Daily Streamflow Research Articles

Streamflow Prediction Using Complex Networks.

The reliable prediction of streamflow is crucial for various water resources, environmental, and ecosystem applications. The current study employs a complex networks-based approach for the prediction of streamflow. The approach consists of three major steps: (1) the formation of a network using streamflow time series; (2) the calculation of the clustering coefficient (CC) as a network measure; and (3) the use of a clustering coefficient-based nearest neighbor search procedure for streamflow prediction. For network construction, each timestep is considered as a node and the existence of link between any node pair is identified based on the difference (distance) between the streamflow values of the nodes. Different distance threshold values are used to identify the critical distance threshold to form the network. The complex networks-based approach is implemented for the prediction of daily streamflow at 142 stations in the contiguous United States. The prediction accuracy is quantified using three statistical measures: correlation coefficient (R), normalized root mean square error (NRMSE), and Nash-Sutcliffe efficiency (NSE). The influence of the number of neighbors on the prediction accuracy is also investigated. The results, obtained with the critical distance threshold, reveal that the clustering coefficients for the 142 stations range from 0.799 to 0.999. Overall, the prediction approach yields reasonably good results for all 142 stations, with R values ranging from 0.05 to 0.99, NRMSE values ranging from 0.1 to 12.3, and the NSE values ranging from -0.89 to 0.99. An attempt is also made to examine the relationship between prediction accuracy and the catchment characteristics/streamflow statistical properties (drainage area, mean flow, coefficient of variation of flow). The results suggest that the prediction accuracy does not have much of a relationship with the drainage area and the mean streamflow values, but with the coefficient of variation of flow. The outcomes from this study are certainly promising regarding the application of complex networks-based concepts for the prediction of streamflow (and other hydrologic) time series.

Improving runoff simulation in the Western United States with Noah-MP and variable infiltration capacity

Abstract. Streamflow predictions are critical for managing water resources and for environmental conservation, especially in the water-short Western United States. Land surface models (LSMs), such as the variable infiltration capacity (VIC) model and the Noah LSM with multiparameterization options (Noah-MP), play an essential role in providing comprehensive runoff predictions across the region. Virtually all LSMs require parameter estimation (calibration) to optimize their predictive capabilities. Here, we focus on the calibration of VIC and Noah-MP models at a 1/16° latitude–longitude resolution across the Western United States. We first performed global optimal calibration of parameters for both models for 263 river basins in the region. We find that the calibration significantly improves the models' performance, with the median daily streamflow Kling–Gupta efficiency (KGE) increasing from 0.37 to 0.70 for VIC, and from 0.22 to 0.54 for Noah-MP. In general, post-calibration model performance is higher for watersheds with relatively high precipitation and runoff ratios, and at lower elevations. At a second stage, we regionalize the river basin calibrations using the donor-basin method, which establishes transfer relationships for hydrologically similar basins, via which we extend our calibration parameters to 4816 hydrologic unit code (HUC)-10 basins across the region. Using the regionalized parameters, we show that the models' capabilities to simulate high and low flow conditions are substantially improved following calibration and regionalization. The refined parameter sets we developed are intended to support regional hydrological studies and hydrological assessments of climate change impacts.

Variability in flood frequency in sub-Saharan Africa: The role of large-scale climate modes of variability and their future impacts

