Ecological Modelling 462 (2021) 109792

Contents lists available at ScienceDirect

Ecological Modelling
journal homepage: www.elsevier.com/locate/ecolmodel

Application of Habitat Suitability Model Coupling with High - precision

Hydrodynamic Processes
Lu Yang a, Jingming Hou a, *, Long Cheng b, Pan Wang b, Zhanpeng Pan a, Tian Wang a,
Yongyong Ma a, Gao Xujun b, Sun Jixin c, Na Liu b
a
State Key Laboratory of Eco-hydraulics in Northwest Arid Region of China, Xi’an University of Technology, Xi’ an 710048, Shaanxi, China
b
Northwest Engineering Corporation Limited, Xi’ an 710065, Shaanxi, China
c
Xi ’an Aerospace Mapping Data Technology Co., Ltd, Shaanxi, China

A R T I C L E I N F O A B S T R A C T

Keywords: The habitat suitability model is coupled with the two-dimensional high-precision hydrodynamic model in this
High-precision hydrodynamic model paper. In the model, finite volume method (FVM) with Godunov type is adopted to solve the two-dimensional
Habitat suitability model hydrodynamic process, which has good robustness and second-order temporal and spatial accuracy; Habitat
Weighted usable area
suitability model is constructed by habitat suitability index based on IFIM theory. Meanwhile, the introduction of
Discharged flow
GPU acceleration technology greatly improves the simulation efficiency. Combined with the results of River2D,
Habitat suitability index
the results of this model have good accuracy. The model was applied to simulate the spawning habitat of
Gymnocypris eckloni in the lower reaches of a hydropower station in the Upper Yellow River, and the reasonable
discharged flow of hydropower station required for a good spawning habitat was 74 m3/s. It increased by
53.377% compared with the average annual discharge in May of the natural river. The evaluation of the
spawning habitat of Gymnocypris eckloni provides a scientific basis for protecting the integrity of river ecosystem
in this paper.

1. Introduction ecological operation and river ecological restoration. The United States
was the first to put forward the concept of eco-environmental water flow
The construction of a large number of water conservancy projects not requirement in the 1870 s, which stipulated that in order to avoid river
only ensures the reasonable allocation of water resources, but also plays ecosystem degradation, rivers should maintain the minimum ecological
a role in other benefits, thus becoming the basic guarantee for the sus­ flow (Covich, 1993; Xu et al., 2004). A large number of calculation
tainable and healthy development of the economy and society (Mao and methods emerged, including hydrology methods, such as 7Q10, Tennant
Dai, 2016). However, rivers are also the natural space for the survival of and Basic Flow method et al., hydraulic methods, such as wetted
various aquatic organisms (Cheng, 2008). With the increasing utiliza­ perimeter method and R2-Cross, and habitat modeling simulation
tion of river water resources, water ecological problems are gradually method, such as Instream Flow Incremental Methodology (Tennant,
emerging, the construction of reservoirs lead to the changes in river 1976; Palau and Alcazar, 1996; Gippel and Stewardson, 1998; Tarme,
hydrological regime, sharp decrease in biodiversity and degradation of 2003; Bovee, 1982).
local ecological functions (Foulger and Petts, 1984; Pringle, 2003; Habitat model simulation methods are also most widely used
Caissie, 2006); Meanwhile, dams block the passage of migratory fish, methods in the world. Instream flow incremental methodology (IFIM)
affecting the communication between species; In addition, the rapid was the earliest applied method among which to evaluate the impact of
change of river level also causes the alternating exposure and inundation the discharge changes on fish habitat, so as to improve ecological
of shoals, which affects the habitat and spawning of fish (Zhang, 2017; environment by regulating discharge (Bovee, 1982; Rhomas, 1998).
Li et al., 2013; Ma et al., 2020). With the proposal of habitat suitability index (HSI), it had been
To alleviate the stress of water conservancy project on river applied in more than 80% of the global habitat models (Yi et al., 2010),
ecosystem, people have taken a series of the measures of reservoir to determine the relationship between indicative target species and

* Corresponding Author.
E-mail address: [email protected] (J. Hou).

https://doi.org/10.1016/j.ecolmodel.2021.109792
Received 18 August 2021; Received in revised form 7 October 2021; Accepted 21 October 2021
Available online 1 November 2021
0304-3800/© 2021 Published by Elsevier B.V.
L. Yang et al. Ecological Modelling 462 (2021) 109792

habitat factors (Bockelmann et al., 2004). A one-dimensional Physical 2. GAST model

