2DHI – Water & Environment (see above), jok@dhi.dk

3Bechtel Systems & Infrastructure Inc., San Francisco, U.S.A, anfindik@bechtel.com

Abstract: This paper presents a 3D eutrophication model established for the Maracaibo system. The model is an add-on module to a hydrodynamic model established for the system, including Lake Maracaibo, Maracaibo Strait, Tablazo Bay and the Gulf of Venezuela. The model represents in detail the important features of the complicated system: the nutrient loading, the production and degradation of phytoplankton, the settling of detritus, oxygen conditions and the important effects from saline intrusions and density stratification in the Lake. The modeling periods using this system model made up a total of 16 months. Furthermore, a 10-year Lake simulation was performed to analyze the effect of interannual variations in the run-off to the Lake on the eutrophication conditions. A short description of the intra-annual variations in the system based on the simulations is given.

 

Keywords: 3D modeling, eutrophication, oxygen depletion, semi-enclosed seas, Maracaibo, denitrification, stratification

A 3D hydrodynamic and eutrophication model has been set-up for the Maracaibo system to analyze the present conditions of the system and to perform scenario assessments of the effects of potential future changes to the system. The model tool is part of an integrated project on environmental remediation options for the Lake (Findikakis et al., 2001). At present Lake Maracaibo suffers from anoxic conditions in the hypolimnion and algae blooming in the surface waters. Furthermore, point source discharges, especially in the Strait, contaminate the water with BOD and coliforms.

The environmental quality of the Lake depends on a large number of physical, chemical, and biological processes in Lake Maracaibo itself and the adjacent bodies of water, i.e. Maracaibo Strait, Tablazo Bay and the Gulf of Venezuela (Herman de Bautista, 1997). To understand the present conditions and predict changes resulting from specific future actions it is necessary to consider the system of the Lake-Strait-Bay-Gulf in its entirety. The major challenge in modeling this system is finding the proper way to deal with the broad range of spatial and temporal scales of its most significant processes. Due to the different time and space scales of the processes affecting the environmental quality in the Lake and the computational requirements for their proper simulation, a combination of linked models was used to meet the objectives of the study.

The present paper deals with the results of the eutrophication modeling. The hydrodynamic basis is described in Hansen et al. (2001a).

The objective of the eutrophication modeling has been to establish a 3D water quality model tool to understand the governing water quality features of the system and their temporal and spatial extent. The steps taken to fulfil this goal build on the hydrodynamic model already established (Hansen et al., 2001a), but include specifically:

• Implementation of loads and boundary setting to the eutrophication module of the hydrodynamic Regional Model and a Lake Subset Model.
• Calibrating the Lake Subset Model against monitoring data from 1997-1998.
• Validating the full Regional Model set-up against monitoring data from December 1998 – March 1999 (Horn et al., 2001b).
• Running the Regional Model for a 12-month period December 1997 – November 1998 and the Lake Subset model for a constructed 10-year period with changing hydrologic and loading input to the Lake.

MIKE 3 is a software package developed at DHI for unsteady 3D Newtonian fluids (Rasmussen et al., 1990). The MIKE 3 EU module (Eutrophication module) includes primary production by phytoplankton, zooplankton grazing, nutrient cycling in water and nutrient fluxes at the water sediment interface, as well as oxygen production and demand processes, including reaeration. The variables or components included in the model are phytoplankton (carbon, nitrogen and phosphorus), chlorophyll-a, zooplankton, detritus (carbon, nitrogen and phosphorus), inorganic nutrients (nitrogen and phosphorus), and dissolved oxygen.

The module was used to simulate the biological and chemical processes, which are important for Tablazo Bay, the Strait and especially the Lake.

The EU module applies the same Cartesian grid representation of the modeling domain as the hydrodynamic model, ie. a dynamically nested set-up with a horizontal resolution of 6750 m in the Gulf and the Lake. Embedded into this domain is a 2250 m resolution domain as a transition to a 750 m resolution domain of the Strait. Finally, a 250 m resolution domain covers Tablazo Bay, see Hansen et al. (2001a). The vertical grid spacing is 1.5 m, with the exception of the top layer, which extends from the actual water surface elevation (varying, but typically –0.5 m to +0.5 m) down to an elevation of –2.25 m.

