mpmaggregate: An Introduction to Aggregating Matrix Population Models

Overview

Matrix population models (MPMs) are widely used to describe population dynamics, and large demographic databases such as COMPADRE and COMADRE now make it possible to study life-history patterns across a wide range of taxa. However, these models often differ substantially in their structure: they may be constructed with different projection intervals, life-cycle stage definitions, and levels of complexity. Such differences can make models difficult to compare directly, even when they describe similar populations. Aggregation methods address this problem by combining stages while preserving key demographic properties of the original model. The mpmaggregate package provides tools for reducing the dimensionality of MPMs under different demographic frameworks, allowing finer-scale models to be compared with coarser models. For example, a 12-stage life-cycle model can be reduced to a simpler 3-stage model that still captures the essential demographic behavior of the population.

Demographic frameworks: $\lambda$ and $R_0$

Aggregation methods can be defined by the demographic quantities they are designed to preserve. In matrix population models, two commonly used measures of growth are the population growth rate $\lambda$ and the net reproductive rate $R_0$. These quantities describe different aspects of population dynamics, and therefore preserving them leads to different frameworks for aggregation.

The quantity $\lambda$ is the dominant eigenvalue of the population projection matrix and describes stable population growth per time step (Caswell, 2001). The quantity $R_0$ measures expected lifetime reproductive output and is interpreted from a cohort-based perspective (Caswell, 2011). In MPMs, $R_0$ is calculated from the next-generation matrix $F(I-U)^{-1}$ (Diekmann et al., 1990; Caswell, 2001).

Although $\lambda$ and $R_0$ share the same population growth threshold, with $\lambda > 1$ if and only if $R_0 > 1$, they summarize different aspects of demography: $\lambda$ emphasizes population trajectories through time, whereas $R_0$ emphasizes lifetime reproductive pathways through the life cycle (Cushing & Yicang, 1994; de-Camino-Beck & Lewis, 2007). The $\lambda$ framework describes population growth through repeated time-step projection of the entire population, so reproduction is incorporated into future population states and therefore compounds over time. By contrast, the $R_0$ framework follows a single cohort through the life cycle and counts the offspring produced by that cohort, but those offspring are not themselves projected forward. In this sense, the $\lambda$ framework includes compounding reproduction, whereas the $R_0$ framework does not (see Figure 1).

This distinction matters for aggregation, because a method that preserves demographic properties in one framework will not generally preserve those same properties in the other. Aggregation in the $\lambda$ framework is appropriate when the goal is to preserve stable population growth and related population-based demographic quantities, whereas aggregation in the $R_0$ framework is appropriate when the goal is to preserve lifetime reproductive output and related cohort-based demographic quantities.

Figure 1. Interpretation of λ and R0 frameworks

Aggregation criteria: standard and elasticity-consistent

Within each framework, users may choose between standard aggregation, which maintains consistency of the stable population growth rate and stable stage distribution (Hooley, 2000; Salguero-Gómez & Plotkin, 2010), and elasticity-consistent aggregation, which additionally maintains consistency in reproductive values and elasticities (Bienvenu et al., 2017). These aggregators are applied in two forms: general-to-general MPM aggregation and Leslie-to-Leslie MPM aggregation (Hinrichsen, 2023; Hinrichsen et al., 2026).

When applying general-to-general MPM aggregation, the projection interval of the aggregated matrix does not change. But when applying Leslie-to-Leslie MPM aggregation, the projection interval changes because, in a Leslie MPM, the projection interval is equal to the age-class width, which increases with the level of aggregation (Hinrichsen et al., 2026).

Package functions

Two primary functions are provided. The function mpm_aggregate() aggregates general stage-structured MPMs (Caswell, 2001) to coarser stage classifications, whereas leslie_aggregate() is designed specifically for age-structured Leslie matrices (Leslie, 1945) and returns a lower-dimensional Leslie matrix. In both cases, aggregation is accompanied by an effectiveness metric that quantifies the degree of agreement between the aggregated and original models. Values close to 1 indicate minimal aggregation error and strong agreement, whereas smaller values reflect increasing approximation.

Together, these tools make it possible to simplify complex matrix models, compare models constructed at different resolutions, and explore how demographic conclusions depend on the level of aggregation.

Loading packages

We begin by loading the mpmaggregate package, which provides the functions used throughout this vignette for aggregating matrix population models under the $\lambda$ and $R_0$ demographic frameworks. The package expm is used to raise a matrix to a power, knitr is used for report generation, and kableExtra is used for table formatting.

library(mpmaggregate)
library(expm)
library(knitr)
library(kableExtra)

Aggregating general-to-general MPM

**Table 1.** Demographic quantities that remain consistent under `mpm_aggregate()` in the λ and R₀ frameworks for both the standard and elasticity-consistent aggregation criteria. Under the λ framework, aggregation maintains consistency in stable growth properties, whereas under the R₀ framework it maintains consistency in cohort-based reproductive quantities. Interstage flow weights the columns of A by their respective stable stage densities (Yokomizo et al. 2024). Here, “cohort” denotes the R₀-framework analogues of the corresponding λ-framework demographic quantities. Notice the symmetry between the λ and R₀ frameworks.
Framework	Criterion	Consistent Demographic Quantities
lambda	standard	λ (stable growth rate) Stable stage distribution Interstage flow matrix
lambda	elasticity-consistent	λ (stable growth rate) Stable stage distribution Reproductive values Elasticities of λ Generation time (T_a)
R0	standard	R₀ (net reproductive rate) Cohort stable stage distribution Cohort interstage flow matrix
R0	elasticity-consistent	R₀ (net reproductive rate) Cohort stable stage distribution Cohort reproductive values Elasticities of R₀ Cohort generation time (T_c)
Note. `mpm_aggregate()` retains the projection interval of the original matrix population model.