habitat simulation model (PHABSIM) based on IFIM uses this index to
simulate the quantitative relationship between discharge and suitable 2.1. Hydrodynamic model
habitat area (Stalnaker et al., 1995; Mark, 2010). Since the accuracy of
habitat factors is very important for the simulation of the river ecolog­ The basic governing equations of the surface hydrodynamic model
ical environment more accurately and the evaluation the quality of adopted in this model are the planar 2D-shallow water equations (SWES)
aquatic habitats, more and more hydraulic models have been intro­ (Hou et al., 2014), which is derived directly from hydrodynamic phys­
duced. In particular, the two-dimensional hydrodynamic model can ical process or simplified from three-dimensional Navier-Stokes(N-S)
accurately simulate the river hydrodynamic characteristics and fish equations. It is widely used in the hydrodynamic simulations of various
habitat, among which the two-dimensional finite element habitat model hydraulic engineering and other engineering (Hou et al., 2013b), to
River2D with average depth is widely adopted (Steffler and Blackburn, describe a large range of flow hydrodynamics with a much larger hori­
2002; Naderi, 2021). zontal dimension than the direction along water depth (Liang et al.,
However, due to the lack of professional maintenance of River2D, 2016). Based on the conservation of mass and momentum, assuming
complicated operation, low accuracy of terrain processing, and the that the hydrostatic pressure distribution is followed, and the kinematic
phenomenon of flashback often occurs in the simulation. In recent years, viscous term, turbulent viscous term, wind stress and Coriolis force are
scholars have introduced the complex and accurate two-dimensional or ignored. The 2D - shallow water equations (SWES) in a matrix form can
three-dimensional fluid dynamics models (Delft3D, Mike21, SSIIM be written as (Liang and Borthwick, 2009; Hou et al., 2014; Hou et al.,
model, etc.) coupled with habitat suitability models to continuously 2017).
improve the spatial and temporal resolution of river channel variables
∂q ∂F ∂G
(Olsen and Stokseth, 1995; Li and Xia, 2011; Belgiorno et al., 2013). But + + =S (1)
∂t ∂x ∂y
for the large scale and complex flow pattern of rivers, these models still
have the shortcomings of insufficient simulation accuracy and long

⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ / 0 √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ ⎤
h uh / vh
− gh∂ z ∂x − C u u2 + v2 ⎦
q = ⎣ uh ⎦, F = ⎣ u2 h + gh2 2 ⎦, G = ⎣ uvh / ⎦, S = ⎣ b
/
f
√ ̅ ,
̅̅̅̅̅̅̅̅̅̅̅̅̅̅ (2)
vh uvh v2 h + gh2 2 − gh∂zb ∂y − Cf v u2 + v2

simulation time, so the two-dimensional high-precision hydrodynamic

model is coupled with habitat suitability model, namely GAST model where, t is time, s; x,y is x− coordinate and y− coordinate, respectively;
(GPU Accelerated Surface Water Flow and Transport Model), to evaluate q is the vector of discharge variables; F,G is x and y directions fluxes,
the habitat quality of single or multi-species aquatic organisms. respectively; S is source terms, which consists of the bed slope source Sb
In GAST model, finite volume method (FVM) with Godunov type is and friction term Sf , Sbx = − gh∂zb /∂x and Sby = − gh∂zb /∂yin the x and
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
adopted to simulate two-dimensional hydrodynamic process, which has y directions, respectively; Sfx = − Cf u u2 + v2 and Sfy = −
good robustness and second-order temporal and spatial accuracy to √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
Cf v u2 + v2 in the x and y directions, respectively; h is water depth, m;
acquire more accurate habitat factor values. And the introduction of
u, v is the velocity in x and y directions respectively, m/s; qx ,qy is the
GPU acceleration technology greatly improves the simulation efficiency.
discharge per unit width in the x and y directions, respectively; the
In terms of simulating the hydrodynamic process, compared with
relation between velocity components and discharges per unit width is
Mike21 FM model, the calculation efficiency is improved by 1.06 to 2.37
reported by qx = uh, qy = vh, m2/s; zb is the bottom elevation of river
times and the calculation accuracy of 5 m accuracy terrain is improved
by 1.07 to 9.56 times in simulating complex flow patterns based on bed, m; Cf is the bed roughness coefficient evaluated herein by gn2 /h1/3
GAST model (Hou and Li, 2018). In addition, most models built are with n being the Manning coefficient, s/m1/3.
CPU-based, GAST model can improve the simulation computing effi­
ciency by 23.88 to 158.72 times by using GPU acceleration technology 2.2. Habitat suitability model
compared with CPU technology (Gong and Hou, 2020). Based on this,
Habitat suitability model used habitat suitability index was coupled, Habitat suitability model assumes that water depth, velocity, the
which makes the model a more powerful tool for evaluating habitat substrate are the main habitat factors based on IFIM theory, and adopts
quality. habitat suitability index (HSI) to quantify the specific behavior of a
In this paper, GAST model was obtained by coupling a two- target organism (Ma, 2020). In this model, the univariate suitability
dimensional high-precision hydrodynamic model with a habitat suit­ curve is the most commonly used to characterize the habitat suitability
ability model. Firstly, the basic governing equations and numerical so­ of target species, whose habitat suitability is defined as 0 (the least
lutions of this model were described, and GPU acceleration technology suitable) to 1(the most suitable) (Armour and Taylor, 1991). On the
was introduced to greatly improve the simulation efficiency. Secondly, basis of establishing habitat suitability index (HSI) curves of each
the lower reach of a proposed hydropower station in the Upper Yellow habitat factor, the combined suitability index (CSI) is calculated from
River was taken as the research object, and Gymnocypris eckloni was the individual suitability of the major habitat factors (also from 0 to 1)
selected as the target fish, the effective spawning habitat of fish under (Jowett and Davey, 2007).
various flow conditions was simulated based on this model. Besides, the The weighted usable area (WUA) was calculated by simulating the
results were validated with that from River2D model. Ultimately, the suitability of target species to hydraulic factors, WUA is the sum of the
minimum discharge required for the spawning habitat of Gymnocypris product of each control unit area and combined habitat suitability index
eckloni downstream of the proposed hydropower station to be in good (CSI) in the study area (Yi et al., 2014), and is expressed as,
condition was determined.

