Next Article in Journal
Performance Improvement of a Drag Hydrokinetic Turbine
Previous Article in Journal
The Impacts of Hydropower Dams in the Mekong River Basin: A Review
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Water
Volume 13
Issue 3
10.3390/w13030272
water-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Baipenzhu Reservoir Inflow Flood Forecasting Based on a Distributed Hydrological Model

by
Shichao Xu
,
Yangbo Chen
*,
Lixue Xing
and
Chuan Li
School of Geography and Planning, Sun Yat-Sen University, Guangzhou 510275, China
*
Author to whom correspondence should be addressed.
Water 2021, 13(3), 272; https://doi.org/10.3390/w13030272
Submission received: 26 November 2020 / Revised: 19 January 2021 / Accepted: 20 January 2021 / Published: 23 January 2021
Browse Figures
Review Reports Versions Notes

Abstract

:
For reservoir basins, complex underlying surface conditions, short flood confluence times, and concentrated water volumes make inflow flood forecasting difficult and cause forecast accuracies to be low. Conventional flood forecasting models can no longer meet the required forecast accuracy values for flood control operations. To give full play to the role of reservoirs in flood control and to maximize the use of reservoir flood resources, high-precision inflow flood forecasting is urgently needed as a support mechanism. In this study, the Baipenzhu Reservoir in Guangdong Province was selected as the study case, and an inflow flood forecast scheme was designed for the reservoir by a physically based distributed hydrological model, the Liuxihe model. The results show that the Liuxihe model has strong applicability for flood forecasting in the studied reservoir basin and that the simulation results are very accurate. This study also found that the use of different Digital Elevation Model (DEM) data sources has a certain impact on the structure of the Liuxihe model, but the constructed models can both simulate the inflow flood process of the Baipenzhu Reservoir well. At the same time, the Liuxihe model can reflect the spatial variation in rainfall well, and using the Particle swarm optimization (PSO) algorithm to optimize the initial model parameters can greatly reduce the uncertainty of the model forecasts. According to China’s hydrological information forecast standards, the Liuxihe model forecast schemes constructed by the two data sources are rated as Grade A and can be used for real-time flood forecasting in the Baipenzhu Reservoir basin.

1. Introduction

Floods are serious natural disasters, and forecasting flood disasters is a very effective non-engineering measure for flood control [1,2,3,4]. At present, the hydrological basin model commonly used in flood forecasting is still the lumped model [5,6,7,8,9]. The representative models are the Stanford model [10], Sacramento model [11], tank model [12], Xinjiang model [13], and ARNO model [14]. These models have achieved good results in forecasting floods of large rivers, but when these models have been applied to forecast the inflow floods of reservoirs, the desired accuracy results have not been reached. There are three main reasons for this: First, for the reservoir basins, the watershed areas are generally small, the flood confluence time is short, and the flood process changes are very sensitive to the influence of precipitation. The traditional lumped model cannot well reflect the spatial changes of rainfall [15,16,17]; second, the lumped model requires long series of data for model parameter calibration, but for small and medium-sized reservoirs, the data sequence is often short and incomplete, which cannot meet the requirements of lumped model modeling [18,19]; third, after the construction of the reservoir, the conditions of runoff generation in the reservoir area have undergone tremendous changes [20,21,22,23]. The lumped model regards the entire basin as a whole and cannot reflect the impact of reservoir impoundment on reservoir inflows. Because of these difficulties, the use of conventional flood forecasting models can no longer meet the accuracy requirements for flood forecasting of reservoir flood control operations. There is an urgent need for efficient and accurate flood forecasting to provide an important scientific basis for reservoir operation decision-making.
In the last decades, because the artificial neural network model can deal with the nonlinearity problem of the inflow data, some scholars have applied it to inflow flood forecasting of the reservoir [24,25,26,27,28]. At the same time, with the continuous development of machine learning technology, some new algorithms have also been widely used in flood forecasting, such as deep learning neural networks [29,30,31,32]. However, these data-driven models mainly focus on the accuracy of the simulation results based on the given datasets and ignore the physical causality between input and output; thus, they face some classical opposition due to reasons inherent in machine learning techniques (e.g., lack of transparency and difficulty of reproducing the results).
With the continuous progress of remote sensing (RS) and geographic information system (GIS) techniques, the development and application of physically based distributed hydrological models (PBDHMs) have become possible [33]. PBDHMs can fully consider the heterogeneity of the underlying surface of a basin and the heterogeneity of the spatial distribution of rainfall in a basin by dividing the basin into refined units, thereby improving the accuracy of flood forecasting in the basin; thus, these PBDHMs have been called a new generation of basin flood forecasting models. The PBDHM blueprint was first presented by Freeze and Harlan [34], and many distributed hydrological models have since been proposed globally. Representative models include the System Hydrologue Européen model (SHE) [35,36], the variable infiltration capacity model (VIC) [37], the gridded, physics-based hydrologic model (Vflo) [38], the raster-based hydrologic model (CASC2D) [39], the water and energy transfer between soil, plants and atmosphere model (WetSpa) [40] and Liuxihe model [33,41], among others.
In order to explore the applicability and stability of the PBDHMs model for forecasting inflow floods in reservoirs, in this study, the Liuxihe model is adopted, and the Baipenzhu Reservoir in Guangdong Province is selected as the research case, ASTER GDEM and SRTM DEM data are adopted with which to construct the Baipenzhu Reservoir inflow flood forecast model, the PSO algorithm is used to optimize the model parameters, and different rainfall distributions are discussed. This study is expected to provide a scientific basis for the application and development of the proposed reservoir inflow flood forecasting scheme.

