5. 1 Introduction
Several studies have reported that the rate of climate change may accelerate owing to both natural and anthropogenic activities. Notably, the Intergovernmental Panel on Climate Change (IPCC) (2007, 2012, and 2013) has addressed these issues in great detail. The IPCC (2013) has reported the global sea-level rise (SLR) scenarios indicating that the global average SLR in the 21st century continues to rise at an increased rate of 2.0 mm/year. Vietnam is a country that is strongly affected by climate change (Nicholls et al., 2008). To deal with this phenomenon, and particularly the SLR, the Ministry of Natural Resources and Environment of Vietnam (2016) reported the climate change scenarios for the SLR based on the IPCC reports. They report that SLR increases the salinity of both surface water and groundwater through seawater intrusion.
The Ho Chi Minh City (HCMC) is one of the top ten cities strongly impacted by the SLR (Nicholls et al., 2008) and frequently face flooding. To protect the city from inundation or submergence, the government commissioned several projects. One of the projects, sea dike construction at the mouth of the Saigon-Dongnai river in the Can Gio Bay as shown in Fig. 1 is now considered the best solution for the inundation by the SLR-induced tidal regime. However, the construction of a sea dike could have a significant impact on the Can Gio mangrove forest that is located in the Can Gio Bay. Nguyen (2014) reported that, although sea dike construction efforts have achieved a high degree of control over the salinity at center of HCMC, salinity reduction would directly impact the Can Gio Bay area, especially the mangrove forest ecosystem abutting it. The positive and negative impacts of a sea dike have been analyzed by many researchers. Several essential studies have been conducted over the past decades in this area, with new and valuable results being obtained (Nguyen et al., 2015;
Nguyen 2014; Ngoc et al., 2013). However, most of them focused on the dike’s capability to control the water level in and flood discharge from the upstream areas of HCMC. Moreover, these studies only identified the changing of water
level under the impact of sea dike construction. To the best of the authors’
knowledge, no studies have focused on the change of salinity in the Can Gio mangrove forest.
Owing its long coastline exceeding 3260 km² and two big river deltas, coupled with its tropical climate, Vietnam fulfills the fundamental requirements needed to establish mangrove ecosystems. More than 60 % of Vietnam’s mangroves are located in the Mekong Delta, and an additional 20 % are located in the Can Gio Bay, Southeast region of Vietnam (Hawkins et al., 2010).
The area of a mangrove ecosystem can expand rapidly in response to regional topographical and climate changes. Mangroves naturally complete their life cycles under salinity conditions (Flowers et al., 1986). A vital factor to mangrove ecosystems is the extent of the SLR. It is complicated to generalize about the effect of the SLR on the mangrove ecosystem. However, all mangrove systems occur somewhere with the different low or high salinity, which makes it clear that they are likely to be significantly influenced by any changes of salinity as well as the change of sea level. Soil salinity, however, characterizes the mangrove habitat and growth of some mangroves have been shown to be maximal under relatively low salinities (Clough, 1984). When the salinity of the soil increases, the mangroves face the problems of rising salt levels in the tissues and decreasing availability of water. The increasing salt levels in the issue may bring about a lessening in the net assimilation rate per unit leaf area (Ball, 1988;
Ball, 2002) and therefore reduce growth.
In the Can Gio Bay, there were a lot of researches related to the mangrove forest. However, most of the studies are associated with the ecosystem of the Can Gio mangrove forest such as Kuenzer and Tuan (2013) that assessed the ecosystem of the Can Gio mangrove biosphere reserve by combining earth-observation- and household-survey-based analyses. Luong (2011) conducted the study on the numerical investigation on the sediment transport trend, and they gave a mathematical model to simulate the sediment transport under current effects by tides and winds. Binh et al. (2008) used multi-temporal remote sensing
data to detect mangrove change and manage them. However, there is no study relating to the change of mangrove forest under the impact of SLR.
Assessing the impact of the SLR and the sea dike construction on salinity distribution in this area is absolutely imperative for the development of the Can Gio mangrove forest. This study aims to develop a comprehensive understanding of the impact of the SLR and the sea dike construction on coastal water salinity of the Can Gio Bay as shown in Fig. 5.1 with the aid of the horizontal two-dimensional hydrodynamic and convective-dispersive models.