2
L. Yang et al. Ecological Modelling 462 (2021) 109792

∑
n ( ) F(q)⋅n is the flux vector normal to the boundary; k is the number of
WUA = CSI HSIvel i , HSIdep i , HSIsub i × ai (3) control cell edges;lk is the length of side k in the control cell i; u,vare the
i=1
velocity in x and y directions respectively, m/s; h is water depth, m; nx
Combined habitat suitability index (CSI) can be obtained by the and ny are the unit outward vectors in the x and y directions of the
following methods, considered boundary, respectively; gis the gravitational acceleration, m/
( ) s2.
CSI HSIvel i , HSIdep i , HSIsub i = HSIvel i × HSIdep i × HSIsub i (4) In this model, the Van Leer approximate Riemann solver with the
contact wave restored (HLLC) is adopted to compute the fluxes of mass
( ) ( )1/3
CSI HSIvel i , HSIdep i , HSIsub i = HSIvel i × HSIdep i × HSIsub i (5) and momentum at the interface, which automatically satisfies the
Godunov scheme and is the most suitable numerical solution for the dry
( ) ( ) and wet boundary (Marche, et al., 2007; Hou et al., 2013a). The bottom
CSI HSIvel i , HSIdep i , HSIsub i = min HSIvel i , HSIdep i , HSIsub i (6) slope source term is treated by the bottom slope flux method, which is
well coordinated with the interface flux and satisfies the condition of full
where, WUA is the weighted usable area, m2; CSI is the combined stability, and combined with non-negative depth reconstruction, the dry
habitat suitability index; i is the number of control cell; ai is the area of and wet boundary problems can also be effectively dealt with (Bermu­
control cell i, m2; veli ,depi ,subi are water depth, velocity, the substrate in dez and Vazquez, 1994), In this simulation, the grid cell judgment
the control cell i, respectively; HSIveli , HSIdepi , HSIsubi are the suitability condition of dry and wet water depth is 0.000 001 m. Explicit - implicit
index of water depth, velocity, the substrate in the control cell i, friction treatment method with better stability is used to calculate the
respectively; Eq. (4) represents the result of comprehensive action of frictional flux term to achieve implicit computation without iteration,
three habitat factors; Eq. (5) represents the compensating influence the calculation accuracy and efficiency are considered (Hou et al.,
among habitat factors when a certain habitat factor is unfavorable; Eq. 2013b). Meanwhile, MUSCL method is used to reorganize the variables
(6) represents CSI is determined by the most unsuitable habitat factor with spatial second order accuracy (Hou et al., 2014). Runge-Kutta
among three habitat factors. method is used to implement time step of time variable to obtain sec­
In this paper, only three habitat factors were considered, namely ond order time accuracy.
water depth, discharge and the substrate. If any of the selected habitat In addition, CFL (Courant Friedrichs Lewy) condition is used to limit
factors were not unsuitable, the habitat of organisms would be greatly the time step to maintain its good stability. The time step formula is as
affected. Thus, From the perspective of the most unfavorable habitat follows (Courant et al., 1967),
factors, Eq. (6) was adopted in this paper to obtain the comprehensive ( )
suitability index. Ri
Δt = CFLmin √̅̅̅̅̅̅̅̅̅̅̅̅̅̅ ̅ √̅̅̅̅̅̅ (11)
u2i + v2i + ghi
3. Numerical solution method
where, Ri is the minimum distance from the center of control cell to the
3.1. Numerical solution of hydrodynamic model boundary; i is the number of control cell; u,v are the velocity in x and y
directions respectively, m/s; h is water depth, m; CFL is called the
In solving two-dimensional shallow water equations (SWES), usually Courant and has a value from 0 to 1. With the increase of Coulomb
finite volume method, finite difference method and finite element number, the convergence speed of grid computing is gradually accel­
method are adopted, a multi-speed discrete Boltzmann method was erated and the stability is gradually decreased. CFL is defined as 0.5 in
proposed in recent years, but it is not widely adopted (Rocca et al., this model.
2020). The cell centered finite volume (CCFV) method of Godunov
scheme is used in this paper (Hou et al., 2015), the study area is divided 3.2. GPU parallel computing technology and model coupling
into triangular or quadrilateral computing units, and the control volume
of cell centered Finite Volume (CCFV) is used as the cell. In control cell i, To solve the problem of large amount of calculation, low calculation
the integral form of the governing equations is expressed as (Hou et al., efficiency and calculation divergence caused by high-precision terrain,
2013a), GPU (Graphic Processing Unit) acceleration technology based on CUDA
∫
∂q
∫ (
∂F ∂G
) ∫ (Compute Unified Device Architecture) is adopted in this model, which
dΩ + + dΩ = SdΩ (7) is developed by NVIDIA and written in C++ language (Liang et al.,
Ω ∂t Ω ∂x ∂y
2016). Compared with CPU, GPUs can provide tens or even hundreds of
Ω

Based on divergence theorem, it becomes, times more power than CPUs in parallel computing as the structure of
∫
∂q
∮ ∫
( GPU is relatively simple, the number of threads and registers is far more
Ω + F(q)⋅ndΓ = Sb + Sf )dΩ (8) than that of CPU; mainstream GPUs are more than 10 times that of CPUs
Ω ∂t
of the same period in terms of single-precision floating-point processing
Γ Ω

In the rectangular control cell in this paper, the flux integral is: power due to the powerful floating-point processing capacity. CPU and
∮ ∑
4
F(q)⋅ndΓ = Fk (q)⋅nk lk (9)
Γ k=1

Regardless of the Coriolis effects, kinetic and turbulent viscous terms

and wind stresses, the flux vector F(q)⋅n of the corresponding interface
can be expressed as,
⎡ ⎤
( ) ( 2 uhn2x/+)νhny
F(q)⋅n = Fnx + Gny = ⎣ u h + gh (
2 nx + uνhny ⎦
/ ) (10)
uνhnx + ν2 h + gh2 2 ny

where, t represent the time; q are the vectors of flow variables; Ω is the
volume of control cell i; Γ is the boundary of control cell; n denotes the
unit outward vector normal to the considered boundary and is defined; Figure 1. CPU and GPU architecture.

