Flood Characteristics Research Articles

Flood risk assessment of coastal cities based on GCW_ISODATA and explainable artificial intelligence methods

Scientific and accurate assessment of risk influencing factors are crucial for flood risk management. This paper aims to propose a new comprehensive framework for flood risk assessment in coastal cities. Firstly, considering the flood characteristics of coastal cities and the impact of floods on urban spatial structure, a flood risk assessment indicator system for coastal cities was established. Secondly, combining game combination weighting and Iterative self-organizing data analysis technique algorithm (GCW_ISODATA) for flood risk assessment. Finally, based on the explainable machine learning techniques, the sensitivity of indicators to flood risk was analyzed. The results indicated that coastal floods are more destructive than rainfall and river floods. Moreover, the indicator weighting and threshold division have a direct impact on the rationality of flood risk, GCW_ISODATA method performs well in Accuracy, F1 score, and AUC, especially with the highest AUC among all methods. Entropy weight method and GCW are significantly superior to Analytic Hierarchy Process (AHP) method, and ISODATA method usually performs better than the K-Means and Natural Break method. Furthermore, the sensitivity of indicators to flood risk reveals that differences in economic, social, and environmental characteristics across regions affect the actual impact of these indicators, leading to the sensitivity of the same indicator to flood risk varies significantly across different regions. It is expected that the framework proposed in this study can be used to explore flood risk impact on other coastal cities.

Estimation of River Flood Damages Among Farmers and Small Business in Developing Countries: Evidence from the Kali Lamong Indonesia

Flooding is a severe disaster in the Kali Lamong river basin (58 Km2). For this reason, detailed flood damage estimates are needed to design an optimal flood management strategy. This study aims to estimate flood damage in the Kali Lamong River Basin. The data collection method used cluster sampling in 15 villages with 870 respondents. We surveyed farmers’ households and micro businesses from October to December 2022 to obtain information on flood damage, flood characteristics, and socio-economics characteristics based on the 27-30 September 2022 flood event. Based on those data, we used regression analysis to develop flood damage functions to know how flood characteristics are related to the damage. Our study found that the economic flood damage in the Kali Lamong river basin is estimated to be US$ 2,3 million for farmer’s households and US$ 10,3 million for micro businesses. This study shows that household flood damages 5,2% more than micro business.

Assessing the characteristics of extreme floods in Nepal

Abstract. This study examines the characteristics (magnitudes, trends, and frequency of occurrences) of extreme floods in Nepal, a country that is at significant risk from floods. Daily discharge data from 1980 to 2015 of three gauging stations (Chisapani of Karnali Basin, Devghat of Narayani Basin, and Chatara of Koshi Basin) were used to assess the largest 1 % of flows, the annual top five high flows, and floods of different return periods (2-, 5-, 10-, 20-, and 100-year). In addition, temporal trend analysis of the flood peaks was carried out using the Mann–Kendall test and Sen's slope estimates. Results show that the magnitudes of the largest 1 % flows range from 6310 to 17 900, from 6967 to 12 100, and from 6080 to 9610 m3 s−1 at Chisapani, Devghat, and Chatara, respectively. The monsoon, especially from mid-June to early September, consistently witnesses over 90 % of 1 % extreme flows, with August registering more than 51 % of these occurrences. July and August combine for 81 % of the top five flow events, predominantly in August. Despite insignificant flow changes at a 95 % confidence level, extreme floods (2-, 5-, 10-, 20-, and 100-year return periods) are concentrated heavily in July and August, with August's second fortnight recording the most flood events. This assessment emphasizes July and August as critical months for extreme floods, aiding Nepalese authorities in planning dynamic resource allocation, disaster response, and effective flood management.

Flood Risk Analysis of Urban Agglomerations in the Yangtze River Basin Under Extreme Precipitation Based on Remote Sensing Technology

Flooding is the most pervasive hydrological disaster globally. This study presents a comprehensive analysis of torrential rain and flood characteristics across three major urban agglomerations (CY, MRYR, and YRD) in the Yangtze River Basin from 1991 to 2020. Utilizing satellite-derived microwave SSM/I data and CHIRPS precipitation datasets, this study examines the impacts of urbanization and climate change on flood risk patterns. The results showed: (1) In 1998, the MRYR had the highest flood risk due to heavy rainfall and poor flood control, but by 2020, risk shifted to the CY with rapid urbanization and more rainfall, while the YRD maintained the lowest risk due to advanced flood control. (2) The relationship between impervious surface area and flood risk varied by region. The CY showed a negative correlation (−0.41), suggesting effective flood mitigation through topography and infrastructure; the MRYR had a slight positive correlation (0.12), indicating increased risks from urban expansion; and the YRD’s weak negative correlation (−0.18) reflected strong flood control systems. This research underscores the imperative of strategic urban planning and effective water resource management to mitigate future flood risks and contributes valuable insights to ongoing efforts in flood disaster prevention and control within the Yangtze River Basin.