Fig. 5.1 The Can Gio Bay study area, showing the location of the Can Gio mangrove forest, shoreline, the Go Cong - Vung Tau sea dike, the Go Cong - Can Gio sea dike, and measurement points of river discharge and water level.
5. 2 Methodology
Nowadays, with the development of numerical models, three-dimensional models are used widely (e.g., Chen et al., 2003; Reza et al., 2016). However, it generally takes a long time to calculate and is complicated to construct the model.
Besides that, because the water depth in the Can Gio Bay as shown in Fig. 5.1 is extensively shallow, and the flow of the Can Gio Bay is driven mostly by tidal flow and flood water inflow, horizontal velocity components are much greater than vertical ones. In addition to that, the impact of the SLR and the sea dike construction on the salinity environment in the Can Gio mangrove forest was evaluated using numerical simulations, in which the evaluation absolutely needed the change of horizontal salinity distribution, but the change of vertical profile of salinity was not necessarily required. Therefore in this study, a horizontal two-dimensional depth-averaged hydrodynamic model for shallow water was effectively utilized for calculating the tidal current (e.g., Hu and Kot, 1997;
Hiramatsu et al., 2005; Dutta et al., 2007; Tabata et al., 2013; Bellos and Tsakiris, 2015). A two-dimensional convective-dispersive model for simulating salinity has been coupled with the hydrodynamic model (e.g., Batu, 1993; Hiramatsu et al., 2005; Stevenson et al., 2007).The Governing equations of the model are below:
Continuity equation was used to calculate the water level:
U h
V h
0t x y
(5.1)
Momentum equations were utilized to calculate velocities in x- and y-directions:
2 2 2 2 2
4/3
2 2
h
U U U U U gn U U V
U V fV g A
t x y x x y h
(5.2)
2 2 2 2 2
4/3
2 2
h
V V V V V gn V U V
U V fU g A
t x y y x y h
(5.3)
A convective-dispersive equation was also used to calculate salinity:
( ) S ( ) S ( ) S ( ) h S ( ) h S
h U h V h h K h K
t x y x x y y
(5.4)
Where is the water level (m); t is the time (s); h is the bottom elevation (m); f is the Coriolis parameter (s-1), indicating the effect of Earth’s rotation; g is the acceleration due to gravity (m/s2); n is the Manning’s coefficient of roughness (s/m1/3), which depends on the bottom roughness of the study area; U and V are the depth-averaged
horizontal velocity components (m/s) in the x- and y-directions of the Cartesian coordinate system, respectively; S is the salinity (psu); and Ah and Kh are the coefficients (m2/s) of eddy viscosity and convective-dispersion in the horizontal direction, respectively, determined by the Smagorinsky model (Smagorinsky 1963):
2 2 1/2
1 2 1
2 2
h h m G
U V U V
A K S A
x x y y
(5.5) Where Sm is the Smagorinsky coefficient and AG is the mesh area (m2).
A finite difference method was used to solve the governing equations by using the staggered grid system. For the continuity equation and momentum equations, the leap-frog scheme was applied. For the convective-dispersive equation, the split operator approach was utilized. In the first step, the convective term was calculated using the first order upwind scheme. After that, the diffusive term was calculated by using the alternating direction implicit method and substituting the value of the results from the first step. The Can Gio Bay is a flat low-lying area, so the wetting-and-drying scheme with a land mask function was utilized to determine tidal flats (Uchiyama, 2004).
5. 3 Boundary Conditions and Data Used
To set the boundary conditions between land and sea, the no-slip boundary condition was introduced for the hydrodynamic model, and the Neumann-type boundary condition was used for the convective-dispersive model.
Table 5.1 shows the data used in the simulations.
Fig. 5.2 Coordinate system.
Table 5.1 Data used in the model.
