Simulation Of Low Flows Research Articles

A process-driven deep learning hydrological model for daily rainfall-runoff simulation

Although deep learning (DL) models, especially long-short-term memory (LSTM), demonstrate greater accuracy than process-based models in rainfall-runoff simulation, the predictions from process-based models are more physical than DL models. The main reason is that DL models have almost no process understanding capabilities like process-based models beyond their fitting capability. In this study, we developed a process-driven DL model under a unified DL architecture to improve the process awareness of DL models. To implement the model, a conceptual hydrological model (EXP-HYDRO) is implanted into a recurrent neural network (RNN) cell as a process driver for providing multi-sub-process variables related to the runoff process, and an Entity-Aware LSTM (EA-LSTM) cell is incorporated as a post-processor layer, resulting in the Process-driven RNN-EA-LSTM (PRNN-EA-LSTM). Under the assistance of the process driver, the model performance of PRNN-EA-LSTM on the 531 catchments from the Catchment Attributes and Meteorology for Large-sample Studies dataset is more robust than the pure DL model, and better than using only EXP-HYDRO as the input of EA-LSTM (i.e., EXP-HYDRO-EA-LSTM). Specifically, the median Nash-Sutcliffe efficiency (NSE) of PRNN-EA-LSTM in local and regional simulation is 0.03 and 0.02 higher than LSTM and 0.01 higher than EXP-HYDRO-EA-LSTM. Additionally, PRNN-EA-LSTM significantly enhances the low flow simulations and reduces the catchments number with negative NSE. This study demonstrates that process-based models can help DL models better represent the rainfall-runoff relationship under a unified architecture. Consequently, integrating the adaptability of process-based models into the DL architecture is anticipated to bolster the process understanding capabilities of DL models.

Large Eddy Simulation of Low Reynolds Number Flow Around a NACA0015 Airfoil with Modified Trailing Edges

The flow field of a low Reynolds number regime is three-dimensional and exceedingly complicated due to numerous forms of vortical phenomena, which has triggered the interest of many researchers. This study aims to investigate the influence of various trailing edge configurations on the aerodynamic characteristics and flow structure of airfoils. Specifically, five different configurations, namely baseline, serration, comb, and comb-serration are analyzed in detail. The study seeks to identify the configuration that provides optimal aerodynamic performance, which can then inform the design of more efficient airfoils for a range of applications, including unmanned aerial vehicle (UAV), wind energy, and automotive design. A Large-eddy simulation numerical model was developed to effectively evaluate unsteady pressure fluctuations and turbulence parameters at the source. The pressure coefficient and skin friction coefficient figures reveal that modifications result in an early negative pressure zone and uneven flow along the airfoil surfaces. This study presents numerical findings on the NACA0015 airfoil at a zero angle of attack. The pressure coefficient distribution along the airfoil surface exhibits a symmetrical pattern, with the highest pressure observed at the leading edge due to flow deceleration. Analysis of the skin friction coefficient confirms the absence of separation zones on the airfoil surface, except for the tip of the trailing edge. The velocity profile demonstrates a consistent and smooth flow, indicating stable and symmetrical conditions across the airfoil. Moreover, the velocity profiles of the baseline airfoil at zero angle of attack do not indicate any signs of flow separation. Notably, the serrated, combed and comb-serrated trailing edge configurations each yield distinctive effects on fluid flow. The serrated trailing edge promotes increased fluid velocity and pressure drop, while the combed configuration induces separation at the root. Meanwhile, the comb-serrated design sustains a continuous fluid flow over the airfoil surface. Overall, these results contribute to a comprehensive understanding of the aerodynamic behavior and flow characteristics influenced by various trailing edge configurations.

Improvement of streamflow simulation by combining physically hydrological model with deep learning methods in data-scarce glacial river basin

