Jebel Ali Harbor in Dubai is one of the largest man-made ports in the world. Due to its bottle-like nature, it is anticipated that little mixing or wave action occurs within the harbor; however, no study has been conducted to investigate the harbor's hydrodynamic regime. In this paper, the circulation pattern in Jebel Ali Harbor is presented. The vertical profile of temperature and salinity shows that the harbor water is fairly well-mixed, justifying the use of a depth-averaged, 2-D model. Modeling the hydrodynamic pattern of Jebel Ali Harbor was carried out using MIKE21 model. Although tide is the main driving force of circulation in the harbor, wind and Coriolis forces were taken into account in the simulation. Simulation results showed that both dead and eddy zones exis. As the harbor receives treated wastewater from several industries located in the area, flushing out of a conservative tracer by the advection-dispersion process was also simulated using MIKE21. Results revealed that the flushing time of a conservative tracer from the harbor varies significantly with the discharge location. Simulated results showed that understanding the hydrodynamic regime of Jebel Ali Harbor is useful for selecting the location and time for discharging tracers into the harbor.

Introduction

Jebel Ali Free Zone (JAFZ) in Dubai is a growing industrial park with an area of about 100 km2 and over 4500 trading, manufacturing, servicing, and distributing companies. The northern part of the Free Zone is under rapid development by existing industries, while the southern part is relatively undeveloped, and will be the location for future industry. Most of the industry within the existing development is light, but the Free Zone is growing rapidly, with a variety of industries under construction, which may lead to additional discharge of “treated” industrial wastewater into the harbor.

The harbor (Figure 1) at JAFZ is the largest man-made port in the world with an approach channel that starts 15 km offshore. The approach channel has a depth of 14–15 m and a width of 280 m reducing to 235 m. It bends after 10 km and becomes the entrance channel. It widens to 300 m at the bend and to 340 m at the entrance channel. There are two basins within the port. The outer 14-m deep basin is 2.3 km long and 600 m wide, while inner basin is 3.7 km long and 425 m wide, with a depth of 11.5 m (Dubai Port Authority (DPA), 2005).

Figure 1.

Location map of Jebel Ali Harbor (Source: Research Triangle Institute (RTI), 2005 after modification).

Figure 1.

Location map of Jebel Ali Harbor (Source: Research Triangle Institute (RTI), 2005 after modification).

Assessment of the harbor's water quality is necessary to protect the marine life and to minimize the impact of industrial activities on the surrounding ecosystem. Currently, the harbor water receives discharge of several treated industrial effluents and treated sewage effluent. All water discharged into the harbor must adhere to the discharged water quality criteria set forth in the Environmental Requirements established by Dubai Ports, Customs and Free Zone Corporation (PCFC, 2003). However, a further expansion of industrial activities may increase discharge flow rates into the harbor. This, in addition to port activities, may increase pollutant loading beyond the capacity of the harbor, despite the fact that all discharges meet the established regulatory limits.

Future harbor water quality may be further complicated due to the bottle-like nature of the harbor, where not much mixing or wave action is expected. Thus, understanding the hydrodynamic nature of the water in the harbor is a first step towards successful assessment of its quality. No previous study has been conducted to address the issue of water circulation in the harbor. Hence, the objective of this study was to simulate the hydrodynamic circulation in the harbor. Another objective was to assess the transport of a conservative tracer under the influence of both advection and dispersion.

Model background

Surface water numerical models are particularly useful for evaluating the effects of engineering works on tidal motion and the quality of tidal water. Numerical modeling is now the most widely used tool for predicting the effects of changes in the modeled processes due to natural and artificial imposed conditions. Moreover, numerical modeling of natural processes assists engineers, planners, environmentalists and scientists to formulate their decisions/judgments or recommendations (Hamza et al., 2004).

Hydrodynamic model

A hydrodynamic (HD) model simulates unsteady water level and flow in response to a variety of functions in lakes, estuaries, bays and coastal areas. The water levels and flows are resolved on a grid system covering the area of interest when provided with the bathymetry, bed resistance coefficients, wind field and hydrographic initial and boundary conditions.