Items/Data Location Period Purpose
DEM Can Gio Bay 2006
Seabed and ground elevations for validation and scenario analyses Observed water
level Vung Tau
2009/08/07
-2009/08/14 Sea boundary for validation
2017/11/01 - 2017/12/30
Sea boundary for validation and scenario analyses
River discharge Phu Xuan, Vam Co, Cua Tieu, Cua Dai, and Thi Vai Rivers
2009/08/07 -2009/08/14
Inflow boundary for validation 2017/11/01 -
2017/12/30
Inflow boundary for validation and scenario analyses
Salinity
Sea Boundary
2017/11/01 - 2017/12/30
Sea boundary for validation and scenario analyses Phu Xuan, Vam Co,
Cua Tieu, Cua Dai, and Thi Vai Rivers
Inflow boundary for validation and scenario analyses Observed time
series of salinity Phu Xuan and Nga Bay
I Stations 2017/12/07 -
2017/12/10 Validation Observed spatial
salinity distribution
Long Tau and Soai Rap
Rivers 2017/12/10 Validation
Firstly, Fig. 5.1 shows a digital elevation model (DEM) of the Can Gio Bay used in the simulations. As is the case with the Chapter 3, the latest digital elevation model (DEM) of the Can Gio Bay with 50 m resolution, which was collected in 2006 by the Division of Science, Technology and International Affairs, South Campus of Thuyloi University through the key national project of DTDL.2011-G/38 (Nguyen, 2014), was used to provide the seabed and ground elevations for simulations of validation and scenario analyses.
The period of salinity simulations was from November 1 to December 31, 2017. Because there was no salinity observation station inside the Can Gio Bay area, thus the observed salinity data recorded from December 7 to December 10, 2017, which were shown in the Chapter 4, were used to validate the salinity
model. The same period was used in the scenario analyses. The simulation period included more than one month before the period for the validation and the scenario analyses, because the model needed the preliminary calculation period to get a stable salinity distribution. In addition, the simulation period from August 7, 2009 to August 14 was also utilized for the validation of the hydrodynamic model as is the case with the Chapter 3.
Next, the hourly observed sea water level for the entire years 2009 and 2017 at the Vung Tau Station was utilized to calculate the harmonic constants at sea boundary for the validation and the scenario analyses. Figure 5.3 shows the observed sea water level at the Vung Tau Station from December 7 to December 11, 2017 as a typical period for the simulations. Those data were recorded by the National Hydro-Meteorological Service of Vietnam and obtained from the Division of Science, Technology and International Affairs, South Campus of Thuyloi University.
The hourly discharges from five rivers, the Phu Xuan, the Vam Co, the Cua Tieu, the Cua Dai, and the Thi Vai Rivers, were used to determine the upstream boundary conditions of the model. As an example, the hourly discharges at five rivers from December 5 to December 11, 2017 are shown in Fig. 5.4. As there were no river discharge stations in those rivers, the river discharge data were archived from the simulations using the observed water levels at those stations by the MIKE 11 Model. As above-mentioned, all of the modules of MIKE software were licensed by South Campus of Thuyloi University, which was certified by DHI. In the model, all rivers and channels in the lower Mekong Delta and the Saigon-Dongnai River system were represented with 4110 branches and reaches, a total length of 24200 km and 39780 cross sections. The model included 22 upstream inflow boundaries, 58 downstream water level boundaries, and rainfall-runoff links to 1682 sub-basins. The domain of upstream model was extended up to the Kratie at the downstream of Cambodian border and the Tonlesap Lake (Ngoc, 2017).
For the salinity simulation, the value of salinity was set to be 35 psu for the open boundary. For upstream boundary, the salinity was 0.5 psu for Phu Xuan
River, 18 psu for the Thi Vai River, 7 psu for the Vam Co River and 20 psu for both Cua Dai and Cua Tieu Rivers. These numbers were chosen by trial-and-error method because there was no salinity station in these areas.
Fig. 5.3 Observed water level at the Vung Tau Station from December 5 to December 11, 2017.
Fig. 5.4 Calculated hourly discharge at the Phu Xuan, the Vam Co, the Cua Tieu, the Cua Dai, and the Thi Vai Rivers from December 5 to December 11, 2017 by
the Mike 11 Model.