Sub-Saharan Africa (SSA) is strongly affected by flood hazards, endangering human lives and economic stability. However, the role of internal climate modes of variability in driving fluctuations in SSA flood occurrence remains poorly documented and understood. To address this gap, we quantify the relative and combined contribution of large-scale climate drivers to seasonal and regional flood occurrence using a new 65-year daily streamflow dataset, sea-surface temperatures derived from observations, and 12 Single Model Initial-condition Large Ensembles (SMILEs) from the Coupled Model Intercomparison Project Phases 5 and 6. We find significant relationships between floods and large-scale climate variability across SSA, with climatic drivers accounting for 30–90 % of the variability in floods. Notably, western, central, and the summer-rain region of southern Africa display stronger teleconnections to large-scale climate variability in comparison to East Africa and the winter-rain region of South Africa, where regional circulation patterns and human activities may play a more important role. In southern and eastern Africa, floods are mainly influenced by teleconnections with the Pacific and Indian Oceans, while in western and central Africa, teleconnections with the Atlantic Ocean and Mediterranean Sea play a larger role. We also find that the number of floods is projected to fluctuate by ± 10–50 % during the 21st century in response to different sequences of key modes of climate variability. We also note that the relative contributions of large-scale climate variability to future flood risks are generally consistent across all SMILEs. Our findings thus provide valuable information for long-term disaster risk reduction and management.

Comparing the hybrid-lumped-LSTM model with a semi-distributed model for improved hydrological modeling

ABSTRACT Accurate hydrological modeling is essential for understanding and managing water resources. This study conducts a comparative analysis of hydrological modeling strategies in a date-scarce region. This study examines lumped (IHACRES), semi-distributed (HEC-HMS), and hybrid-lumped/long short-term memory (LSTM) models, aiming to assess their performance and accuracy in a data-scarce region. It investigates whether lump models can accurately simulate flow and evaluates the impact of combining lump models with machine learning to enhance accuracy, compared to semi-distributed models. The IHACRES model underestimates discharge, but its commendable NSE during calibration (0.628) and validation (0.681) signifies reliable simulation. The HEC-HMS model accurately depicts daily streamflow but struggles with extreme events, showcasing limitations in predicting maximum flows. The hybrid-lumped/LSTM model exhibits improved accuracy over IHACRES. Despite some underestimation, it mitigates IHACRES limitations during extreme events. However, challenges persist in simulating high flows, emphasizing the necessity for further refinement. The findings contribute to the discourse on merging machine learning with traditional hydrological models in data-scarce regions. The hybrid model offers promise but underscores the need for ongoing research to optimize performance, especially during extreme events. This study provides valuable insights for advancing hydrological modeling capabilities in complex watersheds.

Multiple data-driven approaches for estimating daily streamflow in the Kone River basin, Vietnam

Multiple data-driven approaches for estimating daily streamflow in the Kone River basin, Vietnam

Optimization of the SWAT+ model to adequately predict different segments of a managed streamflow hydrograph

ABSTRACT Complete representation of rainfall–runoff responses in complex, large watersheds using a single-objective parameterization approach in watershed models is often unachievable. In this study we present a calibration approach for the SWAT+ model that independently fits model parameters for different flow segments of the hydrograph. The approach is demonstrated for the Feather River, California, USA, using daily streamflow from the Lake Oroville Reservoir outlet gage. Results show that when model parameters were independently fitted for different flow segments the KGE, NSE, PBIAS, and RSR values improved to 0.96, 0.99, −3.3, and 0.10, respectively, compared to 0.72, 0.66, −9.30, and 0.53, respectively, achieved under a multiobjective and full hydrograph (average hydrograph) calibration. The results highlight that when considering the average hydrograph and flow duration curves, a more balanced representation of both poorly and well-performing segments is achieved, emphasizing the importance of segment-specific parameterization and multi-objective evaluation for accurately representing different flow conditions.

A novel insight on input variable and time lag selection in daily streamflow forecasting using deep learning models

This study investigates the feasibility of using hybrid models namely Convolutional Neural Network (CNN)-Long Short-Term Memory (LSTM) and Convolutional Neural Network (CNN)-Gated Recurrent Unit (GRU), for short-to-medium range streamflow forecasting in the Mahanadi River basin in India. The performance of these hybrid models is compared with that of standalone models. It investigates the impact of selected parameters and associated time lags on the model performance and offers valuable insights into the use of hybrid models for runoff simulation. The hybrid CNN-LSTM model proves to be robust in capturing the overall time series and the typical high peak flows in both the correlation-based and constant lag cases. Also, the upstream discharges play a significant role in improving the streamflow forecasting. Furthermore, the consideration of all input variables with a constant time lag equal to the basin lag time may yield better flood forecasts, even in cases where computational resources are limited.

