Figure 1 - available via license: Creative Commons Attribution 4.0 International
Content may be subject to copyright.
Locations and rain gauging nets of the Qingjian River catchment, Qiushui River catchment, Tuwei River catchment and Kuye River catchment.
Source publication
Flood forecasting in semiarid regions is always poor, and a single-criterion assessment provides limited information for decision making. Here,
we propose a multicriteria assessment framework called flood
classification–reliability assessment (FCRA) that combines the absolute
relative error, flow classification and uncertainty interval estimated by...
Context in source publication
Context 1
... four selected study catchments are all key tributaries located in the middle reaches of the Yellow River, China (Fig. 1). The maximum and minimum areas of the catchments are 1989 and 8706 km 2 , respectively. The average an- nual temperature ranges from 6 to 14 • C. The average annual precipitation ranges from 1010 to 1150 mm, and 65 % to 80 % is concentrated in summer ( Li et al., 2019;Li and Huang, 2017;Xiao et al., 2019). The rainfall is generally ...
Similar publications
Dynamic control of flood-limited water level (FLWL) is an effective method to realize floodwater utilization, which can alleviate the contradiction between flood control and conservation during flood seasons. In this study, a method of identifying upper bound of FLWL is presented simultaneously considering pre-release for water supply in rainless p...
Citations
... Evapotranspiration (ET) is a predominant element of the earth's hydrological system (Irmak et al., 2008), which illustrates the energy interchange phenomenon that takes place among the biosphere, hydrosphere, and atmosphere (Fan et al., 2018;Li et al., 2019). ET signifies a combined event of evaporation and transpiration, and it is an influential contributor to atmospheric water (Krishna, 2019). ...
... The three-source runoff partition method adopted in Strategy P4 has been widely used in the XAJ hydrological model [36,43,44]. Unlike the flexible hybrid runoff generation model used in this study, the XAJ model operates based on the saturationexcess runoff generation mechanism, resulting in surface runoff when there is no available soil moisture storage [45]. This is significantly different from the constructed flexible hybrid runoff generation model used in this study, in which the surface runoff occurs when the precipitation rate exceeds the saturated hydraulic conductivity of the surface soil [2]. ...
... The three-source runoff partition method adopted in Strategy P4 has been widely used in the XAJ hydrological model [36,43,44]. Unlike the flexible hybrid runoff generation model used in this study, the XAJ model operates based on the saturation-excess runoff generation mechanism, resulting in surface runoff when there is no available soil moisture storage [45]. This is significantly different from the constructed flexible hybrid runoff generation model used in this study, in which the surface runoff occurs when the precipitation rate exceeds the saturated hydraulic conductivity of the surface soil [2]. ...
The partition of surface runoff and infiltration is crucial in hydrologic modeling. To improve the flood prediction, we designed four strategies to explore the influences of the runoff partition method on the flexible hybrid runoff generation model. The runoff partition strategies consist of a hydrological model without the runoff partition module, a two-source runoff partition method, an improved two-source runoff partition method considering the heterogeneity of the subsurface topography and land cover, and a three-source runoff partition method. The Xin’anjiang hydrological model was used as the modeling framework to simulate a six-hourly stream flow for the Xun River watershed in Shaanxi Province, China. And the saturation-excess runoff generation and infiltration-excess runoff generation mechanisms were combined to construct the flexible hybrid runoff generation model. The performances of the four strategies were compared and analyzed based on the continuous flow discharge as well as the flood events. The runoff components analysis method was used to test the model’s conformity with the reality of the watershed. The results showed that the three-source runoff partition method was not applicable to the flexible hybrid runoff generation model because it overestimated the surface runoff and almost ignored the subsurface stormflow runoff. The improved two-source runoff partition method outperformed the others as it considered the heterogeneity of the watershed.
... The modelling protocol involved four main components, namely, (i) use of appropriate graphical evaluation plots (hydrographs and scatter plots); (ii) use of a set of multi-objective model performance statistics; (iii) analysis of different properties of the simulated time series (peak flows, low flows and hydrograph shape); and (iv) global comparison of the SHETRAN predictions with the ones produced by MIKE SHE [2,11,13,29,31,33,34]. The latter evaluation component (iv) is not uncommon in water resources modelling evaluation (for instance, Dehotin, Vázquez, Braud, Debionne and Viallet [51], Li, et al. [53], Bizhanimanzar, et al. [54]). ...
Despite recent progress in terms of cheap computing power, the application of physically based distributed (PBD) hydrological codes still remains limited, particularly, because some commercial-license codes are expensive, even under academic terms. Thus, there is a need for testing the performance of free-license PBD codes simulating complex catchments, so that cheap and reliable mechanistic modelling alternatives might be identified. The hydrology of a geologically complex catchment (586 km2) was modelled using the free-license PBD code SHETRAN. The SHETRAN evaluation took place by comparing its predictions with (i) discharge and piezometric time series observed at different locations within the catchment, some of which were not taken into account during model calibration (i.e., multi-site test); and (ii) predictions from a comparable commercial-license code, MIKE
SHE. In general, the discharge and piezometric predictions of both codes were comparable, which encourages the use of the free-license SHETRAN code for the distributed modelling of geologically complex systems.
