In Section 1.4 we introduced the exponential differential equation dP/dt=kP,
Suppose that the population changes only by the occurrence of births and deaths—there is no immigration or emigration from outside the country or environment under consideration. It is customary to track the growth or decline of a population in terms of its birth rate and death rate functions defined as follows:
β(t)
δ(t)
Then the numbers of births and deaths that occur during the time interval [t,t+Δt]
Hence the change ΔP
so
The error in this approximation should approach zero as Δt→0,
in which we write β=β(t), δ=δ(t),
Population explosion Suppose that an alligator population numbers 100 initially, and that its death rate is δ=0
(with t in years). Then upon separating the variables we get
Substitution of t=0, P=100
For instance, P(10)=2000/10=200,
In situations as diverse as the human population of a nation and a fruit fly population in a closed container, it is often observed that the birth rate decreases as the population itself increases. The reasons may range from increased scientific or cultural sophistication to a limited food supply. Suppose, for example, that the birth rate β
that is,
where a=β0−δ0
If the coefficients a and b are both positive, then Eq. (2) is called the logistic equation. For the purpose of relating the behavior of the population P(t) to the values of the parameters in the equation, it is useful to rewrite the logistic equation in the form
where k=b
Logistic model In Example 4 of Section 1.3 we explored graphically a population that is modeled by the logistic equation
To solve this differential equation symbolically, we separate the variables and integrate. We get
If we substitute t=0
Finally, this equation is easy to solve for the population
at time t in terms of the initial population P0=P(0).
The finite limiting population noted in Example 2 is characteristic of logistic populations. In Problem 32 we ask you to use the method of solution of Example 2 to show that the solution of the logistic initial value problem
is
Actual animal populations are positive valued. If P0=M,
However, if P0>M,
In either case, the “positive number” or “negative number” in the denominator has absolute value less than P0
Thus a population that satisfies the logistic equation does not grow without bound like a naturally growing population modeled by the exponential equation P′=kP.
Limiting population Suppose that in 1885 the population of a certain country was 50 million and was growing at the rate of 750,000 people per year at that time. Suppose also that in 1940 its population was 100 million and was then growing at the rate of 1 million per year. Assume that this population satisfies the logistic equation. Determine both the limiting population M and the predicted population for the year 2000.
We substitute the two given pairs of data in Eq. (3) and find that
We solve simultaneously for M=200
about 153.7
The logistic equation was introduced (around 1840) by the Belgian mathematician and demographer P. F. Verhulst as a possible model for human population growth. In the next two examples we compare natural growth and logistic model fits to the 19th-century U.S. population census data, then compare projections for the 20th century.
Percentage growth rate The U.S. population in 1800 was 5.308 million and in 1900 was 76.212 million. If we take P0=5.308
Thus our natural growth model for the U.S. population during the 19th century is
(with t in years and P in millions). Because e0.026643≈1.02700,
Logistic modeling The U.S. population in 1850 was 23.192 million. If we take P0=5.308
in the two unknowns k and M. Nonlinear systems like this ordinarily are solved numerically using an appropriate computer system. But with the right algebraic trick (Problem 36 in this section) the equations in (10) can be solved manually for k=0.000167716, M=188.121.
The table in Fig. 2.1.4 compares the actual 1800–2010 U.S. census population figures with those predicted by the exponential growth model in (9) and by the logistic model in (11). Both models agree well with the 19th-century figures. But the exponential model diverges appreciably from the census data in the early decades of the 20th century, whereas the logistic model remains accurate until 1940. By the end of the 20th century the exponential model vastly overestimates the actual U.S. population; indeed it predicts a U.S. population of nearly 1.5 billion in the year 2010, over 3.6 times the actual value. The logistic model, on the other hand, underestimates the U.S. population, but with a percentage error of less than 43%.
