Water Level Forecasts Research Articles

Flood Stage Forecasting at the Gurye-Gyo Station in Sumjin River Using LSTM-Based Deep Learning Models

Instances of flood damage caused by extreme storm rainfall due to climate change and variability have been showing an increasing trend. Particularly, a flood forecasting and warning system has been recognized as an important nonstructural measure for flood damage reduction, including loss of life. Flood forecasting and warning have been performed by the forecasts of flood discharge and flood stage using the physically based rainfall-runoff models. However, recently, studies involving the application of a machine learning-based flood forecasting models, which addresses the limitations of extant physically based flood stage forecasting models, have been performed. We may require various case studies to determine more accurate methods. Therefore, this study performed the real-time forecasting of the river water level or stage at the Gurye station of the Sumjin river with lead times of 1, 3, and 6 h by applying a long short-term memory (LSTM)-based deep learning model. In addition, the applicability of the LSTM model was evaluated by comparing the results with those from widely used models based on support vector machine and multilayer perceptron. Consequently, we noted that the LSTM model exhibited a relatively better forecasting performance. Therefore, the applicability of the LSTM model should be extensively studied for flood forecasting applications.

Прогнозирование стока рек России методом экстраполяции гидрографа

Forecasting the flow of Russian rivers by hydrograph extrapolation / Borsch S.V., Koliy V.M., Semenova N.K., Simonov Yu.A., Khristoforov A.V. // Hydrometeorological Research and Forecasting, 2021, no. 2 (380), pp. 77-94. An automated system has been developed based on the hydrograph extrapolation method, which allows the year-round daily forecasting of water level and streamflow for the Russian rivers with up to 10-day lead time. The forecast of discharges or water levels is expressed by a linear formula depending on their values on the date of the forecast issue and five previous days. The forecasting scheme limits the possible minimum and maximum values of the discharge or water level based on historical data. Forecast schemes were obtained for 2776 river gauges. The time period from 2010 to 2019 with daily observations of discharge and water level was used. The forecast verification shows that this method can be successfully applied to large rivers with smooth hydrographs. Keywords: daily discharge and water levels, short- and medium-term forecasts, hydrograph extrapolation method, forecast verification, maximum lead time of satisfactory forecasts, self-learning of an automated system for preparing and issuing forecasts

A Method of Multi-Stage Reservoir Water Level Forecasting Systems: A Case Study of Techi Hydropower in Taiwan

Reservoirs in Taiwan often provide hydroelectric power, irrigation water, municipal water, and flood control for the whole year. Taiwan has the climatic characteristics of concentrated rainy seasons, instantaneous heavy rains due to typhoons and rainy seasons. In addition, steep rivers in mountainous areas flow fast and furiously. Under such circumstances, reservoirs have to face sudden heavy rainfall and surges in water levels within a short period of time, which often causes the water level to continue to rise to the full level even though hydroelectric units are operating at full capacity, and as reservoirs can only drain the flood water, this results in the waste of hydropower resources. In recent years, the impact of climate change has caused extreme weather events to occur more frequently, increasing the need for flood control, and the reservoir operation has faced severe challenges in order to fulfil its multipurpose requirements. Therefore, in order to avoid the waste of hydropower resources and improve the effectiveness of the reservoir operation, this paper proposes a real-time 48-h ahead water level forecasting system, based on fuzzy neural networks with multi-stage architecture. The proposed multi-stage architecture provides reservoir inflow estimation, 48-h ahead reservoir inflow forecasting, and 48-h ahead water level forecasting. The proposed method has been implemented at the Techi hydropower plant in Taiwan. Experimental results show that the proposed method can effectively increase energy efficiency and allow the reservoir water resources to be fully utilized. In addition, the proposed method can improve the effectiveness of the hydropower plant, especially when rain is heavy.