SWAT model to simulate hydrological responses to land use and land cover changes in a tropical catchment: A study in Sri Lanka

A significant concern is the modification and conversion of the land cover in river basins as a consequence of different influences on ecosystems. A hydrological investigation was performed in the Kelani River basin in Sri Lanka using the Soil and Water Assessment Tool (SWAT) to analyze the impact of changes in land use and land cover (LULC) on the hydrological response. In order to calibrate (2003–2012) and validate (2013–2020) the model, input data from four hydrological stations, namely Glencose, Hanwella, Norwood, and Kitulgala, were utilized. Three distinct LULC scenarios were analyzed and compared to the existing urban land pattern. Scenario I involves the conversion of urban land into pasture, Scenario II comprises the conversion of urban land into rubber plantations, and Scenario III entails the conversion of urban land into bare soil. In accordance with the study, the SWAT model's calibration in the Kelani River was successful. The evaluation indices for both the calibration and validation periods were satisfactory at Norwood (R2 = 0.76, NE = 0.88), Hanwella (R2 = 0.83, NE = 0.57), Glencourse (R2 = 0.78, NE = −0.32), and Kitulgala (R2 = 0.69, NE = 0.59). This finding suggests that the SWAT model demonstrated better accuracy in predicting streamflow at a station with high flow compared to stations with low or medium flow conditions. However, the calibrated model did not perform as expected when used to forecast daily streamflow at Glencourse, where the evaluation indices showed an R2 of 0.78 and NE of −0.32. Compared to the current urban land cover, scenario III had the greatest impact on runoff, potential evapotranspiration (PET), and sediment yield, with increases of 14.1%, 15.2%, and 26.3%, respectively. As a consequence of the significant degree of land degradation that was observed throughout the duration of the study, regions in the northern and western parts of the basin have been highlighted as requiring immediate attention for the establishment of vegetation.

Evaluation of the support vector regression (SVR) and the random forest (RF) models accuracy for streamflow prediction under a data-scarce basin in Morocco

Streamflow prediction is a key variable for water resources management. It becomes more important in semi-arid regions such as the Tensift river basin in Morocco, where water resources are facing a severe drought and the demand is continuously increasing. The present analysis focuses on evaluating Machine Learning techniques, namely support vector regression (SVR) and Random Forest (RF) against the multiple linear regression (MLR) for daily streamflow forecasting in the mountainous sub-basin of Rheraya between 2003 and 2016. The results show that SVR performed best, followed by RF and MLR. In measurable terms and regarding mean performance, SVR exhibited the higher Nash–Sutcliffe efficiency score (NSE = 0.59) and a lower root mean squared error (RMSE = 1.18 m3s-1\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\ ext {m}^3\\,\ ext {s}^{-1}$$\\end{document}) compared to RF (NSE = 0.53, RMSE = 1.18 m3s-1\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\ ext {m}^3\\,\ ext {s}^{-1}$$\\end{document}) and MLR (NSE = 0.54, RMSE = 1.01 m3s-1\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\ ext {m}^3\\,\ ext {s}^{-1}$$\\end{document}). Furthermore,the available time series was too short to properly capture the full range of streamflow variability, which reduced the prediction performance outside of the calibration conditions. These findings suggest that ML algorithms, particularly SVR, can provide accurate streamflow estimation useful for water resources management when trained on a representative period. The results highlight the capacity of Machine Learning algorithms, specifically SVR, to augment streamflow prediction for enhanced water resource management in arid regions.