Year | Actual U.S. Pop. | Exponential Model | Exponential Error | Logistic Model | Logistic Error |
---|---|---|---|---|---|
1800 | 5.308 | 5.308 | 0.000 | 5.308 | 0.000 |
1810 | 7.240 | 6.929 | 0.311 | 7.202 | 0.038 |
1820 | 9.638 | 9.044 | 0.594 | 9.735 | −0.097 |
1830 | 12.861 | 11.805 | 1.056 | 13.095 | −0.234 |
1840 | 17.064 | 15.409 | 1.655 | 17.501 | −0.437 |
1850 | 23.192 | 20.113 | 3.079 | 23.192 | 0.000 |
1860 | 31.443 | 26.253 | 5.190 | 30.405 | 1.038 |
1870 | 38.558 | 34.268 | 4.290 | 39.326 | −0.768 |
1880 | 50.189 | 44.730 | 5.459 | 50.034 | 0.155 |
1890 | 62.980 | 58.387 | 4.593 | 62.435 | 0.545 |
1900 | 76.212 | 76.212 | 0.000 | 76.213 | −0.001 |
1910 | 92.228 | 99.479 | −7.251 |
90.834 | 1.394 |
1920 | 106.022 | 129.849 | −23.827 |
105.612 | 0.410 |
1930 | 123.203 | 169.492 | −46.289 |
119.834 | 3.369 |
1940 | 132.165 | 221.237 | −89.072 |
132.886 | −0.721 |
1950 | 151.326 | 288.780 | −137.454 |
144.354 | 6.972 |
1960 | 179.323 | 376.943 | −197.620 |
154.052 | 25.271 |
1970 | 203.302 | 492.023 | −288.721 |
161.990 | 41.312 |
1980 | 226.542 | 642.236 | −415.694 |
168.316 | 58.226 |
1990 | 248.710 | 838.308 | −589.598 |
173.252 | 75.458 |
2000 | 281.422 | 1094.240 | −812.818 |
177.038 | 104.384 |
2010 | 308.745 | 1428.307 | −1119.562 |
179.905 | 128.839 |
The two models are compared in Fig. 2.1.5, where plots of their respective errors—as a percentage of the actual population—are shown for the 1800–1950 period. We see that the logistic model tracks the actual population reasonably well throughout this 150-year period. However, the exponential error is considerably larger during the 19th century and literally goes off the chart during the first half of the 20th century.
In order to measure the extent to which a given model fits actual data, it is customary to define the average error (in the model) as the square root of the average of the squares of the individual errors (the latter appearing in the fourth and sixth columns of the table in Fig. 2.1.4). Using only the 1800–1900 data, this definition gives 3.162 for the average error in the exponential model, while the average error in the logistic model is only 0.452. Consequently, even in 1900 we might well have anticipated that the logistic model would predict the U.S. population growth during the 20th century more accurately than the exponential model.
The moral of Examples 4 and 5 is simply that one should not expect too much of models that are based on severely limited information (such as just a pair of data points). Much of the science of statistics is devoted to the analysis of large “data sets” to formulate useful (and perhaps reliable) mathematical models.
We next describe some situations that illustrate the varied circumstances in which the logistic equation is a satisfactory mathematical model.
Limited environment situation. A certain environment can support a population of at most M individuals. It is then reasonable to expect the growth rate β−δ
The classic example of a limited environment situation is a fruit fly population in a closed container.
Competition situation. If the birth rate β
This might be a reasonable working hypothesis in a study of a cannibalistic population, in which all deaths result from chance encounters between individuals. Of course, competition between individuals is not usually so deadly, nor its effects so immediate and decisive.
Joint proportion situation. Let P(t) denote the number of individuals in a constant-size susceptible population M who are infected with a certain contagious and incurable disease. The disease is spread by chance encounters. Then P′(t)
Spread of rumor Suppose that at time t=0, 10
Substituting P0=10
Then substitution of t=1, P=20
that is readily solved for
With P(t)=80,
which we solve for e−100kt=136.
thus after about 4 weeks and 3 days.
Consider a population P(t) of unsophisticated animals in which females rely solely on chance encounters to meet males for reproductive purposes. It is reasonable to expect such encounters to occur at a rate that is proportional to the product of the number P/2 of males and the number P/2 of females, hence at a rate proportional to P2.
(where M=δ/k>0
Note that the right-hand side in Eq. (13) is the negative of the right-hand side in the logistic equation in (3). We will see that the constant M is now a threshold population, with the way the population behaves in the future depending critically on whether the initial population P0
Doomsday vs. extinction Consider an animal population P(t) that is modeled by the equation
We want to find P(t) if (a) P(0)=200;
To solve the equation in (14), we separate the variables and integrate. We get
Substitution of t=0
Note that, as t increases and approaches T=ln (4)/0.06≈23.105,
Substitution of t=0
Note that, as t increases without bound, the positive denominator on the right in (16) approaches +∞.
