Flood Simulation Research Articles

Rates of bedrock canyon incision by megafloods, Channeled Scabland, eastern Washington, USA

Pleistocene outburst floods from the drainage of glacial Lake Missoula carved bedrock canyons into the Columbia Plateau in eastern Washington, USA, forming the Channeled Scabland. However, rates of bedrock incision by outburst floods are largely unconstrained, which hinders the ability to link flood hydrology with landscape evolution in the Channeled Scabland and other flood-carved landscapes. We used long profiles of hanging tributaries to reconstruct the pre-flood topography of the two largest Channeled Scabland canyons, upper Grand Coulee and Moses Coulee, and a smaller flood-eroded channel, Wilson Creek. The topographic reconstruction indicates floods eroded 67.8 km3, 14.5 km3, and 1.6 km3 of rock from upper Grand Coulee, Moses Coulee, and Wilson Creek, respectively, which corresponds to an average incision depth of 169 m, 56 m, and 10 m in each flood route. We simulated flood discharge over the reconstructed, pre-flood topography and found that high-water evidence was emplaced in each of these channels by flow discharges of 3.1 × 106 m3 s−1, 0.65−0.9 × 106 m3 s−1, and 0.65−0.9 × 106 m3 s−1, respectively. These discharges are a fraction of those predicted under the assumption that post-flood topography was filled to high-water marks for Grand and Moses Coulees. However, both methods yield similar results for Wilson Creek, where there was less erosion. Sediment transport rates based on these discharges imply that the largest canyons could have formed in only about six or fewer floods, based on the time required to transport the eroded rock from each canyon, with associated rates of knickpoint propagation on the order of several km per day. Overall, our results indicate that a small number of outburst floods, with discharges much lower than commonly assumed, can cause extensive erosion and canyon formation in fractured bedrock.

Does a convection-permitting regional climate model bring new perspectives on the projection of Mediterranean floods?

Abstract. Floods are the primary natural hazard in the French Mediterranean area, causing damages and fatalities every year. These floods are triggered by heavy precipitation events (HPEs) characterized by limited temporal and spatial extents. A new generation of regional climate models at the kilometer scale have been developed, allowing an explicit representation of deep convection and improved simulations of local-scale phenomena such as HPEs. Convection-permitting regional climate models (CPMs) have been scarcely used in hydrological impact studies, and future projections of Mediterranean floods remain uncertain with regional climate models (RCMs). In this paper, we use the CNRM-AROME CPM (2.5 km) and its driving CNRM-ALADIN RCM (12 km) at the hourly timescale to simulate floods over the Gardon d'Anduze catchment located in the French Mediterranean region. Climate simulations are bias-corrected with the CDF-t method. Two hydrological models, a lumped and conceptual model (GR5H) and a process-based distributed model (CREST), forced with historical and future climate simulations from the CPM and from the RCM, have been used. The CPM model confirms its ability to better reproduce extreme hourly rainfall compared to the RCM. This added value is propagated on flood simulation with a better reproduction of flood peaks. Future projections are consistent between the hydrological models but differ between the two climate models. Using the CNRM-ALADIN RCM, the magnitude of all floods is projected to increase. With the CNRM-AROME CPM, a threshold effect is found: the magnitude of the largest floods is expected to intensify, while the magnitude of the less severe floods is expected to decrease. In addition, different flood event characteristics indicate that floods are expected to become flashier in a warmer climate, with shorter lag time between rainfall and runoff peak and a smaller contribution of base flow, regardless of the model. This study is a first step for impact studies driven by CPMs over the Mediterranean.

