Hydrological Forecasting Research Articles

Medium- and Long-Term Precipitation Forecasting Method Based on Data Augmentation and Machine Learning Algorithms

Accurate medium and long-term precipitation forecasting plays a vital role in disaster prevention and mitigation and rational allocation of water resources. In recent years, there are various methods for medium- and long-term precipitation forecasting based on machine learning algorithms. However, machine learning has a high demand for the size of sample data. Therefore, this article proposes a data augmentation algorithm based on the K-means clustering algorithm and synthetic minority oversampling technique (SMOTE), which can effectively enhance sample information. Besides, through constructing random forest (RF), extreme gradient boosting (XGB), recurrent neural network (RNN), and long short-term memory (LSTM) are, respectively, constructed as the models to forecast monthly grid precipitation of the Danjiangkou River Basin. This study aims to improve the accuracy of medium- and long-term precipitation forecasting. The main results are the following two aspects: 1) in most years, the anomaly correlation coefficient and Pg score of SMOTE-km-XGB and SMOTE-km-RF exceed that of XGB and RF; furthermore, compared with the other three methods, SMOTE-km-XGB method is more suitable for precipitation forecasting in the studied basin in this article; and 2) the forecasting results of two deep learning methods (RNN and LSTM) show that the sample data processed by the K-means clustering algorithm and SMOTE data augmentation algorithm have not achieved considerable results in deep learning. This study improves the accuracy of precipitation forecast by expanding and balancing the information of sample data, and provides a new research idea for improving the accuracy of medium- and long-term hydrological forecasting.

Optimizing Hydrological Quantitative Precipitation Forecast (HQPF) based on Machine Learning for Rainfall Impact Forecasting

Optimizing Hydrological Quantitative Precipitation Forecast (HQPF) based on Machine Learning for Rainfall Impact Forecasting

Probability assessment of the Kyiv reservoir overflow

Reservoirs are an integral part of the world’s hydraulic infrastructure and form the basis of modern water management in most countries including Ukraine. However, reservoirs are also sources of an essential danger to the environment, infrastructure, and population. The potential danger and risks to the population living near reservoirs especially downstream may be no less than to people living near nuclear facilities or chemical plants, with which experts and the public usually associate problems of technogenic safety. Moreover, statistics show that about a third of all accidents on dams and levees occurred due to overflow of reservoirs when upstream water levels exceeded allowable values.There are 1103 reservoirs in Ukraine with a total water volume of about 55,500 million m3. The Kyiv reservoir is the third one by volume and water surface area in the country. In addition, the reservoir is created by one of the longest dams in the world; the total dam length of the reservoir reaches 70 km.Admittedly, the overflow of a reservoir can be caused by an extreme flood with inflow parameters exceeding the capacity of hydraulic structures. The challenge is that the capacity of water passage structures may be insufficient both due to the inaccuracy of the hydrological forecast and because of faults, poor functioning, or failures of the hydraulic structures during a design flood. In particular, long-term forecasts of floodwater discharges maxima of the inflow into the Kyiv reservoir based on using various probability distribution functions show the essential divergence of the obtained results. As well, as practice shows, the unavailability of some water passage tracts of the reservoir can reach several months in a year. Sometimes repair works were performed even during floods.The aim of the study consisted of probabilistic forecasting the emergency situation on the Kyiv reservoir as a result of its uncontrolled overflow through the possible inaccuracy of the hydrological forecast concerning an actual water inflow into the reservoir and due to failures of water passage hydraulic structures during floods. To achieve the study aim the following tasks were solved: (1) there was proposed a method of hydrological forecasting, which allows taking into account results of long-term forecasts of floodwater discharges maxima based on using various probability distribution functions and fuzzy modelling; (2) there was performed hydrological forecasting of floodwater discharges maxima of the Dnieper affecting the condition of the Kyiv reservoir based on the actual data collected the Vyshgorod water level gauge; (3) there was assessed the probability of the Kyiv reservoir overflow taking into account the occurrence possibility of a shortage of the capacity of water passage structures with using the failure and fault tree method. Totally, six incompatible hypothetical emergency situations at the Kyiv reservoir were considered. The calculations showed the total probability of the Kyiv reservoir overflow equal to 3.84*10–4 (year–1), which is acceptable to guarantee the hydrological safety of infrastructure and the population.