In a 2-D HD model, conservation of mass and momentum are integrated over the vertical dimension to describe the flow and water level variations at all grid points (Danish Hydraulic Institute (DHI), 2003a). The continuity equation is given as:

formula
where: ζ is the water level (m); p and q are flux densities in x and y directions (m3 s− 1 m− 1); t is time (s); x and y are space coordinates (m). The x-and y-momentum are given by Eq. (2) and Eq. (3), respectively:
formula
formula
where: h is water depth (m); C is Chezy resistance (m1/2 s− 1); f is the wind friction factor (dimensionless), V, Vx, and Vy are wind speed and components in x and y directions (m s− 1), respectively; Ω is Coriolis parameter (s− 1); pa is atmospheric pressure (kg m− 1 s−2); ρw is the density of water (kg m−3); and τxx, τxy and τyy are components of effective shear stress (N m−2). The above equations are based on the following shallow water assumptions: (1) the characteristic horizontal length-scale is much larger than the characteristic vertical length-scale and (2) the characteristic vertical velocity is small in comparison with the characteristic horizontal velocity (Jin, 1993).

Advection-dispersion model

The advection-dispersion (AD) model simulates the spreading of a substance in an aquatic environment under the influence of fluid transport and associated natural dispersion processes. The substance may be treated conservatively or with decay. Discharge quantities and substance concentrations at source and sink points can be included together with a decay rate. The governing equations for a 2-D AD model are given as (DHI, 2003b):

formula
where: c is substance concentration (mg l− 1); u and v are horizontal velocity components in x and y directions (m s− 1), respectively; Dx and Dy are dispersion coefficients in x and y directions (m2 s− 1), respectively; F is the linear decay coefficient (s− 1); Qs is the source/sink discharge per unit horizontal area (m3 s− 1 m-2); cs is substance concentration in the source/sink discharge (mg l− 1).

Data collection

Bathymetry, tidal level, salinity, air temperature, water temperature and wind data were collected and analyzed. The data were used in the model when required.

Bathymetry

Bathymetry plays a key role in the hydraulic behavior of any body of water. The bathymetric data of Jebel Ali Harbor were obtained from Jan de Nul Dredging Ltd (2004). Horizontal coordinates were given in easting and northing projected in UTM Zone-40 and the bed levels were given in local chart datum (CD). The bathymetry was generated by linear interpolation with inverse distance weight and is presented in Figure 2.

Figure 2.

Bathymetry of Jebel Ali Harbor showing locations of data collection (St1-St9) and discharge points (DP1-DP5).

Figure 2.

Bathymetry of Jebel Ali Harbor showing locations of data collection (St1-St9) and discharge points (DP1-DP5).

Tidal level and flux

A continuous record of tidal level at the control tower (see location point in Figure 2) was obtained from Dubai Port Authority. The tidal level at two other locations, one at the dead-end of the outer basin (St4 of Figure 2) and another at the dead-end of the inner basin (St6 of Figure 2) were collected by an automatic pressure gauge for 9 days from 20–28 December, 2004. Tidal levels at the inner and outer basins are almost the same, but there is a little difference between the tidal levels in the basins as compared with that measured at the control tower (Figure 3).

Figure 3.

Observed tidal levels at two locations in the harbor.

Figure 3.

Observed tidal levels at two locations in the harbor.

The tidal levels in the outer and inner basins were analyzed in terms of water level slope as an indicator of energy loss due to bed friction. Slopes in the outer basin were calculated by dividing water level differences between the control tower and the outer basin by the distance between two locations, which is 2.5 km. Similarly, the slopes in the inner basin were calculated by taking the distance between the control tower and the inner basin as 3.5 km. It was found that slopes in the outer and the inner basins follow the same trend. The maximum slope is 6 cm km− 1 with an average value of 2 cm km− 1. Also, slopes vary in spring and neap tides. On the average, the slope is 3 cm km− 1 during the spring tide whereas 1 cm km− 1 during the neap tide.

To calibrate the HD model, the tidal flux (discharge) through the main channel at cross section X-1 (Figure 2) was measured on the 19th and 21st of December 2004. An Acoustic Doppler Current Profiler (ADCP) device mounted on a boat was used to measure three dimensional flow velocities at different depths and the data were processed off-line to calculate total flow through the section. Total flow was calculated by the area-velocity method neglecting the vertical component of the velocity.