Advanced floodplain mapping: HEC-RAS and ArcGIS pro application on Swat River

AbstractFloods stand out as one of the most devastating environmental threats worldwide resulting in the tragic loss of human lives and significant damage to our essential infrastructure. This study focuses on creating floodplain maps for the two important reaches of Swat River in the Khyber Pakhtunkhwa Province of Pakistan, namely, Kalam–Khwazakhela and Khwazakhela–Chakdara reaches. The Advanced Land Observing Satellite Phased Array Type L-band Synthetic Aperture Radar (ALOS PALSAR) 12.5 m Digital Elevation Model (DEM) data has been used for this purpose. Furthermore, the sensitivity of the flood model was assessed for the Flood 2022 event, considering simulated flood depth, extent, and velocity in relation to various terrains derived from the 12.5-m ALOS PALSAR data. To estimate extreme flows for different return periods (2, 5, 10, 25, 50, and 100 years), the Log Normal (LN), Log Pearson III (LP3), and Generalized Extreme Value (GEV) distributions were employed for the frequency analysis. The GEV distribution turned out to be the best fit for modeling the Swat River for both the Chakdara and Khwazakhela gauge stations. To predict flood levels for the peak floods identified through frequency analysis and for the specific return periods the Hydrologic Engineering Center's River Analysis System (HEC-RAS) 2D simulations were performed. Subsequently the outcomes of this model were used to create floodplain maps using Geographic Information System Professional (ArcGIS Pro) software customized for various return periods. The analysis revealed a linear increase in flood inundation area with longer return periods. The floodplain maps developed hold significant importance for the governing authorities of the Swat region. These maps serve as essential tools for implementing proactive measures against potential infrastructure damage, thereby protecting against economic losses and enhancing public safety. This research effectively combines hydrological modeling and geospatial technology offering practical solutions for managing flood risks. It serves as a valuable guide for making well-informed decisions and promoting sustainable development in flood-prone areas.

Analysis of the Distribution Pattern of Remaining Oil and Development Potential after Weak Gel Flooding in the Offshore LD Oilfield.

The LD oilfield is one of the representative offshore oilfields. After weak gel flooding, the recovery rate is significantly improved. However, the oilfield is then in a medium- to high-water content stage, presenting a complex distribution of the remaining oil. The measures for further enhanced oil recovery (EOR) are uncertain. As a result, it is necessary to clarify the distribution pattern and development potential of the remaining oil during the high-water content period after weak gel flooding. In this study, an online nuclear magnetic resonance (NMR) oil displacement experiment and microscopic oil displacement experiment were conducted, and the mechanisms of weak gel flooding and the distribution pattern of the remaining oil were clarified in the LD oilfield. Additionally, high-multiple water flooding and numerical simulation experiments were conducted to analyze the development potential after weak gel flooding. The results show that the effect of weak gel flooding was more significant in the core of 1500 mD, with an increase in oil recovery of 9% compared to 500 mD. At a permeability of 500 mD, the degree of crude oil mobilization in micropores and small pores caused by weak gel flooding was improved by 29.64% and 23.48%, respectively, compared with water flooding. At 1500 mD, the degree of crude oil mobilization in small pores caused by weak gel flooding was increased by 37.79% compared to water flooding. After weak gel flooding, the remaining oil was primarily distributed in medium and large pores. Microscopically, the remaining oil was dominated by cluster residual oil, accounting for 16.49%, followed by columnar, membranous, and blind-end residual oil. High multiple water flooding experiments demonstrated that weak gel flooding could significantly reduce development time. The ultimate oil recovery efficiency of 500 mD and 1500 mD reached 71.85% and 80.69%, respectively. Numerical simulation results show that the ultimate oil recovery efficiency increased from 62.04% to 71.3% after weak gel flooding. This indicated that the LD oilfield still had certain development potential after weak gel flooding. The subsequent direction for enhanced oil recovery focuses mainly on mobilizing oil in medium pores or clustered remaining oil. This will play a crucial role in further exploring methods for utilizing the remaining oil and increasing the recovery rate.

Compilation method of a catalogue of reasonable worst-case rainfall series for flash flood simulations of short, convective rainstorms

In this study, a categorisation method was developed and applied to identify, select and rank reasonable worst-case rainfall storm events with time-series resolution of one minute and duration of up to one hour. The method yields a catalogue of 24 reasonable worst-case rainfall series, i.e. potentially catastrophic events that could be used for safe-to-fail flash flood model scenarios, thus encouraging the testing of critical infrastructure and pedestrian safety for people trapped in floods. The identified 24 events were shown to be considerably more extreme than corresponding design storms, with a return interval of 100 years. In the selection of the extremes, special care was taken to consider different hazard attributes, which are associated with intrinsically different observed hyetograph shapes, including rectangular, short and long triangle, and double peak shape categories. Hydrodynamic simulations were carried out using the 2D robust shallow water model hms++ to quantify the flooding impact of the 24 rainfall scenarios in a small urbanised catchment, which enabled the ranking of the scenarios based on different hazard attributes. Overall, the rectangular and long-triangle events showed the highest total discharge, and the rectangular events comprise both the longest duration for hazard velocity and deep flooding. The long-triangle events presented the highest peak discharge and peak water depths. The short-triangle ones showed the highest flashiness (shortest time to peak), whereas the double peak events yielded the longest time to peak. The ranking emphasised the importance of the hyetograph shape characteristics of extreme storms and which needs to be taken into account in flash flood management. The novel categorisation method is easily transferable to other regions, if high-resolution rain series are available.