Flood forecasting using an improved NARX network based on wavelet analysis coupled with uncertainty analysis by Monte Carlo simulations: a case study of Taihu Basin, China

Abstract Reliable flood forecasting can provide a scientific basis for flood risk assessment and water resources management, and the Taihu water level forecasting with high precision is essential for flood control in the Taihu Basin. To increase the prediction accuracy, a coupling model (DWT-iNARX) is established by combining the discrete wavelet transformation (DWT) with improved nonlinear autoregressive with exogenous inputs network (iNARX), for predicting the daily Taihu water level during the flood season under different forecast periods. And the DWT-iNARX model is compared with the back-propagation neural network (BP) and iNARX models to assess its capability in prediction. Meanwhile, we propose an uncertainty analysis method based on Monte Carlo simulations (MCS) for quantifying model uncertainty and performing probabilistic water level forecast. The results show that three models achieve good simulation results with higher accuracy when the forecast period is short, such as 1–3 days. In overall performance, iNARX and DWT-iNARX models show superiority in comparison with the BP model, while the DWT-iNARX model yields the best performance among all the other models. The research results can provide a certain reference for the water level forecast of the Taihu Lake.

An early warning system for wave-driven coastal flooding at Imperial Beach, CA

Waves overtop berms and seawalls along the shoreline of Imperial Beach (IB), CA when energetic winter swell and high tide coincide. These intermittent, few-hour long events flood low-lying areas and pose a growing inundation risk as sea levels rise. To support city flood response and management, an IB flood warning system was developed. Total water level (TWL) forecasts combine predictions of tides and sea-level anomalies with wave runup estimates based on incident wave forecasts and the nonlinear wave model SWASH. In contrast to widely used empirical runup formulas that rely on significant wave height and peak period, and use only a foreshore slope for bathymetry, the SWASH model incorporates spectral incident wave forcing and uses the cross-shore depth profile. TWL forecasts using a SWASH emulator demonstrate skill several days in advance. Observations set TWL thresholds for minor and moderate flooding. The specific wave and water level conditions that lead to flooding, and key contributors to TWL uncertainty, are identified. TWL forecast skill is reduced by errors in the incident wave forecast and the one-dimensional runup model, and lack of information of variable beach morphology (e.g., protective sand berms can erode during storms). Model errors are largest for the most extreme events. Without mitigation, projected sea-level rise will substantially increase the duration and severity of street flooding. Application of the warning system approach to other locations requires incident wave hindcasts and forecasts, numerical simulation of the runup associated with local storms and beach morphology, and model calibration with flood observations.

Deep learning hybrid model with Boruta-Random forest optimiser algorithm for streamflow forecasting with climate mode indices, rainfall, and periodicity

Long-term forecasting of any hydrologic phenomena is essential for strategic environmental planning, hydrologic and other forms of structural design, agriculture, and water resources management. Climate mode indices, utilising machine learning methods, are frequently considered as predictor variables in order to forecast several different hydrological variables. In this study, a feature selection algorithm based on two different deep learning models, i.e., long short-term memory and a gated recurrent unit, is applied to improve the forecasting capability of streamflow water levels at six gaugingstations in the Murray Darling Basin of Australia. This paper therefore aggregates the significant antecedent lag memory of climate mode indices, rainfall, and the monthly factor based on the periodicity as the predictor variables to attain significantly accurate stream water level forecasts. This novel method identifies an improved relationship between the stream water level and climate mode indices through the aggregation of the significant lagged datasets capturing the historical features to predict the future streamflow water level. The boruta feature selection algorithm (BRF) was then applied in a two phase process before and after attaining the significant lagged inputs to screen the optimum predictor variables. The merits of the forecast models were evaluated through different performance evaluation criteria. The results show that the accumulated significant lagged inputs based on climate mode indices, along with the rainfall and periodicity factors are seen to provide improved forecasting of the SWL over the non-BRF deep learning approaches where no prior feature selection was applied. The hybrid LSTM method (i.e., BRF-LSTM model) achieved a unique advantage in terms of SWL forecasting, particularly attaining over 98% of the predictive errors lying within a band of +/-0.015 m with relatively low relative errors (RRMSE ≈1.30% and RMAE ≈ 0.882%), outperforming all of the benchmark models. It is also found that the periodicity factor has a potential influence on the accuracy of the forecast models for the four monitored study stations. This study concludes that the newly developed hybrid deep learning approaches, coupled with the BRF feature selection, provide improved forecasting performance. The hybrid approach developed in this paper can therefore be used to provide a strong provide predictive response algorithm for the hydrological variables that were influenced by the low-frequency variability of the climate model indices in respect to streamflow water level.

