Hydrological Forecasting Research Articles

Comparison of the performance of SWAT and hybrid M5P tree models in rainfall-runoff simulation.

Stream flow forecasting is a crucial aspect of hydrology and water resource management. This study explores stream flow forecasting using two distinct models: the Soil and Water Assessment Tool (SWAT) and a hybrid M5P model tree. The research specifically targets the daily stream flow predictions at the MH Halli gauge stations, located along the Hemvati River in Karnataka, India. A 14-year dataset spanning from 2003 to 2017 is divided into two subsets for model calibration and validation. The SWAT model's performance is evaluated by comparing its predictions to observed stream flow data. Residual time series values resulting from this comparison are then resolved using the M5P model tree. The findings reveal that the hybrid M5P tree model surpasses the SWAT model in terms of various evaluation metrics, including root-mean-square error, coefficient of determination (R2), Nash-Sutcliffe efficiency, and degree of agreement (d) for the MH Halli stations. In conclusion, this study shows the effectiveness of the hybrid M5P tree model in stream flow forecasting. The research contributes valuable insights into improved water resource management and underscores the importance of selecting appropriate models based on their performance and suitability for specific hydrological forecasting tasks.

Flow discharge coefficient of crest weirs with covered vegetation

The flow data of mountainous rivers is very important in hydrological forecasting. Crest weirs with smooth surface have been widely used and studied, while there is little information on the weir covered with vegetation. In this experiment, the flow coefficient reduction are systematically investigated by different density flexible artificial vegetation, in addition to the changes of discharge, crest length and height. The equivalent roughness height of vegetation on the weir was calculated. It was found that the flow coefficient with vegetation is smaller than smooth crest. According to the ratio of upstream water head to the length of vegetation weir crest, the experimental flow can be divided into broad-crested weir flow rate of H/L < 0.4 and narrow-crested weir flow rate of H/L > 0.4. The flow coefficient reduction rate of both types of weir flow rates varies to a certain extent. Experimental data showed that the crest weir height H/P played a different law for the broad-crested weir and narrow-crested weir. Taking into account the influence of vegetation, crest length and height, empirical formula for flow coefficient was proposed for flow estimation correction.

Randomized block quasi-Monte Carlo sampling for generalized likelihood uncertainty estimation

Abstract Although hydrological model forecasts aid water management decisions, they normally have predictive uncertainties. Generalized likelihood uncertainty estimation (GLUE) is popular for constructing predictive uncertainty bounds (PUBs). GLUE is based on simple Monte Carlo sampling (SMCS), a technique known to be ineffective in establishing behavioural simulations. This study introduced randomized block quasi-Monte Carlo sampling (RBMC). In RBMC, each parameter's range is divided into a stipulated number of sub-blocks (Snb). Parameters' values are separately generated in each sub-block. Finally, the sub-blocks are shuffled while maintaining the sequence of generated values in each sub-block. When Snb is equal to the number of simulations, RBMC reduces to SMCS. Otherwise, each Snb leads to a separate RBMC configuration or sampling scheme. The number of RBMC-based behavioural solutions was often found to be greater than that of SMCS, in some cases, by up to 33.6%. The width of the 90% confidence interval on 95th percentile flow based on SMCS was often larger than those of RBMC, sometimes by up to 23.4%. PUBs were found to vary in widths among RBMC configurations, thereby revealing the influence of the choice of a sampling scheme. Thus, GLUE based on RBMC is recommended to take into account the said influence.

Neglect of Potential Seasonal Streamflow Forecasting Skill in the United States National Water Model

AbstractUsing data from the NASA Soil Moisture Active/Passive mission, Koster et al. (2023, https://doi.org/10.1038/s41467‐023‐39318‐3) conclude that, for medium‐scale basins in the contiguous United States, a quarter of interannual variability in springtime streamflow is explained by interannual anomalies in late‐fall soil moisture. This lagged relationship can be leveraged for seasonal hydrologic forecasting, but only if effectively captured by existing prediction models. Here, we extend the analysis in Koster et al. (2023, https://doi.org/10.1038/s41467‐023‐39318‐3) to diagnose systematic errors present in the United States National Water Model (NWM). Results demonstrate that the NWM tends to underestimate both the trans‐winter temporal memory of 0–1 m soil moisture as well as the correlation between 0 and 1 m soil moisture and streamflow—thereby reducing the NWM's ability to leverage vertically averaged soil moisture as a source of hydrologic predictability.

