An introduction to using population matrix models in R

After completing this assignment, you should know how to

• develop a population matrix model using R

• understand the components of a population matrix model
• use a population matrix model to project population size in the future, estimate λ, and
determine the stable age distribution
• perform an elasticity analysis and interpret the values

To understand the dynamics of populations – for example, how fast they are growing (or
declining) – we must be able to estimate demographic parameters like rates of survival and
reproduction. These rates are often dependent on the age (or stage) of the organism (Gotelli
2001). The study of these vital statistics is known as demography. Ecologists have many tools to
aid in describing and analyzing the demographic parameters (like birth and death rates) of
populations. Life tables are one example of such a tool – they provide an age- or stage-specific
schedule of survival and reproduction for a particular population. Here we will learn about other
tools for representing and working with these demographic parameters to describe population
dynamics: life cycle diagrams and matrix models.

Overview of matrix models

Life cycle diagrams

These simple diagrams provide a visualization of the age structure and vital rates of a population.
Each circle represents the age class (or stage). Each arrow represents transitions between age
classes or stages.

Pi = probability that an individual in age class i will survive to age class i + 1

Fi = fertility rate of each age class i

Figure 1. Example of a life cycle diagram for a

population with four age classes. Adapted from
Donovan and Welden (2002).

Matrix models

Information from a life cycle diagram can be translated into the form of a “matrix,” a rectangular
array of values. The elements of the matrix (these values) represent the vital rates that
characterize a population. Using matrix algebra, one can determine  (the finite rate of increase),
the stable age distribution, project future population sizes, and estimate other demographic

1
parameters based on a population matrix. There are two types of matrix models used to describe
populations: a Leslie matrix can be used to summarize information about the age-specific
survival and fertility rates of a population. A similar model that uses stages instead of ages is
called a Lefkovitch matrix. It may not always be possible (or useful) to determine the age of an
organism; instead we may group an organism’s life history into stages (like egg, larva, adult, for
insects; seedling, small rosette, large rosette, reproductive, for a plant, etc.) (Gotelli 2001).

Translating information in a life cycle diagram to a Leslie (or Lefkovitch) matrix

Let’s say we have a population with four age classes like that in the life cycle diagram above
(Fig. 1) We could set up a Leslie matrix from the vital rates shown in this diagram, using the
conventions illustrated in Fig. 2.

Figure 2. Example of assembling a Leslie matrix based on the vital rates shown in Fig. 1. (Adapted from Donovan
and Welden 2002).

Population projection using a Leslie matrix

We can represent the current population by a single column vector with the number of
individuals in each age class (N0). Using matrix multiplication, we can find the number of
individuals in each age class in the next time step, represented by another vector (N1) (Fig. 3).
But how do you multiply a matrix by a vector? There are special rules for this. You can do it by
hand as illustrated in Fig. 4.

Figure 3. Using matrix multiplication to project population size. (Adapted from Donovan and Welden 2002).

2
Figure 4. Example of multiplying a matrix by a vector to project population size. (Adapted from Donovan and
Welden 2002).

We can then use our two population vectors, N0 and N1 to determine  (lambda), the finite rate of
increase. Just add the elements of each vector to determine the population size at each time point.

N0 = 20 + 40 + 21 + 9 = 90
N1 = 75.3 + 14 + 24 + 6.3 = 119.6

Then we use the equation  = Nt+1 / Nt:  = N1/N0 = 119.6/90 = 1.329

Of course, doing this by hand, especially to project population size farther into the future would
be very tedious. Next we will learn how to use R to work with these matrix models.

Using R to project population size, determine, , and more

You will practice the process of constructing and manipulating a matrix model using R (in
RStudio), with the help of the R package popbio (Stubben and Milligan 2007). You will get an
introduction to these methods using a tool called “swirl.” Before proceeding, you should read
and follow the steps for installing and starting swirl in this document: “Getting started with
swirl.” You should install the course “Population_matrix_models” according to the instructions
in that file. Start swirl and go to the first lesson of the course, “Setting up matrix.”

Lesson 1: Setting up matrix

Follow the prompts in the swirl lesson which will first introduce you to setting up matrices in
general. In the second half of the swirl lesson, you will set up a Leslie matrix model (Fig. 5b)
that corresponds to the life cycle diagram shown in Fig. 5a.

3
Figure 5. (a) Life cycle diagram for a hypothetical population with four age classes. (b) Leslie matrix
corresponding to this life cycle diagram.