Development of Short-term Flood Forecast Using ARIMA

The aim of this study is to use the Box-Jenkins method to build a flood forecast model by analysing real-time flood parameters for Pengkalan Rama, Melaka river, hereafter known as Sungai Melaka. The time series was tested for stationarity using the Augmented Dickey-Fuller (ADF) and differencing method to render a non-stationary time series stationary from 1 July 2020 at 12:00am to 30th July 2020. A utocorrelation (ACF) and partial autocorrelation (PACF) functions was measured and observed using visual observation to identify the suitable model for water level time series. The parameter Akaike Information Information Criterion (AIC) and the Bayesian Information Criterion (BIC) were used to find the best ARIMA model (BIC). ARIMA (2, 1, 3) was the best ARIMA model for the Pengkalan Rama, with an AIC of 5653.7004 and a BIC of 5695.209. The ARIMA (2, 1, 3) model was used to produce a lead forecast of up to 7 hours for the time series. The model's accuracy was tested by comparing the original and forecast sequences by using Pearson r and R squared. The ARIMA model appears to be adequate for Sungai Melaka, according to the findings of this study. Finally, the ARIMA model provides an appropriate short-term water level forecast with a lead forecast of up to 7 hours. As a result, the ARIMA model is undeniably ideal for river flooding.

A Novel Method for Regional Short-Term Forecasting of Water Level

The water level forecasting system represented by the hydrodynamic model relies too much on the input data and the forecast value of the boundary, therefore introducing uncertainty in the prediction results. Tide tables ignore the effect of the residual water level, which is usually significant. Therefore, to solve this problem, a water level forecasting method for the regional short-term (3 h) is proposed in this study. First, a simplified MIKE21 flow model (FM) was established to construct the regional major astronomical tides after subdividing the model residuals into stationary constituents (surplus astronomical tides, simulation deviation) and nonstationary constituents (residual water level). Harmonic analysis (HA) and long short-term memory (LSTM) were adopted to forecast these model residuals, respectively. Finally, according to different spatial background information, the prediction for each composition was corrected by the inverse distance weighting (IDW) algorithm and its improved IDW interpolation algorithm based on signal energy and the spatial distance (IDWSE) from adjacent observation stations to nonmeasured locations. The developed method was applied to Narragansett Bay in Rhode Island. Compared with the assimilation model, the root-mean-square error (RMSE) of the proposed method decreased from 12.3 to 5.0 cm, and R2 increased from 0.932 to 0.988. The possibility of adding meteorological features into the LSTM network was further explored as an extension of the prediction of the residual water level. The results show that the accuracy was limited to a moderate level, which is related to the difficulty presented by using only wind features to completely characterize the regional dynamic energy equilibrium process.

Coupling a hybrid CNN-LSTM deep learning model with a Boundary Corrected Maximal Overlap Discrete Wavelet Transform for multiscale Lake water level forecasting

Developing accurate lake water level (WL) forecasting models is important for flood control, shoreline maintenance and sustainable water resources planning and management. In this study, improved accuracy of forecasts (up to three months) of Lake Michigan and Lake Ontario WLs was achieved by coupling Boundary Corrected (BC) Maximal Overlap Discrete Wavelet Transform (MODWT) data preprocessing, with a hybrid Convolutional Neural Network (CNN) Long-Short Term Memory (LSTM) deep learning (DL) model. Hybrid DL-based model performance was compared to that of BC-MODWT machine learning (ML) [e.g., Random Forest (RF) and Support Vector Regression (SVR)] models. For each lake, all models were calibrated using 70% of the monthly WL (in meters) data series (January 1918 to February 1988), with the remaining 30% (March 1988 to December 2018) serving for validation. In both standalone and wavelet-machine learning models, a hybrid correlation-based feature selection (CFS)-particle swarm optimization (PSO) search method served to select input variables among candidate WL lags of up to twelve months, whereas for the CNN-LSTM DL models, input variable selection was carried out automatically by the CNN structure. For the MODWT-based ML and DL models, input time series were decomposed using a BC-MODWT approach. Scaling coefficients were developed through several mother wavelet approaches (i.e., Haar, Daubechies, Symlets, Fejer-Korovkin and Coiflets) with different filter lengths (up to twelve) and decomposition levels (up to seven). Model performance was evaluated using several visual and statistical metrics, including the correlation coefficient, r; the root mean standard error, RMSE; and Willmot’s Index, WI. The CNN-LSTM DL model outperformed the standalone SVR and RF models. For all time horizons, coupled MODWT-based CNN-LSTM models outperformed standalone and hybrid models in WL forecasting. Not all wavelet family/filter length/decomposition combinations improved standalone models; however, the proposed BC-MODWT-CNN-LSTM model implementing the Haar mother wavelet (for Lake Michigan — one-month ahead: r = 0.994, RMSE = 0.04 m, WI = 0.996; two-months ahead: r = 0.979, RMSE = 0.07 m, WI = 0.989; three-months ahead: r = 0.957, RMSE = 0.102 m, WI = 0.976; for Lake Ontario — one-month ahead: r = 0.956, RMSE = 0.082 m, WI = 0.978; two-months ahead: r = 0.864, RMSE = 0.141, WI = 0.912; three-months ahead: r = 0.755, RMSE = 0.182 m, WI = 0.841) outperformed standalone ML and BC-MODWT-ML-based models. Accordingly, the BC-MODWT-CNN-LSTM model can be viewed as a potentially useful approach to increase the accuracy of lake WL forecasts.