... learning rate and numbers of hidden units) with trial and error as high computational costs make autocalibration prohibitive. The third limitation is that physically-based distributed hydrological models such as MIKE SHE (Li et al., 2019;Refsgaard et al., 2010) are not included in our ensemble member list. The last limitation is that only four catchments are tested for the proposed method and BMA in this study, more generalized conclusions can be made through comparisons across a large number (e.g., hundreds) of sites (Boulesteix et al., 2018). ...
Multi-model ensembles enable assessment of model structural uncertainty across multiple disciplines. Bayesian Model Averaging (BMA) is one of the most popular ensemble averaging approaches in hydrology but its predictions are easily impacted by the type of ensemble members selected. Here, we propose a regression-based ensemble approach, namely a Variational Bayesian Long Short-Term Memory network (VB-LSTM) to address this issue. In this approach, a state-of-the-art variational inference (VI) algorithm that is faster and more scalable than Bayesian Markov chain Monte Carlo (MCMC) is employed to approximate the posterior distributions of thousands of parameters in the LSTM networks. To interpret the behavior of deep learning methods, the Permutation Feature Importance (PFI) algorithm is introduced to understand the degree to which VB-LSTM relies on each ensemble member. Twenty conceptual hydrological models are considered to evaluate BMA and VB-LSTM in four catchments from China. Four scenarios with different ensemble members are established to investigate the impacts of ensemble members on model results. Our results show that compared with BMA, VB-LSTM improves deterministic and probabilistic predictions by approximately 10%–30% in terms of Mean Absolute Error (MAE), Sharpness and Continous Ranked Probability Score (CRPS). In addition, the VB-LSTM predictions are more robust and less impacted by the selection of ensemble members. Furthermore, our study encourages the use of Bayesian deep learning in hydrology as an alternative to other approaches tackling model structural uncertainty.
... Among various types of rainfall-runoff models, conceptual models are widely used due to their simple structure and few data requirements [5][6][7][8][9]. Rainfall, as one of the key inputs to conceptual rainfall-runoff models, is extremely vital to the accuracy of the model outputs [10][11][12][13][14][15]. Nowadays, rain gauges are the most traditional and direct way to obtain reliable rainfall data at a fine temporal resolution. ...
... Therefore, the mixed runoff that couples both infiltration excess and saturation excess mechanisms needs to be studied [46,[48][49][50]. For this purpose, the vertically mixed runoff (VMR) model was developed by [48,51] and has been successfully applied to several catchments in China [13,38,46]. The VMR model assumes that the two runoff generation mechanisms exist simultaneously in the vertical direction at the same time where the Horton runoff occurs when rainfall intensity is greater than the infiltration capacity, and the Dunne runoff occurs when the soil moisture capacity is satisfied [52]. ...
... Runoff is estimated by combining the rainfall pdf (EDD) with the VMR model. The VMR model is a conceptual rainfall-runoff model that has been widely used in China since 1997 [13,48,51,67]. This model is developed from the XAJ model [41] and postulates that two runoff generation mechanisms (i.e., infiltration excess mechanism and saturation excess mechanism) exist simultaneously in the vertical direction (see Figure 1); thus, it can be applied to both humid and arid areas theoretically. ...
Rainfall is an important input to conceptual hydrological models, and its accuracy would have a considerable effect on that of the model simulations. However, traditional conceptual rainfall-runoff models commonly use catchment-average rainfall as inputs without recognizing its spatial variability. To solve this, a seamless integration framework that couples rainfall spatial variability with a conceptual rainfall-runoff model, named the statistical rainfall-runoff (SRR) model, is built in this study. In the SRR model, the exponential difference distribution (EDD) is proposed to describe the spatial variability of rainfall for traditional rain gauging stations. The EDD is then incorporated into the vertically mixed runoff (VMR) model to estimate the statistical runoff component. Then, the stochastic differential equation is adopted to deal with the flow routing under stochastic inflow. To test the performance, the SRR model is then calibrated and validated in a Chinese catchment. The results indicate that the EDD performs well in describing rainfall spatial variability, and that the SRR model is superior to the Xinanjiang model because it provides more accurate mean simulations. The seamless integration framework considering rainfall spatial variability can help build a more reasonable statistical rainfall-runoff model.
... Environmental changes caused by climate variability and anthropogenic activities affect the hydrological processes, with two important indicators of changes in streamflow and sediment runoff [1][2][3][4][5]. In most of the watersheds in North China, water resource availability is low because the water demand in arid areas is far greater than the amount of water available [6][7][8][9][10]. ...