3
L. Yang et al. Ecological Modelling 462 (2021) 109792

Figure 2. GPU acceleration technology implementation process and model coupling

GPU architecture as shown in Figure 1. In actual execution, the program distribution during the main spawning period (May) was studied to
code of GPU acceleration technology based on CUDA includes host code reflect the habitat conditions of the river. The average annual discharge
running on CPU and device code running on GPU, while the data of natural river in May was 434 m3/s; During the operation of hydro­
transfer between CPU and GPU is implemented by cudamemcpy func­ power station, the average annual discharge of normal (392.8 m3/s) and
tion (Smith and Liang, 2013). The CPU and GPU architecture is shown in dry years (352.7 m3/s) in May should be considered as the discharged
Figure 1. flow of hydropower station to ensure the fish breeding in the down­
Based on CUDA architecture the C++ language is used to read the stream channel. In addition, the minimum discharge of the river reach
grid, initial and boundary conditions and output the results and to under the dam should be no less than 63.6 m3/s to maintain aquatic
realize GPU high-speed parallel loop computing calculation which is ecological stability and river landscape. When the high-precision two-
included data initialization, memory space allocation, flux calculation dimensional hydrodynamic model was used to simulate the distribution
and model coupling, etc. of water depth and velocity in river, the discharge was the upstream
This simulation is performed on a PC equipped with graphics card of boundary condition and the corresponding water level was the down­
NVIDIA GeFrce RTX 2080. The GPU acceleration technology imple­ stream boundary condition. Water level downstream under different
mentation process and model coupling is shown in Figure 2. discharges was shown in Table 1.

4. Application and validation of model 4.2. Suitable curve of habitat factors

4.1. Overview and Basic Data Through the investigation of aquatic experts, the ecological habits
and hydraulics characteristics of the habitat of Gymnocypris eckloni were
To promote the local economic development of Qinghai province, it collected and summarized, and the spawning period from early May to
is proposed to build a hydropower station on the main stream of the June of it was obtained. May was used as the main spawning period in
Upper Yellow River, which is about 5 km upstream of Laja town, Maqin this study and focused on the suitability of it to the velocity, water depth
county, Qinghai province. The dam site of the hydropower station and the substrate. The suitability of the velocity, water depth and the
controls a basin area of 98,346 km2 with mean annual discharge is 530 substrate of Gymnocypris eckloni during spawning period is: the suitable
m3/s, and the normal water level for it is 3275 m, the corresponding water depth 0.1 to 2.0 m and the optimal water depth is 0.4 to 1.2 m; the
storage capacity is 1.482 billion m3, of which the regulated storage ca­ suitable velocity is 0.25 to 1.5 m/s and the optimal velocity is 0.5 to 1.0
pacity is 706 million m3. The reach of 1~ 8 km downstream of the dam m/s; the substrate condition of this river reach is sand and gravel mixed
of this hydropower station is a more important fish habitat and a major with sediment, which shows that the suitable substrate condition is sand
spawning area based on the results of fish habitat survey. To protect the gravel and fine sand and distributed in this reach.
downstream river ecological environment, the river reach was taken as Habitat suitability index (HSI) was adopted to quantify the fish
the study area, which twists and turns and is like Chinese character “ji” habitat preference, which was expressed as a value between 0 and 1, the
(meaning a few) in plane, the location of it was shown in Figure 3. higher the value, the better the adaptability. Based on this, the suit­
As an advanced species of aquatic organisms, fish habitat indicates ability index of hydraulic habitat parameters was determined and the
the quality of river ecological environment. In this reach, Gymnocypris suitability curves of major hydraulic factors was drawn of Gymnocypris
eckloni was selected as a target fish for aquatic organisms and its habitat eckloni during spawning period, as shown in Figure 4.

4
L. Yang et al. Ecological Modelling 462 (2021) 109792

Figure 3. Location of the study reach.

