Refine
Year of publication
Document Type
- Article (72)
- Postprint (8)
- Monograph/Edited Volume (7)
- Conference Proceeding (2)
- Part of a Book (1)
- Other (1)
Is part of the Bibliography
- yes (91)
Keywords
- climate change (4)
- Climate change (3)
- Lake Malawi (3)
- Shire River (3)
- drought intensity (3)
- drought projections (3)
- events (3)
- meteorological drought (3)
- Andes (2)
- Compound events (2)
Institute
- Institut für Umweltwissenschaften und Geographie (91) (remove)
Extreme Regenereignisse von kurzer Dauer im Bereich von Stunden und darunter rücken aufgrund der dadurch bedingten Schäden durch Sturzfluten und auch wegen ihrer möglichen Intensivierungen durch den anthropogenen Klimawandel immer stärker in den Fokus. Die vorliegende Studie untersucht auf Basis von teilweise sehr langen (> 50 Jahre) und zeitlich hochaufgelösten Zeitreihen (≤ 15 Minuten) mögliche Trends in Starkregenintensitäten für Stationen aus schweizerischen und österreichischen Alpenregionen sowie für das Emscher-Lippe-Gebiet in Nordrhein-Westfalen. Es wird deutlich, dass es eine Zunahme der extremen Niederschlagsintensitäten gibt, welche gut durch die Erwärmung des regionalen Klimas erklärt werden kann: Die Analysen langfristiger Trends der Überschreitungssummen und Wiederkehrniveaus zeigen zwar erhebliche Unsicherheiten, lassen jedoch eine Zunahme in einer Größenordnung von 30 % pro Jahrhundert erkennen. Zudem wird in diesem Beitrag, basierend auf einer "mittleren" Klimasimulation für das 21. Jahrhundert, für ausgewählte Stationen der Emscher-Lippe-Region eine Projektion für extreme Niederschlagsintensitäten in sehr hoher zeitlicher Auflösung beschrieben. Dabei wird ein gekoppeltes räumliches und zeitliches "Downscaling" angewendet, dessen entscheidende Neuerung die Berücksichtigung der Abhängigkeit der lokalen Regenintensität von der Lufttemperatur ist. Dieses Verfahren beinhaltet zwei Schritte: Zuerst werden großräumige Klimafelder in täglicher Auflösung durch Regression mit den Temperatur- und Niederschlagswerten der Stationen statistisch verbunden (räumliches Downscaling). Im zweiten Schritt werden dann diese Stationswerte mithilfe eines sogenannten multiplikativen stochastischen Kaskadenmodells (MC) auf eine zeitliche Auflösung von 10 Minuten disaggregiert (zeitliches Downscaling). Die neuartige, temperatursensitive Variante berücksichtigt zusätzlich die Lufttemperatur als erklärende Variable für die Niederschlagsintensitäten. Dadurch wird der mit einer Erwärmung zu erwartende höhere atmosphärische Feuchtegehalt, welcher sich aus der Clausius-Clapeyron-Beziehung (CC) ergibt, mit in das zeitliche Downscaling einbezogen.
Für die statistische Auswertung der extremen kurzfristigen Niederschläge wurden die oberen Quantile (99,9 %), Überschreitungssummen (ÜS, P > 5 mm) und 3-jährliche Wiederkehrniveaus (WN) einer Dauerstufe von ≤ 15-Minuten betrachtet. Diese Auswahl erlaubt die gleichzeitige Analyse sowohl von Extremwertstatistiken als auch von deren langfristigen Trends; leichte Abweichungen von dieser Wahl beeinflussen die Hauptergebnisse nur unwesentlich. Nur durch die Hinzunahme der Temperatur wird die beobachtete Temperaturabhängigkeit der extremen Quantile (CC-Scaling) gut wiedergegeben. Bei Vergleich von Beobachtungsdaten und Gegenwartssimulationen der Modellkaskade zeigt das temperatursensitive Verfahren konsistente Ergebnisse. Im Vergleich zu den Entwicklungen der letzten Jahrzehnte werden für die Zukunft ähnliche oder sogar noch stärkere Anstiege der extremen Niederschlagsintensitäten projiziert. Dies ist insofern bemerkenswert, als diese anscheinend hauptsächlich durch die örtliche Temperatur bestimmt werden, denn die projizierten Trends der Niederschlags-Tageswerte sind für diese Region vernachlässigbar.