Robust streamflow simulation at glacial basins is essential for the improvement of water sustainability assessment, water security evaluation, and water resource management under the rapidly changing climate. Therefore, we proposed a hybrid modelling framework to link the SWAT+ model considering glacial hydrological processes (GSWAT+) with Gated Recurrent Unit (GRU) neural networks to improve the model simulations and to establish a framework for the robust simulation and forecast of high and low flows in glacial river basins, which could be further used for the explorations of extreme hydrological events under a warming climate. The performance of different models (GSWAT+, GRU, and GRU-GSWAT+, respectively) were thoroughly investigated based on numerical experiments for two data-scarce glacial watersheds in Northwest China. The results suggested that the hybrid model (GRU-GSWAT+) outperformed both the individual deep learning (DL) model (GRU) and the conventional hydrological model (GSWAT+) in terms of simulation and prediction accuracy. Notably, the proposed hybrid model considerably enhanced the simulations of low and high flows that the conventional GSWAT+ failed to capture. Furthermore, utilizing suitable data integration (DI) schemes on feature and target sequences can substantially help to strengthen model stability and representativeness for monthly and annual streamflow sequences. Specifically, introducing one order differential method and decomposition approach, such as ensemble empirical signal decomposition (EEMD) and complete EEMD with adaptive noise (CEEMDAN), into feature and target sequences enriched the learnable ancillary information, which consequently strengthened the predictive performance of the proposed model. Overall, the proposed hybrid model with the suitable DI scheme has the potential to significantly enhance the accuracy of streamflow simulation in data-scarce glacial river basins. This hybrid model not only upheld the fundamental physical principles from the GSWAT+ model, but also considerably mitigated the accumulated bias errors, which caused by the shortage of climate data and inadequate hydrological principles, by using DL based model and DI schemes.

Multi-model analysis of historical runoff changes in the Lancang-Mekong River Basin – Characteristics and uncertainties

Runoff changes are critical to the sustainable water resource in the Lancang-Mekong River Basin (LMRB). Changes in the LMRB’s runoff over the past decades are unclear because of inadequate streamflow observations. The advancement of global hydrological models (GHMs) has facilitated the understanding of runoff change worldwide. However, it is required to evaluate the performance of GHMs in simulating historical runoff change in the LMRB before assessing the runoff changes. This study aims to conduct a multi-model analysis of temporal-spatial changes in runoff in the LMRB for 1971–2010 using ten GHMs from the Inter-Sectoral Impact Model Intercomparison Project (ISIMIP) and evaluate the corresponding uncertainties among models. Results show that the model ensemble mean has the best performance than the individuals when compared with the reference data. Based on the model ensemble mean, large spatial heterogeneity of runoff is found in the LMRB, with an overall slightly positive trend (8.03%). Besides, the models perform better in estimating the trends of high flow than low flow. As to the trend of runoff in the wet and dry seasons, about 32% (70%) of the basin became drier (wetter) in the dry (wet) season. Meanwhile, 17% of the basin has experienced a trend of drier dry seasons and wetter wet seasons. Overall, our results highlight the uncertainty of the runoff changes in the LMRB in the low flow simulation, particularly requiring more attention in future model improvement. The complex change patterns of the runoff suggest the importance of accurate runoff observations and projections for better water management.

Comparison of the Calibrated Objective Functions for Low Flow Simulation in a Semi-Arid Catchment

Low flow simulation by hydrological models is a common solution in water research and application. However, knowledge about the influence of the objective functions is limited in relatively arid regions. This study aims to increase insight into the difference between the calibrated objective functions by evaluating eight objectives in three different classes (single objectives: KGE(log(Q)) and KGE(1/Q); multi objectives: KGE(Q)+KGE(log(Q)), KGE(Q)+KGE(1/Q), KGE(Qsort)+KGE(log(Qsort)) and KGE(Qsort)+KGE(1/Qsort); Split objectives: split KGE(Q) and split (KGE(Q)+KGE(1/Q))) in Bahe, a semi-arid basin in China. The calibrated model is Xin An Jiang, and the evaluation is repeated under varied climates. The results show a clear difference between objective functions for low flows, and the mean of KGE and logarithmic transformed-based KGE in time series (KGE(Q)+KGE(log(Q))) presents the best compromise between the estimation for low flows and general simulation. In addition, the applications of the inverse transformed-based KGE (KGE(1/Q)) and the Flow Duration Curve-based series (Qsort) in objectives are not suggested.