An open online simulation strategy for hydrological ensemble forecasting

Hydrological ensemble forecasting is crucial for flood forecasting. However, the centralized architecture of hydrological ensemble forecasting systems requires time- and labour-intensive downloads and the installation of executable models and methods. This usage pattern impedes the reusability of forecasting models and techniques. To address these limitations, we propose an open online simulation strategy with three components: model sharing and integration, data sharing and adaptation, and parameter optimization and recommendation. The model sharing and integration method helps researchers publish forecasting models as web services for online simulation and integration. A reusable data sharing and adaptation method is established for managing and processing data to meet model requirements. In addition, the optimization and recommendation methods are intended to assist researchers in optimizing and recommending model parameters online based on the characteristics of different research regions. Finally, a prototype system and case study are constructed to verify the strategy's feasibility and capability.

Adapting subseasonal-to-seasonal (S2S) precipitation forecast at watersheds for hydrologic ensemble streamflow forecasting with a machine learning-based post-processing approach

Accurate and reliable precipitation predictions made by dynamical forecast models could provide crucial information for human socioeconomic activities by enabling hydrologic forecasts at the Subseasonal-to-Seasonal (S2S) timescale. To utilize available S2S precipitation predictions for hydrologic forecasts, post-processing techniques have been applied to adapt the raw S2S precipitation to local watersheds. However, conventional statistical-based post-processing techniques are more focused on correcting the forecast bias, but rather limited in improving the predictive skill of available S2S precipitation forecasts. In this study, we combine the Random Forest classifiers (RF) with the Bias Correction and Spatial Disaggregation (BCSD) to adapt the 10-member ensemble precipitation forecast from the NASA Goddard Earth Observing System model version 5 (GEOS5) at 4 watersheds located in the NCEI South climate region of the United States. The adapted S2S precipitation is further applied for streamflow forecast by forcing a classical lumped hydrologic model. The performance of S2S precipitation as well as the corresponding streamflow predictions are benchmarked with the randomly resampled precipitation and the corresponding Ensemble Streamflow Prediction (ESP) framework-generated streamflow predictions. Evaluation statistics of Kling-Gupta Efficiency (KGE), Continuous Ranked Probability Skill Score (CRPSS), Reliability, Resolution, and Sharpness are employed to evaluate the predictive skill of precipitation and streamflow both deterministically and probabilistically. Our results indicate that dynamical S2S precipitation after forecast adaptation leads to consistently higher deterministic skill over ESP at all forecast lead times and across study watersheds. However, at longer forecast lead times beyond 10–15 days, S2S precipitation with a limited ensemble size does not present higher probabilistic skill than ESP. Our results shows that the joint application of RF and BCSD improves the predictive skill of the raw S2S precipitation at study watersheds in contrast to BCSD. Further, the added predictive skill of S2S precipitation brought by RF propagates into streamflow predictions, predominantly at longer forecast lead times exceeding 10 days. Overall, our results highlight the potential success of future work to apply other data-driven approaches to adapt the raw precipitation to local watersheds for more accurate and reliable streamflow forecasts at the S2S timescale.

Review of Recent Developments in Hydrologic Forecast Merging Techniques

Accurate forecasting in hydrologic modeling is crucial for sustainable water resource management across various sectors, where predicting extreme flow phases holds particular significance due to their severe impact on the territory. Due to the inherent uncertainties in hydrologic forecasting, relying solely on a single rainfall–runoff model may not provide reliable predictions. To address this challenge, over the years, researchers have developed and applied hydrologic forecast merging (HFM) techniques that combine multiple models or ensembles to enhance forecast accuracy, reduce uncertainty, and increase confidence in the forecast. This review summarizes the progress in HFM techniques since the early 1990s and covers developments and applications in flow simulation, uncertainty analysis, monthly and seasonal streamflow predictions, ensemble forecasts, flood forecasting, and climate change analysis. The findings indicate that while HFM techniques outperform individual models regarding forecasting efficiency, their performance across applications is not uniform. Among the different methods, Bayesian model averaging (BMA) is the most popular due to its ability to reduce uncertainty and provide accurate and reliable forecasts in deterministic and probabilistic simulations. With their application simplicity, regression techniques are also robust and efficient as they perform competitively well across different model-merging applications. While specific techniques, such as model-dependent weighted averaging and neural network methods, effectively reduce forecast uncertainty, there is still room for improving forecast accuracy across different lead times. Future research can focus on advanced HFM techniques for estimating optimal weights in time-varying domains and overcoming limitations, like simulating low flows in seasonally dry catchments.