Evaluating the significance of wetland representation in isotope-enabled distributed hydrologic modeling in mesoscale Precambrian shield watersheds

Evaluation of regional-scale models against both streamflow and stable isotopes of water is a relatively new and evolving area of distributed hydrological modeling. While applications of isoWATFLOOD, an isotope-enabled distributed hydrological model are increasing, evaluation of the impacts of model structural features, such as wetland connectivity, on simulation of runoff components is still limited. Wetlands are important water bodies that impact streamflow generation and its spatial variation in Precambrian shield watersheds in parts of Canada and elsewhere. This study investigates the efficiency of isoWATFLOOD in simulating streamflow and instream δ18O and the effects of model percent connected wetlands (%CW) in a mesoscale (∼12,000 km2) Precambrian Shield watershed in Northeastern Ontario, Canada. Model calibration and validation are performed using daily streamflow for five hydrometric gauging stations and biweekly to monthly instream δ18O from 11 river sampling locations from across the watershed. Five model versions with different %CW are generated, calibrated, and validated separately. Results show all models can simulate the timing and pattern of streamflow and δ18O. %CW impacts simulated streamflow during both calibration and validation periods, calibrated parameter sets and simulated internal hydrological processes. Model performance varies spatially with no single value of %CW providing best performance across sub-catchments with 40 % CW providing best mean streamflow simulations in calibration. Increasing %CW (10 % to 40 %) results in higher contribution of lower soil zone water to streamflow that markedly improves baseflow simulation. This study builds on recent isotope-enabled hydrologic simulation capacity of isoWATFLOOD, contributing new examination of model representation of wetland connectivity, a critical component of Precambrian Shield watersheds and evidence of how stable isotopes can contribute to addressing the challenge of equifinality. Results suggest the need to develop model capability for representation of spatial variation in wetland connectivity to recognize this important aspect of heterogeneity in watersheds influencing hydrologic function.

A comparative study of daily streamflow forecasting using firefly, artificial bee colony, and genetic algorithm-based artificial neural network

A comparative study of daily streamflow forecasting using firefly, artificial bee colony, and genetic algorithm-based artificial neural network

Frequency curves of high and low flows in intermittent river basins for hydrological analysis and hydraulic design

Upper and lower percentiles of Flow Duration Curves (FDCs) of daily streamflow data were investigated to develop frequency curves. Upper percentiles with exceedance probability of 1, 5 and 10% (Q1, Q5, Q10) were used for high flows, and lower percentiles with non-exceedance probability of 10, 5 and 1% (Q90, Q95, Q99) for low flows. Median value (Q50) was covered to represent the average conditions of streamflow. A mixed frequency analysis based on the total probability theorem taking zero values into account was applied for the lower percentiles of FDC. Case studies were performed for three intermittent Streamflow Gauging Stations (SGSs) from Kucuk Menderes River Basin in western Turkey. An overall assessment of results shows that the best-fit probability distribution function does not change from one SGS to another considerably for low flows while each SGS has its own probability distribution function for high flows. Upper and lower percentiles, and median value were calculated at various return periods by using the identified probability distribution functions. The calculated values were plotted in the form of frequency curves of high flow percentiles and low flow percentiles. The frequency curves have a practically significant potential use in hydrological analysis, water resources management and hydraulic design under high and low flow conditions. They are yet open to further development for regionalization and their applicability can be extended to ungauged sites in river basins.

Interpretable and explainable hybrid model for daily streamflow prediction based on multi-factor drivers.