A Calibration‐Free Groundwater Module for Improving Predictions of Low Flows

AbstractGroundwater modules are critically important to the simulation of low flows in physically based land surface models (LSMs) and conceptual rainfall‐runoff models (HBV). Here, we develop a Groundwater for Ungauged Basins (GrUB) module that uses only physically based properties for which data are widely available, thus allowing its application without the need for calibration. GrUB is designed to be computationally simple and readily adaptable to a wide variety of LSMs and rainfall‐runoff models. We assess the performance of GrUB in 84 United States watersheds by incorporating it into HBV, a popular rainfall‐runoff model. We compare predictions of low flows by the native (calibrated) HBV groundwater module with those by the (uncalibrated) GrUB module and find that GrUB generates error metrics that are equivalent or superior to those generated by the (calibrated) HBV groundwater module. To assess whether predictions by GrUB are robust to changes in the structure and parameterization of the overlying hydrologic model, we run tests for two artificial scenarios: Slow Recharge with rates of percolation below 0.1 mm/day, and Fast Recharge with rates of percolation of up to 1,000 mm/day. GrUB proves to be robust to these extreme changes, with mean absolute error (MAE) of predictions of low flows only increasing by an average of up to 17%, while average MAE increases by up to 158% when the same tests are performed on HBV without the GrUB module. We suggest GrUB as a potential tool for improving predictions of low flows in LSMs, as well as rainfall‐runoff models when calibration data are sparse.

Use of Abstraction and Discharge Data to Improve the Performance of a National‐Scale Hydrological Model

AbstractAcross the UK, water abstracted from ground, surface, and tidal stores is regulated through a system of licenses to protect both the sources and the environment. Similar permits are required for discharging wastewater to rivers or onto the ground. These abstractions and discharges can have a significant impact on UK Rivers, but measurements are not readily available, which discourages their use in hydrological models of river flows. However, these very unique data sets provide a means to improve the performance of spatially distributed hydrological models, particularly during periods when abstraction regulations change and at ungauged river locations. To demonstrate this, point source abstraction and discharge measurements across England have been transformed into 1 × 1 km resolution gridded data and used with an enhanced formulation of the Grid‐to‐Grid (G2G) hydrological model where these processes are mathematically represented. A comparison of G2G‐simulated and gauged river flows for 605 catchments across England between 1999 and 2014 indicates that model simulation of river flows is generally improved at gauged locations downstream of abstraction/discharge sites. The main improvement is in the simulation of low flows, for which the median performance is improved by 10.7%, however, the impact on simulation of high river flows is more modest (1.5% improvement). These results demonstrate the potential gains available to the international hydrological and land‐surface modeling community from using records of actual water use (where available) in models, in place of more widely used national statistics.

Hierarchical variable fidelity evolutionary optimization methods and their applications in aerodynamic shape design

This paper proposes two hierarchical evolutionary optimization methods based on variable fidelity analysis and search space contraction for aerodynamic shape design, i.e. hierarchical evolutionary Pareto and Nash games. One of these techniques is used in the optimization process, namely the advantages of high and low fidelity flow simulation. The high-fidelity model provides solution accuracy while the low-fidelity model reduces the computational cost. Especially, the search space contraction and the population size reduction are introduced in the process of transition from the optimization on low fidelity simulation to the optimization on high fidelity simulation, so that the optimization based on high fidelity simulation can get the high-precision optimal solution with a relatively less Central Processing Unit(CPU) cost. They are applied to the single objective natural laminar wing shape design at transonic flow and the multidisciplinary shape optimization of a hypersonic air-breathing vehicle respectively. The optimization results show that regardless of a single objective or multi-objective/multidisciplinary optimization problem, the new hierarchical optimization methods proposed in this paper can improve the optimization efficiency by 5-10 times.

Effect of Shear Displacement and Stress Changes on Fracture Hydraulic Aperture and Flow Anisotropy

Fluid flow in fractures has the potential to drastically change the economic, environmental, and safety risks associated with a subsurface operation. This work focuses on the analysis of experimental shear fracturing and subsequent shear displacement of a shale specimen that contains multiple pre-existing natural fractures. A triaxial direct shear apparatus with concurrent µCT imaging allowed continuous monitoring of sample permeability and changes in fracture geometry during fracture creation, displacement, and changes in stress. Steady-state, low Reynolds number fluid flow simulations were used to determine the changes of each individual fracture’s hydraulic aperture due to displacement and effective stress. The development of anisotropy in the hydraulic aperture of a generated shear fracture was determined for shear displacements of 0.38, 0.74, and 2.04 mm. At high confining stress (30 MPa), shear displacement significantly reduced the hydraulic aperture of the shear fracture to below flow detectable limits and X-ray imaging resolution (23 × 23 × 23 µm3). The shear generated fracture was found to be more sensitive to changes in effective stress compared to the pre-existing natural fractures. Existing models derived from fluid flow equations accounting for stochastic fracture roughness were able to partially fit the simulated flow properties of the pre-existing fractures, but no model was found to fit the flow properties of the newly formed shear fracture. These results are the first report of simultaneous changes in the geometry and flow properties of surrounding, parallel natural fractures as a shear fracture is generated with varying stress and shear displacement conditions.