Veränderung der Abflüsse
(2005)
From 6 to 9 August 2012, intense rainfall hit the northern Philippines, causing massive floods in Metropolitan Manila and nearby regions. Local rain gauges recorded almost 1000mm within this period. However, the recently installed Philippine network of weather radars suggests that Metropolitan Manila might have escaped a potentially bigger flood just by a whisker, since the centre of mass of accumulated rainfall was located over Manila Bay. A shift of this centre by no more than 20 km could have resulted in a flood disaster far worse than what occurred during Typhoon Ketsana in September 2009.
Spatial patterns as well as temporal dynamics of soil moisture have a major influence on runoff generation. The investigation of these dynamics and patterns can thus yield valuable information on hydrological processes, especially in data scarce or previously ungauged catchments. The combination of spatially scarce but temporally high resolution soil moisture profiles with episodic and thus temporally scarce moisture profiles at additional locations provides information on spatial as well as temporal patterns of soil moisture at the hillslope transect scale. This approach is better suited to difficult terrain (dense forest, steep slopes) than geophysical techniques and at the same time less cost-intensive than a high resolution grid of continuously measuring sensors. Rainfall simulation experiments with dye tracers while continuously monitoring soil moisture response allows for visualization of flow processes in the unsaturated zone at these locations. Data was analyzed at different spacio-temporal scales using various graphical methods, such as space-time colour maps (for the event and plot scale) and binary indicator maps (for the long-term and hillslope scale). Annual dynamics of soil moisture and decimeterscale variability were also investigated. The proposed approach proved to be successful in the investigation of flow processes in the unsaturated zone and showed the importance of preferential flow in the Malalcahuello Catchment, a datascarce catchment in the Andes of Southern Chile. Fast response times of stream flow indicate that preferential flow observed at the plot scale might also be of importance at the hillslope or catchment scale. Flow patterns were highly variable in space but persistent in time. The most likely explanation for preferential flow in this catchment is a combination of hydrophobicity, small scale heterogeneity in rainfall due to redistribution in the canopy and strong gradients in unsaturated conductivities leading to self-reinforcing flow paths.
Transferability of data-driven models to predict urban pluvial flood water depth in Berlin, Germany
(2023)
Data-driven models have been recently suggested to surrogate computationally expensive hydrodynamic models to map flood hazards. However, most studies focused on developing models for the same area or the same precipitation event. It is thus not obvious how transferable the models are in space. This study evaluates the performance of a convolutional neural network (CNN) based on the U-Net architecture and the random forest (RF) algorithm to predict flood water depth, the models' transferability in space and performance improvement using transfer learning techniques. We used three study areas in Berlin to train, validate and test the models. The results showed that (1) the RF models outperformed the CNN models for predictions within the training domain, presumable at the cost of overfitting; (2) the CNN models had significantly higher potential than the RF models to generalize beyond the training domain; and (3) the CNN models could better benefit from transfer learning technique to boost their performance outside training domains than RF models.
Transferability of data-driven models to predict urban pluvial flood water depth in Berlin, Germany
(2023)
Data-driven models have been recently suggested to surrogate computationally expensive hydrodynamic models to map flood hazards. However, most studies focused on developing models for the same area or the same precipitation event. It is thus not obvious how transferable the models are in space. This study evaluates the performance of a convolutional neural network (CNN) based on the U-Net architecture and the random forest (RF) algorithm to predict flood water depth, the models' transferability in space and performance improvement using transfer learning techniques. We used three study areas in Berlin to train, validate and test the models. The results showed that (1) the RF models outperformed the CNN models for predictions within the training domain, presumable at the cost of overfitting; (2) the CNN models had significantly higher potential than the RF models to generalize beyond the training domain; and (3) the CNN models could better benefit from transfer learning technique to boost their performance outside training domains than RF models.