To illustrate general-to-general MPM aggregation, we begin with a simple 3x3 stage structured $\mathbf{A}$ COMPADRE MatrixID = 247940 (Salguero‐Gómez et al., 2015), for a Marula (Sclerocarya birrea) population, with a seedling, juvenile, and adult stage (Emanuel et al., 2005). Here, the projection interval is 1 year, and is unchanged by general-to-general MPM aggregation.

We define a grouping of stages using the object groups, which specifies how the original stages are combined into broader classes. In this example, the first stage is left unchanged, while the second and third stages are aggregated into a single group. Internally, this grouping is converted to a partition matrix using mpm_partition(), which is then used to construct the aggregated model.

# Example matrices used in the general-to-general MPM aggregation examples below
#population projection matrix for Sclerocarya   birrea
#MatrixID = 247940

# Not run: requires Rcompadre and internet access

#these matrices can be retrieved with:
#library(Rcompadre)
#compadre <- cdb_fetch("compadre")
#mpm <- compadre[compadre$MatrixID == 247940, ]
#A <- matA(mpm)[[1]]
#U <- matU(mpm)[[1]]
#F <- matF(mpm)[[1]]


A <- matrix(
  c(0.0, 0.0000, 12.46,
    0.4, 0.9160,  0.00,
    0.0, 0.0371,  0.99),
  nrow = 3,
  byrow = TRUE
)
#survival/transitions matrix
U <- matrix(
  c(0.0, 0.0000,  0.00,
    0.4, 0.9160,  0.00,
    0.0, 0.0371,  0.99),
  nrow = 3,
  byrow = TRUE
)
#reproduction matrix
F <- matrix(
  c(0.0, 0.0000, 12.46,
    0.0, 0.0000,  0.00,
    0.0, 0.0000,  0.00),
  nrow = 3,
  byrow = TRUE
)

# Stage aggregation used in later examples:
# group vegetative + reproductive stages together,
# leaving the seedling stage separate
groups <- list(
  c(1),
  c(2,3)
)

The object groups defines how stages are aggregated. Internally, this is converted to a partition matrix using mpm_partition().

Lambda framework: standard aggregation

We first aggregate in the $\lambda$ framework using standard aggregation (Hooley, 2000; Salguero-Gomez & Plotkin, 2010). This option is designed to maintain consistency in the dominant eigenvalue ($\lambda$) and the corresponding stable stage distribution, while producing a lower-dimensional matrix consistent with the grouping defined in groups. Consistency of interstage flows is also maintained, and this condition can be used to derive the standard aggregator (Hinrichsen et al., 2026).

The call below returns (i) the aggregated projection matrix and (ii) an effectiveness value $\rho^2$, which summarizes how closely the aggregated model reproduces the dynamics of the original model. Values of $\rho^2$ close to 1 indicate near-perfect agreement.


# General-to-general MPM aggregation
# in the lambda framework, standard aggregation
res_lambda_std <- mpm_aggregate(
  matA = A,
  groups = groups,
  framework = "lambda",
  criterion = "standard"
)

# Aggregated projection matrix
res_lambda_std$matA_agg
#>      [,1]      [,2]
#> [1,]  0.0 1.2324547
#> [2,]  0.4 0.9567499

# Effectiveness of aggregation
res_lambda_std$effectiveness
#> [1] 0.1572928

# ---- Check: lambda consistency ----

# Expected
spectral_radius(A)
#> [1] 1.327977

# Lambda of aggregated model
spectral_radius(res_lambda_std$matA_agg)
#> [1] 1.327977

# ---- Check: stable stage distribution consistency ----

# P is the partitioning matrix that groups stages
P <- mpm_partition(groups=groups,n=3)

# w is the stable stage distribution
w <- stable_stage(A=A)

# Expected
c(P%*%w)
#> [1] 0.4813464 0.5186536

# stable stage distribution of aggregated model
w_agg <-stable_stage(res_lambda_std$matA_agg)
w_agg
#> [1] 0.4813464 0.5186536

# ---- Check: interstage flows consistency ----

# Expected
P%*%A%*%diag(w)%*%t(P)
#>           [,1]      [,2]
#> [1,] 0.0000000 0.6392171
#> [2,] 0.1925385 0.4962218

# Interstage flows of aggregated model
res_lambda_std$matA_agg%*%diag(w_agg)
#>           [,1]      [,2]
#> [1,] 0.0000000 0.6392171
#> [2,] 0.1925385 0.4962218

Lambda framework: elasticity-consistent aggregation

Elasticity-consistent aggregation additionally maintains consistency of reproductive values. This approach can produce survival probabilities greater than 1 in the aggregated matrix (Bienvenu et al. 2017), a behavior that does not occur under standard aggregation. Here, consistency of interstage flows is sacrificed in order to maintain consistency of the elasticities of ($\lambda$) with respect to matrix entries.


