In recent years, data-driven modeling techniques have gained numerous applications in hydrology and water resources studies. River runoff estimation and forecasting is one of the research fields in which these techniques have several applications. In the current study, four data-driven modeling techniques of multiple linear regression, K-nearest neighbors, artificial neural networks, and adaptive neuro-fuzzy inference systems have been used to form runoff forecasting models and then their results have been evaluated. Also, effects of using some different scenarios to select predictor variables have been studied. It was evident from the results that using flow data related to one or two months ago in the predictor variables dataset can improve the accuracy of the results. In addition, comparison of general performances of the modeling techniques showed superiority of KNN models results among the studied models. The selected KNN model presented best performance with a linear correlation coefficient equal to 0.84 between observed flow data and predicted values and a RMSE equal to 2.64.

In this study, applicability of successive-station prediction models, as a practical alternative to streamflow prediction in poor rain gauge catchments, has been investigated using monthly streamflow records of two successive stations on Çoruh River, Turkey. For this goal, at the first stage, based on eight different successive-station prediction scenarios, feed-forward back-propagation (FFBP) neural network algorithm has been applied as a brute search tool to find out the best scenario for the river. Then, two other artificial neural network (ANN) techniques, namely generalized regression neural network (GRNN) and radial basis function (RBF) algorithms, were used to generate two new ANN models for the selected scenario. Ultimately, a comparative performance study between the different algorithms has been performed using Nash-Sutcliffe efficiency, squared correlation coefficient, and root-mean-square error measures. The results indicated a promising role of successive-station methodology in monthly streamflow prediction. Performance analysis showed that only 1-month-lagged record of both stations was satisfactory to achieve accurate models with high-efficiency value. It is also found that the RBF network resulted in higher performance than FFBP and GRNN in our study domain.

River flow is one of the most important components of the hydrological cycle, which depends on several climatic factors and its accurate estimation is used in various fields of water resources management. Therefore, in the present study, random forest (RF) and support vector machine (SVM) models were used to predict the monthly streamflow of the Maroon River in the period of 1981- 2017. One of the important steps in the application of artificial intelligence models is the definition of input patterns and determining the effective variables in the modeling process. The Shannon entropy method was used to select the most efficient inputs among precipitation, evaporation, and minimum, maximum, and average temperatures. The results showed that the total weight of precipitation and evaporation was more than 85 percent. In the next step, three different structures were developed for modeling. In the first case, climate-based patterns were defined that used meteorological data as input. In the second case, nonlinear periodicity was added to the climate-based patterns, and in the third case, the climate-based input data were decomposed using five mother wavelet functions, and W-RF and W-SVM hybrid models were created. The performance evaluation of the standalone RF and SVM models showed that by considering the periodic term, the accuracy is somewhat increased compared to the climate-based inputs, but the analysis of the data with wavelet theory significantly reduced the modeling error. In the meantime, the performance of the two models W-RF and W-SVM was very close to each other, but according to the violin plot, the W-SVM model is suggested as the most suitable option for predicting the monthly streamflow of the Maroon River.

Quantifying water resources is essential for developing evidence-based management strategies. Hydrological models play a great role in estimating streamflow, particularly in regions with limited flow measurement infrastructure. This study evaluates the integration of the GR4J conceptual hydrological model with Machine Learning (ML) techniques, Random Forest (RF), Extreme Learning Machine (ELM), eXtreme Gradient Boosting (XGB), and Long Short-Term Memory (LSTM) networks to improve daily streamflow prediction in the Bilate River watershed. Though GR4J captures general hydrological trends, its limitations in modeling nonlinear dynamics and extreme flows necessitate advanced approaches by augmenting GR4J’s simulated outputs with climate input features to train the ML models. The integrated models GR4J-RF, GR4J-ELM, GR4J-XGB, and GR4J-LSTM combine GR4J’s physical interpretability with ML’s capability to capture complex and nonlinear relationships, addressing the shortcomings of both the conceptual and ML methods. Findings of the study demonstrate significant improvements over standalone GR4J, with GR4J-LSTM and GR4J-XGB achieving the highest test performance (NSE of 0. 77, KGE of up to 0. 86), GR4J-RF excelling in training fit (train NSE of 0. 87) with gaps in generalization, and GR4J-ELM offering computational efficiency with comparable performance (test NSE of 0. 74). These findings highlight the potential of integrated modeling to improve streamflow prediction in data-limited regions, supporting applications such as flood prediction and drought monitoring.