Systematic Modeling Errors Undermine the Application of Land Data Assimilation Systems for Hydrological and Weather Forecasting

Abstract Due to recent advances in the development of land data assimilation systems (LDAS) and the availability of high-quality, satellite-based surface soil moisture (SSM) retrieval products, we now have unambiguous evidence that the assimilation of SSM retrievals, or their proxy, can improve the precision (i.e., correlation versus truth) of surface state estimates provided by a land surface model (LSM). However, this clarity does not yet extend to the estimation of LSM surface water fluxes that are key to hydrologic and numerical weather forecasting applications. Here, we hypothesize that a key obstacle to extrapolating realized improvements in water state precision into comparable improvements in water flux accuracy (i.e., mean absolute error) is the presence of water state–water flux coupling strength biases existing in LSMs. To test this hypothesis, we conduct a series of synthetic fraternal twin data assimilation experiments where realistic levels of state–flux coupling strength bias—involving both evapotranspiration and runoff—are systematically introduced into an assimilation LSM. Results show that the accuracy of the resulting water flux analysis is sharply reduced by the presence of such bias, even in cases where the precision of soil moisture state estimates (e.g., SSM) is improved. The rescaling of SSM observations prior to their assimilation (i.e., the most common approach for addressing systematic differences between LSMs and assimilated observations) is not always a robust strategy for addressing these errors and can, in certain circumstances, degrade water flux accuracy. Overall, results underscore the critical need to assess, and correct for, LSM water state–water flux coupling strength biases during the operation of an LDAS. Significance Statement Land data assimilation is the process by which land surface model estimates of water states (e.g., soil moisture) and water fluxes (e.g., runoff and evapotranspiration) are improved via the incorporation of observations. Over the past decade, substantial improvements have been made in the precision of land surface model states via the assimilation of satellite-based soil moisture information. However, to date, these improvements have not yet been extended into water flux estimates like runoff and evapotranspiration. This is a critical shortcoming since advances in important weather and hydrologic forecasting applications are dependent on the improved estimation of such fluxes. We demonstrate that this shortcoming is linked to the inability of existing land surface models to accurately describe the impact of variations in water states on water fluxes and propose strategies for overcoming this issue in future land data assimilation systems.

A Novel mRMR-RFE-RF Method for Enhancing Medium- and Long-Term Hydrological Forecasting: A Case Study of the Danjiangkou Basin

A Novel mRMR-RFE-RF Method for Enhancing Medium- and Long-Term Hydrological Forecasting: A Case Study of the Danjiangkou Basin

Previsões de vazões sub sazonais para barragens hidrelétricas no Sistema Interligado Elétrico Brasileiro

ABSTRACT Inflow prediction on sub-seasonal timescale have the potential for important contributions to the management of water resources in hydroelectric dam operations. These forecasts challenge the limitations of the medium-term and extend it, bridging a long-standing technical-scientific gap in the forecasting field. In Brazil, the use of sub-seasonal hydrological predictions can boost the hydroelectric production of the National Interconnected System (SIN), since inflow forecast in reservoirs of up to 2 weeks are routinely used using a rain-flow model. This study aimed at the statistical evaluation of hydrological forecasts of up to 6 weeks using a hydrological-hydrodynamic model on a continental scale associated with ensemble precipitation forecasts generated by an atmospheric model, producing future streamflow in the continent basins, and consequently at the SIN’s hydroelectric dams. The statistical evaluation was based on deterministic scores typically used by the SIN operating agent, and additionally we assessed the skill of forecasts based on atmospheric models in relation to simpler forecasts based on the climatology of observed inflows. The performance of the forecasts varies according to the season and geographic location, that is, depending on different hydrological regimes. The best performances were obtained in dams located in the southwest and central-west regions, which have well-defined seasonality, while dams in the south showed greater sensitivity in metrics according to the season. The study presented serves as a technical-scientific contribution for agents and decision makers who seek to improve water resource management by incorporating extended forecasts into the operational chain.

