Analyzing Hydrodynamic Changes in Dubai Creek, UAE: A Pre- and Post-Extension Study

Abstract

This paper presents a comparative study that examines the effects of the Dubai Creek extension on its hydrodynamics and water flushing dynamics. Dubai Creek (Khor Dubai) is a 24 km long artificial seawater stream located in the emirate of Dubai. The creek has experienced the impact of the rapid urbanization of Dubai and a major 13 km extension project, which connected the creek to the Arabian Gulf from the other side. In this paper, two-dimensional hydrodynamic and flushing models were created using Delft3D Flexible Mesh (2021.03) to investigate the water circulation and water quality of the creek before and after the extension. The hydrodynamic models were calibrated and validated to accurately simulate water levels and currents with correlation values close to 1 and very small RMSE and bias. Flushing models were created to simulate water renewal along the creek. The results of the flushing models showed a significant improvement in the flushing characteristics of pollutants in terms of the residence times of the extended creek (Existing Creek) model compared to the old one (Old Creek). This improvement emphasized the positive impact of the creek extension project on the local aquatic ecosystem and its overall water quality.

1. Introduction

Many water bodies worldwide experience pollution due to excessive pollutant discharge caused by different human activities. In the United Arab Emirates (UAE), rapid population growth and large-scale urban development, particularly along the Arabian Gulf coast, have placed significant environmental pressure on coastal ecosystems. In Dubai, the economy has undergone a dramatic shift towards tourism, particularly after the 1997 oil price decline, driving substantial coastal development [1]. This shift, while fostering economic growth, has significantly impacted coastal water quality.

One example of this is Dubai Creek, where recent large-scale infrastructure projects have altered water circulation patterns and introduced new water quality challenges. Dubai Creek, originally a 14 km long artificial seawater channel, has been extended as part of a 2016 project, connecting the creek to the Arabian Gulf from a second inlet and increasing its length to approximately 24 km, with an average width of 300 m [2]. This extension, in addition to the rapid urbanization along its banks, has increased pollution risks, highlighting the need for proactive environmental planning and comprehensive water quality assessments to ensure long-term sustainability. Figure 1 illustrates the original path of Dubai Creek, along with the new extension, showcasing the scale of the modifications.

1.1. Hydrodynamic and Water Renewal Modeling

A model is a representation of reality, which is simulated as accurately as possible based on the model objectives and the availability of data. With the continuous development of numerical modeling software, the use of coastal models became preferable, as it provides very accurate results at a low cost and in a short time [3,4]. Hydrodynamic modeling is an efficient way of representing the dynamics of a water body. It is a computational method used to simulate the behavior of coastal waters using mathematical equations and numerical techniques. These models consider principles such as conservation of mass, momentum, and energy to predict the flow, circulation, and interaction of the water body with its surroundings.

Hydrodynamic modeling has a wide range of applications, from environmental studies to engineering designs. Additionally, hydrodynamic models can be coupled with other models to investigate critical issues like flood risk management, sediment transport, and the dispersion of pollutants in aquatic environments [5]. For example, coupling hydrodynamic models with water quality models enables the prediction of flushing potential in water systems. Flushing refers to the exchange and renewal of water between a water body and the adjacent sea, which serves as an indicator of water quality [3,6,7]. Proper flushing is essential for maintaining good water circulation, as poorly flushed systems tend to accumulate pollutants, leading to stagnation and odorous conditions [8,9].

Before the new extension of the creek, there were concerns about water quality due to poor water exchange with the open sea, which hindered the flushing out of trapped pollutants. Therefore, this study assesses the flushing characteristics of Dubai Creek by modeling a numerical tracer and measuring its residence time. Residence time is a fundamental concept that quantifies the average amount of time a water mass spends within a defined boundary [10]. Residence time can be used for understanding the movement and behavior of substances within water bodies such as rivers, lakes, estuaries, or even coastal areas. A longer residence time indicates that water stays within the water body for a longer period, which can have significant implications for ecological and chemical processes.