2. Methods and Materials

2.1. Liuxihe Model

The Liuxihe model is a PBDHM proposed mainly for watershed flood forecasting [41]. In this model, the reservoir unit is specially set up for the water-covered area so that it has a strong pertinence for inflow flood forecasting of the reservoir. The model divides the basin into several grid cells, calculates the evapotranspiration and production flow on the cell scale, and then converges cell-by-cell to the outlet of the basin through the confluence network. The confluence is divided into slope confluences and river confluences using the kinematical wave approximation and diffusive wave approximation, respectively, for their calculations. The framework of the Liuxihe model operation is shown in Figure 1. At the same time, the Liuxihe model proposes an automatic optimization method for model parameters based on the PSO algorithm [42]. In practical applications, only a representative measured flood is needed to optimize the model parameters, which can greatly improve the performance of the model. The use of refined confluence calculation methods and high-efficiency parameter optimization technology has enabled the Liuxihe model to achieve good results in forecasting floods in small and medium-sized rivers and in reservoir inflow forecasting in China [43,44,45,46,47,48,49].

2.2. Baipenzhu Reservoir Basin

The Baipenzhu Reservoir is located in Guangdong Province, China, between the longitude lines of 115°2′11″ E and 115°24′56″ E and the latitude lines 23°0′5″ N and 23°23′15″ N (Figure 2). The drainage area above the dam site is 856 km2. The reservoir serves multiple functions, including flood regulation, power generation, water supply, irrigation and recreation. The watershed belongs to a subtropical maritime monsoon climate zone characterized by high temperatures, high humidity and abundant water. The rainfall of the region is mainly affected by frontal rain and typhoon rain and is characterized by large magnitudes, high intensities, a long flood season and uneven distribution in time and space, which cause great pressure on river basin flood control [50]. Although the Baipenzhu Reservoir has a relatively large storage capacity, in the event of a large flood, if the inflow flood of the reservoir cannot be accurately predicted, a large amount of water will still be discarded. This discarded water would cause a waste of precious water resources; further, it would also cause an increase in the pressure of flood control downstream. Scientifically carrying out flood forecasting of the Baipenzhu Reservoir and improving the accuracy of inflow flood forecasting are the keys to making full use of the flood control potential and power generation potential of the Baipenzhu Reservoir.
The extent of the Baipenzhu Reservoir is significantly large, accounting for 5.84% of the whole watershed area at a normal storage level of 75 meters. If, when forecasting floods at the Baipenzhu Reservoir, the conventional flood forecast method used for river sections at dam sites is adopted, the impact of the wide water surface formed by reservoir impoundment cannot be considered, and inflow flood forecast error will occur. To improve the prediction accuracy of reservoir inflow floods, it is necessary to adopt a flood forecasting model that can reflect the impact of reservoir impoundment on reservoir inflows.

2.3. DEM Data

The data comprising the physical characteristics of a basin that are required for the construction of the Liuxihe model include DEMs, land-use/cover maps and soil maps. Among them, the DEM data are the basis of the Liuxihe model structure, and the DEM data quality directly affects the slope, basin area, river length and other significant geographic parameters of the basin, so they are also the most important data used in the construction of the Liuxihe model. At present, there are many DEM datasets that can be downloaded for free on a global scale. This study will select two data sources that are commonly used—ASTER GDEM3 (hereafter referred to as ASTER3) and SRTM3.
ASTER3 is jointly developed by Japan’s METI and NASA of the US and is distributed to the public for free (https://earthdata.nasa.gov/). This data product is generated based on the “Advanced spaceborne thermal emission and reflection radiometer model” with a spatial resolution of 30 m, which is rescaled to a resolution of 90 m. The SRTM3 data come from NASA (https://www.nasa.gov/) and are mapped by the US Space Shuttle Endeavor interferometric radar with a spatial resolution of 90 m.
Table 1 records the topographic elevation, slope and watershed area data of the Baipenzhu Reservoir basin from the two different DEM data sources. The results show that there are only slight differences in elevation and slope between the two data sources. Meanwhile, the watershed range extracted by ASTER3 is closer to the true value. The range extracted by SRTM3 is slightly smaller, at only 843.88 km2, which is 1.41% less than the actual area. The DEMs of the Baipenzhu Reservoir watershed generated by ASTER 3 and SRTM3 are shown in Figure 3.

2.4. Land-Use and Soil-Type Data

The land-use/cover type data were downloaded from the USGS land-use type database (http://landcover.usgs.gov/), and the soil-type data were downloaded from the FAO soil-type database (http://www.isric.org). The downloaded land-use and soil-type data were at a resolution of 1000 m × 1000 m and, therefore, were rescaled to a resolution of 90 m.
Due to the deviation between the watershed scope extracted by the different DEMs, the proportion of land-use type and soil type in the watershed indirectly have certain influences. Table 2 and Table 3 show the land-use type and soil-type data corresponding to the two Liuxihe models, respectively. According to the analysis, the land-use and soil types in the study area have not changed. There are 7 types of land-use in the basin, including evergreen needle leaf forest, evergreen broadleaf forest, bush, sparse woods, seaside wetlands, slope grassland and farmland; there are 5 types of soils, including Ferric Acrisols (CN10005, CN30043, CN30053, CN30075), Haplic Acrisols (CN10033, CN10039), Haplic Alisols (CN10169), Humic Acrisols (CN30147, CN30149) and Cumulic Anthrosols (CN30423, CN30673). The difference in the land-use type and soil-type data between the two models is very small. For brevity, this paper only shows the land-use and soil types of the ASTER 3 Liuxihe model in the Baipenzhu Reservoir watershed, as shown in Figure 4.

2.5. Hydrological Data

The Baipenzhu Reservoir area has 2 hydrological stations, Baipenzhu and Baokou, and 5 rainfall stations, Shijian, Xintang, Gaotan, Mashan and Heduobu. In this study, 41 measured flood process datasets in the basin from 1986 to 2017 were collected and sorted, including precipitation data from the rain gauges and discharge data from the Baipenzhu hydrological stations, all of which took had hourly temporal resolutions. The Liuxihe model is a distributed hydrological model with physical significance, which derives model parameters physically from terrain characteristics. Therefore, only one flood event is needed for parameter optimization [41]. The number 20030610 flood was selected for parameter optimization, and the remaining 40 floods were used for model verification. The spatial interpolation of basin rainfall is carried out by using the Thiessen polygon method, and the cell grid areal rainfall is generated according to the rainfall of rain gauges.

