How Do Multiple Kernel Functions in Machine Learning Algorithms Improve Precision in Flood Probability Mapping?

With climate change, hydro-climatic hazards, i.e., floods in the Himalayas regions, are expected to worsen, thus, likely to affect humans and socio-economic growth. Precisely, the Koshi River basin (KRB) is often impacted by flooding over the year. However, studies on estimating and predicting floods still lack in this basin. This study aims at developing flood probability map using machine learning algorithms (MLAs): gaussian process regression (GPR) and support vector machine (SVM) with multiple kernel functions including Pearson VII function kernel (PUK), polynomial, normalized poly kernel, and radial basis kernel function (RBF). Historical flood locations with available topography, hydrogeology, and environmental datasets were further considered to build flood model. Two datasets were carefully chosen to measure the feasibility and robustness of MLAs: training dataset (location of floods between 2010 and 2019) and testing dataset (flood locations of 2020) with thirteen flood influencing factors. The validation of the MLAs was evaluated using a validation dataset and statistical indices such as the coefficient of determination (r2: 0.546~0.995), mean absolute error (MAE: 0.009~0.373), root mean square error (RMSE: 0.051~0.466), relative absolute error (RAE: 1.81~88.55%), and root-relative square error (RRSE: 10.19~91.00%). Results showed that the SVM-Pearson VII kernel (PUK) yielded better prediction than other algorithms. The resultant map from SVM-PUK revealed that 27.99% area with low, 39.91% area with medium, 31.00% with high, and 1.10% area with very high probabilities of flooding in the study area. The final flood probability map could add a greatt value to the effort of flood risk mitigation and planning processes in KRB.

Estimating the probability of compound floods in estuarine regions

The quantification of flood risk in estuarine regions relies on accurate estimation of flood probability, which is often challenging due to the rareness of hazardous flood events and their multi-causal (or “compound”) nature. Failure to consider the compounding nature of estuarine floods can lead to significant underestimation of flood risk in these regions. This study provides a comparative review of alternative approaches for estuarine flood estimation – namely, traditional univariate flood frequency analysis applied to both observed historical data and simulated data, as well as multivariate frequency analysis applied to flood events. Three specific implementations of the above approaches are evaluated on a case study – the estuarine portion of Swan River in Western Australia – highlighting the advantages and disadvantages of each approach. The theoretical understanding of the three approaches, combined with findings from the case study, enable the generation of guidance on method selection for estuarine flood probability estimation, recognizing issues such as data availability, the complexity of the application/analysis process, the location of interest within the estuarine region, the computational demands, and whether or not future conditions need to be assessed.

Scenario-Based Real-Time Flood Prediction with Logistic Regression

This study proposed a real-time flood extent prediction method to shorten the time it takes from the flood occurrence to an alert issuance. This method uses logistic regression to generate a flood probability discriminant for each grid constituting the study area, and then predicts the flood extent with the amount of runoff caused by rainfall. In order to generate the flood probability discriminant for each grid, a two-dimensional (2D) flood inundation model was verified by applying the Typhoon Chaba, which caused great damage to the study area in 2016. Then, 100 probability rainfall scenarios were created by combining the return period, duration, and time distribution using past observation rainfall data, and rainfall-runoff–inundation relation databases were built for each scenario by applying hydrodynamic and hydrological models. A flood probability discriminant based on logistic regression was generated for each grid by using whether the grid was flooded (1 or 0) for the runoff amount in the database. When the runoff amount is input to the generated discriminant, the flood probability on the target grid is calculated by the coefficients, so that the flood extent is quickly predicted. The proposed method predicted the flood extent in a few seconds in both cases and showed high accuracy with 83.6~98.4% and 74.4~99.1%, respectively, in the application of scenario rainfall and actual rainfall.

The FLOod Probability Interpolation Tool (FLOPIT): A Simple Tool to Improve Spatial Flood Probability Quantification and Communication

Understanding flood probabilities is essential to making sound decisions about flood-risk management. Many people rely on flood probability maps to inform decisions about purchasing flood insurance, buying or selling real-estate, flood-proofing a house, or managing floodplain development. Current flood probability maps typically use flood zones (for example the 1 in 100 or 1 in 500-year flood zones) to communicate flooding probabilities. However, this choice of communication format can miss important details and lead to biased risk assessments. Here we develop, test, and demonstrate the FLOod Probability Interpolation Tool (FLOPIT). FLOPIT interpolates flood probabilities between water surface elevation to produce continuous flood-probability maps. FLOPIT uses water surface elevation inundation maps for at least two return periods and creates Annual Exceedance Probability (AEP) as well as inundation maps for new return levels. Potential advantages of FLOPIT include being open-source, relatively easy to implement, capable of creating inundation maps from agencies other than FEMA, and applicable to locations where FEMA published flood inundation maps but not flood probability. Using publicly available data from the Federal Emergency Management Agency (FEMA) flood risk databases as well as state and national datasets, we produce continuous flood-probability maps at three example locations in the United States: Houston (TX), Muncy (PA), and Selinsgrove (PA). We find that the discrete flood zones generally communicate substantially lower flood probabilities than the continuous estimates.

Spatial Prediction of Future Flood Risk: An Approach to the Effects of Climate Change

Preparation of a flood probability map serves as the first step in a flood management program. This research develops a probability flood map for floods resulting from climate change in the future. Two models of Flexible Discrimination Analysis (FDA) and Artificial Neural Network (ANN) were used. Two optimistic (RCP2.6) and pessimistic (RCP8.5) climate change scenarios were considered for mapping future rainfall. Moreover, to produce probability flood occurrence maps, 263 locations of past flood events were used as dependent variables. The number of 13 factors conditioning floods was taken as independent variables in modeling. Of the total 263 flood locations, 80% (210 locations) and 20% (53 locations) were considered model training and validation. The Receiver Operating Characteristic (ROC) curve and other statistical criteria were used to validate the models. Based on assessments of the validated models, FDA, with a ROC-AUC = 0.918, standard error (SE = 0.038), and an accuracy of 0.86% compared to the ANN model with a ROC-AUC = 0.897, has the highest accuracy in preparing the flood probability map in the study area. The modeling results also showed that the factors of distance from the River, altitude, slope, and rainfall have the greatest impact on floods in the study area. Both models’ future flood susceptibility maps showed that the highest area is related to the very low class. The lowest area is related to the high class.