−2
−1 0 1
Year of 2017
12/05 12/06 12/07 12/08 12/11
12/10
12/09
Water level (m)
Vung Tau
−10000 0 10000
Year of 2017
12/05 12/06 12/07 12/08 12/11
12/10
12/09
Discharge (m3 /s) Phu Xuan River
−10000 0 10000
Year of 2017
12/05 12/06 12/07 12/08 12/11
12/10
12/09
Discharge (m3 /s) Cua Dai River
−10000 0 10000
Year of 2017
12/05 12/06 12/07 12/08 12/11
12/10
12/09
Discharge (m3 /s) Vam Co River
−10000 0 10000
Year of 2017
12/05 12/06 12/07 12/08 12/11
12/10
12/09
Discharge (m3 /s) Cua Tieu River
−100 0 100
Year of 2017
12/05 12/06 12/07 12/08 12/11
12/10
12/09
Discharge (m3 /s) Thi Vai River
5. 4 Validation of the Model
5. 4. 1 Validation of the Hydrodynamic Model
The parameters set for the model, which were the same values as those in Table 3.2 in the Chapter 3, are shown in Table 5.2. A time step t of 1 s was used for simulations. The area was divided into a total of 1400000 meshes with a grid size of 50 m. The Coriolis parameter f of 2.6 × 10−5 s−1 was utilized based on the geographic latitude = 10o at the Can Gio Bay. The Manning’s coefficient n, the threshold depth dth, and the minimum depth dmin to be 0.02 s/m1/3, 0.2 m, and 0.1 m, respectively were identified through trial-and-error processing. The Nash-Sutcliffe coefficient ENS and the root mean square error ERMSE evaluated the performance of model simulations. The validated results are shown in Fig. 3.10 in the Chapter 3. The good agreements were obtained in a comparison between simulated and observed data.
Table 5.2 Parameter values.
Parameter Explanation Value Unit
t Time 1.0 s
x, y Grid size 50 m
f Coriolis parameter 2.6 × 10-5 1/s
Sm Smagorinsky coefficient 0.2
AG Mesh area 50 × 50 m2
g Acceleration due to gravity 9.81 m/s2
n Manning’s coefficient of roughness 0.02 s/m1/3
dth Threshold depth 0.2 m
dmin Minimum depth 0.1 m
5. 4. 2 Validation of the Convective-Dispersive Model
Figure 5.5 shows the location of the Phu Xuan and the Nga Bay I Stations for the temporal salinity time series in the Period I in the Chapter 4. To validate the model performance for convective-dispersion of salinity, the observed time series of salinity at both stations were utilized: at the Phu Xuan Station on December 7 to 8, 2017 and at the Nga Bay Station on December 8 to 10, 2017. The salinities were recorded approximately 2 m beneath the water surface at two stations. The Phu Xuan Station was located in the upstream (closed-off section) of Can Gio Bay area. The Nga Bay Station was located in the middle of the mangrove forest and played an essential key for assessing changes in salinity regarding the tidal regime and the upstream discharge in this area. Furthermore, the salinity vertical profiles observed along the Pathway I in the Period I shown in Figs. 4.11 and 4.12, and Table 4.4 in the Chapter 4 were also utilized for the validation data by integrating a vertical profile to get a depth-averaged salinity. The salinity vertical profiles at totally 21 points were obtained along the Pathway I in the observation.
However, three points among 21 points were not used for the validation, because three points were located in erratic terrain areas and considered to have little representativeness as the validation data at the points.
Fig. 5.5 Location of the Phu Xuan Station, the Nga Bay I Station and 18 salinity validation points along of the Pathway I in the Can Gio Bay area. The red circles
show the Phu Xuan and the Nga Bay I Stations.
Fig. 5.6 Tidal level at the Vung Tau Station and the observed and the simulated salinities at the Phu Xuan Station from 09:30 December 7 to 08:30 December 8, 2017.
Fig. 5.7 Tidal level at the Vung Tau Station and the observed and the simulated salinities at the Nga Bay I Station from 14:00 December 8 to 12:00 December 10, 2017.
Fig. 5.8 Observed and simulated salinities of 18 points along the Pathway I.