Enforcing Water Balance in Multitask Deep Learning Models for Hydrological Forecasting

Abstract Accurate prediction of hydrological variables (HVs) is critical for understanding hydrological processes. Deep learning (DL) models have shown excellent forecasting abilities for different HVs. However, most DL models typically predicted HVs independently, without satisfying the principle of water balance. This missed the interactions between different HVs in the hydrological system and the underlying physical rules. In this study, we developed a DL model based on multitask learning and hybrid physically constrained schemes to simultaneously forecast soil moisture, evapotranspiration, and runoff. The models were trained using ERA5-Land data, which have water budget closure. We thoroughly assessed the advantages of the multitask framework and the proposed constrained schemes. Results showed that multitask models with different loss-weighted strategies produced comparable or better performance compared to the single-task model. The multitask model with a scaling factor of 5 achieved the best among all multitask models and performed better than the single-task model over 70.5% of grids. In addition, the hybrid constrained scheme took advantage of both soft and hard constrained models, providing physically consistent predictions with better model performance. The hybrid constrained models performed the best among different constrained models in terms of both general and extreme performance. Moreover, the hybrid model was affected the least as the training data were artificially reduced, and provided better spatiotemporal extrapolation ability under different artificial prediction challenges. These findings suggest that the hybrid model provides better performance compared to previously reported constrained models when facing limited training data and extrapolation challenges.

Spatiotemporal Analysis for Rainfall Prediction Using Extreme Learning Machine Cluster

Rainfall prediction is an essential study as a guideline for water resources management to manage disasters. Still, earlier research cares much about temporal information, only considering a single spatial location. The earth’s land surface has a large area of spatial location, so to manage spatial information simultaneously as temporal, we use spatiotemporal data to analyze rainfall prediction more accurately. This study uses the spatiotemporal Extreme Learning Machines (ELM) Cluster to forecast rainfall using CHIPRS data from satellites and stations. Data consists of spatial two dimensions and temporal data from 1981 to 2020. The dataset for the experiment contains 480 months. We use focal operation for data preprocessing to the nearest neighbor value. Moreover, the ELM cluster can manage every spatial location by sharing the output weight of ELM, so there is no spatial information left behind. Then, comparing the spatiotemporal Extreme Learning Machines Cluster among SVR, Linear Regression, Gaussian, Ridge, and Lasso are used to predict the data on those timescales. The results indicate that spatiotemporal ELM-Cluster can accurately forecast rainfall. Using ELM-Cluster in hydrological rainfall forecasting is encouraging, and the model can practically be used. Evaluation using MAE with a score of 66.77 and RMSE, 83.77, getting the fastest training with only 28.9 seconds compared to the other methods due to the ELM Cluster does not have backpropagation with spatial improvement.

Long-term forecasting methods of dates of ice break-up and disappearance at the Dnipro Cascade Reservoirs by teleconnection indicators

Long-term forecasting of the ice regime at the Dnipro Cascade Reservoirs is very important, first of all, for the operation of hydroelectric power stations. In addition, such forecasts also ensure the work of other sectors of the economy, namely shipping, fisheries, municipal engineering, etc. At present, the Department for hydrological forecasting of the Ukrainian Hydrometeorological Center uses techniques developedin the middle and in thesecond half of the 20th century to compile long-term forecasts and advices about the ice regimeof water bodies of Ukraine. These techniqueswere developed only for rivers;for reservoirs,they are currently lacking.&#x0D; The main objective of our research is to develop methods for long-term forecasting of ice break-up and disappearance at the Dnipro Cascade Reservoirs by teleconnection indicators. The research was based on the recordsfor 35 water gauges located on 6 reservoirs of the Dnipro Cascade, including the destroyed Kakhovka reservoir. The dates of the ice break-up and disappearance were used for the period from the beginning of observations at each water gauge to 2020. Moreover, the teleconnection indicators were used, namely 34 atmospheric indices, sea surface temperature indices, teleconnection indices and patterns determined by the National Weather Service (NWS) of the National Oceanic and Atmospheric Administration of United States of America (NOAA USA).&#x0D; Methods of the long-term forecasting of the ice break-up and disappearance at the reservoirs of the Dnipro Cascade have been developed by searching for the best correlation or regression relationship between dates and teleconnection indicators. The probable of accepted errorsof the developed methods are in the range of 62—71 %, which corresponds to satisfactory method quality that makes it possible to recommend them for use.