Postprocessing of hydrometeorological ensemble forecasts based on multisource precipitation in Ganjiang River basin, China

Flood forecasting is a challenging task with major concerns that include forecast uncertainty and lead time. This paper aims to comprehensively address the data correction issue for hydrological models by developing an integrated postprocessing framework for hydrometeorological ensemble forecasts based on three types of ensemble precipitation forecasts (ECMWF, GEFS, and CFSv2). This framework mainly comprises the Ensemble Preprocessor (EPP) and the Bayesian model averaging (BMA) scheme: (1) the EPP eliminates various deviations from the ensemble precipitation forecasts, with the canonical event model extracting useful predictive information and extending the lead time by linking the forecasts seamlessly, and (2) the BMA scheme develops more skillful and reliable probabilistic hydrological forecasts from ensemble streamflow forecasts generated by three hydrological models (XAJ, VIC and DTVGM) driven by the postprocessed ensemble precipitation forecasts. The Ganjiang River Basin was selected as a case study to examine the capacity and efficiency of our developed framework. We found that the performance of the integrated postprocessing of hydrometeorological ensemble forecasts benefitted from a reasonable design of canonical events, especially when the observed precipitation was factored into the canonical events. Streamflow forecasts with increased lead times and accuracy were obtained by employing postprocessed precipitation forecasts instead of raw forecasts as the input for the hydrological models. We also found that the BMA scheme further improved the forecast effect by generating more skillful and reliable forecasts than a single hydrological model, which was verified by a series of statisticalindices and uncertainty analysis. Our results underscore the importance of considering both input and output for hydrological models in hydrological forecasting.

Convolutional neural network-based statistical post-processing of ensemble precipitation forecasts

Raw forecasts from numerical weather prediction models suffer from systematic bias and cannot be directly used in applications such as hydrological forecasting. Statistical post-processing methods can be used to remove the bias and achieve reliable ensemble forecasts. However, traditional post-processing methods generally use local raw forecasts as the only predictor, which limits their ability to extract spatial information from raw forecasts. In this paper, we develop a convolutional neural network (CNN)-based post-processing method for precipitation forecasts to make use of spatial information and atmospheric circulation variables as auxiliary predictors. The results show that the proposed CNN-based post-processing model outperforms traditional methods in forecast accuracy and reliability, especially for heavy rain. The improvements of the CNN-based model relative to a state-of-the-art joint probability model can reach 5% and 8% in terms of Brier skill score for heavy rain and rainstorms at the lead time of 1 day. The improvements can be attributed to the use of auxiliary predictors such as total column water forecasts in the proposed model. Moreover, the CNN-based model transcends a fully connected network (FCN)-based post-processing model, which illustrates the benefits of using CNNs to extract spatial information. The results illustrate the proposed CNN-based post-processing model is able to utilize spatial information and various auxiliary predictors to improve forecast skill and quantify forecast uncertainty, which is important for further applications such as hydrological forecasting.

Hydrological time series prediction by extreme learning machine and sparrow search algorithm

Abstract In general, accurate hydrological time series prediction information is of great significance for the rational planning and management of water resource systems. Extreme learning machine (ELM) is an effective tool proposed for the single-layer feedforward neural network in regression and classification problems. However, the standard ELM model falls into a local minimum with a high probability in hydrological prediction problems since the randomly assigned parameters (like input-hidden weights and hidden biases) often remain unchanged in the learning process. For effectively improving the prediction accuracy, this paper develops a hybrid hydrological forecasting model where the emerging sparrow search algorithm (SSA) is firstly used to determine the satisfying parameter combinations of the ELM model, and then the Moore–Penrose generalized inverse method is chosen to analytically obtain the weight matrix between the hidden layer and output layer. The proposed method is used to forecast the long-term daily runoff series collected from three real-world hydrological stations in China. Based on several performance evaluation indexes, the results show that the proposed method outperforms several ELM variants optimized by other evolutionary algorithms in both training and testing phases. Hence, an effective evolutionary machine-learning tool is developed for accurate hydrological time series forecasting.