A novel flood conditioning factor based on topography for flood susceptibility modeling

Flood is one of the most devastating natural hazards. Employing machine learning models to construct flood susceptibility maps has become a pivotal step for decision-makers in disaster prevention and management. Existing flood conditioning factors inadequately account for regional characteristics of flood in the depiction of topography, potentially leading to an overestimation of flood susceptibility in flat areas. Addressing this gap, this study proposes a novel flood conditioning factor, local convexity factor (LCF), to enhance the accuracy of flood susceptibility modeling. Initially, LCF is computed based on a standard normal Gaussian surface to highlight elevation variations in local terrain. Subsequently, LCF is applied to flood susceptibility modeling using seven machine learning models across four distinct basins. Comparative analysis is conducted between flood susceptibility maps with and without the application of LCF to evaluate its impact on flood susceptibility modeling. The results demonstrate that the proposed LCF can enhance the accuracy of flood susceptibility modeling to varying degrees, across the four basins investigated. The Fujiang basin exhibited the most substantial improvement, with its AUC improved from 0.861 to 0.886, Producer’s Agreement improved from 0.869 to 0.899, and Overall Agreement improved from 0.778 to 0.811. Comparation with hydrodynamic inundation maps shows that particularly in relatively flat terrain areas, flood susceptibility maps incorporating LCF offer more precise delineation between flood-prone and non-flood-prone zones. This research holds potential for widespread application in the prediction of flood susceptibility using machine learning models, providing a novel perspective for enhancing their accuracy

Personal factors influencing emergency evacuation decisions under different flash flood characteristics

Personal factors influencing emergency evacuation decisions under different flash flood characteristics

The Effect of Dam Break Speed on Flood Evolution in a Downstream Reservoir of a Cascade Reservoir System

The dam break flood is one of the potential causes of catastrophic events in cascade hydropower hub groups. Investigating the movement patterns of dam break flooding among reservoir groups under different dam break speeds is crucial for flood prevention and emergency response. In this study, the evolution characteristics of dam break floods were investigated in a cascading reservoir system, focusing on different break speeds of the upstream dam. The results indicate that the dam break speed determines the concavity or convexity of the water level curve changes in the upstream reservoir. Accordingly, dam breaks are classified into three modes: instant dam break, fast dam break, and slow dam break. An approximate critical speed has been identified to differentiate between the fast dam break and slow dam break. Further investigation into the evolution patterns of dam break floods in downstream reservoirs under different break modes was conducted. Correspondingly, the flood peak discharge and peak arrival time of the dam break floods vary differently with break speed under different break modes. Finally, a theoretical analysis for the flood peak discharge at the dam site during gradual dam break at a certain speed was established, which is able to predict the over-dam flood peak discharge in fast and slow dam break modes. This study is based on a combination of laboratory flume experiments and three-dimensional numerical simulations. This study has theoretical significance for the reinforcement of public infrastructure safety and the prevention of natural disasters.

Enhancing precision of flood estimation in EOS-04 SAR imagery: a statistical approach

Floods pose a recurrent and devastating natural disaster in India, and accurate and timely assessment of their extent is crucial for effective disaster management and mitigation efforts. In this regard, this study proposes a novel approach for flood inundation estimation through the statistical analysis of two key geospatial datasets: the Normalized Difference Flood Index (NDFI) and temporal Standard Deviation images. To establish a baseline under ordinary conditions, two products were derived. The first product is the mean image during the non-flood season, and the second is the temporal Standard Deviation image for the entire year. These images are derived from Sentinel-1A Synthetic Aperture Radar (SAR) data using Google Earth Engine (GEE). The NDFI is calculated by comparing the mean image to the image obtained during flood event, allowing the identification of flood-affected areas. The process involves constructing Kernel Density Estimation (KDE) plots for NDFI and Standard Deviation, from which a 95% density ellipse is generated using the covariance matrix and eigenvalues. Multiple SAR scenes from diverse regions in India are analyzed individually, yielding density ellipses for each location. The innovative step lies in synthesizing these individual ellipses into the best fit ellipse through the convex hull method. This best fit ellipse encapsulates the general flood characteristics observed across disparate geographic regions, providing a unified representation. The derived best fit ellipse serves as a powerful tool for generating near-real-time flood inundation maps, with an overall accuracy of over 98.2%. Thus, the fusion of statistical insights and historical SAR information enables rapid near-real-time flood mapping, with enhanced accuracy of flood extent predictions.