Streamflow time series data typically exhibit nonlinear and nonstationary characteristics that complicate precise estimation. Recently, multifactorial machine learning (ML) models have been developed to enhance the performance of streamflow predictions. However, the lack of interpretability within these ML models raises concerns about their inner workings and reliability. This paper introduces an innovative hybrid architecture, the TCN-LSTM-Multihead-Attention model, which combines two layers of temporal convolutional networks (TCN) followed by one layer of long short-term memory (LSTM) units, integrated with a Multihead-Attention mechanism for predicting streamflow with streamflow causation-driven prediction samples (RCDP), employing local and global interpretability studies through Shapley values and partial dependency analysis. The find_peaks method was used to identify peak flow events in the test dataset, validating the model's generality and uncovering the physical causative patterns of streamflow. The results show that (1) compared to the LSTM model with the same hyperparameter settings, the proposed TCN-LSTM-Multihead-Attention hybrid model increased the R2 by 52.9%, 2.5%, 43.1%, and 10.7% respectively at four stations in the test set predictions using RCDP samples. Moreover, comparing the prediction results of the hybrid model under different samples in Hengshan station, the R2 for RCDP increased by 5.06% and 1.22% compared to streamflow autoregressive prediction samples (RAP) and meteorological-soil volumetric water content coupled autoregressive prediction samples (MCSAP) respectively. (2) Historical streamflow data from the preceding 3days predominantly influences predictions due to strong autocorrelation, with flow quantity (Q) typically emerging as the most significant feature alongside precipitation (P), surface soil moisture (SSM), and adjacent station flow data. (3) During periods of low and normal flow, historical data remains the most crucial factor; however, during flood periods, the roles of upstream inflow and precipitation become significantly more pronounced. This model facilitates the identification and quantification of various hydrodynamic impacts on flow predictions, including upstream flood propagation, precipitation, and soil moisture conditions. It also elucidates the model's nonlinear relationships and threshold responses, thereby enhancing the interpretability and reliability of streamflow predictions.

Exploring the potential of deep learning for streamflow forecasting: A comparative study with hydrological models for seasonal and perennial rivers

Improving streamflow prediction plays a significant role in flood warning, mitigation and development purposes. Therefore, this paper aims to compare the prediction capability of a useful conceptual hydrologic model hydrologiska byråns vattenbalansavdelning (HBV) enhanced and equipped with an automatic calibration procedure (Genetic Algorithm) along with two well-known deep learning (DL) algorithms, including Long-Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU). This study was concentrated on the daily streamflow of two river basins with diverse climatic and hydrological conditions, namely the Halilrood River with fragmented and intermittent flows (located in Iran) and the Tweed River with a permanent stream flow (Located across the border of Scotland and northern England). These were particularly selected to address both seasonal and perennial riverine cases. For reproducing the streamflow, 10-year records of the daily discharge, precipitation, and temperature (from satellite and ground resources), and evapotranspiration were used as the main input data. The accuracy of the implemented models was assessed using several indices. The findings show that 1) the LSTM and GRU show better accuracy than the employed hydrologic model for estimating daily discharges (especially in the seasonal river basin); 2) the DL approaches need only meteorological data inputs (precipitation and temperature) and are independent of involved uncertainties in hydrological modeling such as physical characteristics of basins; and 3) best streamflow simulations were achieved by the LSTM (best agreement with the observations). In general, LSTM has the highest priority to be connected with online hydrological rainfall-runoff simulations and dynamical forecast models.

Daily Streamflow Forecasting Using Networks of Real-Time Monitoring Stations and Hybrid Machine Learning Methods