Accelerating flash flood simulations: An efficient GPU implementation for a slim shallow water solver

Flash floods are an increasing hazard to human infrastructure and life. Effective disaster management and mitigation require accurate and fast predictions for decision-making. The Flood Inundation Parallel Computation (FIP) software presented in this paper shows how to fully exploit the tremendous computational capabilities of graphics processing units (GPUs) to accelerate shallow water solvers beyond the current state-of-the-art. The time efficient explicit shallow water scheme for structured grids RMG, is introduced and implemented in FIP. The optimized GPU implementation of FIP achieves balanced memory- and instruction throughputs of up to 80%. Validation and performance tests using laboratory and historical flood cases are presented that demonstrate the accuracy and effectiveness of FIP in predicting flood inundation. Our implementation achieves a fourfold speedup in comparison to state-of-the-art approaches. It enables faster-than real-time simulations of areas of over 600 km2 at 1 m resolution on consumer-grade GPUs.

Fast urban inundation simulation with RIM2D for flood risk assessment and forecasting

IntroductionUrban pluvial flooding is a growing concern worldwide as consequence of rising urban population and climate change induced increases in heavy rainfall. Easy-to-implement and fast simulation tools are needed to cope with this challenge.MethodsThis study describes the development of the parsimonious, GPU-accelerated hydraulic model RIM2D for urban pluvial flood simulations. This is achieved by considering the built-up urban area as flow obstacles, and by introducing capacity-based approaches to consider urban drainage by infiltration on pervious surfaces and sewer drainage from roofs and sealed surfaces. The model performance was analyzed by simulating 8 heavy rainfall events in a test area in the city of Dresden, Germany. For these events detailed discharge measurements of sewer discharge are available, providing a unique dataset for evaluating the sewer drainage simulation, which is of high importance for realistic pluvial inundation simulations in urban areas.Results and discussionWe show that the model simulates the temporal dynamics of the sewer discharge and the sewer volume within acceptable ranges. Moreover, the erratic variation of the simulated to measured sewer discharge suggests that the deviations from the measurements are caused by the precipitation input rather than the model simplifications. We conclude that RIM2D is a valid tool for urban inundation simulation. Its short simulation runtimes allow probabilistic flood risk assessments and operational flood forecasts.

Study on the Influence of Temporal and Spatial Resolution of Rainfall Data on Watershed Flood Simulation Performance

Study on the Influence of Temporal and Spatial Resolution of Rainfall Data on Watershed Flood Simulation Performance

Hydrologic Evaluation of the Global Precipitation Measurement Mission over the U.S.: Effect of Spatial and Temporal Scales

This research focuses on the effects of spatio-temporal resolutions of global satellite precipitation on simulated flood events characteristics. The analysis is carried out by spatially, temporally, and spatiotemporally upscaling fine-scale precipitation forcings from the ground-radar-based Multi-Radar Multi-Sensor system (MRMS) to the resolution of the Integrated Multi-satellitE Retrievals for GPM (IMERG) satellite precipitation product. These upscaled products were then run through the Ensemble Framework for Flash Flood Forecasting hydrologic modeling framework over the U.S. to assess how the different spatial and temporal resolutions impact the simulated outputs of flood event magnitude, duration, and timing. It was found that the quantitative uncertainties generated by the IMERG Early (IMERG-E) simulations overpowered those associated with resolution across all flood characteristics. When compared to native MRMS simulations, however, the upscaled precipitation estimates tended to underestimate peak discharges and event durations, associated with distinct negative water balance errors. It was also found that the upscaled precipitation simulations exhibited a scaling effect with regards to error contribution of peak discharge, with higher contributions at smaller basin sizes and decreasing contributions as basin sizes increased. Looking at how upscaling can affect the modeled output desired allows for improved understanding of which resolutions lead to the greatest changes in simulation accuracy, increasing the potential global use and utility of current or future satellite products, especially in regions where high-resolution ground radars are sparse.