Towards Impact-Based Flood Forecasting and Warning in Malaysia: A Case Study at Kelantan River

Impact-based forecasting and warning services aim to bridge the gap between producers and users of warning information by connecting and increasing synergies between the components of effective early warning systems. In this research, we proposed an automated warning message based on colour codes to trigger risk mitigation actions at the local level in the flood-exposed communities of Kelantan river basin, Malaysia. With a community-based approach for different groups of users (i.e. sectors), flood-impact scenarios were determined from past events and related to colour codes. These were developed into impact-based forecasting and warnings that can connect water levels, through the colour code, to localised guidance information tailored to sectors’ needs on how to respond to the expected flood. Overall, the colour coded impact-based warnings were found to be an easy and understandable way to link water level forecasts to the necessary risk mitigation actions. However, further investigation is needed to validate these findings under real-time conditions. IBF has vast potential in Malaysia, but its integration requires significant institutional changes, such as an inter-facing agency (long term) or team (short term), adjusted policy frameworks (standing orders on disasters), and new resource allocations for skills development and technological innovation from national to local levels. Overall, this paper aims to offer the first insight into impact-based forecasting and warning services in Malaysia to trigger further research and project developments.

Flood Disaster and Early Warning: Application of ANFIS for River Water Level Forecasting

Intensively monitoring river water level and flows in both upstream and downstream catchments are essential for flood forecasting in disaster risk reduction. This paper presents a developed flood river water level forecasting utilizing a hybrid technique called adaptive neuro-fuzzy inference system (ANFIS) model, employed for Kelantan river basin, Kelantan state, Malaysia. The ANFIS model is designed to forecast river water levels at the downstream area in hourly lead times. River water level, rainfall, and river flows were considered as input variables located in upstream stations, and one river water level in the downstream station is chosen as flood forecasting point (FFP) target. Particularly, each of these input-output configurations consists of four stations located in different areas. About twenty-seven data with fifteen minutes basis recorded in January 2013 to March 2015 were used in training and testing the ANFIS network. Data preprocessing is done with feature reduction by principal component analysis and normalization as well. With more attributes in input configurations, the ANFIS model shows better result in term of coefficient correlation ( ) against artificial neural network (ANN)-based models and support vector machine (SVM) model. In general, it is proven that the presented ANFIS model is a capable machine learning approach for accurate forecasting of river water levels to predict floods for disaster risk reduction and early warning.

A Committee Evolutionary Neural Network for the Prediction of Combined Sewer Overflows

Combined Sewer Overflows (CSOs) are a major source of pollution and urban flooding, spilling untreated wastewater directly into water bodies and the surrounding environment. If overflows can be predicted sufficiently in advance, then techniques are available for mitigation. This paper presents a novel bi-model committee evolutionary artificial neural network (CEANN) designed to forecast water level in a CSO chamber from 15 min to 6 h ahead using inputs of past/current CSO level data, radar rainfall data and forecast forecasted rainfall data. The model is composed of two evolutionary artificial neural network (EANN) models. The two models are trained and optimised for wet and dry weather conditions respectively and their results combined into a single response using a non-linear weighted averaging approach. An evolutionary strategy algorithm is employed to automatically select the optimal artificial neural network (ANN) structure and parameter set, allowing the network to be tailored specifically for different CSO locations and forecast horizons without significant human input. The CEANN model was tested and evaluated on real level data from 4 CSOs located in Northern England and the results compared to three other ANN models. The results demonstrate that the CEANN model is superior in terms of accuracy for almost all forecast horizons considered. It is able to accurately forecast the dry weather and wet weather level, predicting the timing and magnitude of upcoming spill events, thus providing information that is of clear use to a wastewater utility.