Microscopic Experiments to Assess the Macroscopic Sweep Characteristics of Carbon Dioxide Flooding

The Lunnan oilfield in the Tarim Basin, one of China’s major onshore oilfields with substantial geological reserves, faces particular challenges due to the complexity of its reservoir environment and the dispersion of remaining oil. Carbon dioxide, a greenhouse gas, presents an opportunity for enhanced oil recovery (EOR) and geological storage. In this context, the use of carbon dioxide for EOR can simultaneously address environmental concerns and improve oil recovery rates. This study focuses on the TI reservoir in the No. 2 well area of the Lunnan oilfield, employing advanced techniques to analyze the micro- and macro-characteristics of carbon dioxide flooding. Results: From the microscopic point of view, carbon dioxide flooding is mainly miscible with crude oil, which has a strong component exchange effect and can be displaced in the form of full pores, and the microscopic displacement efficiency is close to 100%. Macroscopically, under the combined injection and production of different injected hydrocarbon pore volume multiples (HCPVs), it is injected at the upper and lower layers of the interlayer and produced far away from the lower layer of the interlayer, with a total recovery rate of 52.83%. With the increase in the HCPV, the recovery increased rapidly at first and then slowly, and the HCPV at the demarcation point was 0.5, while the oil production rate increased in a wave-like manner and then decreased rapidly, and the HCPV at the breakthrough point of TI gas was 0.5. However, when the upper and lower layers far away from the interlayer are injected at the same time, the upper and lower layers of the interlayer are produced at the same time, and the total recovery rate can reach 83.02%. With the increase in the HCPV, the recovery rate increases rapidly at first and then slowly, and the HCPV at the turning point is 6.52. The oil production rate increases in a wave-like manner, then decreases rapidly, rises rapidly, and then decreases slowly in a wave-like manner. The HCPV at the breakthrough point of TI gas is 0.63, and the HCPV at the injection–production transition point is 0.63. The total recovery rate of carbon dioxide miscible displacement can reach 88.68% under the condition of separate injection and combined production with different injected hydrocarbon pore volume multiples. With the increase in the HCPV, the recovery increased rapidly at first and then slowly. The HCPV at the demarcation point was 6.5, the oil production rate increased in a wave-like manner, then decreased rapidly, increased rapidly, and then decreased slowly in a wave-like manner. The HCPV at the breakthrough point of TI gas was 0.63, and the HCPV at the injection–production transition point was 6.5. The research results provide data support for the physical reality of the microscopic and macroscopic sweep characteristics of carbon dioxide flooding in the Lunnan oilfield, Tarim Basin.

Miscibility Evaluation for Reinjection of CO2 Flooding Associated Gas in Jilin Oilfield

Abstract In view of large amounts of associated gas produced in the late stage of CO2 flooding in Jilin Oilfield, the production characteristics of CO2 flooding associated gas are divided into three stages (i.e. before injected gas breakthrough, during injected gas breakthrough and after injected gas breakthrough). Taking the formation pressure as miscibility evaluation criteria, the miscibility effect of the produced CO2 flooding associated gas from different stages is evaluated. According to different gas production stages, the reinjection method of mixing the associated gas with pure CO2 under different proportions is proposed to elevate the utilization degree of the produced associated gas. The results indicate that at the very beginning of gas production process, produced gas is mainly composed of CH4 (58%) and C2~C3 (29%) with small amounts of N2 and C4~C6. During this stage, the associated gas can not reach miscibility with the oil by direct reinjection. It needs to be mixed with pure CO2, also the blending ratio of the associated gas should not exceed 6%. In the second stage, the compositions of the produced gas are mainly CH4 (34%) and CO2 (40%). During this stage, the blending ratio should not exceed 11%. At the final stage, the main composition of the produced gas is CO2 (94%). The associated gas can be directly reinjected into the reservoir during this stage.