Quick large-scale spatiotemporal flood inundation computation using integrated Encoder-Decoder LSTM with time distributed spatial output models

Accurate spatiotemporal flood simulations are essential for making informed decisions regarding flood release in affected regions, such as flood detention areas. Traditional spatiotemporal flood simulation approach uses partial differential equation (PDE) models (or physics-based models), which need high computational time. Although many machine learning (ML) models for inundation are increasingly being used to emulate the PDE models to address this issue, utilizing conventional ML models to achieve large-scale spatiotemporal flood prediction (i.e., simulation output in tens of thousands of grids and time steps over the whole flood event) remains a significant challenge. Therefore, we developed a new inundation model (IM) using encoder-decoder long short-term memory (ED-LSTM) with Time Distributed Spatial output model (ED-LSTM-TDS) that can acquire accurate spatially distributed flood information more rapidly. In the new IM framework, each ED-LSTM-TDS is built to simultaneously generate output for multiple (K) cell grids and multiple ED-LSTM-TDS models are built for prediction at all grids of entire PDE model. This study is the first of its kind to employ the ED-LSTM-TDS method to address spatiotemporal flood inundation simulation problems for flood detention areas. A 1994 km2 flood detention area in northeastern China was used as a case study. ED-LSTM-TDS exhibited better performance in predicting flood characteristics (e.g., water depth, velocity) than alternative methods, including ordinary LSTM, artificial neural network (ANN), and multiple linear regression (MLR). In addition, we investigated the trade-off relationship between the accuracy of flood characteristic prediction and the computation time of the proposed model by considering different numbers of K cell grids in each ED-LSTM-TDS model. The final proposed inundation model could accurately predict the spatiotemporal flood characteristics within 1.5 min to acquire the same information that required approximately 1 h by the PDE models. Such rapid and accurate prediction by the proposed model is important for evacuation planning, upstream flood control decisions and flood damage reduction.

An efficient 2-D flood inundation modelling based on a data-driven approach

Study regionThe plain areas of Yilan County in Taiwan. Study focusThe present study employed hourly precipitation and inundation simulation data from a 5-km resolution HiRAM-WRF (High-Resolution Atmospheric Model-Weather Research and Forecasting) downscaled model and a 40-m resolution SOBEK model to construct a data-driven two-dimensional flood inundation model (D2DFIM) based on the support vector regression (SVR) approach. New hydrological insights for the regionThe dataset comprised 149 rainfall events and corresponding inundation events. We extracted the inundated and interpolated rainfall grid points distributed across Yilan County, Taiwan, from the SOBEK and HiRAM-WRF models. The D2DFIM (data-driven two-dimensional flood inundation model) utilized the parallel computing approach in the training phase, saving 90% of the elapsed time. The comparisons between the data-driven and physical-based models indicate that the D2DFIM could perform better in flood simulation if the input features comprised previous inundation depth and the last three time steps of rainfall information. Regarding the minimal required number of training samples, we found almost no significant difference in the flooding simulation once the training samples equalled or exceeded 50. Additionally, the elapsed time of D2DFIM is 9–10 times faster than that of SOBEK. We validated the well-developed D2DFIM using three historical inundated events induced by typhoons in Yilan County; the flood extents derived from the D2DFIM agreed well with the survey results.

Flood risk assessment, a case study in an arid environment of Southeast Morocco

Abstract Since the early 1980s, several regions of Morocco have experienced alternating series of wet and dry periods related to climate change, with significant consequences for the country’s economy. This study aims to map flood hazards by using a hydraulic modeling approach, with a specific focus on the Oued (river) Zaouiat Sidi Hamza passing through Zaouiat Sidi Hamza township belonging to the High Ziz watershed (province of Midelt, region of Drâa-Tafilalet, Morocco). Hydraulic modeling and flood simulation were carried out followed by mapping of the flood hazard. The Oued Zaouiat Sidi Hamza River can cause torrential flooding issues, particularly during periods of heavy storms, impacting agricultural land and certain buildings on both river banks. All the cultivated areas developed alongside the river are at risk of flooding. Moreover, undermining and erosion of the river’s banks lead to rising high-water levels and flooding agricultural terraces. The use of geographic information systems and hydraulic modeling tools is essential to gain a deeper understanding of this situation. Also, the mapping of lineaments and fractures in the study area, carried out by satellite image processing, helped to quantify the directional influence of these lineaments on the flow of the hydrographic network, on the orientation of the circulation of surface water, and on the storage of groundwater. According to this study, the most vulnerable areas to flooding and erosion are those that receive the most water in the basin and have steep slopes. Decision-makers can use the results to produce development plans tailored to combat these risks and raise awareness among the population concerned, improving water resource evaluation and forecasting, as well as improving agricultural production.