4.3. Parameter calibration

Table 1
Downstream water level of the study reach under different discharges
To reasonably select the roughness value of reach, the discharge of
Discharge Water level Discharge Water level 230 m3/s was used for calibration. Based on the simulated and the
(m3/s) (m) (m3/s) (m) measured water level of each section, the simulated water level is in
0 0 318.0 3072.3 good agreement with the measured when the comprehensive roughness
26.5 3070.4 352.7 3072.5 of the reach is 0.05 s/m1/3, the results were shown in Figure 5. Thus, the
53.0 3070.8 392.8 3072.6
63.6 3070.9 434.0 3072.8
manning coefficient n = 0.05s/m1/3.
74.0 3071.0 530.0
3073.1
79.5 3071.0 689.0 3073.6
106.0 3071.2 848.0 3073.8 4.4. Model validation
159.0 3071.5 1060.0 3074.2
212.0 3071.8 – – Due to the lack of actual monitoring data on the Weighted Usable
Area (WUA) of spawning habitat of Gymnocypris eckloni in the study
area, the simulation results of River2D model were adopted to verify the

Figure 4. Suitability curves of major hydraulic factors of Gymnocypris eckloni during spawning period (a) Water depth; (b) Velocity;(c) Substrate

5
L. Yang et al. Ecological Modelling 462 (2021) 109792

May, the average annual discharge of normal (392.8 m3/s) and dry years
(352.7 m3/s) in May and and the proposed minimum discharge of 74
m3/s was considered to further determine the reasonable discharged
flow of hydropower station. The high-precision two-dimensional hy­
drodynamic model (GAST model) was adopted to simulate the distri­
bution of water depth and flow velocity in the study area. The study
reach was divided into 404,544 rectangular grids with a cellsize of 5m
using ArGIS software to generate high-precision terrain files for GAST
models. The distribution of water depth and flow velocity was shown in
Figure 7.
Based on the simulation results of water depth and velocity under
different discharges, the average water depth and velocity increased
Figure 5. Comparison between simulated and measured water levels of each
significantly when the discharge increased from 74 m3/s to 352.7 m3/s,
section in the study area
while there was little difference between 352.7 m3/s, 392.8 m3/s and
434 m3/s. When the discharge was 74 m3/s, the average water depth
rationality of the simulation results of GAST model. River2D is called the was about 1.20 m and the average velocity was about 0.38 m/s; while
two-dimensional depth averaged model of river hydrodynamics and fish the discharge was 434 m3/s, the average water depth was about 2.31 m
habitat, which uses the finite element method to numerically solve the and the average velocity was about 0.59 m/s. It could thus be seen that
depth-average model equations. The discharges of 26.5 m3/s, 53 m3/s, the average water depth was about 48.05 % of the natural May 434 m3/
74 m3/s, 79.5 m3/s, 106 m3/s, 159 m3/s, 212 m3/s, 318 m3/s, 434 m3/s, s, and the average velocity was about 35.59 % of that when the
530 m3/s, 689 m3/s, 848 m3/s, 1060 m3/s were selected to calculate the discharge was 74 m3/s.
effective habitat area (which was represented by the Weighted Usable
Area) of Gymnocypris eckloni. The simulation results based on GAST
5.2. Combined habitat suitable index
model and River2D model were shown in Table 2, and the relationship
between the discharges were obtained based on the simulation of GAST
Due to the substrate condition of this river reach is sand and gravel
model and River2D model, as shown in Figure 6.
mixed with sediment, it meets the substrate condition of Gymnocypris
As could be seen from Table 2 and Figure 6, the relationship between
eckloni spawning, the habitat suitable index of the substrate in this river
the weighted available area of spawning habitat and the discharges
reach is 1. Based on the simulation results of three hydraulic factors,
based on GAST model and River2D model was almost identical. the
namely water depth, velocity and substrate, a combined habitat suit­
results showed that the effective habitat area begins to increase with the
ability index (CSI) was obtained by coupling fish habitat simulation
increase of the discharge, and when it reached the peak, the effective
method to reflect the habitat distribution of Gymnocypris eckloni during
habitat area began to decrease with the increase of the discharge.
spawning period, the distribution of it was shown in Figure 8.
This phenomenon is mainly caused that as the discharge increased,
It could be seen that the spawning habitat area increased when the
so done the water depth and velocity, that is, the area of the suitable
discharge decreased from 434 m3/s to 74 m3/s, and the distribution
water depth and velocity for Gymnocypris eckloni to spawn also
location shifted from two banks to the center of river. In Figure 6 (a),
increased. but after reaching a discharge, as the discharge continued to
under the average annual discharge in May of natural river of 434 m3/s,
increase and the velocity was too fast and the water was too deep, was
the spawning habitats were mainly distributed on both sides of river and
greater than that of Gymnocypris eckloni spawning suitable water depth
the spawning sites were more concentrated at the bend of river; In
and velocity. This moment, the effective habitat area tends to decrease.
Therefore, there was a peak in the weighted available area at a discharge
of 74 m3/s. Compared with River2D model, the effective spawning
habitat area of Gymnocypris eckloni in the study reach obtained by GAST
model was slightly larger, which was due to more accurate velocities and
water depths obtained by GAST model under high resolution terrain
conditions.

5. Results and analysis

5.1. Distribution of water depth and velocity

According to the relationship between the weighted available area of

Gymnocypris eckloni spawning habitat and the discharges, the weighted
available area reaches the maximum when the the discharge was 74 m3/ Figure 6. Relationship between discharges and WUA based on GAST model
s. Therefore, the average annual discharge of natural river (434 m3/s) in and River2D model

Table 2
Changes of WUA under different discharges based on GAST model and River2D model
Discharge (m3/s) WUA (m2) Discharge (m3/s) WUA (m2)
GAST model River2D model GAST model River2D model

0.0 0.00 0.00 159.0 162,398.63 159,325.00

26.5 166,946.31 166,888.00 212.0 149,124.00 146,428.00
53.0 198,289.38 198,282.00 318.0 144,280.98 141,787.00
63.6 200,484.70 198,530.00 530.0 116,742.50 114,779.00
74.0 202,492.08 198,775.00 689.0 84,897.14 83,253.00
79.5 197,268.02 196,433.00 848.0 64,853.66 64,696.00
106.0 183,545.24 181,059.00 1,060.0 55,410.42 55,084.00

