. The mental coping and preparation after floods is further influenced by a good information provision and a social environment, which encourages a positive attitude towards private mitigation. As an overall recommendation, approaches for an integrated flood risk management in Germany should be followed that also take flash floods into account and consider psychological characteristics of affected households to support and promote private flood mitigation. Targeted information campaigns that concern coping options and discuss current flood risks are important to better prepare for future flood hazards in Germany. N2 - Hochwasser zählen zu den schadensträchtigsten Naturgefahren, die in Europa und Deutschland vorkommen. In Deutschland traten in den letzten Jahren einige sehr starke Hochwasser und Überflutungen auf, die die Einstufung von Hochwassern als gefährliche Naturgewalt bestätigten. Private Haushalte leiden unter finanziellen und persönlichen Verlusten und sind sogar teilweise mehrfach betroffen. Folgenreiche Hochwasser, die im Gedächtnis blieben, waren insbesondere das Elbe-Hochwasser im Sommer 2002 sowie Überschwemmungen mit Schwerpukten an Elbe und Donau im Juni 2013. Im Mai und Juni 2016 kam es zu heftigen Unwettern über Zentraleuropa, während insbesondere Süddeutschland von Starkregen und Sturzfluten betroffen war. Hierbei wurden vereinzelte Ortschaften in Baden-Württtemberg (vor allem Braunsbach) und Bayern (vor allem Simbach am Inn) von extremen Sturzfluten beeinträchtigt und Bauwerke stark beschädigt. Als Reaktion auf die Flusshochwasser 2002 und 2013 wurde unter anderem das aktuelle Hochwasserrisikomanagement in Deutschland so angepasst, dass neben übergeordneten und technischen Hochwasserschutzmaßnahmen auch auf lokaler Ebene Maßnahmen ergriffen werden müssen. Diese umfassen Hochwasservorsorgemaßnahmen, die betroffene Haushalte selbst implementieren sollen. Neben strukturellen Maßnahmen wie z.B. der Verlegung von Heizung, Elektronik und Öltank in nicht-gefährdete Stockwerke sowie dem Schutz des Gebäudes vor Eindringen von Wasser, können auch nichtstrukturelle Maßnahmen, wie z.B. eine angepasste Wohnraumnutzung und das Verwenden von geeigneter Inneneinrichtung, ergriffen werden, um Hochwasserschäden signifikant zu verringern. Bis heute ist es jedoch unklar, aus welchen Gründen sich die betroffenen Menschen für Hochwasservorsorgemaßnahmen entscheiden und wie die individuelle Motivation, Maßnahmen zu implementieren, verstärkt werden kann. Neben dem Wissen um die eigene Hochwassergefährdung ist anzunehmen, dass die Selbsteinschätzung in Bezug auf einen wirksamen Umgang mit Hochwassern ausschlaggebend für die Motivation zur Vorsorge ist. Außerdem kann davon ausgegangen werden, dass verschiedene Hochwassertypen wie Flusshochwasser und Sturzfluten mit ihren unterschiedlichen Dynamiken unterschiedliche Auswirkungen auf die mentale Bewältigung und somit auch auf das Vorsorgeverhalten hervorrufen. Die vorliegende Arbeit hat demnach zum Ziel, Flusshochwasser und Sturzfluten in Deutschland miteinander zu vergleichen, wobei der Fokus auf schadenstreibenden Faktoren und psychologischen Auswirkungen auf betroffene Haushalte liegt. Weiterhin sollen damit verbundenes Vorsorgeverhalten untersucht und gegebenenfalls Handlungsempfehlungen für das Hochwasserrisikomanagement abgeleitet werden, das einerseits psychologische Charakteristika und andererseits Sturzfluten als signifikante Naturgefahr in Deutschland miteinbezieht. Hierbei werden sozio-ökonomische, zwischenmenschliche und psychologische Variablen von Haushalten ausgewertet, die 2013 und 2016 von Flusshochwassern und Sturzfluten betroffen waren. Dabei kommen verschiedene Methoden (Regressionen, Bayessche Statistik) und Modelle (Protection Motivation Theory) zum Einsatz, um Verbindungen zwischen den Variablen aufzeigen. Die Ergebnisse veranschaulichen erstens, dass Flusshochwasser und Sturzfluten zwar unterschiedliche Schäden an Gebäuden aufgrund verschiedener Flutdynamiken hervorrufen können, was sich bei Betroffenen jedoch nicht in unterschiedlichen psychologischen Auswirkungen widerspiegelt. Vielmehr ist die jeweilige Stärke und Schwere des Hochwassers entscheidend für charakteristische Ausprägungen von psychologischen Variablen. In diesem Falle sorgt eine stärkere Flut dafür, dass häufiger an das jeweilige Ereignis gedacht wird, während die Motivation zur Eigenvorsorge in solchen Fällen erhöht scheint. Zweitens sind ein soziales Umfeld, in dem bereits Vorsorgemaßnahmen implementiert wurden, sowie hilfreiche Informationen für geeignete Maßnahmen, deren Kosten und Aufklärung über das aktuelle Hochwasserrisiko förderlich für die Motivation, private Vorsorge zu betreiben. Ein aktuelles Hochwasserrisikomanagement sollte demnach