Performance dependence of multi-model combination methods on hydrological model calibration strategy and ensemble size

The multi-model combination is a technique to improve the performances of hydrological streamflow simulations. An area that has not been investigated much is the performance dependence of combination techniques on the hydrological model calibration strategy and ensemble size. This study aims at investigating the joint effect of the hydrological models, calibration strategies and ensemble sizes on combination abilities for selecting the most appropriate multi-model combination method. The ensemble members were constructed by applying four hydrological models and four objective functions over 383 catchments in China. The ensemble members were combined by using nine commonly used methods, which are Equal Weights (EWA), Akaike Information Criterion (AICA), Bayes Information Criterion (BICA), Bates and Granger (BGA), Granger Ramanathan A, B, and C (GRA, GRB, and GRC), Bayesian Model Averaging (BMA) and Multi-model Super Ensemble (MMSE). The GRC is found as the best multi-model combination method for hydrological simulations. Adding ensemble members by either multiple hydrological models or calibration strategies could help to improve the simulation abilities. Specifically, the increase of ensemble members can obviously enhance the performance of multi-model combinations when the ensemble size is less than six, while only limited improvement is achieved when the ensemble size is more than nine. The combination of ensemble members with various calibration strategies is hard to compensate for the weakness of hydrological model structures. As well, the application of a single calibration strategy in ensemble members only emphasizes single discharge periods and neglects other important discharge periods. This study found that various models with different objective functions are more robust and efficient. The combination performs better than any individual model in terms of Nash–Sutcliffe efficiency (NSE) for approximately 70% catchments, but the multi-model combination is less efficient in terms of low-flow simulations.

Component Combination Test to Investigate Improvement of the IHACRES and GR4J Rainfall–Runoff Models

Rainfall–runoff models are not perfect, and the suitability of a model structure depends on catchment characteristics and data. It is important to investigate the pros and cons of a rainfall–runoff model to improve both its high- and low-flow simulation. The production and routing components of the GR4J and IHACRES models were combined to create two new models. Specifically, the GR_IH model is the combination of the production store of the GR4J model and the routing store of the IHACRES model (vice versa in the IH_GR model). The performances of the new models were compared to those of the GR4J and IHACRES models to determine components improving the performance of the two original models. The suitability of the parameters was investigated with sensitivity analysis using 40 years’ worth of spatiotemporally different data for five catchments in Australia. These five catchments consist of two wet catchments, one intermediate catchment, and two dry catchments. As a result, the effective rainfall production and routing components of the IHACRES model were most suitable for high-flow simulation of wet catchments, and the routing component improved the low-flow simulation of intermediate and one dry catchments. Both effective rainfall production and routing components of the GR4J model were suitable for low-flow simulation of one dry catchment. The routing component of the GR4J model improved the low- and high-flow simulation of wet and dry catchments, respectively, and the effective rainfall production component improved both the high- and low-flow simulations of the intermediate catchment relative to the IHACRES model. This study provides useful information for the improvement of the two models.

Mitigating Prediction Error of Deep Learning Streamflow Models in Large Data‐Sparse Regions With Ensemble Modeling and Soft Data

AbstractPredicting discharge in contiguously data‐scarce or ungauged regions is needed for quantifying the global hydrologic cycle. We show that prediction in ungauged regions (PUR) has major, underrecognized uncertainty and is drastically more difficult than previous problems where basins can be represented by neighboring or similar basins (known as prediction in ungauged basins). While deep neural networks demonstrated stellar performance for streamflow predictions, performance nonetheless declined for PUR, benchmarked here with a new stringent region‐based holdout test on a US data set with 671 basins. We tested approaches to reduce such errors, leveraging deep network's flexibility to integrate “soft” data, such as satellite‐based soil moisture product, or daily flow distributions which improved low flow simulations. A novel input‐selection ensemble improved average performance and greatly reduced catastrophic failures. Despite challenges, deep networks showed stronger performance metrics for PUR than traditional hydrologic models. They appear competitive for geoscientific modeling even in data‐scarce settings.