Two model setups were applied for the eutrophication modeling:

• The Regional Model. This set-up was used to study the details in time and space of the nutrient, production and oxygen conditions driven by the exchange flow in the Bay and Strait and the internal mixing conditions in the Lake. This model set-up was used for simulation periods of up to one year.
• A Lake Subset Model. This model is a restricted subset of the Regional Model, covering only the Lake. The Lake Subset Model was used for the initial calibration of the various eutrophication processes and finally for the 10-year simulation. Flows, salt fluxes and eutrophication state variable concentrations at the Lake boundary obtained from the MIKE 3 Regional Model simulations are used as boundary conditions in the Lake Subset Model.

The forces driving the eutrophication module of the Regional Model include:

• Time series of concentrations at the open Gulf boundary between Punta Perret and Amuay, taken from literature and monitoring by the Instituto Para el Control y la Conservacion de la Cuenca del Lago de Maracaibo (ICLAM) in the outermost part of the Bay.
• Time series of riverine and point source load established as part of the Maracaibo study by other study team members
• Time series of photosynthetic active light intensity

The initial distribution of the EU components within the model domain was established based on monitoring data collected by ICLAM.

Table 1 shows the load figures for the Lake. It is assumed that the river load follows the temporal variation of the volume discharge.

Data from ICLAM’s regular quarterly water quality surveys extended with some specific study surveys, covering about 25 stations in the Lake, Strait and Bay in the period 1997 - March 1999 have been applied to calibrate and validate the eutrophication module.

The calibration of the eutrophication model complex was done using data from December 1997 to November 1998. The simulated and measured concentrations of oxygen, chlorophyll, inorganic N and inorganic P in surface and bottom water at Station C11 in the central part of the Lake are illustrated in Figure 1 for the entire calibration period. Figure 2 show an example of typical measured and simulated profiles through the depth.

In general terms the model reproduces the observed concentration pattern over the depth and the variation through time. In the first part of the calibration period intrusions of saline and oxygenated water occur that result in oxygen concentration in the bottom water of the central part of the Lake increasing up to approximately 4 mg/l. From the end of March the oxygen declines to zero and anoxic conditions prevail in the bottom water at Station C11 for the rest of the calibration period.

??The distribution of the intrusion of saline oxygenated water is seen from the S-N and W-E transects from 28 February 1998 illustrated in Figure 3. The DO concentration just above the bottom at the same date is illustrated in Figure 4. Furthermore, Figures 3 and 4

Fig.1  Time series of surface and bottom concentrations at the center of the Lake (C11) through the 1-year simulation, compared with measurements

show an example of the simulated low oxygen concentration in the entire bottom water of the Lake during the stagnation period starting in April 1998. These simulation results are in accordance with observations.

Fig. 2  Profile comparisons of DO, IN and IP from the center of the Lake (C11) at 20 September 1998

For the nutrients (N and P) the simulated surface concentrations are close to the measured values. The model simulates the observed general pattern with an increasing concentration profile towards the bottom. In general, the model simulated the correct concentration levels in the bottom water although some discrepancies between measured and simulated concentrations can be observed. The differences can, to some extent, be ascribed to horizontal differences within the Lake.

Only very few data concerning chlorophyll concentrations from Lake Maracaibo exist for use for the calibration. The available data display differences over depth and time suggesting that probably only the general chlorophyll level, which the model should be able to simulate, can be extracted from these data. Based on measurements from November 1997, an initial surface value of approximately 6 m g/l was chosen for the start of the simulation. During the calibration period chlorophyll data are only available for 2, 17, and 30 November 1998. Surface water concentrations on these dates were approximately 7, <0.4 and 2 m g/l, respectively. Because of the relatively few chlorophyll data during the calibration period, chlorophyll levels measured in January – March 1999 were also taken into account for the calibration of the model. In January – March 1999 all measurements were below 3 m g/l and in most cases below the detection limit (of 0.4- 0.8 m g/l). Based on these data, the simulated chlorophyll level of approximately 1-2 m g/l is considered satisfactory.