Residence time is often used interchangeably with flushing time. A short residence time indicates rapid flushing in the water body. In flushing studies, the residence time is assessed by modeling a conservative tracer in the area of interest. The currents induced by the variation in water levels in the water system cause the tracer to disperse. Hence, residence time is measured as the time it takes the conservative tracer introduced in the water system to fall below a certain criterion [11]. Two common metrics for residence time are the “half-life” residence time (T₅₀) and the “E-folding” residence time (T₃₇). In this paper, the selected criterion is “E-folding” residence time (T₃₇), which is defined as the time taken for the mean concentration of a conservative tracer introduced to a water body to reach 37% of its initial value [12,13,14]. The concentration inside the creek is computed numerically using Equation (1), which is derived from exponential decay processes:

$$c\left(t\right)=c_0.e^{{\displaystyle \frac{-t}{T_f}}},$$

(1)

where $C_0$ is the initial concentration, $c\left(t\right)$ is the concentration at a specific time $t$, and $T_f$ is the residence time.

1.2. Water Quality of Dubai Creek

Several studies have been conducted to assess the water quality of Dubai Creek and its surrounding area. An environmental capacity modeling study conducted in 2002 in the Dubai Coastal Region, United Arab Emirates suggests that the urban development around Dubai Creek has caused a change in the water quality and sediment characteristics in the creek [15]. The outcomes of the study suggest that the nutrients in the water have shown lower levels downstream as compared to upstream. These findings indicate the need for continuous monitoring and improvement of the creek’s water quality, which could be supported through numerical water quality models.

In 2015, a one-dimensional HEC-RAS hydrodynamic model coupled with a water quality model was developed for Dubai Creek [16]. The model was used to run different simulations to evaluate different water quality parameters along the creek, including algae, dissolved oxygen, nitrate, and orthophosphate. The study concluded that human activities and urbanization affected the water quality and the aquatic ecosystem of the creek.

Similarly, a GIS-based spatiotemporal study investigated the variability of water quality data collected from stations in Dubai Creek in 2012 and 2013 [17]. The GIS-based regression analysis model was created to study the vulnerability of Dubai Creek to eutrophication, specifically the relationship between chlorophyll-a and nutrients in the water. The study suggests that the occurrence of eutrophication was more probable in the top half than in the bottom half of the creek. Hence, the degradation of the creek’s water quality is caused as a byproduct of the rapid development in the Dubai Creek area.

Another contributing factor to the pollution in Dubai Creek is the water discharged from the Al-Aweer sewage treatment plant, which is nitrogen-rich [18]. Additionally, nearby green areas are fertilized with nitrogen-rich fertilizers, which also pollutes the creek, as leaching occurs through stormwater runoff [19]. Ali et al. (2017) conducted a study on water quality assessment using a satellite-based monitoring system by employing two DubaiSat-1 images of the creek (acquired in 2010 and 2011), which were used to create spectral-based models of chlorophyll-a [20]. The study suggests that the effluent of the treatment plant is rich in nutrients such as nitrogen that influence algae growth and eutrophication. The presence of large amounts of algae in the water reduces the amount of dissolved oxygen in the water [21]. Therefore, these studies indicate that the water circulation in the creek is poor and that a major pollution factor could be the poor flushing of the creek.

Efforts to improve the hydrodynamics of Dubai Creek were undertaken with its extension and connection to the Arabian Gulf in 2016. Karanam et al. (2018) developed a model to simulate the hydrodynamics of the extended Dubai Creek using MIKE 3FM [22]. Their study indicated poor flushing of the creek before the extension, as evidenced by a numerical model showing a high residence time of approximately 45 days. However, the extension led to significant improvements, with the model demonstrating up to a 33% enhancement in flushing, reducing the residence time to around 33 days [22]. The study, however, did not detail the residence time at various locations along the creek, instead presenting an overall flushing improvement. Additionally, the study used T₅₀, or the “half-life” residence time, rather than the “E-folding” residence time (T₃₇).

In this study, hydrodynamic and flushing models were constructed using Delft3D Flexible Mesh (Delft3D FM) to compare results and further explore the flushing characteristics of the creek at different locations before and after the extension. Unlike Karanam’s study, this paper utilizes the “E-folding” residence time (T₃₇) instead of the “half-life” residence time (T₅₀). The former is favored in hydrodynamic models and scientific research [23,24,25], especially in systems with exponential decay characteristics. Delft3D FM, developed by Deltares, was selected for its advanced capabilities and post-processing features in simulating currents, water levels, and other parameters [5].