Effects of streamflow isotope sampling strategies on the calibration of a tracer‐aided rainfall‐runoff model

AbstractIsotopes are increasingly used in rainfall‐runoff models to constrain conceptualisations of internal catchment functioning and reduce model uncertainty. However, there is little guidance on how much tracer data is required to adequately do this, and different studies use data from different sampling strategies. Here, we used a 7‐year time series of daily stable water isotope samples of precipitation and streamflow to derive a range of typical stream sampling regimes and investigate how this impacts calibration of a semi‐distributed tracer‐aided model in terms of flow, deuterium and flux age simulations. Over the 7 years weekly sampling facilitated an almost identical model performance as daily, and there were only slight deteriorations in performance for fortnightly sampling. Monthly sampling resulted in poorer deuterium simulations and greater uncertainty in the derived parameter sets ability to accurately represent catchment functioning, evidenced by unrealistic reductions in the volumes of water available for mixing in the saturation area causing simulated water age decreases. Reducing sampling effort and restricting data collection to 3 years caused reductions in the accuracy of deuterium simulation, though the deterioration did not occur if sampling continued for 5 years. Analysis was also undertaken to consider the effects of reduced sampling effort over the driest and wettest hydrological years to evaluate effects of more extreme conditions. This showed that the model was particularly sensitive to changes in sampling during dry conditions, when the catchment hydrological response is most non‐linear. Across all dataset durations, sampling in relation to flow conditions, rather than time, revealed that samples collected at flows >Q50 could provide calibration results comparable to daily sampling. Targeting only extreme high flows resulted in poor deuterium and low flow simulations. This study suggests sufficient characterization of catchment functioning can be obtained through reduced sampling effort over longer timescales and the targeting of flows >Q50.

Optimality of antecedent precipitation index and its application

The decay constant k in the antecedent precipitation index (API) formula is generally determined arbitrarily and empirically, which causes great confusion in API calculation and comparison of different research results. In response to these problems, two hypotheses related to API (that is, k has the properties of optimality and multivaluedness) are proposed in this study. The daily runoff and daily precipitation of a typical watershed in the Yellow River Basin from January1956–December 2013 is taken as an objective investigated. The results show that the API has a stable antecedent day t0. When antecedent day t ≥ t0, the API remains stable; and when t ≤ t0, there are the optimal value of k (kopt) and the optimal t (topt) in the kopt scenario. The kopt ranges between 0.906 and 0.957, which is different from the traditional empirical single value suggested between 0.80 and 0.98. The existence of kopt verifies the optimality of k or API (Hypothesis 1). The kopt varies with t, which verifies that the k or API has the property of multivaluedness (Hypothesis 2). The two hypotheses of API exist objectively and are related to the runoff generation and soil moisture content of the watershed. The correlation coefficient and its T statistic, sliding average, approximate entropy (ApEn), and four error indicators are used to determine the optimal pair among four pairs of low flow and API. Based on the two hypotheses and the optimal pair, the daily low flow from January 2011 to December 2013 is simulated. The simulation results from the power function are better than those from the logarithmic function, and the kopt scenario is better than the other k scenarios. The simulation accuracy is high and the results are reliable. The contributions of this study lie in (I) the findings of two properties of API (i.e., the optimality and multivaluedness of k); (II) the proposal of an optimal standard for the selection of k, which overcomes the shortcomings of traditional arbitrary selection of k; and (III) the successful application of the two properties and the complexity (ApEn) to low flow simulation, which has theoretical value and practical significance.

Evaluating a landscape-scale daily water balance model to support spatially continuous representation of flow intermittency throughout stream networks