Flood simulation and impact analysis of Xun River under backwater condition

Abstract The intersection of river channels is easy to cause the backside of the main flow, which greatly influences the change of hydrological conditions of river channels. To study the changing characteristics of river flood under the condition of uplift, a one-dimensional hydrodynamic model was established to simulate the evolution process of river flood under the condition of seeking river uplift. Factors such as distance, height, flood speed, inundation area, inundation depth, and flood speed were selected to analyze the changing rules of the influence of river bullying. The results show that the influence degree of the river following is closely related to the size of the flood. Under the same earthquake magnitude, the backwater distance, backwater height, inundation area, inundation depth, and inundation speed of sub-river flood increase with the increase of flood magnitude, while the flood speed decreases. When the magnitude of the flood is once in 200 years, the flood elements reach the maximum, the reflux distance is 8, 368 m, the reflux height is 3.2 m, and the velocity is reduced by 43.2%. This study can provide relevant data for the flood control work in the Sun River basin and provide a reference for the scheme design of tributary interchanges.

Pore-Scale Mechanism Analysis of Enhanced Oil Recovery by Horizontal Well, Dissolver, Nitrogen, and Steam Combined Flooding in Reducer Systems with Different Viscosities for Heavy Oil Thermal Recovery

Horizontal well, dissolver, nitrogen, and steam (HDNS) combined flooding is mainly applied to shallow and thin heavy oil reservoirs to enhance oil recovery. Due to the lack of pore-scale mechanism studies, it is impossible to clarify the oil displacement mechanism of each slug in the process combination and the influence of their interaction on enhanced oil recovery (EOR). Therefore, in this study, HDNS combined flooding technology was simulated in a two-dimensional visualization microscopic model, and three viscosity reducer systems and multi-cycle combined flooding processes were considered. In combination with an emulsification and viscosity reduction experiment, two-dimensional microscopic multiphase seepage experiments were carried out to compare the dynamic seepage law and microscopic occurrence state of multiphase fluids in different systems. The results showed that the ability of three viscosity reducers to improve viscosity reduction efficiency in HDNS combined flooding was A > B > C, and their contributions to the recovery reached 65%, 41%, and 30%, respectively. In the system where a high viscosity reduction efficiency was shown by the viscosity reducer, the enhancements of both sweeping efficiency and displacement efficiency were primarily influenced by the viscosity reducer flooding. Steam flooding collaborated to improve displacement efficiency. The thermal insulation characteristics of N2 flooding may not provide a gain effect. In the system where a low viscosity reduction efficiency was shown by the viscosity reducer, the steam flooding was more important, contributing to 57% of the sweeping efficiency. Nitrogen was helpful for expanding the sweep area of the subsequent steam and viscosity reducer, and the gain effect of the thermal insulation steam chamber significantly improved the displacement efficiency of the subsequent steam flooding by 25%. The interaction of each slug in HDNS combined flooding resulted in the additive effect of increasing production. In actual production, it is necessary to optimize the process and screen the viscosity reducer according to the actual conditions of the reservoir and the characteristics of different viscosity reducers.

Investigating the influence of urban morphology on pluvial flooding: Insights from urban catchments in England (UK)

As climate change intensifies, cities globally are experiencing more severe rainfall and frequent pluvial floods. Urban expansion is altering the permeability of the land, thus increasing the risk of flooding. This study investigates the impact of urban morphology on pluvial floodwater distribution in 15 urban catchments across England, UK, to provide an analysis of how urban morphology influences flood magnitude. Using a cellular automata-based model, pluvial flood simulations were conducted for catchments characterized by diverse urban morphologies. Then a series of machine learning models were adopted to reveal the relationships between the morphological characteristics of urban configurations (e.g., building footprints, impervious surfaces, street network, topography) and pluvial flooding. These models were used to identify and quantify the effects of key urban morphological indicators on pluvial flooding. The results indicate that, although the total area of impervious surfaces plays the most significant role in floodwater distribution, the edge density (ED) of building footprints and impervious surfaces also influences this process. Synthetic experiments with an exemplary urban fabric show that decreasing “ED of building footprint” and increasing “ED of impervious surface” can mitigate flood volume by up to 6.3 % at 100 % drainage efficiency and 7.8 % at 50 % efficiency. The results of this study are anticipated to aid urban planners and policymakers in developing strategies for implementing flood-resilient cities.