6
L. Yang et al. Ecological Modelling 462 (2021) 109792

Figure 7. Distribution of water depth and velocity under different discharges (a) Average annual discharge in May of natural river of 434 m3/s; (b) The discharge in
May of normal year of 392.8 m3/s; (c) The discharge in May of dry year of 352.7 m3/s; (d) Proposed minimum discharge of 74 m3/s

Figure 6 (b) and (c), under the discharge in May of normal year of 392.8 WUA was 132,022.88 m2 under the average annual discharge in May of
m3/s and the discharge in May of normal year of 352.7 m3/s, the natural river of 434 m3/s; After the operation of hydropower station, the
spawning habitats distribution were consistent with Figure 6 (a), there WUA was 136,627.75 m2, 143,027.48 m2 and 202,492.08 m2 under the
was not much difference between them. In Figure 6 (d), under the discharge in May of normal year of 392.8 m3/s, under the discharge in
proposed minimum discharge of 74 m3/s, the spawning habitats shifted May of dry year of 352.7 m3/s and the proposed minimum discharge of
to the center of river, mainly distributed in the center of river, which 74 m3/s, respectively. Compared with in May of natural river, the
caused this change was that study river reach belonged to a wide and effective habitat area in May of normal year and in May of dry year, with
shallow one, most of river sections had broad and flat undersides. With an increase of 3.488 % and 8.335 %, respectively; while e proposed
the decreasing of river discharge, the water depth, velocity and cross- minimum discharge of 74 m3/s, with an increase of 53.377 %.
section width decreased. As a result, the effective habitat area in the To sum up, if the discharge of 392.8 m3/s or 352.7 m3/s is taken as
original suitable area for spawning decreased, and the central area of the minimum ecological discharge, the increase of effective habitat area
river changed from the original unsuitable area to the suitable area. was not significant; but the discharge of 74 m3/s is selected as the
Meanwhile, the Weighted Usable Area (WUA) of spawning habitat of minimum ecological discharge, the effective habitat area increased
Gymnocypris eckloni were calculated under different discharges. The significantly. Compared with the discharge in May of natural river of

7
L. Yang et al. Ecological Modelling 462 (2021) 109792

Figure 8. Distribution of WUA under different discharge. (a) Average annual discharge in May of natural river of 434 m3/s; (b) The discharge in May of normal year
of 392.8 m3/s; (c) The discharge in May of dry year of 352.7 m3/s; (d) Proposed minimum discharge of 74 m3/s

434 m3/s, the effective habitat area increased by 70,469.20 m2, with an was applied to simulate the spawning habitat of Gymnocypris eckloni in
increase of 53.377%. Thus, the effective habitat area suitable for the lower reaches of a hydropower station in the Upper Yellow River, the
spawning is in a good condition under the discharge of 74 m3/s. main conclusions are as follows:

6. Conclusion I River2D model and GAST model were used to simulate the
weighted available area of Gymnocypris eckloni spawning habitat
GAST model, which was obtained by coupling a two-dimensional under different discharges in the study area. In the case of water
high precision hydrodynamic model with a habitat suitability model, depth, velocity and the substrate, it was shown that GAST model

