Hydrological, Chemical and Biological Processes of Transformation and Transport of Contaminants in Aquatic Environments (Proceedings of the Rostov-on- Symposium, May 1993). IAHSPubl. no. 219, 1994. 251

Hydrodynamic and water quality modelling of the Lower Don River,

ANATOLY M. NTKANOROV Hydrochemical Institute, 198 Stachki av., Rostov-on-Don, 344104 Russia ROSEMARIE C. RUSSO Environmental Research Laboratory, US Environmental Protection Agency, Athens, Georgia, USA MARINA G. YERESCHUKOVA & E. ZIA HOSSE1NIPOUR Hydrochemical Institute, 198 Stachki av., Rostov-on-Don, 344104 Russia ROBERT B. AMBROSE Environmental Research Laboratory, US Environmental Protection Agency, Athens, Georgia, USA

Abstract The Don River and in particular the lower reach is severely polluted. A modular modelling package, WASP4, was applied to the Lower Don River to investigate processes controlling the hydrodynamics and water-quality of the river. These models were used to evaluate the hydrologie dynamics and the transport by and transformation of contaminants in the river to provide water quality management alternatives that are environmentally and ecologically sound and economically feasible. Results of hydrodynamic and water quality simulations and related discussion are presented in this paper.

INTRODUCTION

Among the major rivers of Russia, the lower Don River is one of the most affected by anthropogenic impacts. The Don River basin is an intensively exploited region with respect to agriculture and industry. Waters of the Lower Don system are used for municipal and industrial needs, and for irrigation of crops. To protect and manage these waters, information is needed about the hydrological regime, the transport and transformation of contaminants within the river, and the status of its ecosystem. Modelling the hydrodynamic and water quality characteristics of the Lower Don river system should help in its understanding and its protection. The 313-km Don River reach from to the Gulf is referred to as the Lower Don River system. This part of the river meanders through the valley in a 11-12 km band that in some places is 22-25 km wide. The river width varies from 400 to 600 m. The average water depth during low flow conditions ranges from 4 to 6 m in the main channel, and decreases to 0.7 m in the shoals. There are about 96 shoals on the Lower Don. The mouth of the Don is downstream of Rostov- on-Don. The river drains about 340 km2. The river sub divides into several channels and creeks forming a delta upon entry into the Sea, the largest channels are called the Stary Don and Bolshaya Kalancha. Other tributaries contribute flow to the Lower Don including Seversky Donets, which is the largest and enters from the right 252 Anatoly M. Nikanorov et al.

bank. In general, the Seversky Donets is the main source of pollution of the Lower Don River. Other tributaries which enter from the right bank, such as Temernik and Tuzlov, also are severely impacted, mostly from municipal waste and industrial discharges of big cities. Data analysis indicates that the most likely problems of the Lower Don system are eutrophication, contamination by organic matter, and pesticides. In addition, the high turbidity may degrade aquatic ecosystems and hence self-purification processes. The objective of this paper is to evaluate a hydrodynamics using model simulations of the Don River reach from Razdorskaya to Taganrog Gulf (see Fig. 1). In addition to the evaluating the hydrodynamic regime, nutrients cycling and phytoplankton dynamics were evaluated in the Lower Don River.

Fig. 1 The scheme of the Lower Don river system.

MODELLING FRAMEWORK

The brief description of the Lower Don River system given above indicates that to solve the enumerated problems, models are needed which can quantify the water quality processes in river. The WASP4 modelling package accounts for many of these processes and was applied in this study (Ambrose et al., 1988). The WASP4 modelling package is based on a mass balance of various solutes in the water body, and contains independent blocks for hydrodynamic, eutrophication, and toxicant contamination simulation. These blocks interact through input/output files created automatically or by users. The latter allows us to use the other programs interactively to simulate additional processes. Considering the hydrological and hydraulic characteristics of Lower Don System, the model must simulate spring high flows throughout its network of channels and creeks. One-dimensional equations of momentum and continuity were used in their integral form using the RIVNET model to simulate the channel network (Hosseinipour & Martin, 1991). Hydrodynamic and water quality modelling of the Lower Don River, Russia 253