Temperature and Salinity Profile

Temperature and salinity data were collected at different depths from several locations (see Figure 2) inside the harbor in March 2004 using multiple sensor underwater profilers. The data are presented in Figure 4. The same profile data were also collected at a point in the Arabian Gulf, 5 km from the harbor entrance, to compare with that collected inside the harbor. Figure 4 shows that harbor water can be assumed to be well-mixed in the vertical direction.

Figure 4.

Temperature (left) and salinity (right) profiles at different locations in the harbor.

Figure 4.

Temperature (left) and salinity (right) profiles at different locations in the harbor.

Meteorological data

Wind and temperature data recorded near the control tower from January 2004 to February 2005 were collected and analyzed for this study. Figure 5a shows that the average wind speed over the year is about 5 m s− 1, with a maximum speed of 10 m s− 1. However, during December and January, the wind speed reaches 18 m s− 1 These winds are referred to in the region as shamal wind. Wind direction, on the other hand, is not fixed during a season. Rather, it changes frequently (Figure 5b). The wind rose (not shown here) shows that wind blows from all directions but the dominant direction is west followed by north-west. During almost one-third of the year, the wind blows from the west with a speed of 2–4 m s− 1 while during only 1% of the year does the wind speed exceed 12 m s− 1. During 12% of the year, the wind is calm with a speed of < 2 m s− 1. For a short period, 5% of the year, the wind blows from the north and during another 5% of the year it blows from the south-west.

Figure 5.

Meteorological conditions at Jebel Ali Harbor, (a) wind speed, (b) wind direction and (c) air and water temperatures.

Figure 5.

Meteorological conditions at Jebel Ali Harbor, (a) wind speed, (b) wind direction and (c) air and water temperatures.

Water and air temperatures near the control tower are presented in Figure 5c. Air temperature varies significantly (≈5°C) between day and night. However, water temperature does not vary from day to night but undergoes gradual seasonal changes. The water temperature is very close to the average of the air temperature, except in the fall and winter seasons, during which the water is warmer than the air by about 2°C.

Model setup and calibration

As the harbor is quite well-mixed in the vertical direction, a depth-integrated 2-D model (MIKE21) was used to set up the hydrodynamic and the advection-dispersion model of Jebel Ali Harbor. The HD model was calibrated against available tidal level data measured in December 2004. The calibrated HD model was used for simulating hydrodynamic processes, which are also the basis of the AD model. The setup of the HD and AD models is discussed below.

The area of Jebel Ali Harbor is about 5 × 5 km2. The entrance to the harbor was selected as the open boundary for the model. The model was constructed with a rectangular grid system of 60 × 60 m2. The dimensions of the grid were selected as a compromise between resolution and computational time. The origin of the model is 24°58′ 03 latitude and 55°01′28 longitude, taking east-west and north-south directions as x and y directions, respectively. There are 96 and 93 grid points in the x and y directions, respectively. The model also requires a topographical description of the model area along with other forces (e.g. wind, Coriolis etc.).

The HD model requires initial and boundary conditions to solve the equations. The closed side boundaries as well as the bottom are considered as no flow boundaries. A constant water level and zero velocities were used as initial conditions at all grid points. The tidal level at the open boundary was used as the boundary condition. The flow direction at the open boundary was always considered to be perpendicular to the boundary.

The predicted tidal level at Jebel Ali Harbor was used as the boundary condition for the HD model. The prediction was carried out using the Admiralty method (DHI, 2003c) facilitated in MIKE21 tools using major tidal constituents (see Table 1) with necessary seasonal corrections of-0.1 during February, March and April and +0.1 during July and August obtained from the Admiralty Tide Tables (ATT, 2003). Tidal levels thus predicted were referenced in mean sea level datum and converted to local chart datum (CD) adding 1.02 m (ATT, 2003a).

Table 1.

Tidal constituents at Jebel Ali Harbor (ATT, 2003).

Constituent nameAmplitude (m)Phase (°)
Principal lunar semidiurnal (M20.43 359 
Principal solar semidiurnal (S20.17 49 
Luni-solar declinational diurnal (K10.25 155 
Lunar declinational diurnal (O10.17 100 
First overtide of M2 (F40.0 
Second overtide of M2 (F60.0 
Constituent nameAmplitude (m)Phase (°)
Principal lunar semidiurnal (M20.43 359 
Principal solar semidiurnal (S20.17 49 
Luni-solar declinational diurnal (K10.25 155 
Lunar declinational diurnal (O10.17 100 
First overtide of M2 (F40.0 
Second overtide of M2 (F60.0 

Currently, there are five points from which treated wastewater are discharged into the harbor (denoted DP1 to DP5 in Figure 2). These points are added to the model as source points. The location and flow rates of the discharge points are presented in Table 2. A constant evaporation rate of 6 mm day− 1 (Meshal and Hassan, 1986) was assumed and used in the model simulation.