Background and ObjectivesIn recent years, the use of artificial intelligence methods, such as artificial neural network models, have become increasingly prevalent in simulating complex natural phenomena, including daily streamflow. The streamflow directly correlates with flood occurrences, and mitigating financial and human losses due to floods is crucial. Accurate streamflow simulation is essential for water resource management and river management. Consequently, in hydrology, deep learning methods have emerged as novel tools to address the longstanding challenge of daily streamflow modeling and are widely used in simulations.Advancements in streamflow modeling with Artificial Intelligence (AI): In recent years, the field of hydrology has witnessed a significant shift toward leveraging AI techniques for streamflow modeling. Among these methods, artificial neural network (ANN) models have gained prominence due to their ability to capture complex relationships within hydrological systems. Streamflow, which represents the flow of water in rivers and streams, is a critical variable for understanding water availability, flood risk, and ecosystem health. By accurately simulating streamflow, researchers and water resource managers can make informed decisions regarding water allocation, flood preparedness, and environmental conservation. Hydrological processes are inherently nonlinear and influenced by various factors such as precipitation, temperature, land cover, and soil properties. Traditional hydrological models often struggle to capture these complexities. However, deep learning methods, including convolutional neural networks (CNNs) and recurrent neural networks (RNNs), offer promising solutions. These models can learn intricate patterns from historical streamflow data, adapt to changing conditions, and provide accurate predictions. As a result, they have become indispensable tools for addressing the longstanding challenge of daily streamflow modeling. Researchers continue to explore novel architectures, data augmentation techniques, and hybrid approaches to enhance the performance and robustness of AI-based streamflow simulations. In summary, the integration of deep learning methods into hydrological research has revolutionized streamflow modeling, enabling more accurate predictions and informed decision-making in water management and flood risk assessment.MethodologyIn this study, we focused on selecting an appropriate input scenario for deep learning models and simulate daily streamflow on the Kashkan River using LSTM (Long Short-Term Memory) and GRU (Gated Recurrent Unit) deep learning methods. Prior to this, deep learning modeling with the GRU approach using native streamflow measurements had not been performed for Kashkan river. The study area is a flood-prone and mountainous region, specifically the western part of Iran, where a hydrological station with a history of flood events is situated on the Kashkan River. We employ four approaches for handling outliers (Mahalanobis, critical interval removal, Z-Score, and no removal) and four different preprocessing techniques for input data to train two models: LSTM and GRU. Ultimately, eight distinct models are generated and validated against historical data. The input features include regional average precipitation, normalized vegetation cover index, surface soil moisture, groundwater flow, and the Kashkan River’s own flow at the hydrological station, with the best features selected using statistical correlation control.FindingsThe results demonstrate that among the deep learning models generated with a 10-day time step, the model with the least error and consistent low error retention in error metrics is observed. Furthermore, the best performance is achieved using different approaches, in the following order: the GRU model with Z-Score-corrected inputs, followed by the Mahalanobis removal approach with average RMSE (Root Mean Square Error) and KGE (Kling-Gupta Efficiency) values of 5.41 and 0.99, respectively, and the critical interval removal approach with RMSE of 6.23 and KGE of 0.7.The results showed that among the deep learning models produced with a time step of 10 days in the model, the lowest amount of error and the persistence of low error can be seen in the error statistics, and among the different approaches used, the best performance is the GRU model with input modified by Z-Score elimination of outlier method, Mahalanobis elimination method with average RMSE and KGE values of 5.41, 0.99, 6.23, and 0.7 in the training phase and 8.17, 0.79, 4.21, and 0.81 in the validation phase and 5.01, 0.68, and 7.21 and 0.52 are in the testing phase. The obtained results do not reject the LSTM method in simulating the river flow, but state that the listed scenarios, especially in the GRU method, have a higher power in dealing with the data and recognizing the complex pattern of daily river flow, taking into account the limitation in use They have seven years of regular daily data, and future research will show how the behavior of GRU and LSTM models will differ if data with higher convergence is used.ConclusionGRU in future studies can make difference by enhanced flood forecasting accuracy, efficient computation and real-time applications, integration with lag time preprocessing, adaptability to changing climate and urbanization. Future studies will be on data driven method in flood prone areas. There remains ample room for future research and innovation. Here are some directions for further exploration: hydrological data fusion, spatially explicit models, uncertainty quantification, climate change resilience.

The precise streamflow estimation has importance in the planning of water resources management, the forecasting of the persistence of drought and the planning of reservoir operation. The lack of long term streamflow data in the most of rivers in Iran is an obstacle to suitable water resources management. The disaggregation method is one of the stochastic methods that are the useful tool in applicable hydrology. The reliable planning and design of hydrological systems need the generation of time series in smaller time scales and various sites. Through this method, hydrological variables can be disaggregated into smaller scales, either in temporal or spatial. The temporal disaggregation is the disaggregation of the annual time into series finer ones like monthly time series. The spatial disaggregation is the disaggregation of the annual discharge of Main River into the discharge of subbranches. In this research, the disaggregating of annual time series into semi-annual and monthly ones were carried out using basic and extended models and the spatial disaggregation were carried out using extended model. The streamflow of some branches of Ouromieh river basin have been used in this model.The results showed that the good agreement of the disaggregation models with normal streamflow series, the high accuracy of the extended results (using RMSE) and the preservation of the statistical properties (mean, standard deviation) between observed and disaggregated time series.