3. Model Implementation

3.1. Liuxihe Model Setup

Based on the ASTER3 and SRTM3 data, in this paper, the Liuxihe model was constructed, and the D8 flow direction method was adopted [51]. According to the cumulative flow threshold and the threshold of the normal storage level, the river units and reservoir units were divided, and the remaining units in the basin were indicated as slope units. There is a great difference in the generation and confluence mechanisms of different units. If the model structure has errors, the simulation results will be distorted. A detailed breakdown is shown in Table 4. Generally, there was little difference in the number of river units and slope units between the two models, but the reservoir units were quite different. Compared with the ASTER 3 model, the SRTM3 model lacked 2582 reservoir units, accounting for 72.7% of the total reservoir units in the SRTM3 model structure. The main reason for this result is that most of the watershed area that was missing in the SRTM3 model consisted of reservoir units.
In the process of river extraction, the Strahler method [52] was used to divide the river into three levels. Referring to remotely sensed images obtained from Google Earth, 23 river nodes were set in both models, the river channels were divided into 37 virtual river sections, and the cross-sectional width, side slope and bottom slope of each virtual river channel were estimated. The results of the divisions of unit classification, channel nodes and virtual reaches are shown in Figure 5.

3.2. Initial Parameter Derivation of the Liuxihe Model

The initial parameters input to the Liuxihe model are determined by the physical characteristics of each unit, which can be generally divided into four categories, namely, topographic parameters, meteorological parameters, soil parameters and land-use parameters.
  • The terrain or topographical parameters include the flow direction and slope, which are directly calculated and determined from the DEM;
  • Of the meteorological parameters, the main one is the evaporation capacity. According to experience, the value of the evaporation capacity of all units is set to 5 mm/d [41];
  • The land-use type parameters include the slope roughness and evaporation coefficient. The evaporation coefficient is a very insensitive parameter. According to the parameterization experience of the Liuxihe model, it is uniformly set at 0.7 [41]. The slope roughness adopts the recommended values in the literature [40,53], as shown in Table 5;
  • The soil parameters include the saturated water content, field capacity, saturated hydraulic conductivity, soil thickness, wilting point and soil characteristics. The soil properties are uniformly set at 2.5 [41], and the remaining parameters are calculated by the soil hydraulic characteristic calculator [54] proposed by Arya et al. The results are shown in Table 6.

3.3. Parameter Optimization of the Liuxihe Model

Particle swarm optimization (PSO) [44] is a global searching algorithm, which was first proposed by American psychologist James Kennedy and electrical engineer Russell Eberhart (1995) during their study on the social and intelligent behaviors of a school of birds in their search for food and better living conditions.
In this study, the PSO algorithm is employed to optimize the initial model parameters. Flood number 20030610 was used for the parameter optimization of the Liuxihe model with different DEM data sources. The population size of the particle swarm in the algorithm was set to 20, the maximum number of iterations was 50, and the total number of computations was 1000. The value range of the inertia factor is [0.1, 0.9], and the value ranges of the learning acceleration factors, C1 and C2, are both [0.5, 2.5]. Table 7 shows the parameter optimization results with different DEM databases in the Liuxihe model. Due to space limitations, Figure 6 only shows the evolutionary results of the objective function value and parameter values in the optimization process of the ASTER3 model parameters. The results show that after 35 evolutionary calculation repetitions, the model objective function value tends to be stable, and the model parameters basically converge to optimal values, indicating that the Liuxihe model parameters have a better convergence rate.

4. Results and Discussion

4.1. Influence of Different DEM Data Sources on Simulation Results

To quantitatively evaluate the impacts of different DEM data sources on the accuracy of reservoir inflow flood forecasting, the ASTER3 and SRTM3 datasets were used to construct the Liuxihe model and the corresponding optimization parameters, respectively. Forty floods were successively simulated and verified. Four indicators, including the Nash–Sutcliffe coefficient, peak flow relative error, water balance coefficient and peak flow duration difference, were selected and determined to evaluate the simulation effect of each flood. The detailed results are shown in Figure 7.
According to the statistical results provided in Figure 7, the average Nash–Sutcliffe coefficient, peak flow relative error, water balance coefficient, and peak flow duration difference of the 40 floods simulated by the ASTER3 model were 0.884, 5%, 0.996 and 0.425 h, respectively. The average Nash–Sutcliffe coefficient, peak flow relative error, water balance coefficient and peak flow duration difference of the SRTM3 model were 0.883, 5.77%, 1.006 and 0.425 h, respectively. In comparison, the simulation result of the model constructed by ASTER3 is better than that of the model constructed by SRTM3, but the difference between the two models is very small, and both of them can simulate the flood process well.

4.2. Influence of Spatial Rainfall Distribution on Simulation Results

Because the reservoir basin is generally small and the flood confluence time is short, precipitation becomes the main driving factor of reservoir inflow floods. However, the spatial distribution of rainfall directly affects the magnitude and time distribution of floods, which play decisive roles in the prediction accuracy of a model. Therefore, according to the distribution of stations in the Baipenzhu Reservoir basin, the basin was divided into three sections into this study, including the upstream, middle and downstream sections: three stations were located in the upstream section (Baipenzhu, Shijian, and Xintang), two stations were located in the middle section. (Baokou and Gaotan), and two stations were located in the downstream section (Mashan and Heduobu).
Based on the data of 40 historical flood events, through statistical of the total rainfall of each flood, and 80% of the maximum rainfall was set as the rainfall center, the spatial distribution of rainfall in the basin can be roughly divided into four types: the upstream and downstream rainfall predominating type (herein referred to as upstream and downstream type), midstream and downstream rainfall predominating type (Hereinafter referred to as the midstream and downstream type), downstream rainfall predominating type (hereinafter referred to as the downstream type) and whole basin rainfall type (hereinafter referred to as the whole basin type). The proportions of each type of flood are shown in Table 8. From the statistical data, the spatial distributions of rainfall comprising the upstream and downstream type, midstream and downstream type, downstream type, and whole basin type accounted for 5%, 25%, 57.5% and 12.5% of the entire flood events, respectively. The rainfall centers are mainly distributed in the middle and lower reaches of the basin, with proportions as high as 70%; this is due to the small catchment area of the reservoir and is one of the main reasons for frequent flooding. When a rainfall center is located in the middle or lower reaches of the basin, the rainfall can easily cause a flood with a large peak height.
To specifically analyze the impacts of different rainfall spatial distribution types on the simulation results, floods number 19860625, 20130623, 19910906 and 20130815 were selected to represent the upstream and downstream, middle and downstream, downstream and whole basin rainfall distribution types, respectively. The rainfall distribution of each representative flood is shown in Figure 8.
In Figure 9, the simulation processes of various types of representative floods are shown. The Liuxihe model constructed by the two different DEMs can simulate floods of the four different rainfall types well, and the simulated flood processes are very similar to those seen in the measured data. The ASTER3 model and the SRTM3 model simulate the upstream and downstream rainfall distribution type, middle and downstream type, downstream type and the whole basin type with high Nash–Sutcliffe coefficients for the four representative floods, with values of 0.947, 0.979, 0.982, and 0.889 and of 0.962, 0.97, 0.98, and 0.821, respectively. For the floods simulated by the SRTM3 model, the flood peak error of the simulated flood number 20130815 reached 23.8%, and the rest of the peak errors were within the 20% error range. The peak flow duration difference was within 2 h. The results show that the Liuxihe model can reflect the spatial variation in rainfall well, which is the greatest advantage of PBDHMs.

