MODELLING OF REACTION TERM OF THE ADVECTION-DIFFUSION-REACTION EQUATION FOR ENVIRONMENTAL PROBLEMS.

Introduction
The Advection Diffusion Reaction Equation (ADRE), + , +Ui = KT (1) t xi xi xi where KT is the transference coefficient and is the reaction term, is not only used to evaluate the hydrodynamics (velocity vectors) of a system, its also used to calculate scalars, such as Temperature, Salinity, Concentration of a Pollutant, Population Dynamics, Equilibrium Variables of the system (BOD, OD), etc. The main difference in the treatment of the reaction term in the ADRE is that its represented by an Ordinary Differential Equation (ODE). Since the methods to solve the ODEs are different from those to solve the former parts of the ADRE, a new choose of method should be made.

Problems of interest
1. Population Dynamics.
The choose model for this dynamics is the Lotka-Volterra model, which consists in differential equations that are used to model predator-prey interactions. The system considers of two entities. The Lotka-Volterra equations for the two dimensional system with exponential growth is defined by the following differential equations: x = Ax Bxy = x( A By ) F ( x(t ), y (t )) = (2) y = Cy + Dxy = y (C + Dx) Where x is the prey population, y is the predator population, A, B, C, and D are positive constants representing the growth and decay rate, and the interspecies interactions given in day-1. In order to extend the two entities system above into a three entities system, a third specie is added to the system. In this case the third specie feeds exclusively on the first predator population. The three species Lotka-Volterra system is defined by:

x = Ax Bxy = x( A By ) F (x(t ), y (t )) = y = Cy + Dxy Eyz = y (C + Dx Ez ) (3) z = Fz + Gzy = z ( F + Gy ) Where z is the new predator population, E, F and G are positive constants representing decay rate and the new interspecies interactions given in day-1.

2. Model for BOD DO.

For water quality modeling these Biological Oxygen Demand (BOD) and Dissolved Oxygen (DO) are variables of mean importance. Nearly all water quality models characterize BOD decay with first order kinetics represented by:
d [BOD ] = k [BOD ] dt Where k is the first order decay rate (day-1). Streeter and Phelps developed the kinetics of DO as: A) Deoxygenation: d [DO ] = k1 [DBO ] dt

(8)

(9)

B) Reareation: d [DO ] = k 2 [C s C ] dt Where Cs is the concentration of saturation of DO and C is the actual DO.

(10)

From equations 9 and 10 in terms of oxygen deficit (D) the resultant equation is: dD (11) = k1 [DBO ] k 2 [D ] dt The analytical solutions for these equations are:

[DBO] = [DBO]0 e kt k [DBO ]0 k t (e e k t )+ [D ] e k t D= 1

i

(12)

k 2 k1

3. Consecutive Reactions.
In many cases the pollutants in ecosystems are reactive compounds that with themselves, or with other pollutants, or with the system compounds. In order to make a closure to this problem, a system of ODEs is developed for a pollutant that reacts twice. The system can be set as: (4) A k1 B k 2 C Where A, B, and C are the concentration of the compound, k1 and k2 are the constant kinetics coefficient given in hr-1. An example of this kind of reactions nitrification:
NH 3 N OXD NO 2 N OXD NO3 N

(5)

For this kind of reactions the equation system can be set as:

d [A] = k1 [A] dt d [B ] = k1 [A] k 2 [B ] dt d [C ] = k 2 [C ] dt

(6)

Which analytical solutions are:

[A] = [A]0 e k t [B ] = [B ]0 e k t + [A]0 k1 (e k t e k t )

1
2

[C ] = [A]0 + [B]0 + [C ]0 [B]0 e k t + [A]0 (k1e k t k 2 e k t )

2 2 1

k 2 k1

(7)

k 2 k1

Where [A]0, [B]0, and [C]0 are the initial concentrations. In order to demonstrate the preservation of mass, the total mass of the system is calculated in all time intervals as:

NT = [A] + [B] + [C ]

Numerical Integration Methods

In this work the methods choose to solve the reaction term are Heuns method and the fourth order Runge-Kutta method. Both methods are improvements to the Eulers method. The Heuns method, also known as second order Runge-Kutta method, is resumed in the following 2 equations: ki +1 = yi + hf (ti , yi ) , h yi +1 = yi + ( f (ti , yi ) + f (ti + h , ki +1 )) , 2 where h represents the time interval.

(13)

Graphically it supposes to substitute the tangent to the curve in the point (t,y) by the line with slope equal to the media of the slopes in (t,y) and (t+h,y+h(f,y)), being the last one the point obtained by the Eulers method. The Runge-Kutta method is formulated as:
1 (k1 + 2k 2 + 2k3 + k 4 ) 6 Where h represents the time interval and k1 = f (t , y ) , h h k 2 = f t + , y + k1 , 2 2

(t , y ) =

(14)

h h k3 = f t + , y + k 2 , 2 2 k 4 = f (t + h, y + hk3 )

The Runge-Kutta method requires four slopes evaluations per step.