Facing Losses of Telemetric Signal in Real Time Forecasting of Water Level using Artificial Neural Networks

Real-time forecasting plays a valuable role in the early warning system framework by reducing damage. However, signal loss in telemetric monitoring networks tends to occur during extreme events, precisely when data are needed for forecasting. We present an original approach, consisting of a tree of artificial neural networks (ANNs), with complete and partial models to deal with signal loss scenarios, where we also tested a new type of filter (GWMA – Gamma-Weighted Moving Average) to aggregate data in time and reduce the number of model inputs. In addition to this filter, we tested UWMA (Uniformly Weighted Moving Average), EWMA (Exponentially Weighted Moving Average) and MD (Moving Difference). Novel concepts were used to reduce ANN internal complexity and to identify a training dataset size corresponding to an ideal amount of information, which does not hinder training. We developed a model to forecast the water level up to 24 h ahead at Encantado, in the Taquari-Antas River basin, southern Brazil. The data period comprises hourly records from 26/11/2015 to 24/04/2019. The verification dataset performances of the partial models are compared to the complete model, indicating no substantial loss. The mean absolute error and the Nash–Sutcliffe of the complete model for the lead times of 4, 10, and 20 h are 5.4, 17.7, and 19.4 cm; and 0.99, 0.95, and 0.92, respectively. Therefore, the ANN tree is confirmed as a viable alternative to cope with signal loss scenarios.

Influence of Hurricane Wind Field Variability on Real‐Time Forecast Simulations of the Coastal Environment

AbstractDynamic conditions occur in the coastal ocean during severe storms. Forecasting these conditions is challenging, and large‐scale numerical models require significant computing power. In this paper, we describe a real‐time modeling system (DUNEX‐RT), developed in support of the During Nearshore Event experiment (DUNEX) off the coast of North Carolina, United States of America. The model is run with wave, current, and water level boundary conditions from larger‐scale models, and provides 36‐h forecasts of significant wave height, depth‐averaged velocity, and water levels every 6‐h using Delft3D‐SWAN. Observations and forecasts run at different times are compared and communicated via an interactive website to verify model performance in real‐time and to visualize uncertainty from changing inputs. Here, we evaluate model sensitivity to inputs from seven different atmospheric hindcasts and two atmospheric forecasts for Hurricane Dorian in September 2019. The results emphasize the importance of accurate wind forcing, with significant differences observed between the output model results for different input atmospheric forcing models and forecasts produced at different times. The best results were achieved using atmospheric forcing from the high resolution rapid refresh model, and overall, DUNEX‐RT had low errors at 33 wave, water level, and current sites across the system. The model results for water levels and significant wave heights were also accurate over a longer period of 49 days. Overall, the good forecast skill achieved for the wide range of conditions over this time results suggest that this high‐resolution regional approach could be applied to forecast conditions in other coastal areas.

Forecasting water level of Glacial fed perennial river using a genetically optimized hybrid Machine learning model

The floods are the significant damaging disasters in the world. Either the causes of floods can be natural or anthropogenic but in both cases people have to bear the burden of devastation. The forecasting of water levels of a flood prone river is must in order to minimize the life and property damages. In this paper, we propose a hybrid machine learning model of Adaptive Neuro Fuzzy Inference System (ANFIS) and Genetic algorithm, which is trained on the temperature, precipitation data of two Meteorological stations and water level of the river. Different ANFIS models with different membership functions and optimization techniques are tested and the best performing model is picked for further modification. The parameters of selected ANFIS model are then optimized using the genetic algorithm to get better results. The ANFIS optimized with genetic algorithm showed better results as compared to the conventional ANFIS.