# General-to-general MPM aggregation
# in the lambda framework, elasticity-consistent aggregation
res_lambda_el <- mpm_aggregate(
  matA = A,
  matF = F,
  matU = U,
  groups = groups,
  framework = "lambda",
  criterion = "elasticity"
)

# Aggregated matrix
res_lambda_el$matA_agg
#>           [,1]     [,2]
#> [1,] 0.0000000 1.232455
#> [2,] 0.2000533 1.142314

# Effectiveness of aggregation
res_lambda_el$effectiveness
#> [1] 0.965565

# ---- Check: lambda consistency ----

# Expected
spectral_radius(A)
#> [1] 1.327977

# Lambda of aggregated model
spectral_radius(res_lambda_el$matA_agg)
#> [1] 1.327977

# ---- Check: stable stage distribution consistency ----

# P is the partitioning matrix
P <- mpm_partition(groups=groups,n=3)

# w us the stable stage distribution 
w <- stable_stage(A=A)

# Expected
c(P%*%w)
#> [1] 0.4813464 0.5186536

# Stable stage distribution of aggregated model
stable_stage(res_lambda_el$matA_agg)
#> [1] 0.4813464 0.5186536

# ---- Check: reproductive values consistency ----

# v is reproductive values (dominant left eigenvector)
v = reproductive_values(A)

# Expected
c(solve(P%*%diag(w)%*%t(P))%*%P%*%(w*v))
#> [1] 0.2548271 1.6915719

# Reproductive values of aggregated model
reproductive_values(res_lambda_el$matA_agg)
#> [1] 0.2548271 1.6915719

# ---- Check: elasticities consistency ----

# Elasticity matrix
EA <- mpm_elasticity(matA=A,framework="lambda")$elasticity

# Expected
P%*%EA%*%t(P)
#>           [,1]      [,2]
#> [1,] 0.0000000 0.1226601
#> [2,] 0.1226601 0.7546798

# Elasticities of aggregated model
mpm_elasticity(matA=res_lambda_el$matA_agg,framework="lambda")$elasticity
#>           [,1]      [,2]
#> [1,] 0.0000000 0.1226601
#> [2,] 0.1226601 0.7546798

# ---- Check: generation time consistency ----

# Expected
generation_time(matF=F,matU=U,framework="lambda")$generation_time
#> [1] 8.152612

# Generation time from aggregated model
generation_time(matF=res_lambda_el$matF_agg,
                matU=res_lambda_el$matU_agg,
                framework="lambda")$generation_time
#> [1] 8.152612

Notice that the [2,1] entry, which is survival probability of the aggregated matrix res_lambda_el$matA_agg exceeds 1, which is a known difficulty with the elasticity-consistent aggregator (Bienvenu et al. 2017).

$R_0$framework: decomposition of $\mathbf{A}$

We now illustrate aggregation under the $R_0$ framework, which is based on the decomposition of the population projection matrix into survival and reproduction components. Rather than working directly with the full projection matrix $\mathbf{A}$, aggregation in this framework is performed using the survival matrix $\mathbf{U}$ and the fertility matrix $\mathbf{F}$.

For this example, we use the same life cycle as above, with $\mathbf{U}$ describing survival and transitions among stages and $\mathbf{F}$ describing reproduction. Together, these matrices satisfy $\mathbf{A} = \mathbf{U} + \mathbf{F}$ .

$R_0$ framework: standard aggregation

We first apply standard aggregation in the $R_0$ framework. In this setting, aggregation is carried out using the survival and reproduction components of the model and is designed to ensure consistency of the net reproductive rate ($R_0$) and its associated cohort stable stage distribution. Also consistent is the cohort interstage flow matrix which weights the columns of $\mathbf{A}$ by their associated cohort stable stage densities.

As before, the function returns the aggregated projection matrix and an effectiveness value that summarizes how closely the aggregated model reproduces the demographic behavior of the original model.


# General-to-general MPM aggregation
# in the R0 framework, standard aggregation
res_R0_std <- mpm_aggregate(
  matU = U,
  matF = F,
  groups = groups,
  framework = "R0",
  criterion = "standard"
)

# Aggregated projection matrix
res_R0_std$matA_agg
#>      [,1]      [,2]
#> [1,]  0.0 9.8145648
#> [2,]  0.4 0.9821656

# Effectiveness of aggregation
res_R0_std$effectiveness
#> [1] 0.7893582

# ---- Check: R0 consistency ----

# Expected
spectral_radius(F%*%solve(diag(3)-U))
#> [1] 220.1267

# R0 from aggregated matrix
spectral_radius(res_R0_std$matF_agg%*%solve(diag(2)-res_R0_std$matU_agg))
#> [1] 220.1267

# ---- Check: cohort stable stage distribution consistency ----

# w is the cohort stable stage distribution from a next generation matrix
w <- stable_stage(solve(diag(3)-U)%*%F)

# Expected for consistency
c(P%*%stable_stage(solve(diag(3)-U)%*%F))
#> [1] 0.04268293 0.95731707

# Aggregated cohort stable stage distribution
w_agg <- stable_stage(solve(diag(2)-res_R0_std$matU_agg)%*%res_R0_std$matF_agg)
w_agg
#> [1] 0.04268293 0.95731707