HYDRODYNAMIC MODEL

The equations of continuity and momentum in the integral form were derived using the same assumptions as the Saint-Venant's equations in the RIVMOD model. In the integral form, these equations were used to simulate the hydrodynamic regimes which change rapidly. The one-dimensional equations are as follows (Kutchment, 1982): dA dQ dt dx (1) 2 2 dQ d ,Q . AdY T/ .,. n Vabs(V). n Bw2^o _~+—(-=-) + sA—+qV = gA(in- L^) + C.JJ Mucosa dt dxK A' S 3t q * V #4/3 st where: Q — water discharge, Y — water depth, A — cross-sectional area, V — water velocity, q — lateral inflow, g — acceleration of gravity, i0 — bottom slope, n — Manning roughness coefficient, R — hydraulic radius, W — wind speed, a — angle between channel direction and wind speed, Cst — wind stress coefficient, t — time variable, x — spatial variable. The following functional relation among the cross-sectional area, wetted perimeter (or hydraulic radius), and water depth was used to solve these equations for water velocity and depth. 2 A = ÛQ + Û! * f (2) 2 P = b0 + bx * Y* where: A — cross-sectional area, P — wetted perimeter (or hydraulic radius), au bt, i — 0,1,2 - constants for geometry of each cross section. To simulate the flow dynamics in branched rivers and/or hydraulic networks, these equations (1) can be solved using different numerical methods and associated boundary conditions. The finite difference method was used herein to describe the channelled river system. The river reach and channels were divided into segments with uniform physical properties and the intersections among segments and external boundaries were called junctions. The implicit approximation scheme was used for which the four variables (water discharges and depths at each end of the segment) were assigned to each segment. The boundary conditions between segments was determined for the water depths and discharges using the approach of Schaffranek et al. (1981) as follows: 254 Anatoly M. Nikanorov et al.

Hk = Hj, V£j£/,. (3) E Qj = o je/,. where Qj} Hk, Hj are discharge, and water depths at the ends of the segments, respectively; and /,• are indices corresponding to junction i. If L segments meet at a junction, then L - 1 linear independent equations were applied to the water depths and one equation was applied to the water discharge. Hence, at this junction, L conditions were described. Additionally, L conditions can be obtained for the other ends of the segments at the junctions using the same approach. One condition was specified at the external boundary. The boundary conditions on the portion of branched network were determined in differently depending on available information for each boundary. The time variant values of water depth or discharge were provided as a table. The time variant depth was provided as a harmonic function of time. The functional relation was provided between discharge and depth. To find approximate solutions for equation (1), the four-point Preissman scheme on a time-space grid network was adopted. As well-known, this scheme was defined by the following formulae: 1 1 n 1+1 m df _ 1 (f * _f +f _f \ ~JT ~ . . vn+1 Jn+1 Jn Jn )

where: 8 — weighing parameter, At - time step, AX — space step corresponding to the segment length, m fn — the function value at the nth point in space and Mh lapse in time. The choice of parameter value 0 allows the use of different approximation schemes from explicit (when 6 = 0) to fully implicit (when 6 = 1) and their combinations (when 0 < 6 < 1). For modelling of the Lower Don River system, a weighing factor of 0.75 was used. For a river reach divided into N segments, 2*N equations may be written by applying equations (1) on each segment. Another 2*N equations can be obtained from the boundary conditions specifying for each junction. Thus, altogether 4*N equations can be written for 4*N. For their simultaneous solution the Newton-Raphson iterative procedure was applied. The hydrodynamic model parameters were identified for the period 1-10 July 1979. Subsequent model runs were conducted for 5-15 February and 1-10 September. The predicted values were consistent with the observations. The difference between the simulated and measured values of water stages were less than 5% and did not increase in time (Table 1). However, little information was available for the surface water elevation at the downstream boundaries. Therefore, these boundary conditions were only approximated. Detailed hydrological data for water stages were available for the reach from Razdorskaya to Rostov-on-Don, only. Consequently, model results were compared to the observed data for this reach only. Hydrodynamic and water quality modelling of the Lower Don River, Russia 255

Table 1 Water stage simulation on the Lower Don river system (1-10 September 1979), in m.

Time Melikchovskaya: Bagaevskaya: Starocherkassk: Aksai: (days) sim. obs. sim. obs. sim. obs. sim. obs.