Statistical post-processing of short-term hydrological ensemble forecasts using the application of the dressing method

Statistical post-processing of short-term hydrological ensemble forecasts using the application of the dressing method

Improvement of methods of hydrological forecasting using geoinformation technologies

This article discusses the possibility of improving hydrological forecasting methods based on a neural network. The hydrological series, its importance and forecasting features are considered. For hydrological forecasting using the MapInfoProfessional geoinformation system, an electronic map has been developed containing information about the rivers of Russia, as well as gauging stations on the Ob River. The electronic map is the basis for creating a module for short-term hydrological forecasting based on an artificial neural network. The features of a neural network, methods of its training and implementation are considered. The developed artificial neural network is a layer of neurons with a linear activation function and a delay line at the input. To predict the levels of hydrological series, real water levels at gauging stations of the Ob River in the Novosibirsk region will be used. The developed module and its capabilities have been tested. The study was carried out on the basis of models of hydrological series, as well as on the basis of levels of real hydrological series. Based on the study, dependence of the root-mean-square error on the number of previous values of series was revealed. The study also shows that it is possible to use a neural network for the current one-step forecasting of levels of hydrological series in conditions of insufficient information about the runoff region and its characteristics.

Meteorological situations that generated exceptional discharges along the Danube River. (Case study- the situation of April 2006).

For Europe, the undisputed importance of the Danube can be rendered by some general data: Its hydrographic basin surface exceeds 817000 km2, i.e. about 10% of that of the continent, its length is 2857 km and its mean multiannual discharge is about 6500 m3/s, thus ranking second to Volga river. Romania is the country with the largest surface situated within Danube’s basin (97.4%), representing 29% of Danube’s hydrographic basin. The water resources of the Danube in Bazias section amount to 173 billion m3, 30 billion m3 of which are technically usable resources.Our analysis aimed at determining those complex meteorological situations at the European continent level that triggered exceptional discharges along the Danube river, resulting in severe flooding, causing in turn heavy damages, fatalities, population evacuations and considerable rehabilitation costs. A complex analysis was performed, of statistical-synoptic type and those complex meteorological situations were identified that determined the occurrence of such disasters. Discharges and levels of the Danube were used along the whole measuring period, data from the archive of the National Meteorological Administration, and data, map and image archives from Wetterzentrale (Kartenarchiv, NCEP, NCAR, AVN etc.). The complex meteorological situations at the level of the European continent that triggered exceptional discharges along the Danube correlate with intense cyclonic activity, of both the Icelandic and the Mediterranean cyclones, with the negative phase of the North-Atlantic Oscillation and with decreasing or minimum solar activity (according to data from NOAA's Space Environment Center). The most disastrous floods occurred in the spring of 2006. The paper is important for meteorologists, in their weather forecasting activity, for hydrologists, in their hydrological forecasting and for the institutions involvedin flood management.

A meta-modeling approach for hydrological forecasting under uncertainty: Application to groundwater nitrate response to climate change

We suggest a quantitative predictive approach based on meta-modeling techniques for long-term forecasting in hydrology. We apply our scalable methodological framework to a representative groundwater body in Emilia-Romagna region of Italy to evaluate the potential impact of climate change on groundwater nitrate concentration. The approach allows to (i) handle the uncertainty associated with the projection of climate variables, (ii) preserve an accurate description of flow and transport processes, and (iii) drastically accelerate computationally intensive simulations, at a time. A long-term climate scenario (i.e. referred to the 30-years 2061–2090) is developed for the study area by considering changes in the natural recharge to groundwater, the baseflow of the main rivers, and the rate of nitrate leaching. Global sensitivity analysis is performed to assess the relative influence of these factors on the quantity of interest (QoI), i.e. the long-term prediction of groundwater nitrate concentration. Our analysis enables also us to compute the spatial statistics and variability of the moments of the QoI over the study area and to perform Monte Carlo simulations to fully characterize the probabilistic behavior of the QoI at selected sensitive locations.