# ---- Check: cohort interstage flows  consistency ----

# Expected
P%*%A%*%diag(w)%*%t(P)
#>            [,1]      [,2]
#> [1,] 0.00000000 9.3956504
#> [2,] 0.01707317 0.9402439

# Aggregated interstage flows
res_R0_std$matA_agg%*%diag(w_agg)
#>            [,1]      [,2]
#> [1,] 0.00000000 9.3956504
#> [2,] 0.01707317 0.9402439

$R_0$ framework: elasticity-consistent aggregation

Elasticity-consistent aggregation in the $R_0$ framework additionally ensures consistent cohort reproductive values and the elasticities of $R_0$ with respect to matrix entries. As in the $\lambda$ framework, this approach can produce survival probabilities greater than 1 in the aggregated matrix. This behavior reflects the constraints imposed by elasticity preservation and does not occur under standard aggregation or in Leslie-to-Leslie aggregation.


# General-to-general MPM aggregation
# in the R0 framework, elasticity-consistent aggregation
res_R0_el <- mpm_aggregate(
  matU = U,
  matF = F,
  groups = groups,
  framework = "R0",
  criterion = "elasticity"
)

# Aggregated projection matrix
res_R0_el$matA_agg
#>           [,1]      [,2]
#> [1,] 0.0000000 9.8145648
#> [2,] 0.2004255 0.9910638

# Effectiveness of aggregation
res_R0_el$effectiveness
#> [1] 0.9134305

# ---- Check: R0 consistency ----

# Expected R0
R0 <- spectral_radius(F%*%solve(diag(3)-U))
R0
#> [1] 220.1267

# R0 from the aggregated model
spectral_radius(res_R0_el$matF_agg%*%solve(diag(2)-res_R0_el$matU_agg))
#> [1] 220.1267

# ---- Check: cohort stable stage distribution consistency ----

# Get partitioning matrix
P <- mpm_partition(groups=groups,n=3)

# Construct cohort projection matrix
Aref = F + U*R0

# Cohort stable stage distribution
w <- stable_stage(A=Aref)

# Expected for consistency
c(P%*%w)
#> [1] 0.04268293 0.95731707

# Aggregated cohort projection matrix
Aref_agg = res_R0_el$matF_agg+ R0*res_R0_el$matU_agg

# Aggregated cohort stable stage distribution
stable_stage(Aref_agg)
#> [1] 0.04268293 0.95731707

# ---- Check: cohort reproductive values consistency ----
v = reproductive_values(Aref)

# Expected
c(solve(P%*%diag(w)%*%t(P))%*%P%*%(w*v))
#> [1] 0.2075074 1.0353341

# Cohort reproductive values for aggregated matrix
reproductive_values(Aref_agg)
#> [1] 0.2075074 1.0353341

# ---- Check: elasticities of R0 consistency ----
# Here elasticities are normalized to sum to 1

# Original elasticities
EA <- mpm_elasticity(matF=F,matU=U,framework="R0")$elasticity

# Expected for consistency
P%*%EA%*%t(P)
#>             [,1]        [,2]
#> [1,] 0.000000000 0.008857022
#> [2,] 0.008857022 0.982285955

# Elasticities of R0 from aggregated matrix
mpm_elasticity(matF=res_R0_el$matF_agg,matU=res_R0_el$matU_agg,framework="R0")$elasticity
#>             [,1]        [,2]
#> [1,] 0.000000000 0.008857022
#> [2,] 0.008857022 0.982285955

# ---- Check: cohort generation time consistency ----

# Expected (from original model)
generation_time(matF=F,matU=U,framework="R0")$generation_time
#> [1] 112.9048

# Cohort generation time from aggregated model
generation_time(matF = res_R0_el$matF_agg, 
                matU = res_R0_el$matU_agg,
                framework="R0")$generation_time
#> [1] 112.9048

Aggregating Leslie-to-Leslie MPM

**Table 2.** Summary of the demographic properties that remain consistent under `leslie_aggregate()` when reducing a Leslie matrix from dimensionality n to m. The quantities the remain consistent depend on both the demographic framework (long-term growth under λ or lifetime reproductive output under R₀) and the aggregation criterion (standard versus elasticity-consistent). The integer k = n/m defines the temporal rescaling by aggregating a Leslie model. Quantities in the λ framework are defined for A^k, while cohort quantities in the R₀ framework are defined for A_k^†. Notice the structural symmetry between the λ and R₀ frameworks.
Framework	Criterion	Consistent Demographic Quantities
lambda	standard	λ (stable growth rate) Stable age distribution Interstage flow matrix of A^k
lambda	elasticity-consistent	λ (stable growth rate) Stable age distribution Reproductive values Elasticities of λ^k with respect to the entries of A^k Generation time (T_a) of A^k
R0	standard	R₀ (net reproductive rate) Cohort stable age distribution consistent of A_k
R0	elasticity-consistent	R₀ (net reproductive rate) Cohort stable age distribution Cohort reproductive values Elasticities of R₀ with respect to A_k Cohort generation time (T_c) for A_k
Note. Aggregation with `leslie_aggregate()` changes the projection interval from Δt to kΔt. ^† A_k denotes the k-step-ahead projection matrix in the R₀ framework.