2. Materials and Methods

2.1. Hydrodynamic Models

Two hydrodynamic models were created using Delft3D FM. The first model is for Dubai Creek before the extension (Old Creek) and the second model is for after the extension (Existing Creek). The required data for model development include bathymetry, current speed and direction, and water-level data. The bathymetry data provided a detailed underwater topography, essential for simulating flow patterns. Current speed and direction were derived from field measurements, ensuring temporal and spatial resolution of water movements. Water-level data from tidal gauges were used to set boundary conditions and validate model outputs. This comprehensive dataset facilitated accurate model creation, calibration, and validation, ensuring accurate simulations of the creek’s hydrodynamic behavior.

Several sources were used to obtain all the necessary bathymetric data for the model. The bathymetry of the offshore area in the Arabian Gulf was obtained from the General Bathymetric Chart of the Oceans, GEBCO, http://www.gebco.net (accessed on 1 February 2023). Additionally, the local bathymetry of near-shore areas in Dubai was obtained through bathymetric surveys and other sources for bathymetry, such as Navionics and Admiralty Charts. Elevation and water-depth data were horizontally referenced to the WGS84-UTM40N coordinate system, and vertically reduced to the Mean Sea Level (MSL). Tidal planes were obtained from an Admiralty Tidal Station, located at Al Maktoum Bridge, which is 5 km into Dubai Creek. The tidal planes extracted from the Admiralty Tide Tables are presented in

Table 1. Tidal Levels at Al Maktoum Bridge Tidal Station [26].

Tide Level	Level (mACD)	Level (mDMD)
Highest Astronomical Tide (HAT)	+2.2	+2.2
Mean Higher High Water (MHHW)	+1.7	+1.7
Mean Lower High Water (MLHW)	+1.4	+1.4
Mean Sea Level (MSL)	+1	+1.1
Mean Higher Low Water (MHLW)	+0.8	+0.9
Mean Lower Low Water (MLLW)	+0.4	+0.5
Lowest Astronomical Tide (LAT)	0	0

[26].

The water level is affected by the astronomical tides throughout the year. The hydrodynamic model in this study was forced by boundary conditions in the form of water-level data. The available time series of water-level data at the Jebel Ali Port and Mina Al Jazeera Port were used as the Southern and Northern boundaries of the model, respectively.

The hydrodynamic model was validated by comparing its output to real-world measurements from two key sources. First, water-level data from a tide gauge at Al Mamzar were compared with the model’s simulated water levels, ensuring that the model accurately captured tidal fluctuations. The correlation between the modeled and observed water levels was used to determine the model’s accuracy. Second, current speed and direction from the model were validated against data collected by an Acoustic Doppler Current Profiler (ADCP) positioned near the study area. Figure 2 shows the location where the water level and current data were collected.

2.2. Model Setup

The model grids are unstructured grids that cover an area of around 2500 m². They span around 110 km from the northern boundary to the southern boundary of the grid, with around 23 km from the coastline vertical to the offshore boundary. The size of the grid cells decreases and resolution increases as the grids reach closer to the study area (Dubai Creek). The grids start with 500 m grid cells and become as small as 25 m inside the creek. The overall quality of the created grids was determined by examining the orthogonality and smoothness of the grids based on specific criteria. Orthogonality refers to the perpendicularity of adjacent cells, while smoothness is determined by the ratio of areas of two adjacent cells. Ideally, orthogonality and smoothness ensure that the adjacent cells in a mesh correspond to the requirements listed below [5,27]:

• The corners of two adjacent cells are positioned on a common circle;
• The center of each cell falls within its boundaries;
• Areas of the cells are equal to each other.

The bathymetric data were imported into the model and interpolated by the triangulation method to cover the whole extent of the grid. Water-level boundary conditions were then placed at three boundaries of the model grid. Mina Al Jazeera Port and Jebel Ali Port were used for the northern and southern edges of the grid, respectively. In addition, a third boundary condition was placed at the offshore edge of the grid. The offshore boundary was divided into seven support points, and each had a water-level time series. The three tidal boundary conditions cover a period from 1 December 2013 until 31 January 2014. Figure 3 illustrates the position of the three modeled boundary conditions and the bathymetry of the Existing Creek hydrodynamic model.