Comparison of River Basin Water Level Forecasting Methods: Sequential Neural Networks and Multiple-Input Functional Neural Networks

To precisely forecast downstream water levels in catchment areas during typhoons, the deep learning artificial neural networks were employed to establish two water level forecasting models using sequential neural networks (SNNs) and multiple-input functional neural networks (MIFNNs). SNNs, which have a typical neural network structure, are network models constructed using sequential methods. To develop a network model capable of flexibly consolidating data, MIFNNs are employed for processing data from multiple sources or with multiple dimensions. Specifically, when images (e.g., radar reflectivity images) are used as input attributes, feature extraction is required to provide effective feature maps for model training. Therefore, convolutional layers and pooling layers were adopted to extract features. Long short-term memory (LSTM) layers adopted during model training enabled memory cell units to automatically determine the memory length, providing more useful information. The Hsintien River basin in northern Taiwan was selected as the research area and collected relevant data from 2011 to 2019. The input attributes comprised one-dimensional data (e.g., water levels at river stations, rain rates at rain gauges, and reservoir release) and two-dimensional data (i.e., radar reflectivity mosaics). Typhoons Saola, Soudelor, Dujuan, and Megi were selected, and the water levels 1 to 6 h after the typhoons struck were forecasted. The results indicated that compared with linear regressions (REG), SNN using dense layers (SNN-Dense), and SNN using LSTM layers (SNN-LSTM) models, superior forecasting results were achieved for the MIFNN model. Thus, the MIFNN model, as the optimal model for water level forecasting, was identified.

Intelligent Data Analysis for Forecasting Threats in Complex Distributed Systems

A threat prediction method based on the intellectual analysis of historical data in complex distributed systems (CDS) is proposed. The relevance of the chosen research topic in terms of considering the flood as a physical process of raising the water level, which is measured at stationary and automatic hydrological posts, is substantiated. Based on this, a mathematical formulation of the problem is formulated, within the framework of which an artificial neural network based on the freely distributed TensorFlow software library is implemented. The analysis of the effectiveness of the implemented artificial neural network was carried out, according to which the average deviation of the predicted water level when forecasting for one day at a stationary hydrological post was 3.323%. For further research on forecasting water levels, an algorithm is proposed for evaluating historical data at automatic posts, which will allow using these data to predict water levels according to the proposed method and at automatic posts. Thus, the neural network allows predicting the flood situation with acceptable accuracy, which allows special services to take measures to counter this threat.

Modelling water levels of northwestern India in response to improved irrigation use efficiency

The groundwater crisis in northwestern India is the result of over-exploitation of groundwater resources for irrigation. The Government of India has targeted a 20 percent improvement in irrigation groundwater use efficiency. In this perspective, and using a regional-scale calibrated and validated three-dimensional groundwater flow model, this article provides the first forecasts of water levels in the study area up to the year 2028, both with and without this improvement in use efficiency. Future water levels without any mitigation efforts are anticipated to decline by up to 2.8 m/year in some areas. A simulation with a 20 percent reduction in groundwater abstraction shows spatially varied aquifer responses. Tangible results are visible in a decade, and the water-level decline rates decrease by 36–67 percent in over-exploited areas. Although increasing irrigation use efficiency provides tangible benefits, an integrated approach to agricultural water management practice that incorporates use efficiency along with other measures like water-efficient cropping patterns and rainwater harvesting may yield better results in a shorter period.

Trend Analysis and Forecasting of Water Level in Mtera Dam Using Exponential Smoothing

Trend Analysis and Forecasting of Water Level in Mtera Dam Using Exponential Smoothing

Prediction of Water Level Using Time Series, Wavelet and Neural Network Approaches

To control water resources in many domains such as agriculture, flood forecasting, and hydro-electrical dams, forecasting water level needs to predict. In this article, a new computational approach using a data driven model and time series is proposed to calculate the forecast water level in short time. Concretely, wavelet-artificial neural network (WAANN) and time series (TS) are combined together called WAANN-TS that encourages the advantage of each model. For this real time project work, Yen Bai station, Northwest Vietnam was chosen as an experimental case study to apply the proposed model. Input variables into the Wavelet-ANN structure is water level data. Time series and ANN models are built, and their performances are compared. The results indicate the greater accuracy of the proposed models at Hanoi station. The final proposal WAANN−TS for water level forecasting shows good performance with root mean square error (RMSE) from 10−10 to 10−11.