In some applications, it is desirable to aggregate an age-structured Leslie matrix while retaining its Leslie form (Hinrichsen 2023; Hinrichsen et al., 2026). The function leslie_aggregate() is designed for this purpose. It aggregates a Leslie matrix with fine-grained age classes into a coarser Leslie matrix of dimensionality ngroups.

To illustrate Leslie-to-Leslie MPM aggregation, we begin with a simple 4x4 Leslie matrix $\mathbf{A}$ COMADRE MatrixID = 248187 (Salguero‐Gómez et al., 2016), for a fathead minnow (Pimephales promelas) population (Spromberg & Birge, 2005).

# New example used for Leslie-to-Leslie aggregation
# These matrices replace the earlier example matrices (A, U, F)
# Leslie population projection matrix for Pimephales promelas (MatrixID = 248187)

# Not run: requires Rcompadre and internet access

# Matrices can be retrieved with:
# library(Rcompadre)
# comadre <- cdb_fetch("comadre",version = "4.23.3.1")
# mpm <- comadre[comadre$MatrixID == 248187, ]
# A <- matA(mpm)[[1]]
# U <- matU(mpm)[[1]]
# F <- matF(mpm)[[1]]



# Leslie population projection matrix
A <- rbind(
  c(0.00, 0.00,  6.00, 62.50),
  c(0.50, 0.00,  0.00,  0.00),
  c(0.00, 0.18,  0.00,  0.00),
  c(0.00, 0.00,  0.12,  0.00)
)

# Reproduction matrix
F <- rbind(
  c(0.00, 0.00,  6.00, 62.50),
  c(0.00, 0.00,  0.00,  0.00),
  c(0.00, 0.00,  0.00,  0.00),
  c(0.00, 0.00,  0.00,  0.00)
)

# Survival/transition matrix
U <- rbind(
  c(0.00, 0.00,  0.00,  0.00),
  c(0.50, 0.00,  0.00,  0.00),
  c(0.00, 0.18,  0.00,  0.00),
  c(0.00, 0.00,  0.12,  0.00)
)

# Partitioning matrix used for the aggregation examples below
# (first two age classes combined, last two combined)
# In practice this matrix is constructed automatically inside leslie_aggregate()
# It is reproduced here only to verify the aggregation results.
P <- rbind(
  c(1,1,0,0),
  c(0,0,1,1)
)

# Dimensions used in the aggregation examples
n <- 4     # dimensionality of the original matrix A
m <- 2     # dimensionality of the aggregated matrix
k <- n / m #(also constructed automatically)

Lambda framework: standard aggregation

We first aggregate the Leslie matrix in the $\lambda$ framework using standard aggregation (Hinrichsen, 2023). As in the general-to-general case, this approach ensures consistency of the stable population growth rate ($\lambda$) and the stable age distribution, while producing a Leslie matrix of reduced dimensionality. The interstage flow of the aggregated matrix is consistent with that of $\mathbf{A}^{k}$.


# Leslie-to-Leslie MPM aggregation
# in the lambda framework, standard aggregation
res_leslie_lambda_std <- leslie_aggregate(
  matA = A,
  ngroups = 2,
  framework = "lambda",
  criterion = "standard"
)


# Aggregated Leslie matrix
res_leslie_lambda_std$matAk_agg
#>            [,1]     [,2]
#> [1,] 0.34695224 12.61661
#> [2,] 0.06802636  0.00000

# Effectiveness of aggregation
res_leslie_lambda_std$effectiveness
#> [1] 0.7870295

# ---- Check: lambda consistency ----

# Expected lambda
spectral_radius(A)^k
#> [1] 1.116002

# Lambda of aggregated model
spectral_radius(res_leslie_lambda_std$matAk_agg)
#> [1] 1.116002

# ---- Check: stable age distribution consistency ----

# Stable age distribution
w <- leslie_stable_age(A=A)

# Expected 
c(P%*%w)
#> [1] 0.9425467 0.0574533

# Aggregated stable age distribution
w_agg <-leslie_stable_age(res_leslie_lambda_std$matAk_agg)
w_agg
#> [1] 0.9425467 0.0574533

# ---- Check: interstage flow consistency ----

# Expected
P%*%(A%^%k)%*%diag(w)%*%t(P)
#>            [,1]      [,2]
#> [1,] 0.32701869 0.7248658
#> [2,] 0.06411802 0.0000000

# Interstage flows from aggregated model
res_leslie_lambda_std$matAk_agg%*%diag(w_agg)
#>            [,1]      [,2]
#> [1,] 0.32701869 0.7248658
#> [2,] 0.06411802 0.0000000

Note. In this case, the dimensionality of the aggregated matrix ($m = 2$) divides the dimensionality of $\mathbf{A}$ ($n = 4$) evenly. More generally, when these dimensionalities are not compatible, a disaggregation step is used internally to expand $\mathbf{L}$ to a compatible dimensionality prior to aggregation (Hinrichsen, 2023).

Lambda framework: elasticity-consistent aggregation

We next apply elasticity-consistent aggregation to the Leslie matrix in the $\lambda$ framework (Hinrichsen et al., 2026). In addition to yielding consistent stable population growth rate and stable age distribution, this approach also yields consistent reproductive values and the elasticities of $\lambda^{k}$ with respect to the entries entries of $\mathbf{A}^{k}$. Unlike elasticity-consistent aggregation for general stage-structured MPMs, Leslie-to-Leslie elasticity-consistent aggregation does not produce survival probabilities greater than 1.