1 2.31 2.33 1.33 1.36 0.78 0.81 0.72 0.68 2 2.33 2.39 1.28 1.32 0.50 0.53 0.31 0.22 3 2.25 2.35 1.15 1.16 0.30 0.26 0.12 0.08 4 2.22 2.33 1.11 1.10 0.26 0.18 0.18 0.04 5 2.23 2.33 1.12 1.12 0.24 0.17 0.19 0.06 6 2.25 2.35 1.14 1.16 0.29 0.22 0.20 0.18 7 2.24 2.33 1.15 1.21 0.39 0.41 0.22 0.26 8 2.22 2.32 1.15 1.22 0.40 0.33 0.24 0.20 9 2.19 2.21 1.17 1.20 0.47 0.47 0.32 0.25 10 2.25 2.31 1.22 1.22 0.54 0.52 0.46 0.36

The model also was verified for the simulation of the channelled system from May to October. The model was calibrated for this period in an average year, 1983. The boundary conditions were specified every ten days. The residuals derived from predicted and observed data were larger than those for the linear part of the river. But, the maximum deviation was less than 10%, which was within the accuracy of the measured values. The increased deviations was caused by the error in specifying boundary condition. But, although relatively large errors were associated with the observed data, the model adequately described the hydrodynamic regime of the river. The model was subsequently used to predict the hydrodynamic response for other types of hydrological events in river.

HYDRODYNAMIC SIMULATION

Hydrodynamic processes affect mixing and transport of chemical and biological compounds in natural waters. The model simulations for hydrochemical and hydrobiological processes require input data on the river discharge and velocity which was available for the Lower Don River. Simulations were conducted for different time periods. Of particular interest for water quality simulations is the growing season from May through October. The model accuracy was a concern because of the of the long time period of the simulation. In the upper boundary of this part of river, daily water discharges and depths were available only until 1987. Also, the impact of short-term variations in the water flow could not be evaluated when the model is run for such a long time period. Therefore, ten-day averages were used to provide boundary conditions. For the downstream boundaries, water discharges and depths could be approximated only. These approximations induced uncertainty in modelled water distribution among the channels at the mouth of the Don River. An uncertainty analysis was conducted to evaluate the effect of the errors in boundary conditions on the simulation. The water discharge and depths were measured with an accuracy of about 10%. Simulations were conducted with the boundary conditions modified to ±10% of the observed. The results of these tests were evaluated using a factors analysis, i.e. the influence of each boundary condition were evaluated individually and in combinations. 256 Anatoly M. Nikanorov et al.

To compare the results of different experiments, the differences between the calibration results and test results were evaluated as follows:

1 N 1 N (5)

where: A - absolute measure of error influence, Xb(tt) — the value of characteristic at time tt, which is obtained from the calibration simulation, X(tt) — the value of characteristic at the time /,-, which is obtained by the test simulation, N — the number of time steps during the simulation. The relative measure is a more representative estimator of uncertainty. The relative measure can be calculated on the basis of following formula (Jorgensen, 1986):

1 N N^Xb(ti) (6) è = - 100% Ap P where: ô — the relative measure, in %, Ap/p - the relative change in the parameter value (here in the boundary value). The results of uncertainty analysis are given in Table 2. The first position corresponds to the boundary condition of Stary Don branch, the second to the Bolshaya Kalancha branch. "-" indicates a decrease greater than or equal to 10%, " + " indicates an increase greater than or equal to 10%, and "0" indicates no change over the calibration simulation. Equations (5) and (6) were applied to estimate the changes in water distribution between channels that were caused by a 10% change in boundary conditions. The results of this analysis are presented in the Table 2. The value A is the percentage difference in water flow for the Stary Don. Results indicate that 32.3% of the water flow in the Lower Don River is transported in the Stary Don.

Table 2 The results of uncertainty analysis.

Plan The part of flow

(-,0) 36.7 4.4 13.6 ( + ,0) 27.1 5.2 16.1 (0,-) 28.0 4.3 13.3 (0,+) 40.0 7.7 23.8 (-,+) 43.2 10.3 33.7 2.6 8.0 (-,-) 29.7 2.1 6.5 ( + ,+) 34.4 11.2 34.7 ( + ,-) 21.1 Hydrodynamic and water quality modelling of the Lower Don River, Russia 257