A novel efficient and robust treatment of the friction source term in 2D shallow water inundation models

The technological advancements of the last few decades have fostered the use of two-dimensional (2D) Shallow water Equations (SWE) flood simulations not only in academic research but also in practical real-world applications and territorial planning. The evaluation of flow resistance due to friction is crucial as it plays a relevant role in a variety of flow conditions. However, problems arise in dam-break and overland flow applications involving very small water depth because the time scale connected to the friction term may be much smaller than the hydrodynamic time scale. To cope with this issue, known as source term stiffness, an implicit treatment of the friction term is frequently adopted in Finite Volume (FV) numerical schemes, but there are cases (null roughness coefficient or null flow velocity) where a division by zero may occur during calculations, leading to a crash of the algorithm. Herein, we propose a reformulation of the implicit friction term that is general, more stable and computationally efficient (with a speed-up of approximately 3 times for the routine running the friction computation) than currently available approaches. The novel implicit method allows to manage special conditions with null roughness or null discharge, without resorting to ad-hoc thresholds, and can be easily implemented in existing schemes with pointwise friction treatment. The novel friction approach is implemented in a purposely simple FV numerical scheme (hydrostatic reconstruction to account for the bed slope terms, first-order accuracy in time and space) - to better focus on the friction term treatment - and it is validated against analytical, synthetic, laboratory and real-world case studies, showing promising capabilities.

Many-Objective Hierarchical Pre-Release Flood Operation Rule Considering Forecast Uncertainty

Flood control operation of cascade reservoirs is an important technology to reduce flood disasters and increase economic benefits. Flood forecast information can help reservoir managers make better use of flood resources and reduce flood risks. In this paper, a hierarchical pre-release flood operation rule considering the flood forecast and its uncertainty information is proposed for real-time flood control. A many-objective optimization model considering the cascade reservoir’s power generation objective, flood control objective, and navigation objective is established. Then, a region search evolutionary algorithm is applied to optimize the many-objective optimization model in a real-world case study upstream of the Yangtze River basin. The optimization experimental results show that the region search evolutionary algorithm can balance convergence and diversity well, and the HV value is 40% higher than the MOEA/D algorithm. The simulation flood control results of cascade reservoirs upstream of the Yangtze River demonstrate that the optimized flood control rule can increase the average multi-year power generation of cascade reservoirs by a maximum of 27.72 × 108 kWh under the condition of flood control safety. The rules proposed in this paper utilize flood resources by identifying runoff forecast information, and pre-release to the flood limit level 145 m before the big flood occurs, so as to ensure the safety downstream and the dam’s own flood control and provide reliable decision support for reservoir managers.

An Investigation of Gas-Fingering Behavior during CO2 Flooding in Acid Stimulation Formations

Summary CO2 flooding is emerging as a pivotal technique used extensively for carbon capture, utilization, and storage (CCUS) strategies. Acid stimulation is one common technique widely used to improve well-formation connectivity by creating wormholes. This work is motivated to investigate the gas-fingering behavior induced by acid stimulation during CO2 flooding. We present an integrated simulation framework to couple the acid stimulation and CO2 flooding processes, in which the two-scale continuum model is used to model the development of wormhole dissolution patterns. Then, sensitivity case simulations are conducted through the equation of state (EOS)–based compositional model to further analyze the CO2 fingering behavior in acid stimulation formations separately under immiscible and miscible conditions. Results demonstrate that for acid stimulation, the typical dissolution patterns and the optimal acid injection rate corresponding to the minimum acid breakthrough volume observed in the laboratory are prevalent in field-scale simulations. For CO2 flooding simulation, the dissolution patterns trigger CO2 fingering (bypassing due to the high conductivity of wormholes) in the stimulated region, and a lateral boundary effect eliminating fingers exerts its influence over the system through transverse mixing. The optimal acid injection rate varies when the focus of interest changes from the minimum acid breakthrough volume to CO2 flooding performance. The best CO2 flooding performance is always observed in uniform dissolution, and the dissolution patterns have a greater influence on the performance under miscible conditions. This work provides technical and theoretical support for the practical application of acid stimulation and CO2 flooding.