2.3. Model Calibration and Validation

At this stage, the process of the hydrodynamic model creation was complete, and the created models needed to be calibrated and validated. Model calibration and validation are necessary to ensure confidence in model predictions. Calibration is an iterative process that involves refining model parameters and boundary conditions by comparing the model outputs with observed data and adjusting factors such as the drag coefficient to achieve better alignment.

Model validation is an integral part of the calibration process. It is defined generically as the process of determining the degree to which a model accurately represents the real world [28]. To quantify model accuracy, some statistical parameters were used to compare observed and modeled water levels and the current velocity and direction.

The hydrodynamic model validation was carried out by using the observed water-level data at the Al Mamzar tide gauge and comparing them with the model data. Similarly, the model’s current speed and direction data at the location of the deployed ADCP were compared against the actual ADCP data to validate the model. Root Mean Squared Error (RMSE), bias, and correlation coefficient (r) are statistical parameters that were used to evaluate the model simulations for the current speed/direction and water level to check the validity of the models.

2.4. Water Renewal/Flushing Models

A specific module in the Delft3D FM is the D-Water Quality module (Delwaq), which is designed for water quality simulations [29]. This specific module was used in this study to assess the water renewal/flushing in the Dubai Creek models through residence time. The output of the hydrodynamic models (currents and water levels) was used as input to the flushing models. In this study, the flushing simulations were carried out considering the assumption that the quality of the water will degrade over time unless there is sufficient exchange of water with an external source. The water outside Dubai Creek was assumed to be of good quality and such that water exchange is beneficial to the water system. The effect of the discharge of any pollutant in the creek was not taken into account. Therefore, the effect of effluent being discharged into the creek or surrounding areas should be studied separately. The flushing setups for the Old and Existing Creek models are shown in Figure 4 and Figure 5, respectively, and they ran for a period of approximately 7 weeks. The following steps were followed in setting up the model:

• Conservative tracer was applied to the study area;
• No decay or transformation processes were applied;
• Initial condition was set as a coverage value of 1 g/m³ (Red) in the creek and 0 g/m³ outside (Blue);
• Boundary condition was set as a constant value of 0 g/m³.

Figure 4. Flushing setup for the Old Creek model.

Figure 4. Flushing setup for the Old Creek model.

Figure 5. Flushing setup for the Existing Creek model.

Figure 5. Flushing setup for the Existing Creek model.

3. Results and Discussion

3.1. Hydrodynamic Model Calibration and Validation

Model calibration and validation were performed for the two created hydrodynamic models; however, the presented results are for a single model, as the calibration process and results were nearly identical. The water level and current data were extracted from the models by placing observation points at the locations of the tide gauge and ADCP. Therefore, two observation points were placed in the models at the location of the Al Mamzar tide gauge and the deployed ADCP.

Figure 6 shows a graph of the water-level comparison between the measured and modeled data at the Al Mamzar tide gauge in the Old Creek model. The graph shows a two-week representation of the simulated period, between 13 and 27 December 2013. The graph presented shows a strong correlation between the measured and modeled data at the Al Mamzar tide gauge for the two models. The minimum and maximum measured water levels at the tide gauge are 0.3 and 2.0 m, respectively. From the modeled data, the minimum and maximum modeled water levels are 0.4 and 2.0 m, respectively. This indicates a strong agreement between the observed data and the model predictions.

Additionally, the correlation between the measured and modeled water-level data is shown in a graph in Figure 7. The data show a strong correlation of 0.92 (92%) along the trendline. Additionally, Root Mean Square Error (RMSE) and bias were calculated to better understand how close the measured and model data were to each other. Lower values of RMSE and bias indicate better agreement between the data sets. The calculated RMSE and bias are 0.15 and 0.09 m, respectively, which indicate that the measured and model data are at a high level of agreement. Hence, the created models are considered to be calibrated and validated and are appropriate to be used as models for Dubai creek.