Identification of influencing factors and risk assessment of underground space flooding in the mountain city

As a consequence of the process of urbanization, there is an increasing reliance on and exploitation of underground spaces. Moreover, these areas are susceptible to urban waterlogging, which can be difficult to drain, potentially leading to significant risks. At present, there are few simulations in hydrodynamic models that consider underground spaces. The risk of inundation of underground spaces remains an uncertain phenomenon. This study conceptualized underground spaces as storage areas, with the aim of simulating the spatial distribution of flood risk for different return periods. We studied the flooding processes and characteristics of three underground space types. The results indicate that the flood evolution simulated by the simplified modeling method that incorporates underground space is more aligned with the actual situation. To further analyze the risk distribution of urban waterlogging, we choose the XGBoost model and SHapley Additive exPlanations (SHAP) method to assess the influence of specific indicators on the occurrence of urban waterlogging in mountainous regions. Our analysis revealed that key factors include density of road (DR), flow velocity (V), density of pipes (DP), elevation (DEM), and inundation height (H), with contribution rates of 23.13 %, 17.06 %, 14.67 %, 10.12 %, and 9.04 %, respectively. This analysis clarified the risk of underground inundation within the study area, which can serve as a reference point for the formulation of flood control and drainage strategies.

Flood risk assessment of a small river with limited available data

Flood risk modeling of small watercourses is challenging when only limited input data are available. Therefore, this study assessed the flood characteristics of a small river (Tarna River: entire watershed-C, upper-VS, middle-TMS, and lower section-TOS) from 1990 to 2019. The assessment focused on modeling, model calibration, and validation using feature event-based time-series data in data-scarce environments. We showed that since the 2000s, the number of high-water levels above 250 cm, and the frequency of three flood types had increased. Flood simulation results showed the largest flooded area in the TMS section, followed by the VS, and then the TOS. The outcomes from the VS, TMS, and TOS sections did not exhibit superior performance compared to the C area. Models performed well for larger flood events, with Kling Gupta Efficiency corresponding well to NRMSE and Nash-Sutcliffe efficiency metrics. Accordingly, flood events characterized by the longest duration and high-water levels yielded outstanding results across all areas, followed by moderate flood events with good accuracy. Normal water level events exhibited significant deviations from the reference across all sections. In summary, despite the event-based modeling challenges in data-limited environments, such models can still mitigate potential flood events and improve decision-making processes.

Impacts of climate and land-use change on flood events with different return periods in a mountainous watershed of North China

Study regionLuanhe River Basin in North China Study focusGlobal warming exacerbates hydrological cycles, leading to increased intensity and frequency of floods. Regional studies on the impacts of climate change and land cover changes on flood events in recent decades are particularly important for flood prevention and risk assessment. In this study, the Peak Over Threshold- Generalized Pareto Distribution (POT-GPD) method is used to fit extreme runoff values in the Luanhe River Basin from 1961 to 2018 to identify floods with different return periods. Six flood behavior indicators are adopted to analyze the flood characteristics and extreme climate indices are also used for breakpoint testing and trend analysis. Various scenarios are established as the input of the SWAT model to study the impacts of climate change and land-use change on the flood events. New hydrological insights for the regionThe results indicate that the 2-year and 20-year return period floods are primarily rapid-rise single-peak floods, dominated by short-duration intense precipitation. The 5-year and 10-year return period floods are mainly multi-peak floods, primarily caused by long-duration precipitation. Floods in the Luanhe River Basin are primarily influenced by climate change, however, flood events do not directly correspond to the return periods of precipitation events. Forest is the most effective land-use type for flood retention. It could reduce flood magnitudes, especially for the 20-year return period floods. However, increasing forest cover does not effectively reduce the occurrence of the 5-year or 10-year floods, i.e. multi-peak floods. With more extreme rainfall events triggered by future climate change, particular attention should be paid to long-duration precipitation events, which may result in more severe disasters and economic losses.