Abstract. There is a growing interest globally in the spatial distribution and temporal dynamics of intermittently flowing streams and rivers, and how this varies in relation to climatic and other environmental factors. However, biases in the distribution of stream gauges may give a misleading impression of spatial-temporal variations in streamflow intermittency within river networks. Here, we developed an approach to quantify catchment-wide streamflow intermittency over long time frames and in a spatially explicit manner, using readily accessible and spatially contiguous daily runoff data from a national-scale water balance model. We examined the ability of the water balance model to simulate streamflow in two hydro-climatically distinctive (subtropical and temperate) regions in Australia, with a particular focus on low-flow simulations. We also evaluated the effect of model time step (daily vs. monthly) on flow intermittency estimation to inform future model selection. The water balance model showed better performance in the temperate region characterised by steady baseflow than in the subtropical region with flashy hydrographs and frequent cease-to-flow periods. The model tended to overestimate low-flow magnitude mainly due to overestimation of gains (e.g. groundwater release to baseflow) during low-flow periods. Modelled patterns of flow intermittency revealed highly dynamic behaviour in space and time, with cease-to-flow events affecting between 29 and 80 % of the river network over the period of 1911–2016, using a daily streamflow model. The daily flow model did not perform better than the monthly flow model in quantifying flow intermittency at a monthly time step, and model selection should depend on the intended application of the model outputs. Our general approach to quantifying spatio-temporal patterns of flow intermittency is transferable to other parts of the world, and it can inform hydro-ecological understanding and management of intermittent streams where limited gauging data are available.

Extreme hydrological phenomena in the forest steppe and steppe zones of Ukraine under the climate change

Abstract. Nowadays, during the period of the global climate change, scientists around the world have noticed an increased incidence of extreme natural disasters. The authors of the study suggested methods of using climate change as a part of a genetic model of maximal floods runoff. This model makes possible the introducing of “climate changes” directly through the maximal stocks of snow and precipitation during the spring flood and runoff coefficients. The object of study is the basin of the Southern Bug – one of the largest rivers in Ukraine, which flows within two geographical zones – forest-steppe and steppe. Overall results using scenario RCP 4.5 showed a decrease of runoff by the end of 2050 from 20 % in the north part of the basin (the forest-steppe zone) to 50 % – in the south (the steppe part of basin). On the other hand, the characteristics of minimal river flow in the Southern Bug basin – winter and summer low waters had been analyzed. One option for low water flow simulation is to study the connection between the drought index (e.g. SPEI) and the minimal water. Studies show that such correlation exists, so knowing the forecast of drought index it is possible to project the value of minimal water discharge, which is the object of the research.

Impact of precipitation data density and duration on simulated flow dynamics and implications for ecohydrological modelling of semi-arid catchments of Southern Africa

Understanding flow regime dynamics is important for the management and conservation of river systems and their related ecosystems under current and future pressures like climate and land use change. For this reason, the accuracy and detail with which hydrological models are able to replicate streamflow and other spatially distributed processes is critical for global and regional water management programs. One of the most critical factors for model performance is the spatial and temporal representation of precipitation data, a procedure known as regionalisation. It is therefore imperative to evaluate the sensitivity of models to data from different stations in order to identify locations that exert the strongest control on simulated streamflow. In this study, a spatially distributed rainfall/runoff model, the J2000, calibrated under near natural river flow conditions, was used to analyse the overall precipitation station impact on simulated streamflow and its flow components. For this purpose, different analysis methods including Regional Sensitivity Analysis (RSA) were used. The study was applied to the Berg River catchment, a water stressed region in the Western Cape, South Africa that was recently affected by serious drought conditions. The results show that during peak flows precipitation stations in the headwater part of the basin had the largest systematic impact, often controlling the data input for only a small proportion of the overall spatial extent of the model, but still having the largest impact on the overall modelled streamflow. In contrast, precipitation stations in lowland areas showed to be more important during low-flow periods, although they often control the data input for a larger part of the modelled catchment. The study highlights the importance of good overall data coverage for improved low-flow simulations, as well as identification of critical areas in regions responsible for peak runoff generation processes. The inclusion of systematic station weights has the potential to improve the ecohydrological representation of rainfall/runoff models in data scarce regions, as well as precipitation regionalisation approaches. The results are indicative of known microclimatic variability in semi-arid Southern Africa, highlighting the significance of catchment heterogeneity and areas of dominant sub-basins for overall runoff generation.