Fig. 3  Simulated DO distribution in the Lake for a S-N transect and a W-E transect, displayed for a intrusion situation (28 February 1998) and a stagnation period (20 September 1998)

Fig. 4  The DO concentration just above the bottom in the situation with saline intrusion (28 February 1998) and stagnation in the Lake (20 September 1998)

The calibrated model was validated against measured values from December 1998 – March 1999. In Figure 5 the measured and simulated oxygen concentrations in surface and bottom water in the central part of Lake Maracaibo (Station C11) are shown for the validation period. Neither the measured values nor the simulated values showed any oxygen in the bottom water during the period up to February 1999. In February and March 1999 the model simulated oxygen concentrations at Station C11 between 1.3 and 3.6 mg/l, whereas totally anaerobic conditions were measured in the bottom water during this period. Figure 5 furthermore suggests that the model overestimates the oxygen concentration at the surface.

A more detailed analysis of the simulation result shows that the discrepancy to a great extent can be explained by horizontal variations. Oxygen levels of the measured levels are simulated in areas where bottom water is brought to the surface due to internal circulation. The generally increased oxygen concentration in the bottom water in March 1999 compared to the concentration measured at station C11 is due to a nearly homogenized water column described by the hydrodynamic model (Hansen et al., 2001a). During this period a survey covering the entire Lake was conducted on March 18 only. At this time total anaerobic conditions occurred only at the two stations (C11 and C9) in the middle of the Lake. At most of the other locations in the central part of the Lake the measured bottom oxygen concentration were above 2 mg/l.

Fig. 5  Selected time series at the center of the Lake (C11) from the validation of the set-up, compared with monitoring data

Nutrient levels of the measured levels were simulated in the surface layer (Figure 5). The model simulated in accordance with measured increasing levels towards the bottom, although it did not simulate the very high bottom concentration of inorganic nitrogen seen on 16 December 1998 and 15 January 1999. These high concentrations are due to the occurrence of a restricted volume close to the bottom with high salinity and reduced mixing with the rest of the water column. The hydrodynamic model simulated a less pronounced increment in the salinity and thereby an increased mixing with the rest of the water masses in the Lake, resulting in the less pronounced nutrient profile close to the bottom.

Based on the calibration and validation results, it was concluded that the eutrophication model, in general, was capable of simulating low or no oxygen levels in the bottom layer when the hydrodynamic model accurately simulates the stratification of the water column. In the validation period this is clearly seen in the period up to February 1999. Therefore, it was concluded that the model correctly simulates the response to the most important water quality variables of changes in hydrodynamic conditions.

Figure 6 illustrates the result of the 10-year simulation by showing the variation at the central Station C11. The modeling period results show that the model is in balance and there is no trend in the calculated concentration during the period.

Fig.6  Simulated DO, IN, IP and productivity development in the 10-year simulation from the center of the Lake(C11)

The general picture seen from the 10-year simulation is an increase in oxygen concentration in the bottom water in December - March as a consequence of the saline intrusion, which brings oxygenated water into the bottom layer of Lake Maracaibo. This is in accordance with the measurements carried out on 8 March 1998, see Figure 1. The model describes (as was also observed) that this is followed by the depletion of oxygen in the bottom water until May, after which time the bottom water remains without oxygen until the next saline intrusion the following December – March.