# Leslie-to-Leslie MPM aggregation
# in the lambda framework, elasticity-consistent aggregation
res_leslie_lambda_el <- leslie_aggregate(
  matA = A,
  ngroups = 2,
  framework = "lambda",
  criterion = "elasticity"
)

# Aggregated Leslie matrix
res_leslie_lambda_el$matAk_agg
#>            [,1]     [,2]
#> [1,] 0.25558254 14.11557
#> [2,] 0.06802636  0.00000

# Effectiveness of aggregation
res_leslie_lambda_el$effectiveness
#> [1] 0.9311134

# get aggregated F and U
F_agg <-matrix(0,2,2)
F_agg[1,] <- res_leslie_lambda_el$matAk_agg[1,]
U_agg <- res_leslie_lambda_el$matAk_agg
U_agg[1,] <-0

# ---- Check: lambda consistency ----

# Expected
spectral_radius(A)^k
#> [1] 1.116002

# Lambda for aggregated model
spectral_radius(res_leslie_lambda_el$matAk_agg)
#> [1] 1.116002

# ---- Check: stable age distribution consistency ----

# Stable age distribution
w <- leslie_stable_age(A=A)

# Expected
c(P%*%w)
#> [1] 0.9425467 0.0574533

# Stable age distribution of aggregated odel
w_agg <-leslie_stable_age(res_leslie_lambda_el$matAk_agg)
w_agg
#> [1] 0.9425467 0.0574533

# ---- Check: reproductive values consistency ----

# Get reproductive values
v <- leslie_reproductive_values(A=A)

# Expected
c(solve(P%*%diag(w)%*%t(P))%*%P%*%(v*w))
#> [1] 0.5990768 7.5773220

# Reproductive values of aggregated model
leslie_reproductive_values(A=res_leslie_lambda_el$matAk_agg)
#> [1] 0.5990768 7.5773220

# ---- Check: elasticities consistency ----

# Get elasticities of lambda^k
EA <- mpm_elasticity(A%^%k,framework="lambda")$elasticity

# Expected
P%*%EA%*%t(P)
#>           [,1]      [,2]
#> [1,] 0.1293158 0.4353421
#> [2,] 0.4353421 0.0000000

# Elasticities of lambda^k for aggregated model
mpm_elasticity(res_leslie_lambda_el$matAk_agg,framework="lambda")$elasticity
#>           [,1]      [,2]
#> [1,] 0.1293158 0.4353421
#> [2,] 0.4353421 0.0000000

# ---- Check: generation time consistency ----

# Survival/transition matrix
U_k <- U%^%k

# Reproduction matrix
F_k <- A%^%k - U_k

# Expected generation time
generation_time(matF=F_k,matU=U_k,framework="lambda")$generation_time
#> [1] 1.770984

# Generation time of aggregated model
generation_time(matF=F_agg,matU=U_agg,framework="lambda")$generation_time
#> [1] 1.770984

$R_0$ framework

Leslie-to-Leslie aggregation can also be carried out under the $R_0$ framework, emphasizing net reproductive rate rather than stable population growth. As in the general-to-general case, aggregation in this framework is based on the underlying survival and fertility components of the Leslie matrix, but the result again retains Leslie form.

$R_0$ framework: standard aggregation

We first apply standard aggregation in the $R_0$ framework. This approach yields consistent net reproductive rate ($R_0$) and the cohort stable age distribution, while producing a coarser Leslie matrix compatible with the specified number of age groups.


# Leslie-to-Leslie MPM aggregation
# in the R0 framework, standard aggregation
res_leslie_R0_std <- leslie_aggregate(
  matA = A,
  ngroups = 2,
  framework = "R0",
  criterion = "standard"
)

# Aggregated Leslie matrix
res_leslie_R0_std$matAk_agg
#>        [,1]     [,2]
#> [1,] 0.3600 12.72321
#> [2,] 0.0672  0.00000

# Effectiveness of aggregation
res_leslie_R0_std$effectiveness
#> [1] 0.7843297

# get F and U for aggregated matrix
F_agg <- matrix(0,2,2)
F_agg[1,] <- res_leslie_R0_std$matAk_agg[1,]
U_agg <- res_leslie_R0_std$matAk_agg
U_agg[1,] <- 0

# ---- Check: generation time consistency ----

# Expected
R0 <- spectral_radius(F%*%solve(diag(4)-U))
R0 
#> [1] 1.215

# R0 for aggregated model
spectral_radius(F_agg%*%solve(diag(2)-U_agg))
#> [1] 1.215


# ---- Check: cohort stable age distribution consistency ----

# Standardized cohort projection matrix
A_ref <- (F + U*R0)/R0

# Standardized cohort projection matrix for aggregated model
A_ref_agg <- (F_agg + U_agg*R0)/R0

# w is the stable age distribution
w <- leslie_stable_age(A=A_ref)

# Expected
c(P%*%w)
#> [1] 0.93703148 0.06296852

# Stable age distribution for aggregated model
w_agg <-leslie_stable_age(A_ref_agg)
w_agg
#> [1] 0.93703148 0.06296852

# ---- Check: cohort interstage flows consistency ----

# Survival/transition matrix (k steps ahead)
U_k <- U%^%k

# Reproduction matrix (k steps ahead)
F_k <- A%^%k - U_k

# A projection matrix (k steps ahead)
A_k <- F_k + U_k

# Expected
P%*%A_k%*%diag(w)%*%t(P)
#>            [,1]      [,2]
#> [1,] 0.33733133 0.8011619
#> [2,] 0.06296852 0.0000000