Monthly Streamflow Forecasting Using Convolutional Neural Network

Monthly streamflow forecasting is vital for managing water resources. Recently, numerous studies have explored and evidenced the potential of artificial intelligence (AI) models in hydrological forecasting. In this study, the feasibility of the convolutional neural network (CNN), a deep learning method, is explored for monthly streamflow forecasting. CNN can automatically extract critical features from numerous inputs with its convolution–pooling mechanism, which is a distinct advantage compared with other AI models. Hydrological and large-scale atmospheric circulation variables, including rainfall, streamflow, and atmospheric circulation factors are used to establish models and forecast streamflow for Huanren Reservoir and Xiangjiaba Hydropower Station, China. The artificial neural network (ANN) and extreme learning machine (ELM) with inputs identified based on cross-correlation and mutual information analyses are established for comparative analyses. The performances of these models are assessed with several statistical metrics and graphical evaluation methods. The results show that CNN outperforms ANN and ELM in all statistical measures. Moreover, CNN shows better stability in forecasting accuracy.

Trends in the development of systems for monitoring the state of water bodies and the development of a Geo-information system Dniester on the basis of an industrial ІоТ

The work shows that inadequate and untimely hydrological forecasting of the level of flood development leads to significant annual economic damage and loss of life. The latter requires constant control and monitoring of the state of water bodies, which is possible with modern automated control and monitoring systems. The analysis showed that the information and measurement systems (AIMS) for flood control, which are operated in Ukraine, their quantity, cost and technical characteristics do not provide the information necessary for predicting the development of floods.
 Analysis of trends in the development of systems for automated control and monitoring of the state of water bodies showed that modern foreign systems that are at the stage of development or testing are based on the use of microcontrollers, "smart" converters of physical quantities, modern communication systems and Internet of Things technology, as well as the use of smartphones with a Raspberry Pi camera for determining the water level in rivers.
 The development with the use of the industrial Internet of Things of the basic version of the geoinformation system for control and monitoring of the state of water bodies GIS-Dniester is given. The structure, components and functions of GIS-Dniester are considered. The tasks of organizing hardware and software for parameterizing and diagnosing "smart" converters, establishing a communication environment for industrial communication, developing algorithms for collecting, processing and transmitting data, studying the characteristics of measuring signals are being solved. The results of the study of the echo profiles of the XPS10 ultrasonic level sensor with “smart” -converters Multiranger 100, as well as the procedure for working with the GIS-Dniester are presented. The use of the developed GIS-Dniester has shown its effectiveness and reliability in operation.

Evaluation of Subseasonal-to-Seasonal (S2S) precipitation forecast from the North American Multi-Model ensemble phase II (NMME-2) over the contiguous U.S.

The second phase of the North America Multi-Model Ensemble (NMME-2) provides globally available Subseasonal-to-Seasonal (S2S) precipitation forecasts with a daily resolution. The S2S precipitation forecasts are getting increasing attention for their potentials in providing hydrometeorological forcing information for water resources planning at an extended range. However, the forecast skills of many existing S2S forecast products will significantly decrease when the lead time increases, hindering their applicability for watershed-scale hydrologic modeling. Therefore, forecast validation and large-scale evaluation are of great importance for water resources planning and hydrological applications. In this study, we comprehensively evaluate the S2S precipitation forecasts from the NMME-2 dataset over the contiguous United States (CONUS) and during the study period from 1982 to 2011. Three aspects of precipitation forecast capabilities are compared and analyzed: bias, skill scores, and the ability to predict extreme precipitation events. The Parameter-elevation Regressions on Independent Slopes Model (PRISM) is used as ground truth reference. Differs from other regional forecast validation study, we further examined and analyzed the dependences of NMME-2 precipitation forecast skills according to different seasonality, geographical locations, and lead times. Results show that the forecast biases are not sensitive to lead times but are seasonally dependent of all NMME-2 models. Overestimations are found in the Western U.S. in cooler seasons while underestimations are observed in the central regions of the U.S. in warmer seasons. The forecast skill of all individual NMME-2 models generally decreases as increases of lead times. The simple model averaging (SMA) of five NMME-2 models demonstrates a higher forecast skill than any individual NMME-2 models. Spatially, the highest forecast skill scores are observed at coastal areas in the Western U.S. with an one-week lead time. As compared to the historical resampled forecasts, NMME-2 also shows better performance in predicting extreme precipitation events above 99% percentiles and below 1% percentiles with higher probability of detections and lower false alarm ratios. The obtained results suggest the great potentials of NMME-2 precipitation forecasts in assisting ensemble hydrologic forecasts at the S2S scale over the CONUS.