To enter the matrix from above, you will first need to create a vector to define the names of the
age or stage classes:

ages = c("Age 1", "Age 2", "Age 3", "Age 4")

The next step is to enter the values of the matrix itself, using the matrix() function. We will
call our matrix “A” and enter the values by rows. Here is the code you will need to use to match
the data in Fig. 5b:

A = matrix(c(0,0.9,1.4,1.1,0.7,0,0,0,0,0.6,0,0,0,0,0.3,0),nrow=4,
byrow=T,dimnames=list(ages,ages))

Here’s a breakdown of this code: After entering the 16 elements of our matrix (make sure you
have 16 numbers!), nrow=4 specifies 4 rows, byrow=T indicates filling by row (instead of
column), dimnames=list(…) labels both rows and columns with the age classes that you defined
above in the vector you created.

For our next step, we will need to define the initial population size of our population, broken
down between our 4 age classes. If we have 20 1-year-olds, 40 2-year-olds, 21 3-year-olds, and 9
4-year-olds, we can define our initial population size vector N0 as

N0=c(20,40,21,9)

You will use this vector to project population size into the future using matrix multiplication, in
the last step of lesson 1 in this swirl course. When you are done, move on to Lesson 2:
Analyzing matrix models.

Lesson 2: Analyzing matrix models.

This lesson makes use of the R package popbio which will allow for more sophisticated
analyses and visualizations using the matrix model you developed in lesson 1. Note that this
lesson is pre-loaded with the matrix model A and the initial population vector N0 that you created
in lesson 1, so you do not need to repeat those steps.

4
Whenever we make use of an R package, we should first make sure we have it installed. If you
do not have the popbio package already, swirl will install it for you. Once installed, you must
load the package with the code library(name of package). In this case, we would enter

library(popbio)
[Note, R already did this for you when you started this swirl lesson, but it’s a good step to
remember when you are working outside of this swirl course!]

In the package popbio, the function pop.projection() allows you to project the
population size any number of time steps into the future. To determine our population size 10
steps into the future based on our matrix A and the initial population size vector N0, you would
use this code:

projA=pop.projection(A,N0,10)

For your reference, the important output of the pop.projection() function includes:
$lambda estimate of  based on the last 2 time steps of the projection
$stable.stage stable age (stage) distribution based on last population estimate
$stage.vectors projects number in each age class/stage at each time step
$pop.sizes total population at each time step
$pop.changes estimate of  at each time step (=Nt+1/Nt)
(Stubben and Milligan 2007)

We can use this output to plot population growth over time. The swirl lesson will first have you
make a quick graph to see the pattern of population growth. Then you will create a nicely
formatted graph with proper x and y labels. When requested, write code to reproduce as best you
can the plot shown in Fig. 6. (Some hints for modifying your plot: pch=16 makes solid circles
instead of the default open points, type="b" allows you to plot both points and lines
connecting them.)

Figure 6. Population size projected over 10 years.

5
You will next be instructed on how to use the eigen.analysis() function to learn more
about the structure and dynamics of the population represented by our Leslie matrix.

For your reference, the important output of the eigen.analysis() function includes:

$lambda1  of population (when stabilized)

$stable.stage stable age (stage) distribution
$sensitivities sensitivity of each vital rate in matrix
$elasticities elasticity of each vital rate in matrix
$repro.value reproductive value: expected contribution of each individual to
current and future reproduction
$damping.ratio rate of convergence to a stable stage distribution
(Stubben and Milligan 2007)

These parameters are all explained further in the swirl lesson. In particular, you will learn how
we can interpret elasticity values to determine which vital rates (survival probabilities or fertility
rates) contribute the most to the growth rate (λ) of the population, which can be useful
information in the management of invasive species or the conservation of threatened species
(Conard 2011, Stevens 2009).

References

Conard, J.M. 2011. Demographic vital rates and population growth: an introduction to projection
matrices and elasticity analysis. Teaching Issues and Experiments in Ecology, Vol. 7:
Practice #3. Available from http://tiee.esa.org/vol/v7/issues/data_sets/conard/abstract.html

Donovan, T.M., and C.W. Welden. 2002. Spreadsheet exercises in ecology and evolution.
Sinauer Associates, Sunderland, MA.

Gotelli, N.J. 2001. A primer of ecology. Sinauer Associates, Sunderland, MA.

Stevens, M.H.H. 2009. A primer of ecology with R. Springer Science+Business Media, New
York, NY.

Stubben, C.J., and B.G. Milligan. 2007. Estimating and analyzing demographic models using the
popbio package in R. Journal of Statistical Software 22(11).