Thus the population in Example 7 either explodes or is an endangered species threatened with extinction, depending on whether or not its initial size exceeds the threshold population M=150.
Figure 2.1.6 shows typical solution curves that illustrate the two possibilities for a population P(t) satisfying Eq. (13). If P0=M
Separate variables and use partial fractions to solve the initial value problems in Problems 1–8. Use either the exact solution or a computer-generated slope field to sketch the graphs of several solutions of the given differential equation, and highlight the indicated particular solution.
dxdt=x−x2, x(0)=2
dxdt=10x−x2, x(0)=1
dxdt=1−x2, x(0)=3
dxdt=9−4x2, x(0)=0
dxdt=3x(5−x), x(0)=8
dxdt=3x(x−5),
dxdt=4x(7−x), x(0)=11
dxdt=7x(x−13),
Population growth The time rate of change of a rabbit population P is proportional to the square root of P. At time t=0
Extinction by disease Suppose that the fish population P(t) in a lake is attacked by a disease at time t=0,
Fish population Suppose that when a certain lake is stocked with fish, the birth and death rates β
where k is a constant. (b) If P0=100
Population growth The time rate of change of an alligator population P in a swamp is proportional to the square of P. The swamp contained a dozen alligators in 1988, two dozen in 1998. When will there be four dozen alligators in the swamp? What happens thereafter?
Birth rate exceeds death rate Consider a prolific breed of rabbits whose birth and death rates, β
Note that P(t)→+∞
Death rate exceeds birth rate Repeat part (a) of Problem 13 in the case β<δ.
Limiting population Consider a population P(t) satisfying the logistic equation dP/dt=aP−bP2,
Limiting population Consider a rabbit population P(t) satisfying the logistic equation as in Problem 15. If the initial population is 120 rabbits and there are 8 births per month and 6 deaths per month occurring at time t=0,
Limiting population Consider a rabbit population P(t) satisfying the logistic equation as in Problem 15. If the initial population is 240 rabbits and there are 9 births per month and 12 deaths per month occurring at time t=0,
Threshold population Consider a population P(t) satisfying the extinction-explosion equation dP/dt=aP2−bP,
Threshold population Consider an alligator population P(t) satisfying the extinction-explosion equation as in Problem 18. If the initial population is 100 alligators and there are 10 births per month and 9 deaths per month occurring at time t=0,
Threshold population Consider an alligator population P(t) satisfying the extinction-explosion equation as in Problem 18. If the initial population is 110 alligators and there are 11 births per month and 12 deaths per month occurring at time t=0,
Logistic model Suppose that the population P(t) of a country satisfies the differential equation dP/dt=kP(200−P)
Logistic model Suppose that at time t=0,
Solution rate As the salt KNO3
What is the maximum amount of the salt that will ever dissolve in the methanol?
If x=50
Spread of disease Suppose that a community contains 15,000 people who are susceptible to Michaud’s syndrome, a contagious disease. At time t=0
Logistic model The data in the table in Fig. 2.1.7 are given for a certain population P(t) that satisfies the logistic equation in (3). (a) What is the limiting population M? (Suggestion: Use the approximation
with h=1
Year | P (millions) |
---|---|
1964 | 24.63 |
1965 | 25.00 |
1966 | 25.38 |
⋮ |
⋮ |
2014 | 47.04 |
2015 | 47.54 |
2016 | 48.04 |
Constant death rate A population P(t) of small rodents has birth rate β=(0.001)P
Constant death rate Consider an animal population P(t) with constant death rate δ=0.01
Population growth Suppose that the number x(t) (with t in months) of alligators in a swamp satisfies the differential equation dx/dt=0.0001x2−0.01x
If initially there are 25 alligators in the swamp, solve this differential equation to determine what happens to the alligator population in the long run.
Repeat part (a), except with 150 alligators initially.
Logistic model During the period from 1790 to 1930, the U.S. population P(t) (t in years) grew from 3.9 million to 123.2 million. Throughout this period, P(t) remained close to the solution of the initial value problem
What 1930 population does this logistic equation predict?
What limiting population does it predict?
Has this logistic equation continued since 1930 to accurately model the U.S. population?