Considering the increased risk of urban flooding and drought due to global climate change and rapid urbanization, the imperative for more accurate methods for streamflow forecasting has intensified. This study introduces a pioneering approach leveraging the available network of real-time monitoring stations and advanced machine learning algorithms that can accurately simulate spatial–temporal problems. The Spatio-Temporal Attention Gated Recurrent Unit (STA-GRU) model is renowned for its computational efficacy in forecasting streamflow events with a forecast horizon of 7 days. The novel integration of the groundwater level, precipitation, and river discharge as predictive variables offers a holistic view of the hydrological cycle, enhancing the model’s accuracy. Our findings reveal that for a 7-day forecasting period, the STA-GRU model demonstrates superior performance, with a notable improvement in mean absolute percentage error (MAPE) values and R-square (R2) alongside reductions in the root mean squared error (RMSE) and mean absolute error (MAE) metrics, underscoring the model’s generalizability and reliability. Comparative analysis with seven conventional deep learning models, including the Long Short-Term Memory (LSTM), the Convolutional Neural Network LSTM (CNNLSTM), the Convolutional LSTM (ConvLSTM), the Spatio-Temporal Attention LSTM (STA-LSTM), the Gated Recurrent Unit (GRU), the Convolutional Neural Network GRU (CNNGRU), and the STA-GRU, confirms the superior predictive power of the STA-LSTM and STA-GRU models when faced with long-term prediction. This research marks a significant shift towards an integrated network of real-time monitoring stations with advanced deep-learning algorithms for streamflow forecasting, emphasizing the importance of spatially and temporally encompassing streamflow variability within an urban watershed’s stream network.

Simbi: historical hydro-meteorological time series and signatures for 24 catchments in Haiti

Abstract. Haiti, a Caribbean country, is highly vulnerable to hydroclimatic hazards due to heavy rainfall, which is partly linked to tropical cyclones. Additionally, its steep slopes generate flash floods, particularly in small catchments. Moreover, the hydrology of this region remains poorly understood and understudied. Unfortunately, there is no accessible database for the scientific community to use in this country. To fill this gap, hydroclimatic data were collected to create the first historical database in Haiti. This database, called Simbi (guardian of rivers, freshwater, and rain in Haitian mythology), includes 156 monthly rainfall series over the period 1905–2005, 59 daily rainfall series over the period 1920–1940, 70 daily streamflow series, and 23 monthly temperature series, not necessarily continuous, over the period 1920–1940. It also provides simulated streamflow series over the period 1920–1940 using the GR2M and GR4J rainfall–runoff models for 24 catchments and 49 attributes covering a wide range of topographic, climatic, geological, land use, hydrogeological, and hydrological signature indices. Simbi is the first open-access hydro-meteorological dataset for Haiti and will contribute to a better knowledge of hydrological risk in Haiti. Several sources of uncertainty associated with Simbi are acknowledged, including data quality (historical data), digitisation of paper archives, identification of relevant rain gauges, and rainfall–runoff models. It is important to consider these uncertainties when using Simbi. The database will be regularly updated to include additional historical data that will be digitised in the future. It will thus contribute toward better knowledge of the hydrology of Haitian catchments and will enable the implementation of various hydrological calculations useful for designing structures or flow forecasting. Simbi is an open-access database and is available for download at https://doi.org/10.23708/02POK6 (Bathelemy et al., 2023).

Indirect effects of revegetation dominate groundwater recharge change at the basin scale

Attribution of groundwater recharge (GR) change is important for understanding subsurface hydrological processes and managing regional groundwater resources. Despite vegetation change directly influencing GR, its indirect effects are mostly incorporated into other effects because of multicollinearity in environmental factors. Taking the 59 hydrological stations from 13 basins in the Loess Plateau of China as an example, GR was estimated with daily streamflow data and recession-curve-displacement method for 1961–2019. Geographical detector was employed to qualitatively attribute the spatial variability of GR, while structural equation modeling and elasticity coefficient method were utilized for qualitative and quantitative analysis of the direct and indirect impacts of environmental factors on the temporal variation of GR. The annual mean GR in 13 basins ranged from 11 to 53 mm/year, and the spatial heterogeneity of GR can be well explained by loess thickness. GR significantly decreased in 68% of hydrological stations by 38% ± 15%. Averaged over the 13 basins, the direct and indirect effects of NDVI change respectively accounted for 10% and 48% of GR reduction, which greatly exceeded other factors. The indirect effects of vegetation change dominated GR change, but they were getting smaller with increasing latitude. This study highlights the importance in considering indirect effects of vegetation change in hydrological variability, and the results can improve the understanding of hydrological variability.

