Integral projection models
Continuous variable determines
Survival
Growth
Reproduction
Easterling, Ellner and Dixon, 2000. Size-specific
elasticity: applying a new structured population
model. Ecology 81:694-708.
The state of the population
0.0 2.5
frequency
Stable distributions for n=50 and n=100
0
1
2
3
4
5
6
alue
size
Relative reproductive value for n=50 and n=1
Integral Projection Model
n( y, t 1) [ p( x, y ) f(x,y)] n( x, t )dx
Integrate over all possible sizes
Number of
size y individuals
at time t+1
Number of
size x individuals
at time t
=
Babies of size y
made by size x individuals
Probability size x individuals
Will survive and become size y individuals
Integral Projection Model
n( y, t 1) [k ( x, y )] n( x, t )dx
Integrate over all possible sizes
Number of
size y individuals
at time t+1
=
Number of
size x individuals
at time t
The kernel
(a non-negative surface representing
All possible transitions from size x to size y)
survival and growth functions
p( x, y) s( x) g ( x, y)
s(x) is the probability that size x individual
survives
g(x,y) is the probability
that size x individuals who survive
grow to size y
survival
s(x) is the probability that size x individual
survives
logistic regression
check for nonlinearity
log(s( x) / 1 s( x)) a bx
growth function
g(x,y) is the probability
that size x individuals who survive
grow to size y
mean
regression
check for nonlinearity
variance
growth function
0.35
probability density
0.3
0.25
0.2
0.15
0.1
0.05
0
-4
-2
0
2
4
6
8
size y, at time t+1
g ( x, y)
1
2 ( x)
e
( y ( x )) 2 / 2 ( x ) 2
Comparison to
Matrix Projection Model
Matrix Projection Model
Populations are structured
Discrete time model
Population divided into
discrete stages
Parameters are estimated
for each cell of the matrix:
many parameters needed
Parameters estimated by
counts of transitions
Integral Projection Model
• Populations are structured
• Discrete time model
• Population characterized by a
continuous distribution
• Parameters are estimated
statistically for relationships:
few parameters are needed
• Parameters estimated by
regression analysis
Comparison to
Matrix Projection Model
Matrix Projection Model
Integral Projection Model
Recruitment usually to a • Recruitment usually to
more than one stage
single stage
• Construction from
Construction from
combining
observed counts
• survival, growth and
fertility functions
into one integral kernel
Asymptotic growth rate
• Asymptotic growth rate
and structure
and structure
Comparison to
Matrix Projection Model
Matrix Projection Model
Analysis by matrix
methods
Integral Projection Model
• Analysis by numerical
integration of the kernel
• In practice: make a big
matrix with small
category ranges
• Analysis then by matrix
methods
Steps
read in the data
statistically fit the model components
combine the components to compute the
kernel
construct the "big matrix“
analyze the matrix
draw the surfaces