3.2. Current Speed and Direction

The current speed and direction measured at the ADCP were used to calibrate the hydrodynamic model. Figure 8 and Figure 9 present graphs showing the measured x and y components of the current speed, respectively, against the Delft3D model data for the period from 9 December 2013 to 3 January 2014. The graphs suggest a strong correlation between the measured and the model current velocities, which are 0.85 and 0.83 for the x and y components, respectively. Additionally, Root Mean Square Error (RMSE) and bias were calculated to better understand how close the measured and model data were to each other. Lower values of RMSE and bias indicate better agreement between the data sets. The calculated RMSE and bias for the x component of the current speed are 0.09 and 0.03 m/s, respectively. Similarly, the calculated RMSE and bias for the y component of the current speed are 0.08 and 0.02 m/s, respectively. The calculated RMSE and bias indicate that the measured and model data are in a high level of agreement. In other words, the strong alignment between the data gathered through observation and the predictions generated by the model implies a high degree of accuracy and reliability in the model’s simulation of real-world conditions.

3.3. Old Creek Flushing Model

The Old Creek has one opening to the open sea and extends around 14 km downstream at Ras Al Khor Wildlife Sanctuary. Hence, the insufficient water circulation in the Old Creek suggests that the flushing might seem to be poor, as suggested in previous research [20]. The flushing model of the Old Creek was run for a period of 53 days, which translates to around 7 weeks. Figure 10 presents the concentration of the introduced conservative tracer along the Old Creek flushing model at a time step of 1 week. The model shows the drop in the tracer concentration along the creek as time progresses. The highest drop in concentration was observed towards the opening of the creek. On the other end of the creek, the drop in the tracer concentration is lower as it is further away from the open sea. At the end of the simulation period (7 weeks), the concentration of the tracer in some areas of the creek reached as high as 65% of the initial concentration.

Observation points (Points A–C) were placed in the Old Creek flushing model, as shown in Figure 11, to examine the residence times at different locations in the creek. Graphs showing the concentration of the conservative tracer over time at the three observation points are shown in Figure 12. The graph of point “A” shows the highest drop in the tracer concentration and comes after points “B” and “C”. The concentration of the tracer in point “A” reaches the “E-folding” residence time criterion of 37% of the initial concentration in approximately 30 days. However, the concentrations of the tracer in points “B” and “C” do not fall below the 37% criteria during the whole simulation period, which is 53 days. At point “C”, which is the furthest point from the open sea, the concentration of the tracer reaches a minimum of 65% which is considered relatively high. In order to flush the whole creek and reach the “E-folding” residence time criteria, a very long simulation time would be necessary. This indicates poor flushing of the Old Creek model and poor water quality in general. The poor flushing in Old Creek is supported by research conducted in the past on the water quality in Dubai Creek [30,31].

3.4. Extended Creek Flushing Model (Existing Creek)

Unlike the Old Creek, which had only one opening to the sea, the Extended Creek is open to the sea from both inlets. The Existing Creek extends for around 22 km (from one end to the other), which means that more water volume needs to be flushed out to renew the water in the system. The flushing model of the Existing Creek was run for 53 days, which translates to around 7 weeks. Figure 13 presents the concentration of the introduced conservative tracer along the Existing Creek flushing model at a time step of 1 week. The model shows the drop in the tracer concentration along the creek as time progresses. The highest drop in concentration is observed near the two openings of the creek. Towards the center of the creek, the drop in the tracer concentration is lower as it is further away from the open sea. At the end of the simulation period (53 days), the concentration of the tracer at the center of the creek reaches as low as 10% of the initial concentration.

Observation points (Points A–E) were placed in the Existing Creek flushing model, as shown in Figure 14, to examine the residence times at different locations in the creek. Graphs showing the concentration of the conservative tracer over time at the five observation points are shown in Figure 15. Observation points “A”, “B”, and “C” in the Existing Creek model were placed at the same location as they were in the Old Creek model. The concentration of the tracer at point “A” reaches the “E-folding” residence time criterion of 37% of the initial concentration in approximately 30 days. Unlike the Old Creek model, points “B” and “C” reach 37% of the initial concentration of the tracer after around 36 and 31 days, respectively. Generally, the Existing Creek model shows lower residence times than the Old Creek model in the common observation points.