The performance of salinity simulation was evaluated by the root mean square error (ERMSE) as shown in Figs. 5.6 and 5.7. At the Phu Xuan Station, the simulated salinity variation from 09:30 December 7 to 08:30 December 8, 2017 was the same trend and similar with the observed variation as shown in Fig. 5.6, and the ERMSE was small with 1.27 psu. Figure 5.7 shows the observed and simulated salinity variation at the Nga Bay I Station from 14:00 December 8 to
0 5 10
−2
−1 0 1
Salinity (psu)
Phu Xuan Station
Water level (m)
09:00 15:00 21:00 03:00 9:00
Observed salinity Water level at Vung Tau
2017/12/07 2017/12/08
Time
Simulated salinity
RMSE= 1.27 psu E
14 16 18 20 22
−2
−1 0 1
Salinity (psu)
Nga Bay Station
Water level (m)
14:00 01:30 13:00 00:30 12:00
Observed salinity Water level at Vung Tau
2017/12/08 2017/12/10
Time 2017/12/09
Simulated salinity
RMSE= 1.04 psu E
0 5 10 15 20 25 30
0 5 10 15 20 25 30
Observed salinity (psu)
Simulated salinity (psu)
12:00 December 10, 2017. At the Nga Bay I Station, the value of the ERMSE was also small at around 1.04 psu. Although the simulated salinity tended to be higher than observed data and had relatively low values in Figs. 5.6 and 5.7, The simulated salinity variation followed a similar trend to that of the observed salinity data, and the disagreement was negligible. The scatter plot of the observed and the simulated salinities with the isoline are shown in Fig. 5.8. The obtained results demonstrated the acceptable agreement between the observed and the simulated salinities in 18 points along the Pathway I in the Can Gio Bay area. These validation results indicated that the model was used to simulate the salinity at the Can Gio Bay area reasonably well, and the model was a sufficient tool for simulating the convective dispersion of salinity in the Can Gio Bay area.
5. 5 Scenarios Analysis
Scenario analyses were conducted to evaluate the impact of the SLR and the sea dike construction on the salinity regime in the Can Gio Bay area.
The simulation period of scenario analyses was from 00:00 on November 1 to 24:00 on December 31, 2017, which was determined based on the Period I of salinity observation in the Chapter 4 and the validation period in this chapter.
The differences among the scenarios were assessed based on the simulation results on December 10, 2017.
The scenarios were set up based on the baseline scenario, which represented the hydrodynamic regime and the salinity regime in the Period I from December 7 to December 10, 2017 by combining the additional conditions of the SLR and the sea dike construction.
To monitor climate change of Vietnam based on the IPCC assessment report, two levels of SLR have been developed and published as the newest version by the Ministry of Natural Resources and Environment (MONRE, 2016).
This report estimated the SLRs for two scenarios of greenhouse gas emission, the medium emission scenario of Representative Concentration Pathway (RCP) 4.5 and the high emission scenario RCP 8.5. According to the report, and the SLRs tend to increase by 0.22 m for the RCP 4.5 and by 0.25 m for the RCP 8.5 by
2050, and by 0.53 m for the RCP 4.5 and by 0.73 m for the RCP 8.5 by 2100, respectively. Considering the worst case scenarios of the MONRE assessment report, the SLRs of 0.25 m in 2050 and 0.73 m in 2100 were introduced into scenarios to simulate the worst possible impacts. The SLRs were incorporated into the hydrodynamic model with the water level rise in the sea boundary condition.
The sea dike project has been proposed in 2010 and debated because of its remarkably high investment cost and unwanted effects. Subsequently, this project became to be associated with long-term planning for addressing dramatic population growth, excessive use of natural resources and rapid climate change.
Therefore, the highest SLR scenarios for 2050 and 2100 were used to evaluate the effects of the sea dike project.
Following the results of state project DTDL.2011-G/38, “Study on integrated measures for flood control in the downstream area of Dong Nai Basin and vicinity areas” (Nguyen, 2014), which has been used officially by the government for reporting purposes and for establishing construction plans, two sea dike plans were proposed: a sea dike connecting the Go Cong to the Vung Tau (the GCVT sea dike) and a sea dike connecting the Go Cong to the Can Gio (the GCCG sea dike), as shown in Fig. 5.1. To evaluate the impact of the sea dike construction, the GCCG sea dike and the GCVT sea dike were introduced into scenarios. In the sea dike plans, the sea dike gates are constructed on the GCCG sea dike and the GCVT sea dike to promote water exchange between inside and outside of dike system. The sea dike gate with a width of 2500 m was considered in the scenarios as is the case with the Chapter 3.
The gate operation mode was also included in the scenarios. To lower water levels around the HCMC, the maximum water level in the Can Gio Bay must be kept lower than three levels: +0.8, +1.0 and +1.1 m depending on the fluctuation of heavy rainfall scenarios in the HCMC (Nguyen, 2014). The gate operation mode was introduced into the setup scenarios based on controlling water levels at the inside and the outside of the gate. The operating conditions of sea dike gates were shown in Table 3.4 in the Chapter 3. In this chapter, to
investigate the most effective operation of the sea dike gate, a water level of +0.8 m, which corresponded to nearly lowest tidal peak, was considered in the scenarios as the maximum water levels inside the sea dike gate.