A picture reflecting this pattern is seen with regard to the nutrient concentrations in the bottom water. The concentration builds up in the bottom water throughout the year until December. In the period December to March the concentrations drop and fluctuations in the concentrations are observed, reflecting the intrusion of saline water with lower nutrient concentrations. An increase in nutrient concentrations starting as early as November is described in the surface water. The set-off of this increase coincided with decreasing temperature in the surface layer. The decreasing surface temperature causes an increased surface layer thickness due to the mixing with mid-depth water of elevated nutrient concentration. This process is ongoing until January/February, when the temperatures in the surface and bottom waters approach the same value (see Hansen et al., 2001a).

The supply of nutrients to the surface water causes increasing productivity within the Lake during the period November - April and, as a consequence of this, an increasing chlorophyll concentration and decreasing Secchi depth or transparency. From the 10-year simulation a spike in the productivity is also observed every year in August, which results in increased chlorophyll. At the same time a less pronounced increase is also seen in the nutrient concentrations. This is caused by a temporary decline in temperature in the surface water and thereby increased mixing with the nutrient-rich mid-depth water layers. The temperature decline is introduced by the meteorological forcing function used for the modeling.

Furthermore, it can be mentioned that the present model calibration provides a net accumulation in the sediment corresponding to 36% of the N load (including denitrification) and 54% of the P load to the Lake. Also, 44 % of the settled organic material in the central part of the Lake is degraded under oxygen consumption.

Based on the calibration and validation results and the 10-years simulation, it is concluded that the eutrophication model is, in general, capable of correctly simulating the response to the most important water quality variables of changes in hydrodynamic conditions. Furthermore, it can be concluded that it is vital to include the 3-dimentional distribution within the Lake and the complicated dynamic hydrodynamic processes to understand and interpret the measured concentrations correctly. This is especially important during the period December - March with the potential intrusion of saline and oxygenated water through the Strait. It is also important to describe the decoupling between the hypolimnion and the epilimnion correctly in time and space to achieve the correct simulation of nutrient and oxygen concentrations and the assimilative capacity of the Lake during the stagnation period. As the model provides a reasonable description of the response to changing hydrodynamic conditions, it is concluded that the established model complex is suitable for use in predictive simulations under various assumptions for future conditions.

The work reported in this paper was part of the Integral Study for the Environmental Restoration of Lake Maracaibo, conducted for PDVSA by Bechtel International. Luis Delgado and Nelson Corrie managed the study for PDVSA. The authors wish to acknowledge the support of PDVSA and the constructive criticism of the initial model results by PDVSA’s Technical Committee overseeing the study. The comments by Dr. Reinaldo Garcia-Martinez of the Venezuelan Central University, are especially appreciated.

Findikakis, Angelos N., J?rg Imberger, Ian Sehested Hansen, Luis A. Delgado, Reinaldo García-Mantínez, and Erich Gundlach, 2001: A Study Of Environmental Remediation Options, For Lake Maracaibo, Venezuela, Proceedings XXIX IAHR Congress, Beijing, China, September 17-21, 2001.

Hansen, Ian Sehested, Jacob Steen M?ller, and Angelos N. Findikakis, 2001a: 3 Dimensional Hydrodynamic Modeling of the Maracaibo System, Venezuela, Proceedings XXIX IAHR Congress, Beijing, China, September 17-21, 2001.

Herman de Bautista, Susana, 1997: Proceso de salinización en el Lago de Maracaibo, Instituto Para el Control y la Conservacion de la Cuenca del Lago de Maracaibo (ICLAM), Maracaibo.

Horn, David A., Peter Yeates, J?rg Imberger, and Angelos N. Findikakis, 2001a: One-Dimensional Water Quality Modelling of Lake Maracaibo, Proceedings XXIX IAHR Congress, Beijing, China, September 17-21, 2001.

Horn, David A., Bernard Laval, J?rg Imberger, and Angelos N. Findikakis, 2001b: Field Study of Physical Processes in Lake Maracaibo, Proceedings XXIX IAHR Congress, Beijing, China, September 17-21, 2001.

Rasmussen, E.B., H.J. Vested, P. Justesen and L.C. Ekebj?rg, (1990), SYSTEM 3: A Three Dimensional Hydrodynamical Model, Danish Hydraulic Institute.