[This problem is based on a computation by Verhulst, who in 1845 used the 1790–1840 U.S. population data to predict accurately the U.S. population through the year 1930 (long after his own death, of course).]
Tumor growth A tumor may be regarded as a population of multiplying cells. It is found empirically that the “birth rate” of the cells in a tumor decreases exponentially with time, so that β(t)=β0e−αt
Solve this initial value problem for
Observe that P(t) approaches the finite limiting population P0exp(β0/α)
Tumor growth For the tumor of Problem 30, suppose that at time t=0
Derive the solution
of the logistic initial value problem P′=kP(M−P), P(0)=P0.
Derive the solution
of the extinction-explosion initial value problem P′=kP(P−M), P(0)=P0
How does the behavior of P(t) as t increases depend on whether 0<P0<M
If P(t) satisfies the logistic equation in (3), use the chain rule to show that
Conclude that P″>0
Approach to limiting population Consider two population functions P1(t)
Logistic modeling To solve the two equations in (10) for the values of k and M, begin by solving the first equation for the quantity x=e−50kM
Logistic modeling Use the method of Problem 36 to fit the logistic equation to the actual U.S. population data (Fig. 2.1.4) for the years 1850, 1900, and 1950. Solve the resulting logistic equation and compare the predicted and actual populations for the years 1990 and 2000.
Logistic modeling Fit the logistic equation to the actual U.S. population data (Fig. 2.1.4) for the years 1900, 1930, and 1960. Solve the resulting logistic equation, then compare the predicted and actual populations for the years 1980, 1990, and 2000.
Periodic growth rate Birth and death rates of animal populations typically are not constant; instead, they vary periodically with the passage of seasons. Find P(t) if the population P satisfies the differential equation
where t is in years and k and b are positive constants. Thus the growth-rate function r(t)=k+bcos 2πt
These investigations deal with the problem of fitting a logistic model to given population data. Thus we want to determine the numerical constants a and b so that the solution P(t) of the initial value problem
approximates the given values P0, P1, …, Pn
then we see that the points
should all lie on the straight line with y-intercept a and slope b (as determined by the function of P on the right-hand side in Eq. (2)).
This observation provides a way to find a and b. If we can determine the approximate values of the derivatives P1′,P2′,…
First plot the points (P1,P1′/P1), (P2,P2′/P2), …
Then use a ruler to draw a straight line that appears to approximate these points well.
Finally, measure this straight line’s y-intercept a and slope b.
But where are we to find the needed values of the derivative P′(t)
suggested by Fig. 2.1.8. For instance, if we take i=0
for the slope at (t1,P1)
Investigation A: Use Eq. (3) to verify the slope figures shown in the final column of the table in Fig. 2.1.9, then plot the points (P1,P1′/P1), …, (P11,P11′/P11)
Year | i | ti | Population Pi | Slope Pi′ |
---|---|---|---|---|
1790 | 0 | −10 | 3.929 | |
1800 | 1 | 0 | 5.308 | 0.166 |
1810 | 2 | 10 | 7.240 | 0.217 |
1820 | 3 | 20 | 9.638 | 0.281 |
1830 | 4 | 30 | 12.861 | 0.371 |
1840 | 5 | 40 | 17.064 | 0.517 |
1850 | 6 | 50 | 23.192 | 0.719 |
1860 | 7 | 60 | 31.443 | 0.768 |
1870 | 8 | 70 | 38.558 | 0.937 |
1880 | 9 | 80 | 50.189 | 1.221 |
1890 | 10 | 90 | 62.980 | 1.301 |
1900 | 11 | 100 | 76.212 | 1.462 |
1910 | 12 | 110 | 92.228 |
Year | World Population (billions) |
---|---|
1975 | 4.062 |
1980 | 4.440 |
1985 | 4.853 |
1990 | 5.310 |
1995 | 5.735 |
2000 | 6.127 |
2005 | 6.520 |
2010 | 6.930 |
2015 | 7.349 |
Investigation B: Repeat Investigation A, but take t=0 in 1900 and use only the 20th-century population data listed in Fig. 2.1.4 to create a logistic model. How well does your model predict the U.S. population in the years 1990–2010?
Investigation C: Model similarly the world population data shown in Fig. 2.1.11. The Population Division of the United Nations predicts a world population of 9.157 billion in the year 2040. What do you predict?
18.221.185.155