4.3. The Influence of Model Parameter Optimization on Simulation Results

To analyze the improvement in the model performance after parameter optimization, the initial model parameters and the optimized model parameters were used to simulate 40 floods. Due to a large amount of data, Table 9 only lists the average statistical indexes of the flood simulation results before and after the optimization of two Liuxihe model parameters with ASTER3 and SRTM3. Through the analysis, it was found that the simulation results obtained with the initial parameters of the Liuxihe models with ASTER3 and SRTM3 were both poor, which indicates that there are certain uncertainties in the model parameters derived from physics and that the distributed model, when used without parameter optimization, has high uncertainty in its flood simulations and forecasts. However, through the optimization of the PSO algorithm, the simulation results of the Liuxihe models with ASTER3 and SRTM3 showed that the average Nash–Sutcliffe coefficient increased by 0.47 and 0.75, the average peak flow error decreased by 47% and 63%, and the average peak flow duration difference error decreased by 3.48 h and 8.6 h, respectively. After the parameter optimization occurred, the model flood simulation effect was obviously improved compared with the initial parameter simulation effect, which indicates that optimization of the distributed model parameters can significantly improve the performance of the model.
After parameter optimization was conducted by the PSO algorithm, the determination coefficients of the Liuxihe model with ASTER3 simulation results were above 80% for 39 of the 40 floods, accounting for 97.5%. The qualification rate of flood peak errors less than 20% was 100%. The peak flow duration difference error was within 3 h, and the pass rate reached 100%. The SRTM3 simulation results showed that there were 38 floods for which the Nash–Sutcliffe coefficient higher than 80% was calculated, accounting for 95%. There were 38 floods with peak errors of less than 20%, accounting for 95%. The peak flow duration difference error was within 3 h, and the pass rate reached 100%. According to the hydrological information forecast standards of China, the model forecast schemes constructed by the two data sources are rated as Grade A, meaning they can be used for real-time inflow flood forecasting in the Baipenzhu Reservoir basin.

5. Conclusions

In this study, the Baipenzhu Reservoir basin is selected as the research case, and a distributed hydrological model, the Liuxihe model, is adopted and used to simulate the inflow flood of the reservoir. Meanwhile, according to three aspects: different DEM data sources, the spatial distribution of rainfall and parameter optimization to explore the applicability and stability of the Liuxihe model for forecasting inflow floods in reservoirs. The results show that:
  • The use of different DEM data sources has a certain influence on the structure of the Liuxihe model. Among different DEMs, there are only slight differences in elevation and slope. The watershed range extracted by ASTER3 is close to the real value, and the watershed area extracted by SRTM3 is 1.41% less than the actual value. Due to the deviation between the watershed area extracted by the different DEMs, the proportion of land-use type and soil type in the watershed has a slight indirect influence. From the perspective of the model structure, there is little difference between the two modeled river units and slope units, but the reservoir units are quite different. The main reason for this result is that most of the watershed area missing from the SRTM3 model consists of reservoir units;
  • The DEMs from different data sources have little effect on the simulation accuracy of the Liuxihe model. The model verification results based on the ASTER3 and SRTM3 data sources show that the average values of the Nash–Sutcliffe coefficient of the 40 flood processes are 0.884 and 0.883, and the average flood peak errors are 5% and 5.77%, respectively; the average peak flow duration difference error is 0.425 h for both the ASTER3 and SRTM3 models. In comparison, the simulation result of the model constructed by ASTER3 is better than that of SRTM3, but in general, the Liuxihe models constructed by the two DEMs can both simulate inflow flood processes in the Baipenzhu Reservoir well;
  • The Liuxihe model can reflect spatial variations in rainfall well. According to the measured data of 40 floods, the spatial distribution of rainfall in the Baipenzhu Reservoir basin can be classified into four types: the upstream and downstream type, midstream and downstream type, downstream type and whole basin type. Floods number 19860625, 20130623, 19910906 and 20130815 were selected as representative floods, which were simulated by the Liuxihe models built with the two DEMs, and the results showed that both models could simulate the flood processes of the four different rainfall spatial distribution types well.
  • The Liuxihe model needs to carry out model parameter optimization. Model parameters derived from physical data are uncertain, and the performance of a model can be significantly improved through parameter optimization. The simulation results of the Liuxihe models built with ASTER3 and SRTM3 show that after parameter optimization is conducted using the PSO algorithm, the average Nash–Sutcliffe coefficient increases by 0.47 and 0.75, the average peak flow error decreases by 47% and 63%, and the average peak flow duration difference error decreases by 3.48 h and 8.6 h, respectively.
  • The Liuxihe model is a PBDHM that can meet the accuracy requirements of inflow flood forecasting for reservoir flood control operations. The simulation results of the Liuxihe models built with ASTER3 and SRTM3 show that the qualification rates of the Nash–Sutcliffe coefficient, peak error and peak flow duration difference error are 97.5%, 100%, and 100% and 95%, 95%, and 100%, respectively. According to the hydrological information forecast standards of China, the Liuxihe model forecasting schemes constructed by the two data sources are rated as Grade A forecasting schemes and can be used for real-time inflow flood forecasting in the Baipenzhu Reservoir basin.