Drought and water deficit is a challenge in arid and semi-arid regions which have recently intensified because of climate change. In recent years, the combined effect of climate change and socioeconomic factors exacerbate desertification processes, especially due to lack of water resources in land, wetlands and lakes, which appears in most parts of the country. Climate change is one of the challenges which lead to many environmental consequences. Trend analysis of river flows is an important issue in water resources planning and management and can provide valuable information. Heretofore, a numerous studies used parametric and non-parametric methods to examine the existence of significant trends in hydro-climatic time series. Most of the studies used non-parametric methods for trend analysis and a few studies used linear regression test. The non-parametric methods were used in this study because the non-parametric methods are distribution-free, robust against outliers, and have a higher power for non-normally distributed data. The Mann-Kendall (MK) method (Mann, 1945; Kendall, 1975) is the most commonly used non-parametric method that has recommended for identification of monotonic trends in different hydrologic and climatologic time series by World Meteorological Organization (WMO). The serial dependence between observations should not exist when the original classic MK test used for trend detection. However, in most of the hydro-meteorological time series, significant autocorrelation with different time lags, in addition to lag-1, may exist among observations. In such a situation, application of the classic version of the MK test for trend analysis could yield unreliable results. As some of previous studies showed that the presence of positive auto-correlation overestimates the significance of both positive and negative trends, whereas negative auto-correlation underestimates the significance of both positive and negative trends. The existence of more than one significant auto-correlation among data called as long-term persistence (LTP). To incorporate the LTP behavior in MK test, Hamed (2008) suggested to remove the effect of all significant serial correlation before applying the classic MK test. The surface water is one of the main resource for providing irrigation demand in Hamadan province. However, in recent decays, because of increasing the farm land area, the available surface water resources cannot provide the agricultural demand for water completely, so the farmers have drilled a lot of wells to extract groundwater for irrigation uses. The overexploitation of groundwater led to severe decline of the water table in most parts of the Hamadan province. In this study, the trend of river flows of the Hamadan province was investigated in monthly, seasonal and annual time scales by using Mann-Kendall non-parametric test, after removing the effect of all significant serial correlation. For this purpose, monthly stream flow data of 17 hydrometric stations during 1985 to 2013 were used. The Sen’ s slope estimator was used to estimate trend line slope. Also, the abrupt change points in the stream flow time series were detected using the Pettitt test. The results showed that in annual time scale all stations had negative trends, as about half of them were significant at the 10 % level or less. The most severe significant negative trend in 1% level belonged to Bujin station with a Z value of-3. 28. At seasonal time scale, the discharges of most rivers were experienced decreasing trend which the summer ranked first. In monthly time scale, among 204 considered series (12*17), only 15 stream flow series showed a significant positive trend (at 10% significance level) and 102 stream flow series have experienced a significant decreasing trend (at 10% significance level) and 87 series had no significant trend. The most significant negative trend of monthly stream flow series belonged to Kooshkabad stations in June with Z value of-4. 45. The maximum number of stations with significant negative trend at monthly time scale at the level of 10% or less belonged to April. The highest slope of the trend line for annual time scale belonged to the Aran station, which was equal to 0. 36 m3/s/yr. In general, trends of river flows in Hamadan province were statistically negative at 10% level. The results of applying the Pettitt test showed that in most stations, the significant change point in annual stream flow time series were occurred between 1995 and 1999. The results of investigating the trend of precipitation across the Hamadan province reveal that there is no negative trend in precipitation, and it seems the main reason of decreasing stream flow in this province is due to water extraction at the upstream of rivers in recent years. The results of the present study may be used by water resources planners to alter surface water allocations based on the trend of river flows.

Drought is known as one of the main natural hazards especially in arid and semi-arid regions where there are considerable issues in regard to water resources management. The focus of the present study is mainly on hydrological aspects of drought. For hydrological drought analysis, streamflow data is used as the key variable to identify drought events with reference to a demand specific threshold level, termed as truncation level. Thus, the objective of the present study is to (a) investigate the hydrological drought characteristics in Nesa River using streamflow data; (b) determine independent drought events, their duration, and severity using the variable truncation level approach; and (c) derive streamflow drought severity index. Based on expedience probabilities, the monthly flow duration curves for Nesa River were derived. These were utilized to estimate different dependable flows, and the values of variable truncation levels were obtained for a 75% probability level for each month. These values were used to distinguish the deficit and surplus flow periods independent drought events identified using the pooling procedure. Since 10 daily flow data were utilized, the minimum deficit flow duration was 10 days. In the following, have been identified some short duration (one or two 10-daily time step) surplus and deficit events. To decide on independent drought despite the short duration inter-event surplus has been used for a pooling procedure known as inter-event time and volume criterion (IC). Eventually, identified independent drought events and also describe their duration, severity, intensity, and DSI. Analysis of independent drought Characteristics in Nesa River indicated that are prolonged dry period in the hydrological regime of this river. In addition, based on DSI, Nesa droughts mostly are in sever category. Hence, it is suggested more realistic reload occurs in management programs of this river including storage, distribution and assign to various resources.