Application of the Improved K-Nearest Neighbor-Based Multi-Model Ensemble Method for Runoff Prediction

Hydrological forecasting plays a crucial role in mitigating flood risks and managing water resources. Data-driven hydrological models demonstrate exceptional fitting capabilities and adaptability. Recognizing the limitations of single-model forecasting, this study introduces an innovative approach known as the Improved K-Nearest Neighbor Multi-Model Ensemble (IKNN-MME) method to enhance the runoff prediction. IKNN-MME dynamically adjusts model weights based on the similarity of historical data, acknowledging the influence of different training data features on localized predictions. By combining an enhanced K-Nearest Neighbor (KNN) algorithm with adaptive weighting, it offers a more powerful and flexible ensemble. This study evaluates the performance of the IKNN-MME method across four basins in the United States and compares it to other multi-model ensemble methods and benchmark models. The results underscore its outstanding performance and adaptability, offering a promising avenue for improving runoff forecasting.

Gauging the ungauged: estimating rainfall in a West African urbanized river basin using ground-based and spaceborne sensors

ABSTRACT Accurate precipitation observations are crucial for hydrological forecasts, notably over rapidly responding urban areas. This study evaluated the accuracy of three gridded spaceborne rainfall products (Integrated Multi-satellitE Retrievals for GPM (IMERG), Meteosat Second Generation Visible (MSG-VIS), and MSG-Infrared (MSG-IR)) and the non-governmental Trans-African Hydro-Meteorological Observatory (TAHMO) gauges across the Odaw catchment (Accra, Ghana) from January 2020-July 2022. IMERG is hardly able to capture the strong spatial variability of rainfall required for flood forecasting, but agrees in annual sums with TAHMO and MSG-IR. MSG-IR has difficulties during the wet season. MSG-VIS, only available during daylight, shows limited accuracy and gives high estimates while other products do not detect rain. TAHMO gauges effectively record high-intensity events and their strong spatial variability, although some (daily) accumulations are doubtful and data gaps exist due to technical issues. These findings assist hydrological modelers in selecting appropriate datasets at suitable spatiotemporal resolutions for their research.

Multimodal deep learning water level forecasting model for multiscale drought alert in Feiyun River basin

Hydrological forecasting is an indispensable tool in intelligent water conservation for flood control and drought mitigation. Due to the influences of human activities and climate variability, accurate water level prediction poses a significant challenge. To address this, we develop a novel hybrid deep architecture, that is Dual-Stage Attention-based Multi-modal Deep Learning (DSAMDL), for reliable and interpretable multi-scale water level forecasting. Unlike previous techniques, multisource data is treated as different modalities in the proposed model. Firstly, we employ the one-dimensional Convolution to capture local trend features, followed by the Bidirectional Long Short-term Memory network to learn long-term dependencies. Subsequently, we design a dual-stage attention mechanism, which assigns contributions in a phased manner to different temporal and spatial. Finally, an adaptive fusion method is applied to enhance overall performance. To validate its accuracy and efficiency, short long-term drought water level forecasting and predictions under various events are implemented over six reservoir stations in the Feiyun River basin, Wenzhou City. Four evaluation metrics, namely RMSE, MAE, CORR, and NSE, are introduced for comprehensive assessment against the start-of-the-art baselines. The results well-document that the DSAMDL framework achieves satisfactory accuracy, with an average improvement of 22.4%, 27.8%, 29.7%, and 11.5% in these four metrics, showcasing the model’s effectiveness in handling complex multiscale drought water level forecasting.