As shown in Table 2, it is important to determine correct boundary values to describe the water distribution among channels. In most cases, overestimation of water depths for a channel causes a decrease in water discharge in that channel, and a subsequent change in the water transport among channels. This effect is most significant at low-flow conditions. During the simulations, effects were noted in July, when the average discharge was at a minimum: the maximum difference between the calibration and test simulation was 60%. Errors in the boundary conditions affected the flow of the entire river, but the effect was less than for individual channels and more pronounced downstream than at the upstream reach. In the middle part of Lower Don River, the deviation between the calibration and test simulations increased, but did not exceed 5%. The test simulations indicate that to provide accurate simulation of hydrodynamic regime at the mouth of Don River, the downstream boundary conditions must to be specified as accurately as possible, and no less accurately than the upstream boundary conditions. Also, the accuracy is most important for the hydrodynamic simulation of low flow. The study of Lower Don River morphometric characteristics showed, that river's widths and depths were uneven and change significantly along the river's channel. The widths vary from 175 m to 400 m. The depths vary from 2.5 m to 5 m, and sometimes reach 8 m. The channel configuration was evaluated from cross-sections constructed at a few points along the river. This analysis indicated that the configuration of the cross sections also changed along the river. The most downstream cross-section where discharge was measured is near Razdorskaya. The predicted water discharge at the mouth of river from the hydrodynamic simulation was extremely important to modelling the water quality there. Results of water discharge simulations for the period from May to October 1983 are presented on Fig. 2. Discharge did not change significantly along the river. Discharge was high during the beginning of the study period, and decreased thereafter, but the deviation from average flow did not exceed 15 % of the mean value. Discharge, therefore, was held constant for the water quality simulation.

WATER QUALITY SIMULATION

The eutrophication simulation was conducted using the WASP4 modelling package (Ambrose et al, 1988). This package provides several choices for eutrophication modelling. Given the paucity of data to describe the dynamics phytoplankton groups, the following state variables were included into the model: total phytoplankton biomass, organic nitrogen and phosphorus concentrations, and concentrations of inorganic phosphorus and nitrogen species including ammonium and nitrate. Diatoms constitute a large part of spring phytoplankton content, and dissolved silica is the primary nutrient affecting their dynamics. However, this nutrient cycle was evaluated in the model, because silica is not a nutrient to other groups of algae. Implicitly, the water quality model assumes that silicon is never depleted enough to limit diatom growth. Water, temperature, and light intensity have a significant effect on phytoplankton dynamics. These parameters were included in the model. 258 Anatoly M. Nikanorov et al. Rostov-on-Don

&

Q) CD Ki Xi V to •H Q H (1) 4J (d June July August September October Dugino

May June July August September October Fig. 2 Simulated water discharges.

Sensitivity analysis is an important component of model development. This analysis provides a means for evaluating sensitive parameters in the model. To qualify the influence of parameter changes on the dynamics of state variables, equation 6 was applied. The following parameters affecting phytoplankton dynamics were chosen for the analysis: maximum growth rate (G), mortality rate (M), decomposition rates of

Table 3 Model sensitivity to the parameters variations.

Variable G Kv #N M L r

Phytoplankton 85.00 0.01 0.00 2.20 9.44 99.91 Ammonia 16.41 0.18 3.64 38.00 18.10 11.78 Nitrate 2.37 0.06 6.01 5.92 2.59 2.83 Organic N 4.38 0.05 0.00 0.29 4.84 3.40 Inorganic P 2.48 6.16 0.00 0.10 2.72 1.85 Organic P 0.41 36.33 0.00 2.19 0.45 7.29 Hydrodynamic and water quality modelling of the Lower Don River, Russia 259 organic phosphorus and nitrogen (KP, K^), light saturation coefficient (L), and carbon- to-chlorophyll ratio (r). Parameter values were varied ± 10% of their base values. The analysis was conducted according a one-factor plan. The maximum values of equation (6) are presented in Table 3. The model was numerically stable for the range of parameters values used in this study. Because the phytoplankton dynamics are most sensitive to the variations in carbon-to-chlorophyll ratio, it would be useful to determine more precisely the species composition of algae, and their elemental composition in a future research effort. Among the factors affecting the phytoplankton dynamics, light intensity was the most important because it was the most limiting factor on this ecosystem. Nitrogen exerts an insignificant influence on phytoplankton dynamics, partly because nitrogen species concentrations in the Lower Don waters were in excess of their requirement. Likewise, phosphorus concentrations had an insignificant effect on phytoplankton dynamics because phosphorous concentrations in excess. The effect of water flow on the modelled ecosystem also was investigated. Routine monitoring data and hydrodynamic simulations indicate that water discharge remained above 400 m3 s"1 in the Lower Don system during the growing season. Occasionally, discharge exceeded 1800 m3 s"1 at the beginning of the period in some years such as 1979, but it typically decreased to about 500-600 m3 s"1. A discharge of 600 m s was used to calibrate the water quality model. The model was subsequently tested using the same input data but varying the discharge to 400 and 1000 m3 s"1. The test simulations indicate that maximum concentrations of phytoplankton biomass were not affected by flow. Also, regardless of the discharge, phytoplankton concentrations remained similar to those of the upstream boundary. The travel time was increased during low flows to evaluate its impact on phytoplankton concentrations and results indicate that slowing the transport was insufficient to cause an intensive bloom of phytoplankton (Fig. 3). The date of maximum phytoplankton concentrations did not match the date on which concentrations began to decrease rapidly, which was easily explained by differences in the discharge. The influence of water redistribution among channels was evaluated with respect to effects on phytoplankton dynamics at the mouth of Don River. As discussed previously,