... The statistical models, Equations (4), (5), and (6), were established and tested with the data in the pre-impact periods (excluding the anomalous data points). Figure 7 shows the calibrated parameters, observed and calculated flood season's runoff/sediment series, the changes of runoff/sediment, and their separated contributions of climate variability and human activities. ...
Streamflow and sediment runoff are important indicators for the changes in hydrological processes. In the context of environmental changes, decreases in both streamflow and sediment (especially in the flood season) are often observed in most of the tributaries of the middle Yellow River in China’s Loess Plateau. Understanding the effect of human activities could be useful for the management of soil and water conservation (SWC) and new constructions. In this paper, changes in streamflow and sediment during the flood season (June–September) of the 1966–2017 period in a typical loess hill and gully landform basin were analyzed. Basin-wide rainfall of the flood season decreased nonsignificantly with an average rate of −0.6 mm/flood season for the whole study period by using the trend-free pre-whitening based Mann–Kendall trend test, while the decreasing rate was weakened on the time scale. A remarkable warming trend (1985–1999) and two decreasing trends (1966–1984 and 2000–2017) were observed, and the overall increasing trend could be found in air temperature series with a rate of 0.01 °C/flood season during the study period. Statistical models were developed to describe the rainfall-runoff and rainfall-sediment processes in the pre-impact period (when the hydrological series was stationary). Furthermore, the relative effects of climate variability and human activities on hydrological changes were quantified. Results proved the dominant role of human activities (versus climate variability) on the reductions of both streamflow and sediment load. The relative contribution of human activities to streamflow decrease was 84.6% during the post-impact period 1995–2017, while the contributions were 48.8% and 80.1% for two post-impact periods (1982–1996 and 1997–2017), respectively, to the reduction of sediment load. Besides, the effect of the exclusion of anomalous streamflow or sediment events on change-point detection was also analyzed. It indicated that the anomalous events affect the detection of change points and should be given full consideration in order to decide whether to remove them in the change-point detection. Otherwise, the full series with anomalous samples will completely affect the attribution results of hydrological changes. We also suggest that large-scale SWC measures with different construction quality and operational life could intercept and relieve most floods and high sediment concentration processes, but may amplify the peaks of streamflow and sediment when the interception capacities are exceeded under the condition of extreme rainstorm events.
Error correction methods are widely used to improve the performance of streamflow modeling. The Dynamic system response curve (DSRC) is a newly developed error correction method that utilizes error feedback correction to obtain updated variables and thus enhances model accuracy. While the DSRC method is simple and practical, its objectivity has not been systematically demonstrated. To address this issue, the paper proposes an experiment to verify its objectivity. First, three hydrological models (Xinanjiang (XAJ), Hydrologiska Byråns Vattenavdelning (HBV) and vertically hybrid yielding (VHY)) were calibrated and three corrected areal mean rainfall series were obtained by using the DSRC-based models (XAJ-DSRC, HBV-DSRC and VHY-DSRC). The corrected rainfall series from one hydrological model was then used as input to other hydrological models, and its performance was compared to those of using the original rainfall series by five performance measures (Nash-Sutcliffe efficiency coefficient (NSE), root mean squared error (RMSE), percent bias (PBIAS), mean absolute error (MAE) and correlation coefficient (R)). Finally, this objectivity verification experiment was applied to different model combinations for further demonstration. Case studies of the Sunshui River and Shiquan catchments in China showed that when the DSRC-corrected rainfall series from either hydrological model was applied to the original or other hydrological models, an increase in NSE and R values and a decrease in RMSE and MAE values were observed. The absolute values of PBIAS were within 10%. Accordingly, the hypothesis that the DSRC method is objective is not rejected based on the results of the experiment.
Error correction methods play an important role in improving the reliability and accuracy of hydrological modeling. The Dynamic system response curve (DSRC) is a novel and effective error correction method, but it may have the problem of over-correction. Therefore, two multi-model integrated error correction models based on DSRC and Bayesian model averaging (BMA) were proposed in this paper, namely DSRC-BMA and BMA-DSRC. The Sunshui River catchment is selected for a case study. First, three hydrological models including Xinanjiang model (XAJ), Hydrologiska Fyrans Vattenbalans modell (HBV) and vertically hybrid yield model (VHY) were employed. Then, a standard BMA model and three DSRC-based models were constructed separately. Finally, two multi-model integrated error correction models (DSRC-BMA and BMA-DSRC) were applied. The performance of these nine models was compared by Nash-Sutcliffe Efficiency coefficient (NSE), root mean squared error (RMSE) and percent bias (PBIAS). Results showed that the DSRC-based models presented better results than the standard BMA method and most DSRC-based models. Moreover, the uncertainty in BMA, DSRC-BMA and BMA-DSRC models were assessed. The 90% confidence interval of the BMA-DSRC model had high containing ratio values and low average relative bandwidth. Overall, the proposed multi-model integrated error correction methods are effective and can be applicable in improving streamflow modeling.