auch Sturzfluten als mögliches Risiko in Deutschland miteinbeziehen und mehr in die Aufklärung und private Unterstützung bei Hochwassern investieren. Ein besseres Verstehen von psychologischen und mentalen Auswirkungen von verschiedenen Hochwassertypen hat den Vorteil, dass Hilfe und Informationskampagnen individuell und effizient gestaltet, Schäden minimiert und Schadensprognosen aufgrund der genaueren Kenntnisse über Vorsorgeverhalten verbessert werden können. KW - floods KW - psychology KW - flash floods KW - Hochwasser KW - Psychologie KW - Sturzfluten Y1 - 2019 U6 - http://nbn-resolving.de/urn/resolver.pl?urn:nbn:de:kobv:517-opus4-434421 ER - TY - THES A1 - Banerjee, Abhirup T1 - Characterizing the spatio-temporal patterns of extreme events BT - from recurrence to prediction N2 - Over the past decades, there has been a growing interest in ‘extreme events’ owing to the increasing threats that climate-related extremes such as floods, heatwaves, droughts, etc., pose to society. While extreme events have diverse definitions across various disciplines, ranging from earth science to neuroscience, they are characterized mainly as dynamic occurrences within a limited time frame that impedes the normal functioning of a system. Although extreme events are rare in occurrence, it has been found in various hydro-meteorological and physiological time series (e.g., river flows, temperatures, heartbeat intervals) that they may exhibit recurrent behavior, i.e., do not end the lifetime of the system. The aim of this thesis to develop some sophisticated methods to study various properties of extreme events. One of the main challenges in analyzing such extreme event-like time series is that they have large temporal gaps due to the paucity of the number of observations of extreme events. As a result, existing time series analysis tools are usually not helpful to decode the underlying information. I use the edit distance (ED) method to analyze extreme event-like time series in their unaltered form. ED is a specific distance metric, mainly designed to measure the similarity/dissimilarity between point process-like data. I combine ED with recurrence plot techniques to identify the recurrence property of flood events in the Mississippi River in the United States. I also use recurrence quantification analysis to show the deterministic properties and serial dependency in flood events. After that, I use this non-linear similarity measure (ED) to compute the pairwise dependency in extreme precipitation event series. I incorporate the similarity measure within the framework of complex network theory to study the collective behavior of climate extremes. Under this architecture, the nodes are defined by the spatial grid points of the given spatio-temporal climate dataset. Each node is associated with a time series corresponding to the temporal evolution of the climate observation at that grid point. Finally, the network links are functions of the pairwise statistical interdependence between the nodes. Various network measures, such as degree, betweenness centrality, clustering coefficient, etc., can be used to quantify the network’s topology. We apply the methodology mentioned above to study the spatio-temporal coherence pattern of extreme rainfall events in the United States and the Ganga River basin, which reveals its relation to various climate processes and the orography of the region. The identification of precursors associated with the occurrence of extreme events in the near future is extremely important to prepare the masses for an upcoming disaster and mitigate the potential risks associated with such events. Under this motivation, I propose an in-data prediction recipe for predicting the data structures that typically occur prior to extreme events using the Echo state network, a type of Recurrent Neural Network which is a part of the reservoir computing framework. However, unlike previous works that identify precursory structures in the same variable in which extreme events are manifested (active variable), I try to predict these structures by using data from another dynamic variable (passive variable) which does not show large excursions from the nominal condition but carries imprints of these extreme events. Furthermore, my results demonstrate that the quality of prediction depends on the magnitude of events, i.e., the higher the magnitude of the extreme, the better is its predictability skill. I show quantitatively that this is because the input signals collectively form a more coherent pattern for an extreme event of higher magnitude, which enhances the efficiency of the machine to predict the forthcoming extreme events. N2 - In den letzten Jahrzehnten hat das Interesse an ‘Extremereignissen’ aufgrund der zunehmenden Bedrohung, die klimabedingte Extreme wie Überschwemmungen, Hitzewellen, Dürren usw. für die Gesellschaft darstellen, zugenommen. Obwohl Extremereignisse in verschiedenen Disziplinen - von der Geowissenschaft bis zu den Neurowissenschaften - unterschiedlich definiert werden, werden sie hauptsächlich als dynamische Ereignisse innerhalb eines begrenzten Zeitrahmens charakterisiert, die das normale Funktionieren eines Systems beeinträchtigen. Obwohl Extremereignisse selten vorkommen, wurde festgestellt, dass verschiedene hydro-meteorologische und physiologische Zeitreihen (z. B. Stromabflussmengen, Temperaturen, Herzschlagintervalle) rekurrentes Verhalten. Das heißt, sie enden nicht an der Lebensdauer des Systems. Das Ziel dieser Arbeit ist es, fortschrittliche Methoden zur Untersuchung verschiedener Eigenschaften von Extremereignissen zu entwickeln. Eine der größten Herausforderungen bei der Analyse solcher extremen Ereignisse ist, dass sie große zeitliche Lücken aufweisen, da die Anzahl beobachteter Extremereignissen gering ist. Bestehende Zeitreihenanalysetools sind daher in der Regel nicht hilfreich, um die zugrundeliegenden Informationen zu entschlüsseln. Ich verwende die Edit-Distanz (ED) Methode, um extremeereignisähnliche Zeitreihen in ihrer unveränderten Form zu analysieren. ED ist eine spezielle Abstandsmetrik, die hauptsächlich zur Messung der Ähnlichkeit/Unähnlichkeit zwischen punktprozessähnlichen Daten entwickelt wurde. Ich kombiniere ED mit Rekurrenzplots, um die Wiederkehr-Eigenschaften von Hochwasserereignissen am Mississippi in den Vereinigten Staaten zu ermitteln. Außerdem werte ich die Wiederkehr-Eigenschaften quantitativ aus, um die deterministische Charakteristik und die serielle Abhängigkeit bei Hochwasserereignissen aufzuzeigen. Anschließend verwende ich dieses nichtlineare Ähnlichkeitsmaß(ED), um die paarweise Abhängigkeit in extremen Niederschlagsereignisreihen zu berechnen. Ich verknüpfe das Ähnlichkeitsmaß mit der Theorie komplexer Netzwerke, um das kollektive Verhalten von Klimaextremen zu untersuchen. In diesem Fall werden die Knoten durch die räumlichen Gitterpunkte des zu untersuchenden raumzeitlichen Klimadatensatzes definiert. Jeder Knoten repräsentiert eine Zeitreihe, die die zeitliche Entwicklung der Klimabeobachtung an diesem Gitterpunkt beschreibt. Schließlich sind die Netzwerkverbindungen Funktionen der paarweisen statistischen Interdependenz zwischen den Knotenpunkten. Zur Quantifizierung der Netztopologie können verschiedene Netzwerkmaße wie unter anderem der Grad, die Betweenness-Zentralität, oder der Clustering-Koeffizient verwendet werden. Wir wenden die oben erwähnte Methodik an, um das raumzeitliche Kohärenzmuster extremer Niederschlagsereignisse in den Vereinigten Staaten und im Einzugsgebiet des Ganges zu untersuchen. Die Ergebnisse zeigen, dass ein Zusammenhang zwischen dem Kohärenzmuster und verschiedenen Klimaprozessen und der Orographie der Region besteht. Die Identifizierung von Vorläufern, die mit dem Auftreten von Extremereignissen in naher Zukunft verbunden sind, ist äußerst wichtig, um die Bevölkerung auf eine bevorstehende Katastrophe vorzubereiten und potenziell resultierende Risiken zu mindern. Deshalb schlage ich ein datenbasiertes Vorhersageverfahren zur Bestimmung der Datenstrukturen vor, die typischerweise vor extremen Ereignissen auftreten. Das Verfahren basiert auf dem Echo-State Netzwerk, einem rekurrenten neuronalen Netz, das dem Reservoir-Computing zugeordnet wird. Im Gegensatz zu früheren Arbeiten, die Vorläuferstrukturen in der Variablen identifizieren, in der sich Extremereignisse manifestieren (aktive Variable), versuche ich die Strukturen anhand anderer dynamischer Variablen (passive Variablen) vorherzusagen. Diese Variablen weichen selbst nicht sonderlich von ihrem eigenen Sollzustand ab, aber sie besitzen eine Aussagekraft gegenüber den Extremereignissen. Meine Ergebnisse zeigen, dass die Qualität der Vorhersage von der Magnitude der Ereignisse abhängt, d.h. je extremer ein Ereignis ist, desto besser kann es vorhergesagt werden. Ich belege quantitativ, dass dieser Zusammenhang darauf basiert, dass die gesammelten Eingangssignale aussagekräftigere Kohärenzmuster für Extremereignisse hoher Magnitude bilden. Dies erhöht die Wirksamkeit des Ansatzes bei der Vorhersage bevorstehender Extremereignisse. KW - extreme events KW - Extremereignisse KW - floods KW - extreme precipitation KW - edit distance KW - recurrence plot KW - complex network KW - echo state network KW - Überschwemmungen KW - Extremniederschläge KW - Edit-Distanz KW - Rekurrenzplot KW - komplexes Netzwerk KW - Echo-State Netzwerk Y1 - 2022 U6 - http://nbn-resolving.de/urn/resolver.pl?urn:nbn:de:kobv:517-opus4-559839 ER -