Table 2.

Discharge from source points at Jebel Ali Harbor.

Point IDEasting (m)Northing (m)Flow rate (m3 hr−1)
DP1 303030 2767330 9.5 
DP2 305830 2765550 9.3 
DP3 306260 2765425 112.5 
DP4 303050 2763185 2.1 
DP5 302780 2765960 4.9 
Point IDEasting (m)Northing (m)Flow rate (m3 hr−1)
DP1 303030 2767330 9.5 
DP2 305830 2765550 9.3 
DP3 306260 2765425 112.5 
DP4 303050 2763185 2.1 
DP5 302780 2765960 4.9 

The hydrodynamic regime of the harbor was simulated with the HD model. The simulated period covered 1st March 2004 to 28th February 2005, with time steps of 60 seconds. The time step was selected based on the stability criterion recommended by DHI (2003c). Although negligible, the Coriolis effect was included in the model.

The quality of a model depends on the capacity of reproducing parameters which are known from measurement. In practice, water level and velocity data are used to compare refining bed resistance and eddy viscosity to calibrate the HD model. Sometimes, bathymetry is also refined as it might be distorted during interpolation. Through the calibration processes conducted in this study, values of Chezy number (bed resistance) and eddy viscosity were selected as 40 and 1.0 m2 s− 1, respectively.

The simulated tidal levels at three locations (mentioned in Section 3.2) inside the harbor were compared with the measured data. Figure 6a shows the comparison between the simulated tidal level at St6 (Figure 2) and that of the measured values. The figure shows that tidal levels and phases are reproduced quite well. Simulated and measured flows through the main channel are compared in Figure 6b. The randomness of measured flow arises from uncertainty related to measurement. Nevertheless, the simulated flow matches the measured flow reasonably well.

Figure 6.

Comparison between simulated and measured (a) tidal level at the dead-end of the inner basin and (b) flow through the main channel of the harbor at cross-section X-1.

Figure 6.

Comparison between simulated and measured (a) tidal level at the dead-end of the inner basin and (b) flow through the main channel of the harbor at cross-section X-1.

An AD model was also set up to simulate the flushing time of conservative substances from the harbor. Equation (4) was solved using an explicit, third-order finite difference scheme (Ekebjærg and Justesenu, 1991). The initial concentration is set as zero unless otherwise mentioned. Zero concentrations at the open boundary and at all source and sink points are used in the presented AD model simulation. A spatially varied dispersion coefficient is used in the model. The procedure for calculating the dispersion coefficient is described below.

Natural channels differ from uniform rectangular ones in three important respects (Fischer et al., 1979): the depth may vary irregularly, the channel is likely to curve, and there may be large sidewall irregularities such as groins or points of land. Fischer et al. (1979) suggested a formula for estimating dispersion coefficients in real streams as:

formula
where: K is the dispersion coefficient (m2 s− 1); ü is the average velocity (m s− 1); W is the width of the channel (m); d is the depth of the channel (m); u* is the shear velocity (m s− 1) (= √ghS); g is the gravitational acceleration = 9.81 (m s−2); h is the hydraulic radius ≈ depth of the channel (m); and S is water surface slope.

Jebel Ali Harbor is directly connected to the sea and no river flows into the harbor. Nevertheless, the flow patterns in the man-made harbor are like streams that alternate their directions with the tide. Variable depth, bends and sidewall irregularities increase the mixing process considered in the above mentioned formula. Although the depth does not vary significantly, which indicates less mixing, ship movement again increases the mixing process. As there are no measurements, it makes little sense to aim for a high degree of accuracy in predicting dispersion coefficients. Therefore, Eq. (5) is considered quite appropriate for estimating the dispersion coefficient.

