Exploring the additional value of class imbalance distributions on interpretable flash flood susceptibility prediction in the Black Warrior River basin, Alabama, United States

Ekmekçioğlu, Ömer; Koç, Kerim; Özger, Mehmet; Işık, Zeynep

doi:10.1016/j.jhydrol.2022.127877

Cited by 31 publications

(18 citation statements)

References 73 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In [86], SHAP analysis showed that forecasting streamflow relies on important precipitation data, besides streamflow inputs, often influencing the mode positively. In Ekmekcio glu et al [87] and Aydin and Iban [88], SHAP analysis showed that precipitation may affect the forecasted result differently depending on the ML model used, which is an expected behavior since different ML approaches process data differently, resulting in different results for an identical task [33,48,49,89].…”

Section: Discussionmentioning

confidence: 99%

A New Graph-Based Deep Learning Model to Predict Flooding with Validation on a Case Study on the Humber River

Oliveira Santos,

Costa Rocha,

Scott

et al. 2023

Water

View full text Add to dashboard Cite

Floods are one of the most lethal natural disasters. It is crucial to forecast the timing and evolution of these events and create an advanced warning system to allow for the proper implementation of preventive measures. This work introduced a new graph-based forecasting model, namely, graph neural network sample and aggregate (GNN-SAGE), to estimate river flooding. It then validated the proposed model in the Humber River watershed in Ontario, Canada. Using past precipitation and stage data from reference and neighboring stations, the proposed GNN-SAGE model could estimate the river stage for flooding events up to 24 h ahead, improving its forecasting performance by an average of 18% compared with the persistence model and 9% compared with the graph-based model residual gated graph convolutional network (GNN-ResGated), which were used as baselines. Furthermore, GNN-SAGE generated smaller errors than those reported in the current literature. The Shapley additive explanations (SHAP) revealed that prior data from the reference station was the most significant factor for all prediction intervals, with seasonality and precipitation being more influential for longer-range forecasts. The findings positioned the proposed GNN-SAGE model as a cutting-edge solution for flood forecasting and a valuable resource for devising early flood-warning systems.

show abstract

Section: Discussionmentioning

confidence: 99%

A New Graph-Based Deep Learning Model to Predict Flooding with Validation on a Case Study on the Humber River

Oliveira Santos,

Costa Rocha,

Scott

et al. 2023

Water

View full text Add to dashboard Cite

show abstract

“…This demonstrates significant potential for the utilisation of this model in LSM. Ekmekcioğlu et al (2022) proposed a novel flash flood susceptibility prediction framework based on CS‐RF, employing the SHAP model to explain the triggering factors of flash floods, thereby improving the model's transparency significantly. These studies collectively highlight the efficacy of interpretable machine learning within the realm of disaster susceptibility mapping, particularly for specific geographical areas.…”

Section: Discussionmentioning

confidence: 99%

Insights into landslide susceptibility in different karst erosion landforms based on interpretable machine learning

Sun,

Wang,

Wen

et al. 2023

Earth Surf Processes Landf

View full text Add to dashboard Cite

The aim of the present study was to assess differences in the conditioning factors and the performance of landslide susceptibility mapping (LSM), employing the SHapley Additive exPlanations (SHAP) model to gain profound insights into the intrinsic decision‐making mechanism of LSM in diverse landforms. Two typical karst erosion landforms were selected as the research areas. Based on 15 conditioning factors, LSMs for the two areas were developed using the Bayesian optimization random forest (RF) and eXtreme Gradient Boosting (XGBoost). The SHAP model was used to explore the landslide formation mechanisms from both global and local perspectives. The results show that the area under the curve (AUC) values of the XGBoost models were 0.791 and 0.761, and the AUC values of the RF models were 0.844 and 0.817, in the two different landform areas, respectively. The RF model's accuracy was higher than that of the XGBoost model in both regions. In the low‐elevation hills area, the primary three conditioning factors were identified as slope, topographic relief and distance from the river. Conversely, in the microrelief and mesorelief low mountain area, the predominant conditioning factors were elevation, distance from the river and distance from the road. Both karst landform areas exhibited a high sensitivity to the distance from the river, indicating its significant interaction with other factors contributing to landslide occurrences. Notably, the RF model demonstrated superior performance compared to the XGBoost model, rendering it a more suitable choice for conducting landslide susceptibility mapping research in karst erosion landform areas. In the present study, a comprehensive explanatory framework based on the RF‐SHAP model was proposed, which enables both global and local interpretation of landslides in various karst landscapes. Such an approach explores the intrinsic decision‐making mechanism of the model, enhancing the transparency and realism of landslide susceptibility prediction results.

show abstract

“…In the XGBoost algorithm, there are strong correlations between successive decision trees [ 42 ]. Each round of prediction is based on the prediction error in the previous round; thus, it is iteratively constructed, which greatly improves the accuracy of the prediction [ 43 ]. Compared with traditional statistical models, it can determine a default direction of a branch for missing data, thereby reducing the resulting error [ 44 ].…”

Section: Methodsmentioning

confidence: 99%

Influencing Factors and Risk Assessment of Precipitation-Induced Flooding in Zhengzhou, China, Based on Random Forest and XGBoost Algorithms

Liu

Zhou

Lin

et al. 2022

IJERPH

View full text Add to dashboard Cite

Due to extreme weather phenomena, precipitation-induced flooding has become a frequent, widespread, and destructive natural disaster. Risk assessments of flooding have thus become a popular area of research. In this study, we studied the severe precipitation-induced flooding that occurred in Zhengzhou, Henan Province, China, in July 2021. We identified 16 basic indicators, and the random forest algorithm was used to determine the contribution of each indicator to the Zhengzhou flood. We then optimised the selected indicators and introduced the XGBoost algorithm to construct a risk index assessment model of precipitation-induced flooding. Our results identified four primary indicators for precipitation-induced flooding in the study area: total rainfall for three consecutive days, extreme daily rainfall, vegetation cover, and the river system. The Zhengzhou storm and flood risk evaluation model was constructed from 12 indicators: elevation, slope, water system index, extreme daily rainfall, total rainfall for three consecutive days, night-time light brightness, land-use type, proportion of arable land area, gross regional product, proportion of elderly population, vegetation cover, and medical rescue capacity. After streamlining the bottom four indicators in terms of contribution rate, it had the best performance, with an accuracy rate reaching 91.3%. Very high-risk and high-risk areas accounted for 11.46% and 27.50% of the total area of Zhengzhou, respectively, and their distribution was more significantly influenced by the extent of heavy rainfall, direction of river systems, and land types; the medium-risk area was the largest, accounting for 33.96% of the total area; the second-lowest-risk and low-risk areas together accounted for 27.09%. The areas with the highest risk of heavy rainfall and flooding in Zhengzhou were in the Erqi, Guanchenghui, Jinshui, Zhongyuan, and Huizi Districts and the western part of Xinmi City; these areas should be given priority attention during disaster monitoring and early warning and risk prevention and control.

show abstract

Exploring the additional value of class imbalance distributions on interpretable flash flood susceptibility prediction in the Black Warrior River basin, Alabama, United States

Cited by 31 publications

References 73 publications

A New Graph-Based Deep Learning Model to Predict Flooding with Validation on a Case Study on the Humber River

A New Graph-Based Deep Learning Model to Predict Flooding with Validation on a Case Study on the Humber River

Insights into landslide susceptibility in different karst erosion landforms based on interpretable machine learning

Influencing Factors and Risk Assessment of Precipitation-Induced Flooding in Zhengzhou, China, Based on Random Forest and XGBoost Algorithms

Contact Info

Product

Resources

About