# Interstage flows for aggregated model
(F_agg + U_agg)%*%diag(w_agg)
#>            [,1]      [,2]
#> [1,] 0.33733133 0.8011619
#> [2,] 0.06296852 0.0000000

$R_0$ framework: elasticity-consistent aggregation

We now apply elasticity-consistent aggregation in the $R_0$ framework. In addition to providing consistent net reproductive rate ($R_0$) and the cohort stable age distribution, this approach provides consistent cohort reproductive values and the elasticities of $R_0$ with respect to the entries of $\mathbf{A}_k^{*}$, the $k$-step–ahead version of $\mathbf{A}$, where $k = n/m$, $n$ is the dimensionality of the original model, and $m$ is the dimensionality of the aggregated model. As in the $\lambda$ framework, the Leslie structure is maintained and survival probabilities cannot exceed 1.


# Leslie-to-Leslie MPM aggregation
# in the R0 framework, elasticity-consistent aggregation
res_leslie_R0_el <- leslie_aggregate(
  matA = A,
  ngroups = 2,
  framework = "R0",
  criterion = "elasticity"
)

# Aggregated Leslie matrix
res_leslie_R0_el$matAk_agg
#>        [,1]    [,2]
#> [1,] 0.2700 14.0625
#> [2,] 0.0672  0.0000

# Effectiveness of aggregation
res_leslie_R0_el$effectiveness
#> [1] 0.9372361

# Get F and U for aggregated matrix
F_agg <- matrix(0,2,2)
F_agg[1,] <- res_leslie_R0_el$matAk_agg[1,]
U_agg <- res_leslie_R0_el$matAk_agg
U_agg[1,] <- 0

# ---- Check: R0 consistency ----

# Expected
R0 <- spectral_radius(F%*%solve(diag(4)-U))
R0 
#> [1] 1.215

# R0 of aggregated model
spectral_radius(F_agg%*%solve(diag(2)-U_agg))
#> [1] 1.215


# ---- Check: cohort stage age distribution consistency ----

# Standardized cohort projection matrix
A_ref <- (F + U*R0)/R0

# Standardized cohort projection matrix for aggregated model
A_ref_agg <- (F_agg + U_agg*R0)/R0

# Cohort stable age distribution
w <- leslie_stable_age(A=A_ref)

# Expected
c(P%*%w)
#> [1] 0.93703148 0.06296852

# Cohort stage age distribution for aggregated model
w_agg <-leslie_stable_age(A_ref_agg)
w_agg
#> [1] 0.93703148 0.06296852

# ---- Check: cohort reproductive values consistency ----

# Cohort reproductive values
v <- leslie_reproductive_values(A_ref)

# Expected
c(solve(P%*%diag(w)%*%t(P))%*%P%*%(v*w))
#> [1] 0.600300 6.947917

# Cohort reproductive values of aggregated model
leslie_reproductive_values(A_ref_agg)
#> [1] 0.600300 6.947917

# ---- Check: elasticities of R0 consistency ----
# These elasticities are normalized to sum to 1

# Survival/transition matrix (k steps ahead)
U_k <- U%^%k

# Reproduction matrix (k steps ahead)
F_k <- A%^%k - U_k

# Cohort projection matrix (k steps ahead)
A_k <- F_k + U_k

# Elasticity of R0 (normalized to sum to 1)
EA <- mpm_elasticity(matF=F_k,matU=U_k,framework="R0")$elasticity

# Expected
P%*%EA%*%t(P)
#>        [,1]   [,2]
#> [1,] 0.1250 0.4375
#> [2,] 0.4375 0.0000

# Elasticity of R0 for the aggregated model
mpm_elasticity(matF=F_agg,matU=U_agg,framework="R0")$elasticity
#>        [,1]   [,2]
#> [1,] 0.1250 0.4375
#> [2,] 0.4375 0.0000

# ---- Check: cohort generation time consistency ----

# Expected
generation_time(matF=F_k, matU = U_k, framework="R0")$generation_time
#> [1] 1.777778

# Generation time of aggregated model
generation_time(matF=F_agg,matU=U_agg, framework="R0")$generation_time
#> [1] 1.777778

Interpretation

The lambda framework ensures consistency of stable growth and stable structure.
The R0 framework ensures consistency of generation-based reproductive dynamics.
Standard aggregation ensures consistency of the core demographic quantity associated with the chosen framework (either $\lambda$ or $R_0$).
Elasticity-consistent aggregation additionally ensures consistency of reproductive values, associated elasticities, and generation times within their respective frameworks.

References

Bienvenu, F., Akçay, E., Legendre, S., & McCandlish, D. M. (2017). The genealogical decomposition of a matrix population model with applications to the aggregation of stages. Theoretical Population Biology, 115, 69–80. https://doi.org/10.1016/j.tpb.2017.04.002

Caswell, H. (2001). Matrix Population Models: Construction, Analysis, and Interpretation (2nd ed.). Sinauer Associates, Sunderland, MA.

