Você está na página 1de 8

CLEANING UP THE GREAT LAKES

Group; The Wreck of the Eigen Fitzgerald


Brody Nagy and Helene Casanova
MATH-2500-W2
November 22nd, 2016

The Problem
This project explores an application of one of the most powerful tools within
differential equations; compartmental analysis. The problem was to derive a system
of equations that is able to model the change in amount of pollution in the Great
Lakes over time. In this problem, it is assumed at the volume and flow rates of each
lake remains constant, the liquids entering each Lake undergo a process of perfect
mixing and pollutions enter or leave only by inflow or outflow of solution. We also
assume the mass of pollutants to be equal to 1 mass unit per cubic mile and that all
input and output rates are in cubic miles per year.
Initial Estimates Using Simple Models
I Drain Time
The purpose of the first group of equations is to set a lower bound for the amount of
time required to remove all of pollutants from each lake. These equations were
constructed by quantifying the change in volume of each lake as it relates to the
outflow rate with respect to time. They are linear expressions and dont account for
concentration or mass of pollutants, they simply describe the amount of time it
would take each lake to be completely drained if the input rate was zero and outflow
rate remained constant. This is a useful model because it illustrates the fastest way
to remove all pollutants from the lakes, and if any of the subsequent models
returned values less than those described here, they would obviously be incorrect.
Each equation was plotted in Desmos and the x-intercept was calculated and used
as the solution.

Superior :S ' ( x ) =3132mi 315 x mi 3

S ' ( x ) =0 when x =208.8 years

Michigan : M ' ( x )=1322 mi338 x mi3

M ' ( x )=0 when x=34.791 years

Huron : H ' ( x )=782 mi3 68 x mi3

H ' ( x )=0 when x=11.5 years

Erie: E ' ( x )=102 mi3 85 x mi3

E ' ( x )=0 when x=1.2 years

Ontario ; O ' ( x )=424 mi399 x mi3

O' ( x )=0 when x=4.283 years

II 5 Separate Tanks
The next group of equations is designed to model each lake as a singular tank,
uncoupled from the other four lakes. This means each lakes can still be modeling
using only one variable. This model uses a constant input rate of clean water with a
pollutant mass of zero and is able to describe the quantity of pollutant in each lake
over time, (assuming perfect mixing). These equations were calculated using the

C1 e ( S ' ( x ) , M ' ( x ) O ' ( x) ) t

characteristic equations

where

C1

is equal to the

amount of pollutant at time zero. The starting quantity of the pollutant in each lake
3

mi

was set a 1 unit per

meaning the initial mass is equal to the volume of each

lake, and the starting concentration is 100%. The equations were then solved for
the amount of time required to reduce the level of pollutant to 50% and 5% of its
original level. This model gives a better idea of what the answers returned by the
final model should look like, however because the equations dont account for the
multivariable inputs, the answers returned are still inaccurate. Each function was
plotted in Desmos and appropriate intercepts were calculated to return these
solutions. (E.g. the Lake Superior function was graphed along with y=1566 and
y=156.6)
(

5 t

Superior :S ( x )=3132e 1044


Michigan : M ( x )=1322 e
(

Huron : H ( x )=782 e e
Erie : E ( x )=102 e

19t
)
661

2 t
)
23

5 t
)
6

Ontario:O ( x )=424 e

S ( x ) =50 , 5 when x=144.729 ,625.509 years


M ( x )=50 ,5 when x=24.114 , 104.22 years
H ( x )=50 ,5 when x =7.971, 34.451 years
E ( x )=50 , 5 when x=.832 ,3.595 years

99t
)
424

O ( x )=50 ,5 when x =2.989 ,12.83 years

Final Model
In order to properly model the Great Lakes as a system of 5 coupled mixing tanks,
the final condition that needed to be accounted for was the fact that the pollution
from one lake flows into the next in chain. The first step was to create a matrix that
described the inputs of each lake with their respective variables. As the input rates
of fresh water in each lake are assumed to have water with a pollutant mass of zero,
they require no further manipulation and this matrix will be sufficient to describe
the flow from one lake to the next. Each value in the matrix describes an input or
output of solution (positive is an input negative is an output) and were developed
from the rate of input/output divided by the lake the rate originates from. For
example, the first row describes the change of the amount of pollutants in Lake
Superior, the input is only freshwater and the output is its only outflow rate divided
by its volume. The same is true for Lake Michigan. Lake Huron has the input of both
Superior and Michigan, and Erie and Ontario follow suit.