Financial Risk Control of Hydro Generation Systems through Market Intelligence and Stochastic Optimization

In the competitive electricity wholesale market, decisions regarding hydro generators are generally made under uncertain conditions, such as pool price, hydrological affluence, and other players’ strategies. From this perspective, this work presents a computational model formulation with associated market intelligence and game theory tools to support a decision-making process in a competitive environment. The idea behind using a market intelligence tool is to apply a stochastic optimization model with an associated conditional value at risk metric defining a utility function, which calculates the weight that the agents attribute to each stochastic variable associated with the problem to be faced. Subsequently, this utility function is used to emulate the other agents’ strategies based on their previous decisions. The final step finds the Nash equilibrium solution between a player and their competitors. The methodology is applied to the monthly allocation of firm energy by hydro generators under the current Brazilian regulatory framework. The results show a change in the generators’ behavior over the years, from risk-neutral agents seeking to maximize their return with 88% of decisions based on spot price forecasts in 2015, to risk-averse agents with 100% of decisions following a factor that is directly impacted by the hydrological affluence forecasts in 2018.

Machine learning for precipitation forecasts post-processing — Multi-model comparison and experimental investigation

AbstractObtaining high-quality quantitative precipitation forecasts is a key precondition for hydrological forecast systems. Due to multisource uncertainties (e.g., initial conditions, model structures and parameters), raw forecasts are subject to systematic biases; hence, statistical post-processing is often required to reduce these errors before the forecasts can proceed to hydrological applications. Machine learning (ML) algorithms are canonical statistical models, and they are diverse in type and variation. It is important to verify and compare their performance in the same scenario (e.g., precipitation post-processing). In this paper, we conduct a large-scale comparison study for the major ML models with diverse model structures and regularization strategies as post-processors for improving the quality of precipitation forecasts. Specifically, we compare the efficiency and effectiveness of 21 ML algorithms on solving this task. Daily reforecast precipitation with lead times up to 8 days from the Global Ensemble Forecast System and corresponding observations are employed to determine the usability of different models in the Yalong River basin in China. The performance of each model is validated by a group of carefully designed experiments and statistical metrics. The results reveal that improvements in model structures are more effective than regularization strategies. Among these algorithms, the optimized extra-trees regressor exhibit the best performance, effectively reduce overestimation and achieve the best skill in forecasting precipitation. Eleven ensemble members and a 2-day time window can be used as predictors to obtain the best model performance. The systematic experiments and findings also offer useful guidelines for other related studies.

Need for system of protection against floods and high waters in mountain and foothill rivers of the Caucasus

The article presents findings on the estimation of fast-forming floods parameters and hydrological forecasts for short-term changes in water content in the foothill and mountain rivers of southern Russia as exemplified by the Krasnodar Krai. The research aims to provide economic facilities and population with better protection against rapidly forming mixed-origin floods. The need to develop a system of protection against floods and high waters on the mountain and foothill rivers of the Caucasus are considered. The floods in the Tuapse region in October 2010, Krymsk in July 2012, Novomikhaylovsky in August 2012 are taken as an example. The causes of floods and high waters on rivers are different. Precipitation is the main and dominant source of floods and high waters (exclusive of natural and anthropogenic discharge breakthrough floods) in river basins. The North-West Caucasus territory and especially its Black Sea coast is distinguished by rain floods at any time of the year. The main reason for floods is rainfall with an intensity of minimum 50-100 mm/day. Considering that 1 mm of precipitation causes the fallout of 1 liter of water over 1 m2 of the earth’s surface, this amount of precipitation is equivalent to the flow of 50-100 liters of water per 1 m2 of mountain slopes. The surface of the slopes cannot immediately absorb and hold such an amount of water, and it rushes down into the rivers’ and streams’ beds. An increase in water content leads to a sharp rise in the water level and flooding of the territories adjacent to the river.

