Refine
Language
- English (3)
Is part of the Bibliography
- yes (3) (remove)
Keywords
- support vector machine (3) (remove)
Institute
- Institut für Umweltwissenschaften und Geographie (3) (remove)
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.
Large-scale crop yield failures are increasingly associated with food price spikes and food insecurity and are a large source of income risk for farmers. While the evidence linking extreme weather to yield failures is clear, consensus on the broader set of weather drivers and conditions responsible for recent yield failures is lacking. We investigate this for the case of four major crops in Germany over the past 20 years using a combination of machine learning and process-based modelling. Our results confirm that years associated with widespread yield failures across crops were generally associated with severe drought, such as in 2018 and to a lesser extent 2003. However, for years with more localized yield failures and large differences in spatial patterns of yield failures between crops, no single driver or combination of drivers was identified. Relatively large residuals of unexplained variation likely indicate the importance of non-weather related factors, such as management (pest, weed and nutrient management and possible interactions with weather) explaining yield failures. Models to inform adaptation planning at farm, market or policy levels are here suggested to require consideration of cumulative resource capture and use, as well as effects of extreme events, the latter largely missing in process-based models. However, increasingly novel combinations of weather events under climate change may limit the extent to which data driven methods can replace process-based models in risk assessments.