The Correlation between Water–Sediment Index and Floodplain Transverse Slope Based on Wavelet Analysis

The floodplain transverse slope is a significant parameter reflecting the degree of development of a secondary suspended river, as well as a crucial index of the flood risk in the river channel. Clarifying the factors that influence the evolution of the floodplain transverse slope has always been a hot and difficult topic for researchers working on the Yellow River management. We took the severe section of the secondary suspended river from Dongbatou to Gaocun in the lower Yellow River as the research object, selecting the annual runoff, annual sediment load, annual sediment coefficient, and the intensity of flood-season flow scouring at the Huayuankou station in the downstream as the water–sediment indexes. The correlation between different water–sediment indexes and the floodplain transverse slope under three modes: interannual, flood season, and flood-season overbank was studied through methods such as cross-wavelet transform and wavelet coherence analysis. The results showed that under the three modes, the annual sediment load and annual sediment coefficient had a high correlation with the evolution cycle of the transverse slope, followed by the intensity of flood-season flow scouring, and the annual runoff had the lowest correlation. Meanwhile, the change in the transverse slope had a good correlation with the flood-season overbank mode, indicating there was a high similarity between the water–sediment characteristics of floodplain flooding and the evolution cycle of the transverse slope; that is, the change in the transverse slope is greatly influenced by floodplain flooding events.

Flood damage models and flood damage factors in a data-scarce river basin, Nigeria

ABSTRACT To evaluate the effectiveness of various adaptation strategies, it is critical to develop a flood damage function that considers building characteristics for future risk evaluation. This study analysed the contributions of some important flood and household characteristics in flood damage processes and developed a new, more accurate, multi-variable flood loss model for flood damage assessment for residential buildings and contents, considering different house types in Ogun River Basin, Nigeria. A detailed household questionnaire survey was conducted to collect data (149 samples). Depth-damage functions were developed to estimate direct flood damage to building structures and contents using the nonlinear logistic and power (concave) models, and the model parameters were estimated using the Gauss–Newton algorithm. The performance of all flood damage regression analysis functions for building structures and building contents under different construction materials and basement floors was evaluated by comparing predicted damages and after-flood survey damages, which demonstrated acceptable comparativeness. In the event of a flood disaster in the area, the models can be used to forecast future damages and devise risk-mitigation strategies. The study’s findings will aid in developing flood protection adaptation measures.

A simple method to automatically remove artificial terrain from airborne LiDAR DTMs in plain areas

Alluvial plains are highly vulnerable to floods and ground disasters. Recent rapid urbanization and climate change have heightened disaster risks in urban areas. Geomorphological maps, crucial for estimating disaster risks, delineate landform boundaries based on patterns of concave breaks of slope and micro-landforms. However, enhancing the accuracy of such maps requires data extraction methods capable of capturing these features with greater precision. Traditional topographic measurements derived from adjacent elevation points in airborne light detection and ranging (LiDAR) digital terrain models (DTMs) fail to accurately represent slope variations on low ground surfaces due to the inclusion of numerous noise-like artificial terrain features. Consequently, analyzing natural terrain becomes challenging. To address this issue, our study devised a method to automatically identify and eliminate noise-like artificial terrain from LiDAR DTMs. We achieved this by removing major artificial terrain features from land-use vector data, creating an edge-preserving smooth DTM, and selectively removing and interpolating only those areas where were large differences between the smooth DTM and the LiDAR DTM. This method minimizes the interpolation of artificial terrain and quotes the LiDAR DTM for other areas, thereby minimizing the data quality loss. It is possible to identify and demarcate topographic boundaries in plains with a longitudinal gradient of approximately 1% or less at a high resolution, which can be used to investigate the relationship between flood and ground characteristics and landform volume in the plains. This method can be easily processed using only QGIS and free open data. This approach enhances the precision of disaster risk estimation and facilitates more effective urban planning in vulnerable areas.

Improving flood forecast accuracy based on explainable convolutional neural network by Grad-CAM method

The advent of deep learning techniques has shown promising improvement in flood forecasting accuracy which are crucial for operating reservoir and mitigating flood damages. However, the practical application of typical deep learning models, such as the Long Short-Term Memory network (LSTM), is hindered by their high complexity, computational cost, and lack of interpretability. This study introduces an Explainable Convolutional Neural Network (ECNN) model that can effectively capture inter-variable relationships and temporal dynamics to address these challenges. The proposed ECNN model was applied to the Lushui basin in China and compared with benchmark models. The interpretability of the ECNN model was analyzed using the Gradient-weighted Class Activation Mapping (Grad-CAM) method. Our findings demonstrated that the ECNN model exhibits exceptional performance in flood forecast tasks, particularly for long-term forecast horizons. Furthermore, precipitation and inflow play a decisive role in flood forecasting with the watershed-specific flood characteristics. The ECNN model is highly promising in real-time flood forecasting with ingenious parallel structure, remarkable accuracy, and interpretability.