A Hybrid Approach Combining Conceptual Hydrological Models, Support Vector Machines and Remote Sensing Data for Rainfall-Runoff Modeling

Understanding catchment response to rainfall events is important for accurate runoff estimation in many water-related applications, including water resources management. This study introduced a hybrid model, the Tank-least squared support vector machine (LSSVM), that incorporated intermediate state variables from a conceptual tank model within the least squared support vector machine (LSSVM) framework in order to describe aspects of the rainfall-runoff (RR) process. The efficacy of the Tank-LSSVM model was demonstrated with hydro-meteorological data measured in the Yongdam Catchment between 2007 and 2016, South Korea. We first explored the role of satellite soil moisture (SM) data (i.e., European Space Agency (ESA) CCI) in the rainfall-runoff modeling. The results indicated that the SM states inferred from the ESA CCISWI provided an effective means of describing the temporal dynamics of SM. Further, the Tank-LSSVM model’s ability to simulate daily runoff was assessed by using goodness of fit measures (i.e., root mean square error, Nash Sutcliffe coefficient (NSE), and coefficient of determination). The Tank-LSSVM models’ NSE were all classified as “very good” based on their performance during the training and testing periods. Compared to individual LSSVM and Tank models, improved daily runoff simulations were seen in the proposed Tank-LSSVM model. In particular, low flow simulations demonstrated the improvement of the Tank-LSSVM model compared to the conventional tank model.

Improving the Horton infiltration equation by considering soil moisture variation

Soil water infiltration simulation is a subject receiving great interest in hydrological cycle modelling. The traditional Horton equation is based on the curve of infiltration capacity-rainfall duration time. However, the infiltration process is directly affected by soil moisture content, rather than rainfall duration. The objective of this study was to determine a relationship between infiltration capacity and soil moisture content in order to improve the Horton infiltration equation. Artificial rainfall-infiltration experiments were used to determine a series of power functions. The improved equation was cross-validated with observations from 32 experiments of multiple rainfall intensities and antecedent soil moisture. The simulation performance and uncertainty of the improved equation were compared with those of the original Horton equation to verify its accuracy. The results showed that infiltration rate decreases nonlinearly as soil moisture increases, and finally approaches a stable infiltration rate when the soil is saturated. Overland flow simulations by the improved Horton equation closely matched the observations from all 32 experiments over a soil moisture range of 0.222–0.349 m3/m3. The simulation performance was rated as good for most of the experiments for both the calibration and validation data sets. Compared with the original Horton equation, the simulation performance of the improved equation clearly improved estimation of infiltration, particularly as quantified by the Nash-Sutcliffe efficiency coefficient (NSE) and the coefficient of determination (R2). The number of simulations with NSE values greater than 0.65 increased 11.59% and 2.50% for the calibration and validation data sets, respectively. The number of simulations with R2 values greater than 0.90 increased 31.14% and 22.50%, respectively. The uncertainty intervals of the improved Horton equation became a little greater than those of the original Horton equation. For all 32 experimental simulations, the average relative length of the uncertainty interval at the 95% confidence level increased from 40.52% with the original Horton equation to 49.17% with the improved Horton equation. The number of observations falling within the 95% confidence interval increased from 92.13% to 95.94% with the improved Horton equation. Most of the observations were accurately simulated using the improved Horton equation. The greatest improvements in simulating overland flow were seen for the experiments with low flow simulations. The study results provide insights into soil infiltration mechanisms, and also provide references to support improved infiltration simulation by considering soil moisture variation.

Evaluation of Soil and Water Assessment Tool and Artificial Neural Network models for hydrologic simulation in different climatic regions of Asia

In this study, a physically-based hydrological model, Soil and Water Assessment Tool (SWAT) and three types of Artificial Neural Network (ANN) models were used to simulate daily streamflow, and results were compared with observed data for performance analysis. The study was carried out in three different river basins with three different climatic characteristics, namely the West-Seti River Basin in a sub-tropical (partially wet) climatic region, Sre Pok River Basin in a tropical (wet) climatic region and Hari Rod River Basin in a semi-arid (dry) climatic region. The SWAT and ANN models were evaluated using statistical indicators such as the correlation coefficient (R2), Nash-Sutcliffe efficiency (NSE), and percentage bias (PBIAS). The performance of ANN models was found to be very good with both R2 and NSE values greater than 0.95 for the training and validation periods in the West-Seti River Basin and Sre Pok River Basin. Whereas, in the Hari Rod River Basin, the performance of the SWAT model was good with both R2 and NSE values greater than 0.60 for the calibration and validation periods. Moreover, the performance of SWAT and ANN models was evaluated based on hydrological indicators (i.e. annual discharge, base flow, Qdry, and Qwet), during different flow periods (very high to very low flow) using flow duration curves (FDCs). The SWAT model was found to be better for low flow simulation and the ANN model performed better for high flow simulation in the three river basins.