M m id e 0 •H m

September October Fig. 3 The effect of water flow on the phytoplankton dynamics. 260 Anatoly M. Nikanorov et al.

significant changes in the distribution of water flow was caused by errors in specifying hydrodynamic boundary conditions. During water quality simulations, however, only slight deviations in biomass (2% of the biomass maximum) were observed. It should be noted that only part of Don River mouth was modeled, and travel time along its channels did not increase sufficiently to cause a phytoplankton bloom. In general, results indicate that the Don River ecosystem is affected by the hydrodynamic regime. But, the river has a high capacity to prevent intensive algal blooms in that reach. The high concentrations of phytoplankton biomass observed in the river are derived from upstream transport. As the sensitivity analysis indicated, light intensity is the most important factor affecting the phytoplankton dynamics. Due to high turbidity in the Lower Don River, light extinction in the water column is extremely high also. The effect of decreasing turbidity on phytoplankton concentrations was investigated. Simulations results indicate that a decrease in light extinction coefficient by 50% causes an phytoplankton biomass concentrations to increase by less than 8%. Decreasing the light extinction coefficient by a factor of 8 caused phytoplankton biomass concentrations to increase by only 18%. Using an extinction coefficient of 0.01, which corresponds to pure water, caused phytoplankton biomass concentrations to increase by only 22% in the channels of the Don River mouth.

CONCLUSION

Effects of light intensity on phytoplankton biomass concentrations in the Lower Don River ecosystem are compensated by effects of the hydrodynamic regime in this part of the river. Short travel time prevented phytoplankton from intensively blooming, even with significant decreases in water turbidity. The simulation of eutrophication demonstrated that the main factor controlling phyto­ plankton dynamics is river discharge. The river has a high capacity for inhibiting intensive phytoplankton blooms in the reach from Razdorskaya to Azov and Dugino. Most of the phytoplankton biomass in the reach is derived from transport from upstream. The water quality simulations provide a broad view of factors affecting the aquatic ecosystem. Because the simulations did indicate that upstream sources had a significant impact on the water quality of the river reach from Razdorskaya to Azov and Dugino, a future study of water quality and ecosystem structure of the upstream reach from Tsimlyansk Reservoir to Razdorskaya should provide sufficiently detailed information to help improve water quality in the Lower Don river system.

REFERENCES

Ambrose, R. B., Wool, T. A., Connoly, J. P. &Schanz, R. W. (1988) WASP4, a hydrodynamic and water quality model — model theory, user's manual, and programmer's guide. USEPA/60O/3-87/039, Athens, Georgia. Hosseinipour, E. Z. & Martin, J. L. (1991) RTVMOD — a one-dimensional hydrodynamic and sediment transport model. Model Theory and User's Manual. Jorgensen, S. E. (1986) Fundamentals of Ecological Modelling. Elsevier. Kutchment, L. S. (1980) Models of River Flow Forming Processes (in Russian). Hydrometeoizdat, Leningrad. Schaffranek, R. W., Baltzer, R. A. & Goldberg, D. E. (1981) A model for simulation of flow in singular and interconnected channels. USGS Techniques of Water Resources Investigations, Book 7, Chapter C3.