Author Contributions

Y.C. was responsible for proposing the original ideal and providing technical guidance; S.X. was responsible for the data compilation, processing, computation and writing; L.X. and C.L. were responsible for the data sorting. All authors have read and agreed to the published version of the manuscript.

Funding

This work was supported by the Natural Science Foundation of China (No. 51961125206).

Data Availability Statement

Data sharing not applicable.

Acknowledgments

We thank the Dongjiang River Basin Administration of Guangdong Province and the Baipenzhu Reservoir Administration Bureau for the data support, evaluation of the results and for their valuable comments.

Conflicts of Interest

The authors declare no conflict of interest regarding the publication of this paper.

References

  1. Xie, L.L.; Liu, P.; Guo, S.L. Optimal design of seasonal flood limited water levels by jointing operation of the reservoir and floodplains. Water Resour. Manag. 2018, 32, 179–193. [Google Scholar] [CrossRef]
  2. Chen, S.T.; Yu, P.S. Real-time probabilistic forecasting of flood stages. J. Hydrol. 2007, 340, 63–77. [Google Scholar] [CrossRef]
  3. Pagano, T.C.; Wang, Q.J.; Hapuarachchi, P.; Robertson, D. A dual-pass error-correction technique for forecasting streamflow. J. Hydrol. 2011, 405, 367–381. [Google Scholar] [CrossRef]
  4. Liu, Z.J.; Guo, S.L.; Zhang, H.; Yang, G. Comparative Study of Three Updating Procedures for Real-Time Flood Forecasting. Water Resour. Manag. 2016, 30, 2111–2126. [Google Scholar] [CrossRef]
  5. El Alfy, M. Assessing the impact of arid area urbanization on flash floods using GIS, remote sensing, and HEC-HMS rainfall–runoff modeling. Hydrol. Res. 2016, 47, 1142–1160. [Google Scholar] [CrossRef] [Green Version]
  6. Jie, M.X.; Chen, H.; Xu, C.Y.; Zeng, Q.; Tao, X.E. A comparative study of different objective functions to improve the flood forecasting accuracy. Hydrol. Res. 2016, 47, 718–735. [Google Scholar] [CrossRef] [Green Version]
  7. Thiboult, A.; Anctil, F. On the difficulty to optimally implement the ensemble kalman filter: An experiment based on many hydrological models and catchments. J. Hydrol. 2015, 529, 1147–1160. [Google Scholar] [CrossRef] [Green Version]
  8. Huang, S.C.; Hattermann, F.F. Coupling a global hydrodynamic algorithm and a regional hydrological model for large-scale flood inundation simulations. Hydrol. Res. 2018, 49, 438–449. [Google Scholar] [CrossRef]
  9. Su, B.N.; Huang, H.; Zhu, W. An urban pluvial flood simulation model based on diffusive wave approximation of shallow water equations. Hydrol. Res. 2018, 50, 138–154. [Google Scholar] [CrossRef]
  10. Crawford, N.H.; Linsley, R.K. Digital Simulation in Hydrology, Stanford Watershed Model IV; Technical Report No. 39; Department of Civil Engineering, Stanford University: Stanford, CA, USA, 1966. [Google Scholar]
  11. Burnash, R.J.C. The NWS River Forecast System-Catchment Modeling, Computer Models of Watershed Hydrology; Singh, V.P., Ed.; Water Resource Publications: Littleton, CO, USA, 1995. [Google Scholar]
  12. Sugawara, M. “Tank model.” Computer Models of Watershed Hydrology; Singh, V.P., Ed.; Water Resources Publications: Littleton, CO, USA, 1995. [Google Scholar]
  13. Zhao, R.J. Flood Forecasting Method for Humid Regions of China; East China College of Hydraulic Engineering: Nanjing, China, 1977. [Google Scholar]
  14. Todini, E. The ARNO rainfall-runoff model. J. Hydrol. 1996, 175, 339–382. [Google Scholar] [CrossRef]
  15. Lin, K.R.; Zhang, Q.; Chen, X. An evaluation of impacts of dem resolution and parameter correlation on topmodel modeling uncertainty. J. Hydrol. 2010, 394, 370–383. [Google Scholar] [CrossRef]
  16. Goyal, M.K.; Panchariya, V.K.; Sharma, A.; Singh, V. Comparative Assessment of SWAT Model Performance in two Distinct Catchments under Various DEM Scenarios of Varying Resolution, Sources and Resampling Methods. Water Resour. Manag. 2018, 32, 805–825. [Google Scholar] [CrossRef]
  17. Kumar, B.; Lakshmi, V.; Patra, K.C. Evaluating the uncertainties in the SWAT model outputs due to DEM grid size and resampling techniques in a large Himalayan River basin. J. Hydrol. Eng. 2017, 22, 120–131. [Google Scholar] [CrossRef]
  18. Kapangaziwiri, E.; Hughes, D.A.; Wagener, T. Incorporating uncertainty in hydrological predictions for gauged and ungauged basins in southern Africa. Hydrol. Sci. J. 2012, 57, 1000–1019. [Google Scholar] [CrossRef] [Green Version]
  19. Worqlul, A.W.; Yen, H.; Collick, A.S.; Tilahun, S.A.; Steenhuis, T.S. Evaluation of CFSR, TMPA 3B42 and Ground-based Rainfall Data as Input for Hydrological Models, in Data-Scarce Regions: The Upper Blue Nile Basin, Ethiopia. Catena 2017, 152, 242–251. [Google Scholar] [CrossRef]
  20. Graf, W.L. Downstream hydrologic and geomorphic effects of large dams on American rivers. Geomorphology 2006, 79, 336–360. [Google Scholar] [CrossRef]
  21. Magilligan, F.J.; Nislow, K.H. Changes in hydrologic regime by dams. Geomorphology 2005, 71, 61–78. [Google Scholar] [CrossRef]
  22. Gregory, K.J. The human role in changing river channels. Geomorphology 2006, 79, 172–191. [Google Scholar] [CrossRef]