Seasonal discharge forecasting for the Upper Danube

Study RegionThe authors contributed with a seasonal forecast system for discharge in the case study area of the Upper Danube basin upstream of Vienna. The basin, with a drainage area of approx. 100.000 km2, is characterized by complex topography dominated by the Alps. Study FocusNew advances in the mean term (seasonal) climate forecasting increase the potential of seasonal hydrological forecasting. In the frame of an European project (Clim2Power, 2020) a climate service for the assessment of seasonal energy production of renewable resources was developed. Seasonal discharge forecasts are generated using a dynamical approach running the hydrological model COSERO with forecasted climate inputs provided by Germany’s National Meteorological Service (Deutscher Wetterdienst – DWD). The seasonal climate forecasts are based on a large ensemble of daily predictions available up to 6 months in the future. These forecasts are downscaled to a spatial resolution of 6 km. The following analysis is based on aggregation of these daily values to monthly and seasonal values. New Hydrological Insights for the RegionDownscaled historical forecasts (hindcasts) and historical discharge observations over the last 15 years are used to assess the skill of the seasonal forecast in several Alpine basins. In general, the results show a limited skill. Therefore, the seasonal discharge forecasting system in this region is still not sufficient to incorporate the results into water resources decision support systems.

The efficacy of seasonal terrestrial water storage forecasts for predicting vegetation activity over Africa

AbstractTerrestrial water storage (TWS) provides important information on terrestrial hydroclimate and may have value for seasonal forecasting because of its strong persistence. We use the NASA Hydrological Forecast and Analysis System (NHyFAS) to investigate TWS forecast skill over Africa and assess its value for predicting vegetation activity from satellite estimates of leaf area index (LAI). Forecast skill is high over East and Southern Africa, extending up to 3–6 months in some cases, with more modest skill over West Africa. Highest skill generally occurs during the dry season or beginning of the wet season when TWS anomalies from the previous wet season are most likely to carry forward in time. In East Africa, this occurs prior to and during the transition into the spring “Long Rains” from January–March, while in Southern Africa this period of highest skill starts at the beginning of the dry season in April and extends through to the start of the wet season in October. TWS is highly and positively correlated with LAI, and a logistic regression model shows high cross-validation skill in predicting above or below normal LAI using TWS. Combining the LAI regression model with the NHyFAS forecasts, 1-month lead LAI predictions have high accuracy over East and Southern Africa, with reduced but significant skill at 3-month leads over smaller sub-regions. This highlights the potential value of TWS as an additional source of information for seasonal forecasts over Africa, with direct applications to some of the most vulnerable agricultural regions on the continent.

Improving Prediction Accuracy of Hydrologic Time Series by Least-Squares Support Vector Machine Using Decomposition Reconstruction and Swarm Intelligence

AbstractAccurate hydrologic forecasting plays a significant role in water resource planning and management. To improve the prediction accuracy, this study develops a hybrid hydrological forecasting...

Verification of two hydrological models for real-time flood forecasting in the Hindu Kush Himalaya (HKH) region

The Hindu Kush Himalayan region is extremely susceptible to periodic monsoon floods. Early warning systems with the ability to predict floods in advance can benefit tens of millions of people living in the region. Two web-based flood forecasting tools (ECMWF-SPT and HIWAT-SPT) are therefore developed and deployed jointly by SERVIR-HKH and NASA-AST to provide early warning to Bangladesh, Bhutan, and Nepal. ECMWF-SPT provides ensemble forecast up to 15-day lead time, whereas HIWAT-SPT provides deterministic forecast up to 3-day lead time covering almost 100% of the rivers. Hydrological models in conjunction with forecast validation contribute not only to advancing the processes of a forecasting system, but also objectively assess the joint distribution of forecasts and observations in quantifying forecast accuracy. The validation of forecast products has emerged as a priority need to evaluate the worth of the predictive information in terms of quality and consistency. This paper describes the effort made in developing the hydrological forecast systems, the current state of the flood forecast services, and the performance of the forecast evaluation. Both tools are validated using a selection of appropriate metrics in measurement in both probabilistic and deterministic space. The numerical metrics are further complemented by graphical representations of scores and probabilities. It was found that the models had a good performance in capturing high flood events. The evaluation across multiple locations indicates that the model performance and forecast goodness are variable on spatiotemporal scale. The resulting information is used to support good decision-making in risk and resource management.