Improving the forecasting accuracy of monthly runoff time series of the Brahmani River in India using a hybrid deep learning model

Abstract Accurate prediction of monthly runoff is critical for effective water resource management and flood forecasting in river basins. In this study, we developed a hybrid deep learning (DL) model, Fourier transform long short-term memory (FT-LSTM), to improve the prediction accuracy of monthly discharge time series in the Brahmani river basin at Jenapur station. We compare the performance of FT-LSTM with three popular DL models: LSTM, recurrent neutral network, and gated recurrent unit, considering different lag periods (1, 3, 6, and 12). The lag period, representing the interval between the observed data points and the predicted data points, is crucial for capturing the temporal relationships and identifying patterns within the hydrological data. The results of this study show that the FT-LSTM model consistently outperforms other models across all lag periods in terms of error metrics. Furthermore, the FT-LSTM model demonstrates higher Nash–Sutcliffe efficiency and R2 values, indicating a better fit between predicted and actual runoff values. This work contributes to the growing field of hybrid DL models for hydrological forecasting. The FT-LSTM model proves effective in improving the accuracy of monthly runoff forecasts and offers a promising solution for water resource management and river basin decision-making processes.

Bayesian estimation of glacier surface elevation changes from DEMs

Accurate estimates of glacier surface elevation changes are paramount for various aspects of the study of the cryosphere, from glacier flow and thickness estimates to hydrological forecasts and projections of sea-level-rise. We present a novel probabilistic framework to filter outliers and estimate uncertainties in glacier surface elevation changes computed from the subtraction of digital elevation models (DEM). Our methodology frames outlier filtering as a Bayesian inference problem, thus characterizing the state of knowledge on glacier surface elevation changes through the posterior distribution as the combination of glacier volume variation observations and prior knowledge arising from previously collected data and/or modeled results. We validate this technique with experiments using Gaussian random fields to generate artificial noise in glacier surface elevation variation observations and show that the model satisfactorily culls the simulated outliers. Surface elevation change estimates are consistent with results computed from widely-used outlier filtering and uncertainty estimation techniques. The Bayesian framework allows unifying DEM error models with physical considerations on glacier surface elevation changes within a simple, statistically coherent model preventing temporal correlation and additional biases in other techniques. On the basis of these results, we discuss the implications of DEM uncertainty and offer suggestions for the glaciological community.

Paired Satellite and NWP Precipitation for Global Flood Forecasting

Abstract Precipitation data are known to be the key driver of hydrological simulations. Hence, reliable quantitative precipitation estimates and forecasts are vital for accurate hydrological forecasting. Satellite-based precipitation estimates from Integrated Multi-satellitE Retrievals for GPM Early Run (IMERG-E) and forecasted precipitation from NASA’s Goddard Earth Observing System Forward Processing (GEOS-FP) have shown values in global flood nowcasting and forecasting. However, few studies have comprehensively evaluated their hydrological performance let alone explored the potential value of combining them. Therefore, this study undertakes a quasi-global evaluation of their utility in real-time hydrological monitoring and 1–5-day forecasting with the Dominant River Tracing-Routing Integrated with Variable Infiltration Capacity (VIC) Environment (DRIVE) model. The gauge-corrected IMERG Final Run precipitation estimates and corresponding hydrological simulation are used as the references. Results showed that the hit bias is the dominant error source of IMERG-E, while the false precipitation is more noticeable in GEOS-FP. In terms of hydrological performance, the GEOS-FP-driven model (DRIVE-FP) performance is close to the IMERG-E-driven model (DRIVE-E) performance on day 1, indicating that GEOS-FP could nicely fill the gap of nowcasting caused by the IMERG-E time latency. For longer lead-time forecasts, the bias tends to diminish in most regions, likely because the under- or overestimation in IMERG-E is generally offset by the distinct types of misestimation in GEOS-FP. The skillful initial hydrological conditions present outperformed forecasts in most regions, except for tropical areas where the accuracy of GEOS-FP prevails. Overall, this study provides a valuable view of the combined use of IMERG-E and GEOS-FP precipitation in the context of hydrological nowcasts and forecasts.

ДОРОЖНАЯ КАРТА СОВРЕМЕННЫХ ГИДРОЛОГИЧЕСКИХ ПРОГНОЗОВ ВОДНОГО РЕЖИМА

An array of information sources related to the current and proposed development of methods and systems of hydrological forecasts has been analyzed. It included not only publications of recent years in the field of mathematical modeling of river flow formation processes, but also fundamental documents of a structural and conceptual type related to international cooperation and division of labor in the field of hydrometeorology. The results of the analysis allow us to identify several main directions of the road map that determines the development of this area of hydrology for the near and long term.