Modeling Land Use Transformations and Flood Hazard on Ibaraki’s Coastal in 2030: A Scenario-Based Approach Amid Population Fluctuations

Coastal areas, influenced by human activity and natural factors, face major environmental shifts, including climate-induced flood risks. This highlights the importance of forecasting coastal land use for effective flood defense and ecological conservation. Japan’s distinct demographic path necessitates flexible strategies for managing its urban development. The study examines the Ibaraki Coastal region to analyze the impacts of land-use changes in 2030, predicting and evaluating future floods from intensified high tides and waves in scenario-based forecasts. The future roughness map is derived from projected land-use changes, and we utilize this information in DioVISTA 3.5.0 software to simulate flood scenarios. Finally, we analyzed the overlap between simulated floods and each land-use category. The results indicate since 2020, built-up areas have increased by 52.37 sq. km (39%). In scenarios of constant or shrinking urban areas, grassland increased by 28.54 sq. km (42%), and urban land cover decreased by 7.47 sq. km (5.6%) over ten years. Our research examines two separate peaks in water levels associated with urban flooding. Using 2030 land use maps and a peak height of 4 m, which is the lower limit of the maximum run-up height due to storm surge expected in the study area, 4.71 sq. km of residential areas flooded in the urban growth scenario, compared to 4.01 sq. km in the stagnant scenario and 3.96 sq. km in the shrinkage scenario. With the upper limit of 7.2 m, which is the extreme case in most of the study area, these areas increased to 49.91 sq. km, 42.52 sq. km, and 42.31 sq. km, respectively. The simulation highlights future flood-prone urban areas for each scenario, guiding targeted flood prevention efforts.

Effects of runoff generation methods and simulation time steps on flood simulation: a case study in Liulin experimental watershed

Effects of runoff generation methods and simulation time steps on flood simulation: a case study in Liulin experimental watershed

Effects of urban drainage inlet layout on surface flood dynamics and discharge

Drainage inlets, the primary conduits for surface floodwater into underground facilities, significantly influence the performance of a drainage system. This study employs a two-dimensional hydrodynamic model to investigate how drainage inlet distribution affects urban flood dynamics and drainage flow discharge. We first validate the model with dataset obtained in a dedicated experiment observing urban surface and drainage flows. We identified two common layouts of drainage inlets, namely straight-line distribution (SLD) and surround type distribution (STD), and based on the two layouts, we simulated detailed surface flood dynamics and drainage discharge under various urban flood conditions and inlet spacing scenarios. The results indicate that SLD exhibits higher drainage capacity in shorter time and greater total drainage volume under urban peripheral flood condition, while STD consistently outperforms in draining urban surface flooding induced by intense rainfall. It is also found that inlet spacing plays a crucial role in affecting drainage efficiency. To be specific, both layouts with medium spacing (l0 = 0.53 for SLD and 0.93 for STD) near flood-prone areas perform better when a sudden-onset peripheral flooding occurred; however, in the case of rainfall-induced flooding, maximizing inlet spacing and strategically distributing them across urban streets proves to be a more effective strategy for draining floodwater. Individual inlets near the ends of SLD and inlets at far-corners of STD shoulder higher drainage loads. This study provides valuable insights for optimizing inlet layout strategies under various urban flood conditions and improving floodwater collection.

Research on Urban Storm Flood Simulation by Coupling K-means Machine Learning Algorithm and GIS Spatial Analysis Technology into SWMM Model

Research on Urban Storm Flood Simulation by Coupling K-means Machine Learning Algorithm and GIS Spatial Analysis Technology into SWMM Model

Machine Learning-Based Flood Forecasting System for Window Cliffs State Natural Area, Tennessee

The prevalence of unforeseen floods has heightened the need for more accurate flood simulation and forecasting models. Even though forecast stations are expanding across the United States, the coverage is usually limited to major rivers and urban areas. Most rural and sub-urban areas, including recreational areas such as the Window Cliffs State Natural Area, do not have such forecast stations and as such, are prone to the dire effects of unforeseen flooding. In this study, four machine learning model architectures were developed based on the long short-term memory, random forest, and support vector regression techniques to forecast water depths at the Window Cliffs State Natural Area, located within the Cane Creek watershed in Putnam County, Tennessee. Historic upstream and downstream water levels and absolute pressure were used to forecast the future water levels downstream of the Cane Creek watershed. The models were tested with lead times of 3, 4, 5, and 6 h, revealing that the model performances reduced with an increase in lead time. Even though the models yielded low errors of 0.063–0.368 ft MAE, there was an apparent delay in predicting the peak water depths. However, including rainfall data in the forecast showed a promising improvement in the models’ performance. Tests conducted on the Cumberland River in Tennessee showed a promising improvement in model performance when trained with larger data.