Based on Eq. (5), the dispersion coefficients in the x (east-west) and y (north-south) directions were calculated assuming a 300-m channel width. The average water surface slope is 2 × 10−5 m m− 1 (Section 3.2). The hydraulic radius, which is here considered to be the average depth of flow, was taken as 12.0 m. The mean velocity at each grid point during ebb tide in spring was calculated from the HD model and used as an average velocity for calculating dispersion coefficient. Thus, the calculated spatially distributed dispersion coefficients (Figure 7) were used in the AD model. Figure 7 shows that at dead-end locations (specifically inner and outer basins), dispersion is very low, and is probably dominated by molecular diffusion.

Figure 7.

Dispersion coefficients in the x-direction (left) and y-direction (right).

Figure 7.

Dispersion coefficients in the x-direction (left) and y-direction (right).

Results and discussion

Simulation of the hydrodynamic regime

Changes in the path of the main flow are a common phenomenon in most tidal channels. Simulated velocity during flooding and ebbing are shown in Figure 8a and Figure 8b, respectively. As can be clearly seen, the main flow follows alternate paths during flooding and ebbing, creating eddy-like circulations in net flow distribution (Figure 8c).

Figure 8.

Flow patterns during (a) flooding, (b) ebbing, and (c) net flow over a tidal cycle.

Figure 8.

Flow patterns during (a) flooding, (b) ebbing, and (c) net flow over a tidal cycle.

There are three eddies, all in the main channel and all anti-clockwise (Figure 8c). The first one is the most dominant and is located in the wider area just inside the entrance. The second one is located at the mouth of the outer basin and the third one is located at the mouth of the inner basin. Knowledge of flow patterns can help to select an appropriate location for wastewater discharge and cooling water intake. Figure 8c further shows that dead-end locations have a very low water circulation and can be considered relatively stagnant.

Simulation of tracer movement

Movement of a conservative tracer was simulated with the AD model. In the simulation, a conservative (non-reactive) tracer was placed into specified areas in the harbor. A pulse concentration of 1.0 mg l− 1 was assumed for the whole water column in three specific areas: (1) the whole harbor; (2) a 60 × 60 m2 grid cell at the dead-end of the outer basin; and (3) a 60 × 60 m2 grid cell at the dead-end of the inner basin. The aforementioned cases were considered in the flushing time simulation. Each case was simulated for 7 consecutive years. Cumulative mass moving out through the open boundary (entrance of harbor) under the influence of advection and dispersion was then simulated and analyzed for each case.

Figure 9a shows one of the examples of breakthrough curves of mass flushed out of the harbor under case (2), with the tracer placed at the dead-end of the outer basin. The figure shows that the majority of the mass leaves the harbor within three years but some of the tracer's mass experiences a long tail.

Figure 9.

Movement of a conservative tracer out of the harbor, (a) simulated movement from the outer basin, and (b) relative mass out for three scenarios.

Figure 9.

Movement of a conservative tracer out of the harbor, (a) simulated movement from the outer basin, and (b) relative mass out for three scenarios.

The percentage of total mass flushed out for each case was calculated at the end of each year and plotted in Figure 9b. The figure shows that the flushing time is very much dependent on the location of the tracer. The selection of a suitable discharge location depends on the objectives set by the regulatory authority. For quick flushing, the entrance channel is the most suitable location; while for slower flushing, the dead-end of the inner basin is the most suitable location. The closer the discharge location to the mouth of the harbor, the faster the flushing out of the tracer.

Eighty five percent of the tracer is flushed out within one year in case (1). On the other hand, a negligible amount of only 0.55% is flushed out during the same period in case (2). In case (3), about half of the tracer is flushed out within a year. In seven years, more than 95% of the tracer is flushed out from the whole harbor and the outer basin, whereas only 50% is flushed out from the inner basin. Slow movement of tracer from the inner channel is responsible for the pronounced tail in reaching a 100% mass removal in case (1).

The results presented above can also be used to estimate the hydraulic residence time of the harbor, i.e. the time water resides in the harbor before being “completely” exchanged by water from the Gulf. For practical purposes, we will assume that “complete” exchange of water means that 95% of harbor water or more has been replaced. Now by considering case (1) above, the hydraulic residence time would be the time needed to flush 95% of the mass of tracer present in the harbor. This time, as mentioned before, is about 7 years. It should be realized that, from a theoretical point-of-view, it takes an infinite time to exchange all the water in the harbor, as is made obvious by the long tail experienced by the tracer (Figure 9a).