Identifying urban pluvial flood-prone areas is necessary but the application of two-dimensional hydrodynamic models is limited to small areas. Data-driven models have been showing their ability to map flood susceptibility but their application in urban pluvial flooding is still rare. A flood inventory (4333 flooded locations) and 11 factors which potentially indicate an increased hazard for pluvial flooding were used to implement convolutional neural network (CNN), artificial neural network (ANN), random forest (RF) and support vector machine (SVM) to: (1) Map flood susceptibility in Berlin at 30, 10, 5, and 2 m spatial resolutions. (2) Evaluate the trained models' transferability in space. (3) Estimate the most useful factors for flood susceptibility mapping. The models' performance was validated using the Kappa, and the area under the receiver operating characteristic curve (AUC). The results indicated that all models perform very well (minimum AUC = 0.87 for the testing dataset). The RF models outperformed all other models at all spatial resolutions and the RF model at 2 m spatial resolution was superior for the present flood inventory and predictor variables. The majority of the models had a moderate performance for predictions outside the training area based on Kappa evaluation (minimum AUC = 0.8). Aspect and altitude were the most influencing factors on the image-based and point-based models respectively. Data-driven models can be a reliable tool for urban pluvial flood susceptibility mapping wherever a reliable flood inventory is available.
Identifying urban pluvial flood-prone areas is necessary but the application of two-dimensional hydrodynamic models is limited to small areas. Data-driven models have been showing their ability to map flood susceptibility but their application in urban pluvial flooding is still rare. A flood inventory (4333 flooded locations) and 11 factors which potentially indicate an increased hazard for pluvial flooding were used to implement convolutional neural network (CNN), artificial neural network (ANN), random forest (RF) and support vector machine (SVM) to: (1) Map flood susceptibility in Berlin at 30, 10, 5, and 2 m spatial resolutions. (2) Evaluate the trained models' transferability in space. (3) Estimate the most useful factors for flood susceptibility mapping. The models' performance was validated using the Kappa, and the area under the receiver operating characteristic curve (AUC). The results indicated that all models perform very well (minimum AUC = 0.87 for the testing dataset). The RF models outperformed all other models at all spatial resolutions and the RF model at 2 m spatial resolution was superior for the present flood inventory and predictor variables. The majority of the models had a moderate performance for predictions outside the training area based on Kappa evaluation (minimum AUC = 0.8). Aspect and altitude were the most influencing factors on the image-based and point-based models respectively. Data-driven models can be a reliable tool for urban pluvial flood susceptibility mapping wherever a reliable flood inventory is available.
Most hydrological studies rely on a model calibrated using discharge alone. However, judging the model reliability based on such calibration is problematic, as it does not guarantee the correct representation of internal hydrological processes. This study aims (a) to develop a comprehensive multi-objective calibration framework using remote sensing vegetation data and hydrological signatures (flow duration curve - FDC, and baseflow index) in addition to discharge, and (b) to apply this framework for calibration of the Soil and Water Assessment Tool (SWAT) in a typical Andean catchment. Overall, our calibration approach outperformed traditional discharge-based and FDC signature-based calibration strategies in terms of vegetation, streamflow, and flow partitioning simulation. New hydrological insights for the region are the following: baseflow is the main component of the streamflow sustaining the long dry-season flow, and pasture areas offer higher water yield and baseflow than other land-cover types. The proposed approach could be used in other data-scarce regions with complex topography.
Most hydrological studies rely on a model calibrated using discharge alone. However, judging the model reliability based on such calibration is problematic, as it does not guarantee the correct representation of internal hydrological processes. This study aims (a) to develop a comprehensive multi-objective calibration framework using remote sensing vegetation data and hydrological signatures (flow duration curve - FDC, and baseflow index) in addition to discharge, and (b) to apply this framework for calibration of the Soil and Water Assessment Tool (SWAT) in a typical Andean catchment. Overall, our calibration approach outperformed traditional discharge-based and FDC signature-based calibration strategies in terms of vegetation, streamflow, and flow partitioning simulation. New hydrological insights for the region are the following: baseflow is the main component of the streamflow sustaining the long dry-season flow, and pasture areas offer higher water yield and baseflow than other land-cover types. The proposed approach could be used in other data-scarce regions with complex topography.