15 /3132
0
0
0
0
0
38 /1322
0
0
0
A= 15/3132
38/1322 68 /782
0
0
0
0
68/782 85/ 102
0
0
0
0
85/102 99/424

This matrix was entered in Matlab and the characteristic polynomial was determined
using the poly function. The roots, Eigen values and vectors were found by using
the roots functions, subtracting the roots from each row times a 5x5 identity
matrix from the matrix A and using the null function respectively. Finally, in order
to determine the constant coefficients, integer-valued Eigenvector were used and
manually assessed. The final matrix equation is described as:

(31.7 ) e(0.8333t )
A=

[ ]

0
0
(
)
( 128.1 ) e 0.2335 t
0
0.5842
0.8116

[] [ ]

0
0
0
0
( 0.0870t )
(
)
0 + ( 64.3 ) e
0.8297 ( 1503.7 ) e 0.0207t
0
0.0967
1
0.5497

[ ]

0
0.8791
0.4341 ( 313
0.469
0.1910

Multiplying through we are able to translate the matrix equation into a system of
scalar equations:

Superior :S (t )=3131.9236exp (.0048t)


Michigan : M ( t )=1321.90267exp(.0287t)

Huron : H ( t )=53.34971exp (.0870t)+652.756exp(.0284t)+182.64324exp (.0048t)

Erie: E ( t )=18.51914exp (.8333 t)6.21781exp(.0870 t)+70.53253exp(.0287 t)+19.14302exp (.0

Ontario:O ( t )=25.72772exp(.8333t)+128.1exp (.2335t)35.34571exp(.0870t)+287.2067e

To describe the concentration of pollutants over time these equations are divided by
the volume of each lake:

Superior :S c ( t )=[ 3131.9236exp (.0048t ) ] /3132


Michigan : M c ( t )=[ 1321.90267exp (.0287t ) ] /1322
Huron : H c ( t )=[ 53.34971exp (.0870t ) +652.756exp (.0284t ) +182.64324exp (.0048t ) ] /782

18.51914exp(.8333 t )6.21781exp(.0870 t )+70.53253exp (.0287 t )+ 19.14302exp(.0048t )


Erie : Ec ( t )=
] / 102

25.72772exp(.8333t)+128.1exp (.2335t)35.34571exp(.0870t )+287.2067exp(.0287t)+


Ontario :O c ( t )=
] / 424

The numerical solver function vpasolve was used to determine when the pollution
level in each lake had reached 50% of its original level, as well as 5% of its original
level. Although it is true that the first two lakes, Superior and Michigan do not need
to be solved in this manor, as they are not coupled to any other system meaning
they can be solved exactly as they were in the Separate Tanks model. Our final
answers for the time required to reduce the mass of pollutants are as follows:

Superior

S ( x ) =50 , 5 when x=144.401, 624.239 years

Michigan

M ( x )=50 ,5 when x=24.149, 104.378 years

Huron

H ( x )=50 ,5 when x =35.332, 321.505 years

Erie

E ( x )=50 , 5 when x=24.693 , 276.593 years

Ontario

O ( x )=50 ,5 when x =21.769 ,250.032 years

The solutions provided above are valid, the most striking difference between the
solution to our final model and those of the simpler models is the large increase of
time required to reduce the pollution in the final 3 lakes. This makes logical sense as
we are now accounting for the fact that all of the pollution from the previous lakes is

entering the subsequent lakes. All solutions fit within the lower bound described by
the Drain Time model.

Our final answers for the time required to reduce the concentration of pollutants
were calculated in an identical manor to the mass of pollutants and are as follows:

Superior

S c ( x )=50 ,5 when x =144.401, 624.239 years

Michigan

M c ( x )=50 , 5 when x=24.149 , 104.378 years

Huron

H c ( x )=50 ,5 when x=35.332 , 321.505 years

Erie

Ec ( x ) =50 , 5 when x=24.693 ,276.593 years

Ontario

Oc ( x )=50 ,5 when x=21.769 , 250.032 years

If you are an observant reader, you may notice that our solutions for concentration
of pollutants appear to be identical to those for mass of pollutants; thats because
they are. This occurs because it is assumed that the mass of pollutants is assumed
to be 1 mass unit per cubic mile, so when the concentration functions are graphed,
they are actually identical to the mass functions with the exception of the yintercept which has been relocated to y=100. This makes perfect sense as the lakes
start completely saturated with pollutants: concentration at t=0 is 100%.

Graphs

Group Members
Brody Nagy
Brody was vital to this project, he developed the first draft of the equations for all
models and also drafted the initial version of the Matlab script and subsequent
revisions. He graphed all functions in Desmos to check the values of the solutions
via intercept calculation as an extra method of assessing the validity of the
solutions. He worked diligently on the initial as well as final reports and completed
all tasks in a timely manner.
Helene Casanova
Helene was just as vital to this project, revising all equations and code to correct the
errors made in the initial versions. Without her contributions, the correct solutions
would not have been calculated. Helene proofread and checked the initial and final
report and created the citations and problem statement. She also submitted the
hardcopy of the final report as Brody was abroad.

Citations

Bender, E. A. (1978). An introduction to mathematical modeling. Retrieved


November 20, 2016.

Differential Equations - Modeling with First Order DE's. (n.d.). Retrieved November
21, 2016, from http://tutorial.math.lamar.edu/classes/de/modeling.aspx

Henriksen, M. (2016). Applications of 1st order ODEs Compartmental Analysis [PDF


document]. Retrieved from https://wit.blackboard.com/bbcswebdav/pid-533742-dtcontent-rid-1837443_1/courses/SPEC_MATH2500_merge_sec01w2_Fall2016/Sec
%203.2%20A%20WS.pdf

Você também pode gostar