8
L. Yang et al. Ecological Modelling 462 (2021) 109792

could well evaluate the habitat extension of fish habitat com­ Gippel, G.J., Stewardson, M.J., 1998. Use of wetted perimeter in defining minimum
environmental flaws. Regulated Rivers: Research & Management 14 (1), 53–67.
bined with the results of River2D model.
https://doi.org/10.1002/(SICI)1099-1646(199801/02)14:1<53::AID-RRR476>3.0.
II Based on the relationship between the weighted available area of CO;2-Z.
spawning habitat and the discharge, the effective habitat area Gong, J.H., Hou, J.M., Xue, Y., et al., 2020. Research on the efficiency of urban
began to increase with the increasing of the discharge, and when stormwater process simulation by GPU [J]. Environmental Engineering 262 (04).
https://doi.org/10.13205/j.hjgc.202004029, 167-172+178(in Chinese).
it reached the peak, the effective habitat area began to decrease Hou, J.M., Li, G.Y, Li, G.D., et al., 2018. Application of efficient high-resolution
with the increasing of the discharge. Thus, the effective habitat hydrodynamic model to simulations of flood propagation. Journal of Hydroelectric
area suitable for spawning ground reached the maximum at the Engineering 37 (2), 96–107 (in Chinese).
Hou, J., Simons, F., Mahgoub, M., Hinkelmann, R., 2013b. A robust well-balanced model
discharge was 74 m3/s. on unstructured grids for shallow water flows with wetting and drying over complex
III The average annual discharge of natural river (434 m3/s) in May, topography. Computer Methods in Applied Mechanics and Engineering 257,
the average annual discharge of normal (392.8 m3/s) and dry 126–149. https://doi.org/10.1016/j.cma.2013.01.015.
Hou, J.M., Wang, R., Jing, H.xiao, Zhang, X., Liang, Q., hua, Di, Y.yan, 2017. An efficient
years (352.7 m3/s) in May and and the proposed minimum dynamic uniform Cartesian grid system for inundation modeling. Water Science and
discharge (74 m3/s) were considered to further determine the Engineering 10, 267–274. https://doi.org/10.1016/j.wse.2017.12.004.
reasonable discharged flow of hydropower station. Compared Hou, J.M., Liang, Q., Zhang, H., Hinkelmann, R., 2015. An efficient unstructured MUSCL
scheme for solving the 2D shallow water equations. Environmental Modelling and
with the discharge in May of natural river of 434 m3/s, under the Software 66, 131–152. https://doi.org/10.1016/j.envsoft.2014.12.007.
discharge of 74 m3/s, the effective habitat area increased by Hou, J.M., Liang, Q., Simons, F., Hinkelmann, R., 2013a. A 2D well-balanced shallow
70,469.20 m2, with an increase of 53.377%, it is far more sig­ flow model for unstructured grids with novel slope source term treatment. Advances
in Water Resources 52, 107–131. https://doi.org/10.1016/j.advwatres.2012.08.003.
nificant than the increase under the discharges of 392.8 m3/s and
Hou, J.M., Liang, Q., Zhang, H., Hinkelmann, R., 2014. Multislope MUSCL method
352.7 m3/s. From this indicated that the effective habitat area applied to solve shallow water equations. Computers and Mathematics with
suitable for spawning is in a good condition under the discharge Applications 68, 2012–2027. https://doi.org/10.1016/j.camwa.2014.09.018.
of 74 m3/s with considering water depth, velocity and the sub­ Jowett, Ian G., Davey, Andrew J.H., 2007. A comparison of composite habitat suitability
indices and generalized additive models of invertebrate abundance and fish
strate as the main habitat factors. Thus, the discharge of 74 m3/s presenceehabitat availability. Transactions of the American Fisheries Society 136,
was selected to the reasonable discharged flow of hydropower 428–444.
station. Li, J., Xia, Z.Q., Dai, H.C., 2013. Effect of the Three Gorges Reservoir initial filling on
downstream habitat suitability of the typical fishes[J]. Journal of hydraulic
engineering 44 (08), 892–900. https://doi.org/10.13243/j.cnki.slxb.2013.08.001
(in Chinese).
Declaration of Competing Interest Li, J., Xia, Z.Q., 2011. Study on instream ecological flow of the middle Yangtze River
based on physical habitat simulation. Journal of hydraulic engineering 42 (06),
678–684. https://doi.org/10.1080/01688638808402819 (in Chinese).
No conflict of interest exits in the submission of this manuscript. The Liang, Q., Xia, X., Hou, J., 2016. Catchment-scale High-resolution Flash Flood Simulation
authors declare that they have no known competing financial interests Using the GPU-based Technology. Procedia Engineering 154, 975–981. https://doi.
org/10.1016/j.proeng.2016.07.585.
or personal relationships that could have appeared to influence the work Liang, Q., Borthwick, A.G.L., 2009. Adaptive quadtree simulation of shallow flows with
reported in this paper. wet-dry fronts over complex topography. Computers and Fluids 38, 221–234.
https://doi.org/10.1016/j.compfluid.2008.02.008.
Mao, J., Dai, H., 2016. Effects of large water conservancy and hydropower projects on
Acknowledgements major aquatic species and regulation. Journal of Hohai University 44, 240–245.
https://doi.org/10.3876/j.issn.1000-1980.2016.03.009 (in Chinese).
Ma, B., Dong, F., Peng, W.Q., et al., 2020. Evaluation of impact of spur dike designs on
This work is partly supported by the Numerical Simulation of Flood
enhancement of aquatic habitats in urban streams using 2D habitat numerical
Process in Urban Areas with Fine Terrain and Lack of Pipe Network Data simulations[J]. Global Ecology and Conservation, 2020 24, e01288. https://doi.org/
(No. 52079106); High Resolution Numerical Simulation of Sediment 10.1016/j.gecco.2020.e01288.
Carrying Capacity Mechanism and Erosion Process of Over surface Flow Mark, Gard, 2010. Response to Williams (2010) on Gard (2009): Comparison of
spawning habitat predictions of PHABSIM and River2D models[J]. International
on Whole Sand Slope (No. 52009104); Chinesisch - Deutsches Mobilit journal of river basin management: JRBM 8 (1), 121–125. https://doi.org/10.1080/
taprogramm: High – Resolution Numerical Simulating and Predicting 15715121003714969.
Methods for Urban Floods (No. M-0427). Marche, F., Bonneton, P., Fabrie, P., et al., 2007. Evaluation of well-balanced bore-
capturing schemes for 2D wetting and drying processes[J]. International Journal for
Numerical Methods in Fluids 53 (5), 867–894. https://doi.org/10.1002/fld.1311.
References Naderi, M.H., Arab, N., Jahandideh, O., et al., 2021. Estimation of Optimal Release Flow
Range from Jamishan Dam Considering the Optimal Instream Ecological Water
Demand for Conservation the Habitat Potential of the Dinavar River[J]. Journal of
Armour, C.L., Taylor, J.G., 1991. Evaluation of the instream flow incremental
Water and Soil Science 35 (2), 203–225. https://doi.org/10.22067/
methodology by United States Fish and Wildlife Service Field Users. Fisheries 16 (5),
jsw.2021.67061.0.
36–43. https://doi.org/10.1577/1548-8446(1991)016<0036:EOTIFI>2.0.CO;2.
Olsen, N.R.B., Stokseth, S., 1995. Three-dimensional numerical modelling of water flow
Bovee, K.D., 1982. A guide to stream habitat analysis using the instream incremental
in a river with large bed roughness. Journal of Hydraulic Research 33 (4), 571–581.
flow methodology. Instream Flow Information Paper. United States Fish and Wildlife
https://doi.org/10.1080/00221689509498662.
Service. FWS/OBS-82/26 12, 1–248. https://doi.org/10.5897/SRE11.355.
Pringle, C.M., 2003. Hydrologic connectivity and the management of biological reserves:
Bockelmann, B.N., Fenrich, E.K., Lin, B., et al., 2004. Development of an ecohydraulics
a global perspective[J]. Ecological Applications 11, 981–998.
model for stream and river restoration[J]. Ecological Engineering 22, 227–235.
Palau, A., Alcazar, J., 1996. The basic flow: An alternative approach to calculate
https://doi.org/10.1016/j.ecoleng.2004.04.003.
minimum environmental instream flows. // Proceeding of 2nd International
Belgiorno, V., Naddeo, V., Scannapieco, D., et al., 2013. Ecological status of rivers in
Symposium on Habitats Hydraulics 547–558.
preserved areas: Effects of meteorological parameters. Ecological Engineering 53,
Rhomas, B.H., 1998. The future of habitat modeling and instream flow assessment
173–182. https://doi.org/10.1016/j.ecoleng.2012.12.039.
techniques[J]. Regulated Rivers: Research and Management 14 (2), 405− 420.
Bermudez, A., Vazquez, M.E., 1994. Upwind methods for hyperbolic conservation laws
https://doi.org/10.1002/(SICI)1099-1646(1998090)14:5<405::AID-RRR510>3.0.
with source terms. Computers and Fluids 23, 1049–1071. https://doi.org/10.1016/
CO;2-0.
0045-7930(94)90004-3.
Rocca, M.L., Miliani, S., Prestininzi, P., 2020. Discrete Boltzmann Numerical Simulation
Cheng, Z.N., 2008. Study on the simulation of river water ecological environment quality
of Simplified Urban Flooding Configurations Caused by Dam Break[J]. Frontiers in
change -fish model[D]. Beijing Institute of Civil Engineering and Architecture. (in
Earth Science 8, 346. https://doi.org/10.3389/feart.2020.00346.
Chinese).
Steffler, P., Blackburn, J., 2002. River2D, Two-Dimensional Depth Averaged Model of
Caissie, D., 2006. The thermal regime of rivers:a review[J]. Fresh Water Biology 51,
River Hydrodynamics and Fish Habitat, Introduction to Depth Averaged Modeling
1389–1406. https://doi.org/10.1111/j.1558-5646.2012.01717.x.
and User’s Manual[J]. river2d users manual.
Covich, A., 1993. Water in crisis: a guide to the world’s fresh water resources[J]. New
Stalnaker, C., Lamb, B.L., Henriksen, J., et al., 1995. The Instream Flow Incremental
York New York Oxford University Press, pp. 40–55.
Methodology: a primer for IFIM [J]. department of interior national biological
Courant, R., Friedrichs, K., Lewy, H., 1967. On the Partial Difference Equations of
service biological report.
Mathematical Physics[J]. Ibm Journal of Research & Development 11 (2), 215–234.
Smith, L.S., Liang, Q., 2013. Towards a generalised GPU/CPU shallow-flow modelling
Foulger, T.R., Petts, G.E., 1984. Water quality implications of artificial flow fluctuations
tool. Computers and Fluids 88, 334–343. https://doi.org/10.1016/j.
in regulated rivers[J]. Science of the Total Environment 37 (2-3), 177–185. https://
compfluid.2013.09.018.
doi.org/10.1016/0048-9697(84)90094-9.