Unveiling hydrological dynamics in data-scarce regions: experiences from the Ethiopian Rift Valley Lakes Basin

Abstract. The hydrological system of the Rift Valley Lakes in Ethiopia has recently experienced changes over the past 2 decades. Potential causes for these changes include anthropogenic, hydro-climatic, and geological factors. The main objective of this study was to utilize an integrated methodology to gain a comprehensive understanding of the hydrological systems and potential driving factors within a complex and data-scarce region. To this end, we integrated a hydrologic model, change point analysis, indicators of hydrological alteration (IHA), and a bathymetry survey to investigate hydrological dynamics and potential causes. A hydrologic model (the Soil and Water Assessment Tool Plus, SWAT+) was parameterized for the gauged watersheds and extended to the ungauged watersheds using multi-site regionalization techniques. The SWAT+ model performed very well to satisfactorily for daily streamflow in all watersheds with respect to the objective functions of the Kling–Gupta efficiency (KGE), the Nash–Sutcliffe efficiency (NSE), and percent bias (PBIAS). The findings reveal notable changes in lake inflows and lake levels over the past 2 decades. Lake Chamo experienced an increase in area of 30.1 km2 (9.5 %), an increase in depth of 4.4 m (30.9 %), and an increase in volume of 7.8×108 m3 (27.2 %). In contrast, Lake Abijata witnessed an extraordinary 68 % decrease in area and a depth decrease of 1.6 m (37.2 %). During the impact period, the mean annual rainfall experienced a decrease of 6.5 % and 2.7 % over Lake Abijata and Lake Chamo, respectively. Actual evapotranspiration decreased by 2.9 % in Lake Abijata but increased by up to 4.5 % in Lake Chamo. Surface inflow to Lake Abijata decreased by 12.5 %, while Lake Chamo experienced an 80.5 % increase in surface inflow. Sediment depth in Lake Chamo also increased by 0.6 m (4.2 %). The results highlight that the changing hydrological regime in Lake Chamo is driven by increased surface runoff and sediment intrusion associated with anthropogenic influences. The hydrological regime of Lake Abijata is affected by water abstraction from feeding rivers and lakes for industrial and irrigation purposes. This integrated methodology provides a holistic understanding of complex data-scarce hydrological systems and potential driving factors in the Rift Valley Lakes in Ethiopia, which could have global applicability.

Assessment of machine learning models to predict daily streamflow in a semiarid river catchment

Assessment of machine learning models to predict daily streamflow in a semiarid river catchment

Evaluation of streamflow predictions from LSTM models in water- and energy-limited regions in the United States

The application of Long Short-Term Memory (LSTM) models for streamflow predictions has been an area of rapid development, supported by advancements in computing technology, increasing availability of spatiotemporal data, and availability of historical data that allows for training data-driven LSTM models. Several studies have focused on improving the performance of LSTM models; however, few studies have assessed the applicability of these LSTM models across different hydroclimate regions. This study investigated the single-basin trained local (one model for each basin), multi-basin trained regional (one model for one region), and grand (one model for several regions) models for predicting daily streamflow in water-limited Great Basin (18 basins) and energy-limited New England (27 basins) regions in the United States using the CAMELS (Catchment Attributes and Meteorology for Large-sample Studies) data set. The results show a general pattern of higher accuracy in daily streamflow predictions from the regional model when compared to local or grand models for most basins in the New England region. For the Great Basin region, local models provided smaller errors for most basins and substantially lower for those basins with relatively larger errors from the regional and grand models. The evaluation of one-layer and three-layer LSTM network architectures trained with 1-day lag information indicates that the addition of model complexity by increasing the number of layers may not necessarily increase the model skill for improving streamflow predictions. Findings from our study highlight the strengths and limitations of LSTM models across contrasting hydroclimate regions in the United States, which could be useful for local and regional scale decisions using standalone or potential integration of data-driven LSTM models with physics-based hydrological models.