Furthermore, additional observation points were introduced in the Existing Creek model to further analyze the overall flushing of the Creek. Point “D” was placed in the center of the creek as this area seems to be the most critical in terms of flushing. Additionally, point “E” was placed in the extended channel of the creek. The highest residence time was obtained at point “D”, which was around 40 days. On the other hand, point “E” showed the shortest residence time, of around 14 days.

A summary of the residence times at the observation points in the Old and Existing Creek flushing models is shown in

Table 2. Residence times at observation points in the Old and Existing Creek models.

Observation Point	Old Creek Model	Existing Creek Model
A	30 days	30 days
B	>53 days	36 days
C	>53 days	31 days
D	–	40 days
E	–	14 days

. The residence times recorded at observation point “A” are very close to each other in the Old and Existing Creek models. Hence, it can be seen that the flushing of the water at point “A” was not significantly affected by the creek extension. However, there was a significant reduction in the residence time recorded at points “B” and “C”. This reduction in residence time indicated a significant improvement in the flushing of water in the Existing Creek model. Additionally, the lower residence time suggests improved water circulation in the Creek. The newly introduced points “D” and “E” in the Existing Creek model also showed enhancement in the overall water flushing of the creek. Hence, the extension of Dubai Creek was vital in improving the water circulation and flushing, which led to an improvement in the overall quality of the water in Dubai Creek.

A noticeable increase in residence time could be seen in the creek with distance from the creek mouth. In the Old Creek model, the residence time reaches its maximum toward the end of the creek in the Ras Al Khor area (Point C). Similarly, in the Existing Creek model, the maximum residence time is reached at that location (Point D). This observed trend is due to the hydrodynamics within the creek and the geometric configuration of the waterway. As water enters the creek from the Arabian Gulf, it encounters a narrowing of the channel width and a reduction in current velocity due to the surrounding topography. This leads to a decrease in the overall water movement and an increase in residence time. Additionally, hydraulic features such as channel bends and localized shallows, and constructed structures in the creek, such as bridges, further contribute to water retention. These features create flow obstructions, turbulence, and areas of recirculation, which slow down the flow and prolong the water’s residence within the creek. In general, a gradient of residence times along Dubai Creek is established, with higher values observed farther inland.

4. Conclusions

This paper presents a comparative study that examines the effects of the Dubai Creek’s extension on its hydrodynamics and water flushing characteristics. Dubai Creek has witnessed substantial changes due to urbanization and a significant extension project linking it to the Arabian Gulf. Two-dimensional hydrodynamic and flushing models were developed using Delft3D Flexible Mesh, employing bathymetric data, field-measured currents, and tidal gauge observations as inputs. The models were calibrated through iterative adjustments to parameters such as roughness coefficients and boundary conditions to align outputs with field measurements. Validation was performed by comparing the model predictions against the observed data to ensure an accurate simulation of water levels and currents. The results of this study have shown significant improvements in the flushing capabilities of the Existing Creek model compared to the Old Creek model. Notably, there is a substantial reduction in the residence time of the conservative tracer in the Existing Creek. At different locations in the Old Creek, the residence time exceeded the whole simulation period, which was 53 days. However, in the Existing Creek model, the residence time was reduced to 36 and 31 days in two different locations. The obtained results indicate the positive influence of the creek extension project on the local hydrodynamics and overall water quality. The created models can serve as versatile tools for various applications. They can guide coastal engineering projects, aiding in the design and optimization of structures along the creek. Similarly, as urbanization continues in Dubai Creek, these models can support local authorities in making informed decisions regarding urban planning and the sustainable management and development of Dubai Creek. Beyond Dubai, these models can be applied to other coastal cities facing similar challenges, such as those in Asia, the Middle East, and other rapidly urbanizing regions. However, it is essential to account for local environmental conditions and hydrodynamic variations. By addressing these considerations, coastal cities can use hydrodynamic models to manage urban expansion effectively, enhance water quality, and ensure ecological and functional sustainability in diverse coastal environments.