  23. Duan, W.X.; Guo, S.L.; Wang, J. Impact of Cascaded Reservoirs Group on Flow Regime in the Middle and Lower Reaches of the Yangtze River. Water 2016, 8, 218. [Google Scholar] [CrossRef] [Green Version]
  24. Coulibaly, P.; Anctil, F.; Bobee, B. Daily reservoir inflow forecasting using artificial neural networks with stopped training approach. J. Hydrol. 2000, 230, 244–257. [Google Scholar] [CrossRef]
  25. Coulibaly, P.; Haché, M.; Fortin, V.; Bobée, B. Improving daily reservoir inflow forecasts with model combination. J. Hydrol. Eng. 2005, 10, 91–99. [Google Scholar] [CrossRef]
  26. Lin, G.F.; Chen, G.R.; Huang, P.Y.; Chou, Y.C. Support vector machine-based models for hourly reservoir inflow forecasting during typhoon-warning periods. J. Hydrol. 2009, 372, 17–29. [Google Scholar] [CrossRef]
  27. Valipour, M.; Banihabib, M.E.; Behbahani, S.M.R. Comparison of the ARMA, ARIMA, and the autoregressive artificial neural network models in forecasting the monthly inflow of Dez dam reservoir. J. Hydrol. 2013, 476, 433–441. [Google Scholar] [CrossRef]
  28. Zhang, D.; Lin, J.; Peng, Q.; Wang, D.; Yang, T.; Sorooshian, S.; Liu, X.; Zhuang, J. Modeling and simulating of reservoir operation using the artificial neural network, support vector regression, deep learning algorithm. J. Hydrol. 2018, 565, 720–736. [Google Scholar] [CrossRef] [Green Version]
  29. Zhou, Y.L.; Guo, S.L.; Chang, F.J. Explore an evolutionary recurrent ANFIS for modelling multi-step-ahead flood forecasts. J. Hydrol. 2019, 570, 343–355. [Google Scholar] [CrossRef]
  30. Chang, L.C.; Chang, F.J.; Yang, S.N.; Kao, I.; Ku, Y.Y.; Kuo, C.L.; Amin, I.; bin Mat, M.Z. Building an Intelligent Hydroinformatics Integration Platform for Regional Flood Inundation Warning Systems. Water 2019, 11, 9. [Google Scholar] [CrossRef] [Green Version]
  31. Chang, L.C.; Chang, F.J.; Yang, S.N.; Tsai, F.H.; Chang, T.H.; Herricks, E.E. Self-organizing maps of typhoon tracks allow for flood forecasts up to two days in advance. Nat. Commun. 2020, 11, 1. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  32. Kao, I.F.; Zhou, Y.L.; Chang, L.C.; Chang, F.J. Exploring a Long Short-Term Memory based Encoder-Decoder framework for multi-step-ahead flood forecasting. J. Hydrol. 2020, 583, 124631. [Google Scholar] [CrossRef]
  33. Chen, Y. Liuxihe Model; Science Press: Beijing, China, 2009; 198p. [Google Scholar]
  34. Freeze, R.A.; Harlan, R.L. Blueprint for a physically-based, digitally simulated, hydrologic response model. J. Hydrol. 1969, 9, 237–258. [Google Scholar] [CrossRef]
  35. Abbott, M.B.; Bathurst, J.C.; Cunge, J.A.; O’Connell, P.E.; Rasmussen, J. An Introduction to the European Hydrologic System-System Hydrologue Europeen, ‘SHE’, a: History and Philosophy of a Physically-based, Distributed Modelling System. J. Hydrol. 1986, 87, 45–59. [Google Scholar] [CrossRef]
  36. Abbott, M.B.; Bathurst, J.C.; Cunge, J.A.; O’Connell, P.E.; Rasmussen, J. An Introduction to the European Hydrologic System-System Hydrologue Europeen, ‘SHE’, b: Structure of a Physically based, Distributed Modeling System. J. Hydrol. 1986, 87, 61–77. [Google Scholar] [CrossRef]
  37. Liang, X.; Lettenmaier, D.P.; Wood, E.F.; Burges, S.J. A simple hydrologically based model of land surface water and energy fluxes for general circulation models. J. Geophys. Res. 1994, 99, 14415–14428. [Google Scholar] [CrossRef]
  38. Vieux, B.E.; Vieux, J.E. VfloTM: A Real-time Distributed Hydrologic Model. In Proceedings of the 2nd Federal Interagency Hydrologic Modeling Conference, Las Vegas, NV, USA, 28 July–1 August 2002. [Google Scholar]
  39. Julien, P.Y.; Saghafian, B.; Ogden, F.L. Raster-Based Hydrologic Modeling of spatially-Varied Surface Runoff. Water Resour. Bull. 1995, 31, 523–536. [Google Scholar] [CrossRef]
  40. Wang, Z.; Batelaan, O.; De Smedt, F. A distributed model for water and energy transfer between soil, plants and atmosphere (WetSpa). Phys. Chem. Earth. 1997, 21, 189–193. [Google Scholar] [CrossRef]
  41. Chen, Y.B.; Ren, Q.W.; Huang, F.H.; Xu, H.J.; Cluckie, L. Liuxihe Model and Its Modeling to River Basin Flood. J. Hydrol. Eng. 2010, 16, 33–50. [Google Scholar] [CrossRef]
  42. Eberhart, R.C.; Shi, Y.H. Particle swarm optimization: Developments, applications and resources. In Proceedings of the 2001 Congress on Evolutionary Computation, IEEE, Seoul, Korea, 27–30 May 2001; pp. 81–86. [Google Scholar]
  43. Xu, H.; Chen, Y.; Zeng, B.; He, J.; Liao, Z. Application of SCE-UA Algorithm to Parameter Optimization of Liuxihe Model. Singap. J. Trop. Geogr. 2012, 32, 32–37. [Google Scholar]
  44. Chen, Y.; Li, J.; Xu, H. Improving flood forecasting capability of physically based distributed hydrological model by parameter optimization. Hydrol. Earth Syst. Sci. 2016, 20, 375–392. [Google Scholar] [CrossRef] [Green Version]
  45. Li, J.; Chen, Y.B.; Wang, H.Y.; Qin, J.M.; Li, J.; Chiao, S. Extending flood forecasting lead time in a large watershed by coupling WRF QPF with a distributed hydrological model. Hydrol. Earth Syst. Sci. 2017, 21, 1279–1294. [Google Scholar] [CrossRef] [Green Version]