While the above discussion focused on the behavior of ideal tracers, the results are also amenable to some discussion in regard to reactive tracers. Discharge of reactive tracers into relatively stagnant water regions could have a varying impact. On the one hand, the chemical will encounter a longer holding time in the harbor, which will reduce the mass of the chemical flushed into the Gulf due to the longer reaction time within the harbor. On the other hand, exposing harbor's regions with “stagnant” water to some reactive chemicals for long durations may have environmental complications within and near that region.

It should be noticed that as the main channel of the harbor is linked to the Gulf, development activities in the Arabian Gulf within the vicinity of the harbor will probably have an impact on the water quality of the harbor itself. Tracers discharged in a close vicinity of the harbor may find their way into the harbor and may reside there a long time once they reach the inner and outer basins.

Conclusions

The main flow regime in Jebel Ali Harbor follows alternate paths during flooding and ebbing, creating eddy-like circulations in net flow distribution. There are three eddy-like circulations, all in the main channel of the harbor and all moving in an anti-clockwise direction. The flushing time of a conservative tracer discharged into the harbor varies from a few months to a few years depending on the discharge location. Thus, selection of a suitable discharge location depends on the objectives of the regulatory authority. For quick flushing, the entrance channel is a suitable location; for slower flushing, the dead-end of the inner basin is the most suitable location.

The gained knowledge about the hydrodynamic regime of Jebel Ali Harbor, as presented in this paper, will be useful for sustainable management of this water body. Results presented herein could, for example, be utilized to assess the harbor water quality under current and future discharge conditions. The results may also be utilized as a basis for establishing maximum pollutant loading rate such that pollutant level in harbor water does not exceed a pre-set water quality limit.

Acknowledgements

This project is funded by the Ports, Customs and Free Zone Corporation at Jebel Ali, Dubai and the Research Sector at the UAE University. The authors are grateful to Dubai Ports Authority for providing meteorological and tidal data and to Jan de Nul Dredging Ltd. for providing bathymetry data. Thanks are also due to Stephen Aston at the ESP Section, UAE University General Requirement Unit, for editing the manuscript.

References

ATT, 2003. Admiralty Tide Tables, Admiralty Charts and Publications, The United Kingdom Hydrographic Office, Vol. 3, NP 203-01
DHI, 2003a. Scientific Documentation, MIKE 21 Hydrodynamic Module, DHI Water and Environment, DHI Software
DHI, 2003b. Scientific Documentation, MIKE 21 Advection-Dispersion Module, DHI Water and Environment, DHI Software
DHI, 2003c. User Guide, MIKE 21 Coastal Hydraulics and Oceanography, DHI Water and Environment, DHI Software
http://www.dpa.co.ae/marine/jebelaliport.htm DPA, 2005. Jebel Ali Port, Dubai Ports Authority, UAE, Download: 20 February 2005
Ekebjærg, L. and Justesenu, P.
1991
.
An explicit scheme for advection-diffusion modelling in two dimensions
.
Comp. Meth. Appl. Mech. Eng.
,
88
(
3
):
287
297
.
Fischer, H. B., List, E. J., Koh, R. C. Y., Imberger, J. and Brooks, N. H.
1979
.
Mixing in Inland and Coastal Waters
,
London
:
Academic Press Inc. (London) Ltd.
.
Hamza, W., Ennet, P. and Tamsalu, R.
2004
.
Simulation of the Egyptian ecosystem functions
.
Maritime Engineering
,
157
:
23
33
.
Jan de Nul Dredging Ltd.
2004
.
Dredging and Land Reclamation, A company commissioned for dredging Jebel Ali Harbor
Personal Communication, June 2004
Jin, X. Y.
1993
.
Quasi-three-dimensional numerical modelling of flow and dispersion in shallow water
,
Ph.D. thesis
Delft
:
Delft University of Technology
.
Meshal, A. H. and Hassan, H. M.
1986
.
Evaporation from the coastal waters of the central part of the Gulf
.
Arab J. Sci. Res.
,
4
(
2
):
649
655
.
PCFC, 2003. Environmental Control Rules and Requirements, third edition, Environment, Health and Safety Department, Ports, Customs and Free Zone Corporation, Dubai, UAE
http://www.rti.org/newsletters/cid/2002aug/startingup.htm RTI, 2005. Mohammed Bin Rashid Technology Park Planned, Center for International Development, August 2002, Download: 20 February 2005