Based on those conditions, nine scenarios were set up by combining different options for the SLR and the sea dike construction with the operation of the sea dike gate as shown in Table 5.3. All scenarios were based on baseline scenario in December 2017 and the same set of model parameters. In detail, BL2017 is the baseline scenario in 2017, while CC2050 and CC2100 denote climate change in the years 2050 and 2100 with corresponding SLRs of 0.25 m and 0.73 m, respectively. In the sea dike construction scenarios, GCVT is the sea dike connecting the Go Cong to the Vung Tau, and GCCG is the sea dike connecting the Go Cong to the Can Gio.
Table 5.3 Scenarios set up based on the SLR and the sea dike construction. The SLR was incorporated into the model with the water level rise in the sea
boundary condition.
No. Scenario Sea dike condition
Sea boundary condition
(SLR)
S.1 BL2017 No dike 2017 (SLR 0.00 m)
S.2 BL2017_CC2050 No dike 2050 (SLR 0.25 m)
S.3 BL2017_CC2100 No dike 2100 (SLR 0.73 m)
S.4 GCCG_2017 Go Cong - Can Gio sea dike 2017 (SLR 0.00 m) S.5 GCCG_CC2050 Go Cong - Can Gio sea dike 2050 (SLR 0.25 m) S.6 GCCG_CC2100 Go Cong - Can Gio sea dike 2100 (SLR 0.73 m) S.7 GCVT_2017 Go Cong - Vung Tau sea dike 2017 (SLR 0.00 m) S.8 GCVT_CC2050 Go Cong - Vung Tau sea dike 2050 (SLR 0.25 m) S.9 GCVT_CC2100 Go Cong - Vung Tau sea dike 2100 (SLR 0.73 m)
5. 6 Results and Discussions
5. 6. 1 Spatial-Temporal Salinity Distribution in Can Gio Mangrove Forest The Can Gio mangroves forest is located at the ending of the Saigon-Dongnai River and the confluence between the upstream of the Saigon-Dongnai River network and the East Sea. This area is a flat low-lying with an average elevation of land from 1.4 m to 2.0 m, and is strongly affected by not only tidal variation but also river inflow discharge. Therefore, the salinity had significant fluctuation in the range of 0 psu – 32 psu. The simulated results of salinity distribution in the baseline (S.1) is shown in Figs 5.9 and 5.10. The location of the Phu Xuan Station, the Nga Bay I Station and 18 referenced points in the Can Gio Bay area was shown in Fig. 5.5. Figures 5.9 and 5.10 indicated that the highest salinity was 32 psu found around the exterior of river mouths, and a dramatic decrease in salinity occurred in the interior of rivers with a fluctuation of 8 psu – 24 psu. But the salinity was low in the upstream around 2 psu – 6 psu at the Phu Xuan Station.
However, the salinity change was uneven among the simulated points. Higher values of salinity distributed at the points 8 and 9 near the Nga Bay I Station with an average salinity of about 22 psu and lower salinities of around 10 psu were found at the points 16 and 18 near the Vam Co intersection. The same trend was also found: a higher value of around 26 psu at the point 11 in the Ganh Rai Estuary and a lower value of about 20 psu at the point 14 in the Dong Tranh Estuary. On the other hand, the oscillation amplitude of salinity was also very different along with locations. The salinity oscillated largely in a range of 1 psu - 6 psu at the point 2 near the Phu Xuan Station, 12 psu – 17 psu at the point 5 in the Long Tau River, 9 psu – 13 psu at the points 16 and 17 near the Vam Co intersection, 18 psu – 21 psu at the point 8 near the Nga Bay Station, 20 psu – 24 psu at the point 9 due to the strong influence of hydrodynamic regime. On the contrary, because of fading impact from the upstream inflow discharge, the region at the points 11 and 14 from the Ganh Rai Estuary to the Dong Tranh Estuary had smaller amplitude of salinity. The magnitude of salinity changed around 18 psu – 19 psu at the points 14 and 25 psu – 26 psu at the point 11.