Solutions
1. Population Dynamics.
By means of using Heuns and Runge-Kutta method to solve the equations 3, and compared with the graphs presented by Jacob Schrum (2005), was found good agreement with his results. In his paper Schrum presents a full analysis of confidence and stability of the method. In order to make a validation, were used same values as Schrum used, the prey population was set as 0.5, the first predator population as 1, and second predator population as 2, as well as the decay, growth and interspecies coefficients were set all as 1. Figures 1 and 2 show the comparison between the results of this work and his.

Fig. 2. Graph presented by Schrum. Fig. 1. Results of Heuns method for 3D Lotka-Volterra equations.

2. Model for BOD DO.

This model was treated by means of a comparison between the analytical and the numerical solutions. The model of Streeter and Phelps was fulfilled with the normal values for the rate constants and oxygen saturation founded in literature (Rates, Constants and Kinetic Formulations in Surface Quality Modeling, second edition, by EPA). There was supposed equilibrium at the beginning and then a source of BOD was settled with a concentration of 50 mg/L. The founded coefficients and rates were: k1 = .01 hr-1 (BODs growth) k2 = 0.0833 hr-1 (Reaeration of the water body) Cs = 8.6 mg/L (concentration of oxygen saturation)

The outputs of the model were:

Fig. 3 Comparison between Analytical and numerical calculations for concentrations of DBO and OD by Heuns method.

Fig. 4 Comparison between Analytical and numerical calculations for concentrations of DBO and OD by RungeKuttas method.

This graph shows good agreement between the numerically calculated concentrations of DBO and OD and the ones calculated by analytical solutions, showing good confidence for both methods. In terms of computational effort the best option will be Heun, because would be less time expensive and show similar results.

3. Consecutive Reactions.
For this problem the time interval of the analytical solution were set as minimum as possible, to reach a better approach. Then, for the numerical solution were choose a set of time intervals, in order to find the optimum one, taking into account the computational time. Since numerical solution gives punctual values, a linear interpolation was carried out in order to get values for the whole reaction time. A comparison between both analytical solution and the different results obtained for the different time intervals was carried out, the choose intervals were 360, 180, 60, 20, and 10 min and also was determined the error for all the points. The model were used for the specific reaction NH 3 N OXD NO2 N OXD NO3 N , where the kinetic coefficients k1 and k2 were found to be 0.15 and .2 respectively and the initial concentrations were set as A = 0.05, B = 0.01, and C = 0. The resultant graphs are following presented.

For Heuns method:

Fig. 5. Comparison between analytical solution and Heun method for a time interval of 10 min.

Fig. 6. Comparison between analytical solution and Heun method for a time interval of 20 min.

Fig. 7. Comparison between analytical solution and Heun method for a time interval of 60 min.

Fig. 8. Comparison between analytical solution and Heun method for a time interval of 180 min.

Fig. 9. Comparison between analytical solution and Heun method for a time interval of 360 min.

The resultant error graphs for the three compounds are the following:

Fig. 10. Error for compound A along the reaction time

Fig. 11. Error for compound B along the reaction time

Fig. 12. Error for compound C along the reaction time

For the Runge-Kutta method are presented the graphs of concentrations:

Fig. 13. Comparison between analytical solution and the Runge-Kutta method for a time interval of 10 min.

Fig. 14. Comparison between analytical solution and the Runge-Kutta method for a time interval of 20 min.

Fig. 15. Comparison between analytical solution and the Runge-Kutta method for a time interval of 60 min.

Fig. 16. Comparison between analytical solution and the Runge-Kutta method for a time interval of 180 min.

Fig. 17. Comparison between analytical solution and the Runge-Kutta method for a time interval of 360 min.

The error graphs for Runge-Kuttas method:

Fig. 18. Error for compound A along the reaction time

Fig. 19. Error for compound B along the reaction time

Fig. 20. Error for compound C along the reaction time

Since the graph for the time interval of 360 minutes seems to be the worst approach in both cases, because the B compound has a slope so much different from the analytical solution, a comparison of this compound were carried out in the different time intervals and methods:

Fig. 21. compound B behavior whit different interval times, using Heuns method.

Fig. 22. compound B behavior whit different interval times, using Runge-Kuttas method.

These comparisons of the B compound were used to determine the optimum time interval and method to develop this kind of studies. The results show that Heuns method with a time interval of 60 min reach to good results and its not a time expensive method in computational time.

References.
Rates, Constants, and Kinetic Formulations in Surface Water Quality Modeling, second edition. Bowie, et. al., U. S. Environmental Protection Agency, 1985. Modeling Tools for Environmental Engineers and Scientists, N. Nirmala Khandan. CRC Press, 2002. Modeling Population Dynamics with Volterra-Lotka Equations, Jacob Schrum. In partial fulfillment of the Mathematics Capstone, 2005. Basic River Quality Models, Dr. Gza Jolnkai. For the International Hydrological Program, Technical Documents in Hydrology, No. 13. UNESCO, Paris, 1997. Water Quality and Flow Simulation in River Kali, India, Dr. R. Jha, Dr. C. S. P. Ojha, Dr. K. K. S. Bhatia, India, 2004.