Caswell, H. (2011). Beyond R0: demographic models for variability of lifetime reproductive output. PloS ONE, 6(6), p.e20809. https://doi.org/10.1371/journal.pone.0020809

Cushing, J.M., & Yicang, Z.(1994). The net reproductive value and stability in matrix population models. Natural Resource Modeling, 8(4), 297-333. https://doi.org/10.1111/j.1939-7445.1994.tb00188.x

de-Camino-Beck, T., & Lewis, M.A. (2007). A new method for calculating net reproductive rate from graph reduction with applications to the control of invasive species. Bulletin of Mathematical Biology, 69(4), 1341-1354. https://doi.org/10.1007/s11538-006-9162-0

Diekmann, O., Heesterbeek, J.A.P., & Metz, J.A.J. (1990). On the definition and the computation of the basic reproduction ratio R0 in models for infectious diseases in heterogeneous populations. Journal of mathematical biology, 28(4), 365-382. https://doi.org/10.1007/BF00178324

Emanuel, P. L., Shackleton, C. M., & Baxter, J. S. (2005). Modelling the sustainable harvest of Sclerocarya birrea subsp. caffra fruits in the South African lowveld. Forest Ecology and Management, 214(1–3), 91–103. https://doi.org/10.1016/j.foreco.2005.03.066

Hinrichsen, R. A. (2023). Aggregation of Leslie matrix models with application to ten diverse animal species. Population Ecology, 65(3), 146–166. https://doi.org/10.1002/1438-390X.12149

Hinrichsen, R. A., Yokomizo, H., & Salguero-Gómez, R. (2026). From theory to application: Elasticity-consistent aggregation of Leslie matrix population models for comparative demography. bioRxiv, preprint. https://doi.org/10.64898/2026.02.04.703802

Hooley, D. E. (2000). Collapsed matrices with (almost) the same eigenstuff. The College Mathematics Journal, 31(4), 297–299. https://doi.org/10.1080/07468342.2000.11974162

Leslie, P. H. (1945). On the use of matrices in certain population mathematics. Biometrika, 33(3), 183–212. https://doi.org/10.2307/2332297

Salguero-Gómez, R., Jones, O. R., Archer, C. R., Buckley, Y. M., Che-Castaldo, J., Caswell, H., Hodgson, D., Scheuerlein, A., Conde, D. A., Brinks, E., de Buhr, H., Farack, C., Gottschalk, F., Hartmann, A., Henning, A., Hoppe, G., Römer, G., Runge, J., Ruoff, T., Wille, J., Zeh, S., Davison, R., Vieregg, D., Baudisch, A., Altwegg, R., Colchero, F., Dong, M., de Kroon, H., Lebreton, J.-D., Metcalf, C. J. E., Neel, M. M., Parker, I. M., Takada, T., Valverde, T., Vélez-Espino, L. A., Wardle, G. M., Franco, M., & Vaupel, J. W. (2015). The COMPADRE Plant Matrix Database: an open online repository for plant demography. Journal of Ecology, 103, 202–218. https://doi.org/10.1111/1365-2745.12334

Salguero-Gómez, R., Jones, O. R., Archer, C. R., Bein, C., de Buhr, H., Farack, C., Gottschalk, F., Hartmann, A., Henning, A., Hoppe, G., Römer, G., Ruoff, T., Sommer, V., Wille, J., Voigt, J., Zeh, S., Vieregg, D., Buckley, Y. M., Che-Castaldo, J., Hodgson, D., Scheuerlein, A., Caswell, H., & Vaupel, J. W. (2016). COMADRE: a global database of animal demography. Journal of Animal Ecology, 85, 371–384. https://doi.org/10.1111/1365-2656.12482

Salguero-Gómez, R., & Plotkin, J. B. (2010). Matrix dimensions bias demographic inferences: implications for comparative plant demography. The American Naturalist, 176, 710–722. https://doi.org/10.1086/657044

Spromberg, J. A., & Birge, W. J. (2005). Modeling the effects of chronic toxicity on fish populations: The influence of life-history strategies. Environmental Toxicology and Chemistry, 24(6), 1532–1540. https://doi.org/10.1897/04-160.1

Yokomizo, H., Fukaya, K., Lambrinos, J. G., Kawai, Y., & Takada, T. (2024). Interstage flow matrices: Population statistics derived from matrix population models. Journal of Ecology, 112(6), 1326–1338. https://doi.org/10.1111/1365-2745.14303

mpmaggregate: An Introduction to Aggregating Matrix Population Models

Richard A. Hinrichsen

Overview

Demographic frameworks: \(\lambda\) and \(R_0\)

Aggregation criteria: standard and elasticity-consistent

Package functions

Loading packages

Aggregating general-to-general MPM

Lambda framework: standard aggregation

Lambda framework: elasticity-consistent aggregation

\(R_0\)framework: decomposition of \(\mathbf{A}\)

\(R_0\) framework: standard aggregation

\(R_0\) framework: elasticity-consistent aggregation

Aggregating Leslie-to-Leslie MPM

Lambda framework: standard aggregation

Lambda framework: elasticity-consistent aggregation

\(R_0\) framework

\(R_0\) framework: standard aggregation

\(R_0\) framework: elasticity-consistent aggregation

Interpretation

References