An integrated model chain for future flood risk prediction under land-use changes

Confidence interval forecasting model of small watershed flood based on compound recurrent neural networks and Bayesian

Flood forecasting exhibits rapid fluctuations, water level forecasting shows great uncertainty and inaccuracy in small watersheds, and the reliability and accuracy performance of traditional probability forecasting is often unbalanced. This study combined Recurrent Neural Networks (RNN) and Bayesian to establish a comprehensive forecasting model framework of RNNs-Bayesian for the forecasting of water level confidence interval, to achieve both reasonable reliability and accuracy. In the Bayesian structure, weight training was used. In the RNNs, base RNN, Long Short-term Memory (LSTM), and Gated Recurrent Unit (GRU) are used for comparative analysis, and experiments are carried out at the point of the Qixi Reservoir in a small watershed in Zhejiang Province of China. We used the multidimensional disaster data input unit for water level forecasting, including hydrology, meteorology, and geography, and 5 days of time windows for forecasting, The comprehensive reliability of LSTM-Bayesian for 0~102 hours flood reached 92.31%, and the comprehensive accuracy reached 89.15%, and confidence interval forecasting using LSTM is the best method, and achieved reasonable balance of reliability and accuracy. Overall, compound RNN could be a good alternative for forecasting hourly streamflow and extreme water level in small watersheds.

Development of a real-time dynamic inundation risk assessment approach on paddy fields during typhoons: Exploration of adaptation strategies and quantification of risks

When typhoons strike, heavy rainfall frequently triggers inundation disasters, causing significant impacts on agricultural crops. Efficiently understanding the effects of typhoon-induced flooding and implementing effective disaster mitigation measures are critical challenges for decision-makers. This study proposes a real-time dynamic inundation risk assessment framework to address these challenges, focusing on adaptation strategies and risk evaluation for paddy fields during typhoons. In the qualitative phase, governance levels and stakeholders are identified to facilitate discussions on key issues and protected targets. Both climatic and non-climatic factors contributing to hazards are considered. In the quantitative phase, hazard levels are defined, and the spatial distribution and vulnerability factors of paddy fields are assessed to quantify exposure and vulnerability. These three components are integrated using a risk assessment matrix to derive real-time dynamic risk levels. Yilan County, Taiwan, was selected as the study area to demonstrate the methodology. Forecasted values from a local AI-based inundation forecasting model served as the hazard component. The results reveal that the proposed approach delivers higher accuracy in predicting dynamic hazard and risk levels during severe inundation events. Furthermore, risk assessments incorporating hazard, exposure, and vulnerability provided more nuanced spatial distributions compared to those considering hazard alone. This highlights the framework's ability to integrate protected asset information, supporting decision-makers in devising timely and effective adaptation strategies, ultimately contributing to agricultural disaster management.

Flood risk transfer analysis based on the "Source-Sink" theory and its impact on ecological environment: A case study of the Poyang Lake Basin, China

Driven by climate change, the frequent occurrence of regional destructive floods poses a grave threat to socio-economic systems and ecological environments. Previous flood risk studies have disregarded risk transfer within a region, resulting in inadequate flood risk assessment and ineffective disaster prevention and mitigation outcomes. Therefore, this study introduced the "Source-Sink" theory into flood disaster field to constructing flood risk transfer model. Flood risk assessment and transfer was conducted in the Poyang Lake Basin, China, where the impacts of the initial and transfer statuses on ecosystem service values were quantified. The results showed that the flood risk in the Poyang Lake Basin was relatively low, with high spatial distribution characteristics in the central-north areas but low in the surrounding areas. High-risk zones were mainly distributed southwest of the Poyang Lake. The lower-risk zones exhibited a contiguous distribution and were surrounded by higher-risk zones. Following the completion of the flood risk transfer, high-risk zones increased significantly; but there were a few zones where the risk was transferred to other zones, thereby lowering their risks. Flood risk transfer occurs primarily in low- and medium-risk zones, with high-risk zones being the most important growth targets. The change in risk transfer was most evident in the area surrounding Poyang Lake, while that in the Upper Gan River Basin was lower and less sensitive to the transfer effect. Accounting for flood disaster risk, the ecosystem service values of the Poyang Lake Basin decreased by 8.18 %, with the most significant impacts observed in the surrounding environment and southwest Poyang Lake. After the completion of the flood risk transfer, the ecosystem service value in the Poyang Lake Basin declined by 24.66 %. This study provides a reference point for flood risk management and sustainable regional development that account for risk transfer.

Tool for fast assessment of stormwater flood volumes for urban catchment: A machine learning approach

Specific flood volume is an important criterion for evaluating the performance of sewer networks. Currently, mechanistic models - MCMs (e.g., SWMM) are usually used for its prediction, but they require the collection of detailed information about the characteristics of the catchment and sewer network, which can be difficult to obtain, and the process of model calibration is a complex task. This paper presents a methodology for developing simulators to predict specific flood volume using machine learning methods (DNN - Deep Neural Network, GAM - Generalized Additive Model). The results of Sobol index calculations using the GSA method were used to select the ML model as an alternative to the MCM model. It was shown that the DNN model can be used for flood prediction, for which high agreement was obtained between the results of GSA calculations for rainfall data, catchment and sewer network characteristics, and calibrated SWMM parameters describing land use and sewer retention. Regression relationships (polynomials and exponential functions) were determined between Sobol indices (retention depth of impervious area, correction factor of impervious area, Manning's roughness coefficient of sewers) and sewer network characteristics (unit density of sewers, retention factor - the downstream and upstream of retention ratio) obtaining R2 = 0. 55-0.78. The feasibility of predicting sewer network flooding and modernization with the DNN model using a limited range of input data compared to the SWMM was shown. The developed model can be applied to the management of urban catchments with limited access to data and at the stage of urban planning.

The application of integrating comprehensive evaluation and clustering algorithms weighted by maximal information coefficient for urban flood susceptibility

Different sub-regions of Zhengzhou city have various levels of sensitivity to flood due to the impact of urbanization. Thus, an accurate flood sensitivities assessment is a key tool for flood prevention and urban planning and development. To successfully link the urban flood sensitivity assessment with the real flood situation, a method combining clustering algorithm with comprehensive evaluation is presented. The proposed method is not affected by the classification standard of sensitivities levels and has a small and undemanding demand for flood data. First, Maximal Information Coefficient between conditional factors and flood is employed to determine the weight. Then, the different results are obtained by three clustering algorithms. Finally, a four-layer evaluation structure weighted by analytic hierarchy process is established to select the best flood susceptibility map. A case study in the Zhengzhou city, China shows that the positive scale amplification strategy is relatively best and the flood sensitivity of sub-regions in Zhengzhou city should be divided into four levels obtained by K-Means clustering. Hence, it supplies the valuable insights for the urban planning and flood mitigation.