  46. Chen, Y.; Li, J.; Wang, H.; Qin, J.; Dong, L. Large-watershed flood forecasting with high-resolution distributed hydrological model. Hydrol. Earth Syst. Sci. 2017, 21, 735–749. [Google Scholar] [CrossRef] [Green Version]
  47. Li, J.; Yuan, D.X.; Liu, J.; Jiang, Y.J.; Chen, Y.B.; Hsu, K.L.; Sorooshian, S. Predicting floods in a large karst river basin by coupling PERSIANN-CCS QPEs with a physically based distributed hydrological model. Hydrol. Earth Syst. Sci. 2019, 23, 1505–1532. [Google Scholar] [CrossRef] [Green Version]
  48. Wang, H.W.; Chen, Y.B. Identifying Key Hydrological Processes in Highly Urbanized Watersheds for Flood Forecasting with a Distributed Hydrological Model. Water 2019, 11, 1641. [Google Scholar] [CrossRef] [Green Version]
  49. Chen, Y.; An, X.; Dou, P.; Zhang, T. Estimating Land Use/Cover of Foshan City in Southern China with Landsat Remote Sensing Imagery for Flood Modeling. In Proceedings of the International Multidisciplinary Scientific GeoConference Surveying Geology and Mining Ecology Management, SGEM, Albena, Bulgaria, 27 June–6 July 2017; Volume 17, pp. 79–86. [Google Scholar]
  50. Tu, X.J.; Singh, V.P.; Chen, X.H.; Chen, L.; Zhang, Q.; Zhao, Y. Intra-annual Distribution of Streamflow and Individual Impacts of Climate Change and Human Activities in the Dongijang River Basin, China. Water Resour. Manag. 2015, 29, 2677–2695. [Google Scholar] [CrossRef]
  51. Jenson, S.K.; Domingue, J.O. Extracting topographic structure from digital elevation data for geographic information system analysis. Photogramm. Eng. Remote Sens. 1988, 54, 1593–1600. [Google Scholar]
  52. Strahler, A.N. Quantitative Analysis of watershed Geomorphology. Trans. Am. Geophys. Union 1957, 35, 913–920. [Google Scholar] [CrossRef] [Green Version]
  53. O’Callaghan, J.F.; Mark, D.M. The extraction of drainage networks from digital elevation data. Comput. Vis. Graph. Image Process. 1984, 27, 323–344. [Google Scholar] [CrossRef]
  54. Arya, L.M.; Paris, J.F. A physicoempirical model to predict the soil moisture characteristic from particle-size distribution and bulk density data. Soil Sci. Soc. Am. J. 1981, 45, 1023–1030. [Google Scholar] [CrossRef]
Water 13 00272 g001 550
Figure 1. The framework diagram of the Liuxihe model operation.
Figure 1. The framework diagram of the Liuxihe model operation.
Water 13 00272 g001
Water 13 00272 g002 550
Figure 2. Map of the Baipenzhu Reservoir watershed.
Figure 2. Map of the Baipenzhu Reservoir watershed.
Water 13 00272 g002
Water 13 00272 g003 550
Figure 3. DEMs of the Baipenzhu Reservoir watershed created with the ASTER3 and SRTM3 databases. (a) DEM from the ASTER 3 database; (b) DEM from the SRTM3 database.
Figure 3. DEMs of the Baipenzhu Reservoir watershed created with the ASTER3 and SRTM3 databases. (a) DEM from the ASTER 3 database; (b) DEM from the SRTM3 database.
Water 13 00272 g003
Water 13 00272 g004 550
Figure 4. Terrain property data of the Baipenzhu Reservoir watershed from the ASTER 3 database. (a) Land-use types; (b) soil types.
Figure 4. Terrain property data of the Baipenzhu Reservoir watershed from the ASTER 3 database. (a) Land-use types; (b) soil types.
Water 13 00272 g004
Water 13 00272 g005 550
Figure 5. Liuxihe model structures under different data sources. (a) Model structure under the ASTER 3 database; (b) model structure under the SRTM3 database.
Figure 5. Liuxihe model structures under different data sources. (a) Model structure under the ASTER 3 database; (b) model structure under the SRTM3 database.
Water 13 00272 g005
Water 13 00272 g006 550
Figure 6. Results of parameter optimization of the Liuxihe model with the particle swarm optimization (PSO) algorithm. (a) Changing curve of the objective function; (b) parameter evolution process.
Figure 6. Results of parameter optimization of the Liuxihe model with the particle swarm optimization (PSO) algorithm. (a) Changing curve of the objective function; (b) parameter evolution process.
Water 13 00272 g006
Water 13 00272 g007 550
Figure 7. Statistical indicators of flood simulation results obtained with different. DEM databases in the Liuxihe model. (a) Nash–Sutcliffe coefficient; (b) peak flow relative error; (c) water balance coefficient; (d) peak flow duration difference.
Figure 7. Statistical indicators of flood simulation results obtained with different. DEM databases in the Liuxihe model. (a) Nash–Sutcliffe coefficient; (b) peak flow relative error; (c) water balance coefficient; (d) peak flow duration difference.
Water 13 00272 g007
Water 13 00272 g008 550
Figure 8. Rainfall distributions of representative floods. (a) Flood number. 19860625, upstream and downstream rainfall distribution type; (b) flood number 20130623, midstream and downstream type; (c) flood number 19910906, downstream type; (d) flood number 20130815, whole basin type.
Figure 8. Rainfall distributions of representative floods. (a) Flood number. 19860625, upstream and downstream rainfall distribution type; (b) flood number 20130623, midstream and downstream type; (c) flood number 19910906, downstream type; (d) flood number 20130815, whole basin type.
Water 13 00272 g008
Water 13 00272 g009 550