9
L. Yang et al. Ecological Modelling 462 (2021) 109792

Tennant, D.L., 1976. Instream flow regiments for fish, wildlife, recreation and related Zhang, L.Y., 2017. A Brief discussion on dam construction and fish conservation in the
environmental resources fisheries management and ecology. Fisheries 1 (4), 6–10. United States - the implementation and consideration of the Endangered Species Act
Tarme, R.E., 2003. A global perspective on environmental flow methodogies for rivers. of the United States[J]. Science and Technology Wind 000 (015), 110. https://doi.
River Research and Applications 19 (5-6), 397–441. org/10.19392/j.cnki.1671-7341.201715098 (in Chinese).
Xu, Z.X., Chen, M.J., Dong, Z.C., 2004. A review on the calculation methods of river
ecological water requirement[J]. Journal of Hohai University (01), 5–9 (in Chinese).
Yi, Y., Wang, Z., Yang, Z., 2010. Two-dimensional habitat modeling of Chinese sturgeon
Further reading
spawning sites. Ecological Modelling 221, 864–875. https://doi.org/10.1016/j.
ecolmodel.2009.11.018. Hubbard, M.E., 1999. Multidimensional Slope Limiters for MUSCL-Type Finite Volume
Yi, Y., Cheng, X., Wieprecht, S., Tang, C., 2014. Comparison of habitat suitability models Schemes on Unstructured Grids. Journal of Computational Physics 155, 54–74.
using different habitat suitability evaluation methods. Ecological Engineering 71, https://doi.org/10.1006/jcph.1999.6329.
335–345. https://doi.org/10.1016/j.ecoleng.2014.07.034.

10