Figure 9. Simulation processes of representative floods. (a) Upstream and downstream rainfall distribution type; (b) midstream and downstream type; (c) downstream type; (d) whole basin type.
Figure 9. Simulation processes of representative floods. (a) Upstream and downstream rainfall distribution type; (b) midstream and downstream type; (c) downstream type; (d) whole basin type.
Water 13 00272 g009
Table
Table 1. Topography, slope and basin characteristics.
Table 1. Topography, slope and basin characteristics.
DatabaseElevation (m)Slope (Degree)Basin
MinMaxMeanMinMaxMeanArea (km2)
ASTER3611274317.49047.6413.17855.55
SRTM3531277330.36050.3613.04843.88
Table
Table 2. Land-use data of the Baipenzhu Reservoir watershed from different DEM databases.
Table 2. Land-use data of the Baipenzhu Reservoir watershed from different DEM databases.
Land UseDatabase
ASTER3 (%)SRTM3 (%)
Evergreen needle leaf forest39.8739.47
Evergreen broadleaf forest42.6243.33
Bush6.796.74
Sparse woods3.273.28
Seaside wetlands1.591.39
Slope grassland1.081.07
Farmland4.794.72
Table
Table 3. Soil-type data of the Baipenzhu Reservoir watershed from different DEM databases.
Table 3. Soil-type data of the Baipenzhu Reservoir watershed from different DEM databases.
Soil TypeDatabase
ASTER3 (%)SRTM3 (%)
Ferric Acrisols62.4161.72
Haplic Acrisols22.3122.81
Haplic Alisols1.041.05
Humic Acrisols8.398.57
Cumulic Anthrosols5.855.86
Table
Table 4. Model structure information under different data sources.
Table 4. Model structure information under different data sources.
DEM DatabaseWatershed UnitReservoir UnitRiver UnitSlope Unit
ASTER 3105,6246132171697,776
SRTM 3104,1833550172698,907
Table
Table 5. The initial values of land-use/cover-related parameters.
Table 5. The initial values of land-use/cover-related parameters.
Land Use/CoverEvaporation CoefficientRoughness Coefficient
Evergreen needleleaf forest0.70.4
Evergreen broadleaf forest0.70.6
Bush0.70.4
Sparse woods0.70.3
Seaside wet lands0.70.2
Slope grassland0.70.1
Farmland0.70.15
Table
Table 6. The initial values of soil-related parameters.
Table 6. The initial values of soil-related parameters.
TypeThickness of Soil Layer (mm)Saturated Water ContentField CapacitySaturated Hydraulic Conductivity of Soil
(mm·h−1)		Soil Characteristic CoefficientWilting Point
CN1000510000.5020.3559.822.50.136
CN1003310000.4510.38.642.50.176
CN100396000.5150.4221.952.50.296
CN1016910000.4380.19235.152.50.109
CN3004322000.4660.3385.372.50.202
CN300538500.4580.3532.812.50.231
CN3007515000.4590.3781.342.50.258
CN3014710000.4430.26214.882.50.149
CN3014913000.4290.21124.132.50.132
CN304236700.4460.2421.872.50.126
CN3067310000.4330.20129.312.50.121
Table
Table 7. Different parameter optimization results with different DEM databases in the Liuxihe model.
Table 7. Different parameter optimization results with different DEM databases in the Liuxihe model.
ParametersSoil Saturated Hydraulic Conductivity/KsSlope Roughness/nManning Coefficient/MannSoil Layer Thickness/ZsSoil Characteristic Coefficient/bThe River Bottom Slope/Bs
SRTM30.6930.5010.5060.7730.5081.497
ASTER30.6740.5040.5220.5850.7980.685
ParametersThe River Bottom Width/BwSaturated Water Content/CsatField Capacity/CfcEvaporation Coefficient/vWilting Percentage/CwSide Slope Grade/Ss
SRTM30.5110.5720.6080.5170.7160.588
ASTER30.5081.081.180.5390.5161.255
Table
Table 8. Percentage of different rainfall distribution types in total floods.
Table 8. Percentage of different rainfall distribution types in total floods.
StatisticsRainfall Distribution Type
Upstream and Downstream TypeMidstream and Downstream TypeDownstream TypeWhole Basin Type
Quantity210235
Percentage (%)52557.512.5
Table
Table 9. Statistical indicators of flood simulation results before and after optimization of different Liuxihe model parameters.
Table 9. Statistical indicators of flood simulation results before and after optimization of different Liuxihe model parameters.
DatabaseParametersNash–Sutcliffe CoefficientPeak Flow
Relative Error (%)		Water
Balance Coefficient		Peak Flow
Duration Difference (Hour)
ASTER3Initial0.41851.810.6343.900
Optimized0.8845.000.9960.425
SRTM3Initial0.13369.040.4239.025
Optimized0.8835.771.0060.425
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Share and Cite

MDPI and ACS Style

Xu, S.; Chen, Y.; Xing, L.; Li, C. Baipenzhu Reservoir Inflow Flood Forecasting Based on a Distributed Hydrological Model. Water 2021, 13, 272. https://doi.org/10.3390/w13030272

AMA Style

Xu S, Chen Y, Xing L, Li C. Baipenzhu Reservoir Inflow Flood Forecasting Based on a Distributed Hydrological Model. Water. 2021; 13(3):272. https://doi.org/10.3390/w13030272

Chicago/Turabian Style

Xu, Shichao, Yangbo Chen, Lixue Xing, and Chuan Li. 2021. "Baipenzhu Reservoir Inflow Flood Forecasting Based on a Distributed Hydrological Model" Water 13, no. 3: 272. https://doi.org/10.3390/w13030272

APA Style

Xu, S., Chen, Y., Xing, L., & Li, C. (2021). Baipenzhu Reservoir Inflow Flood Forecasting Based on a Distributed Hydrological Model. Water, 13(3), 272. https://doi.org/10.3390/w13030272

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop