The Central Idea
Population, ecological, and epidemiological models all describe how quantities change through flows between states or through interactions among groups. A general structure is
\[ \boxed{\text{rate of change} =\text{inflows}-\text{outflows}+\text{interactions}}. \]The same modeling language can describe a growing population, predator-prey dynamics, competing species, susceptible and infected people, or the spatial spread of organisms and disease.
State Variables
A state variable records a quantity needed to describe the system at a particular time. Examples include
- total population \(N(t)\);
- prey and predator populations \(X(t)\) and \(Y(t)\);
- susceptible, infected, and recovered populations \(S(t)\), \(I(t)\), and \(R(t)\);
- population density \(u(x,t)\) across space;
- resource concentration, habitat quality, or environmental capacity.
Choosing state variables is one of the most consequential modeling decisions.
Parameters
Parameters summarize mechanisms or rates. Common examples include
- birth and death rates;
- intrinsic growth rate \(r\);
- carrying capacity \(K\);
- interaction coefficients;
- transmission rate \(\beta\);
- recovery rate \(\gamma\);
- migration or diffusion coefficients;
- intervention effectiveness.
Parameters may be measured directly, estimated from data, or treated as uncertain quantities.
Conservation and Accounting
Many biological models are accounting models. For a population \(N(t)\),
\[ \boxed{\frac{dN}{dt} =\text{births}-\text{deaths} +\text{immigration}-\text{emigration}}. \]This balance-law viewpoint helps prevent missing or double-counting flows.
Exponential Population Growth
If the per-capita net growth rate is constant, then
\[ \boxed{\frac{dN}{dt}=rN}. \]The solution is
\[ \boxed{N(t)=N_0e^{rt}}. \]When \(r>0\), the population grows exponentially. When \(r<0\), it declines exponentially.
Per-Capita Growth Rate
Dividing the exponential model by \(N\) gives
\[ \boxed{\frac{1}{N}\frac{dN}{dt}=r}. \]Thus, exponential growth assumes a constant proportional growth rate rather than a constant absolute increase.
Doubling Time
For exponential growth, the doubling time satisfies
\[ 2N_0=N_0e^{rT_d}. \]Therefore,
\[ \boxed{T_d=\frac{\ln 2}{r}}. \]Doubling time translates an abstract rate parameter into a more intuitive time scale.
Limits of Exponential Growth
Exponential growth is useful when resource limitation and density effects are negligible over the period being modeled. It becomes unrealistic when crowding, finite habitat, competition, depletion, or other constraints become important.
A model can therefore be locally useful without being valid indefinitely.
Logistic Growth
The logistic model introduces a carrying capacity \(K\):
\[ \boxed{\frac{dN}{dt} =rN\left(1-\frac{N}{K}\right)}. \]When \(N\ll K\), the population grows approximately exponentially. As \(N\) approaches \(K\), growth slows.
Carrying Capacity
The parameter \(K\) represents the population level at which the logistic model predicts zero net growth:
\[ \frac{dN}{dt}=0 \qquad\text{when}\qquad N=K. \]Carrying capacity should not automatically be interpreted as a permanent fixed biological constant. It can change with habitat, climate, resource supply, technology, or behavior.
The Logistic Solution
For initial population \(N_0>0\), the logistic equation has solution
\[ \boxed{ N(t)= \frac{K} {1+\left(\frac{K-N_0}{N_0}\right)e^{-rt}} }. \]
The curve is sigmoidal when \(0
Maximum Logistic Growth
In the logistic model,
\[ \frac{dN}{dt} =rN\left(1-\frac{N}{K}\right). \]The growth rate is largest at
\[ \boxed{N=\frac{K}{2}}. \]At that population,
\[ \boxed{ \left(\frac{dN}{dt}\right)_{\max} =\frac{rK}{4} }. \]Harvesting Models
Resource populations may be modeled with harvesting. A constant-harvest model is
\[ \boxed{ \frac{dN}{dt} =rN\left(1-\frac{N}{K}\right)-H }. \]The harvest rate \(H\) changes equilibrium structure and can eliminate positive equilibria if it is too large.
Thresholds and Collapse
Biological models often contain thresholds. A parameter change may have little qualitative effect until a critical value is crossed, after which equilibria, stability, or long-term survival change sharply.
These threshold effects are important in conservation, resource management, invasive-species control, and epidemiology.
Age-Structured Populations
Total population alone may be insufficient when birth and survival depend strongly on age or life stage.
A discrete age-structured model can be written
\[ \boxed{\mathbf{n}_{t+1}=L\mathbf{n}_t}, \]where \(\mathbf{n}_t\) contains age-class abundances and \(L\) is a projection matrix containing fertility and survival terms.
Long-Term Growth in Matrix Models
Repeated multiplication gives
\[ \mathbf{n}_t=L^t\mathbf{n}_0. \]Under suitable conditions, the dominant eigenvalue of \(L\) determines the long-run multiplicative growth factor, while the associated eigenvector determines the stable age distribution.
Interacting Populations
Ecological systems often require several state variables because species influence one another.
General two-species models take the form
\[ \boxed{ \frac{dX}{dt}=f(X,Y), \qquad \frac{dY}{dt}=g(X,Y) }. \]The functions \(f\) and \(g\) encode growth, competition, predation, cooperation, or other interactions.
Lotka-Volterra Predator-Prey Model
A classical predator-prey model is
\[ \boxed{ \frac{dX}{dt}=\alpha X-\beta XY } \]and
\[ \boxed{ \frac{dY}{dt}=\delta XY-\gamma Y }. \]Here \(X\) is prey, \(Y\) is predator, and the interaction term \(XY\) models encounters between the two populations.
Why Interaction Terms Often Use Products
If individuals from two groups mix approximately at random, the number of encounters can be modeled as proportional to the product of their population sizes:
\[ \text{encounter rate}\propto XY. \]This same product structure appears in predator-prey models, chemical kinetics, and basic epidemic transmission models.
Competition Models
Two populations competing for shared resources can be modeled by
\[ \frac{dX}{dt} =r_1X \left( 1-\frac{X+\alpha Y}{K_1} \right), \] \[ \frac{dY}{dt} =r_2Y \left( 1-\frac{Y+\beta X}{K_2} \right). \]The coefficients \(\alpha\) and \(\beta\) convert the influence of one species into equivalent competitive pressure on the other.
Mutualism and Facilitation
Not all interactions are harmful. Positive interaction terms can represent mutualism or facilitation, where one population increases the growth or survival of another.
Care is required because unrestricted positive feedback can produce unrealistic unbounded growth unless limiting mechanisms are also included.
Equilibria
An equilibrium occurs when all state variables stop changing:
\[ f(X^*,Y^*)=0, \qquad g(X^*,Y^*)=0. \]Equilibria may represent extinction, persistence of one population, or coexistence of several populations.
Stability
Stability asks what happens after a small disturbance from equilibrium. If the system tends to return, the equilibrium is locally stable. If perturbations grow, it is unstable.
Linearization and eigenvalue analysis provide standard tools for local stability analysis.
Phase-Plane Analysis
For two-variable ecological systems, a phase plane plots one state variable against another rather than plotting both against time.
Nullclines, equilibrium points, trajectory direction, and closed orbits can reveal qualitative behavior that is difficult to see from formulas alone.
Epidemiological Compartments
Compartment models divide a population into epidemiological states. A basic SIR model uses
- \(S(t)\): susceptible;
- \(I(t)\): infectious;
- \(R(t)\): removed or recovered.
The total population is
\[ \boxed{N=S+I+R}. \]The SIR Model
A standard closed-population SIR model is
\[ \boxed{ \frac{dS}{dt} =-\beta\frac{SI}{N} } \] \[ \boxed{ \frac{dI}{dt} =\beta\frac{SI}{N}-\gamma I } \] \[ \boxed{ \frac{dR}{dt} =\gamma I }. \]Transmission moves individuals from susceptible to infectious, while recovery or removal moves them from infectious to removed.
Interpretation of \(\beta\) and \(\gamma\)
The parameter \(\beta\) summarizes transmission opportunity and probability under the model's mixing assumptions. The parameter \(\gamma\) is the per-capita rate of leaving the infectious compartment.
If infectious duration is exponentially distributed with mean duration \(D\), then
\[ \boxed{\gamma=\frac{1}{D}}. \]The Basic Reproduction Number
In the simplest SIR model, the basic reproduction number is
\[ \boxed{R_0=\frac{\beta}{\gamma}}. \]It represents the expected number of secondary infections generated by one typical infectious individual in a fully susceptible population under the assumptions of the model.
The Epidemic Threshold
Early in an outbreak, \(S\approx N\), so
\[ \frac{dI}{dt} \approx(\beta-\gamma)I. \]Infection initially grows when
\[ \boxed{R_0>1}. \]The threshold is a model-based statement, not a universal constant independent of population structure and assumptions.
The Effective Reproduction Number
As susceptibility changes, a simple effective reproduction number is
\[ \boxed{ R_{\text{eff}}(t) =R_0\frac{S(t)}{N} }. \]When \(R_{\text{eff}}<1\), incidence tends to decline in the basic model.
SEIR Models
When infection includes a latent period before infectiousness, an exposed compartment \(E\) may be added:
\[ S\rightarrow E\rightarrow I\rightarrow R. \]A simple SEIR model includes
\[ \frac{dE}{dt} =\beta\frac{SI}{N}-\sigma E, \]where \(1/\sigma\) is the mean time spent in the exposed compartment under the model.
Births, Deaths, and Waning Immunity
Longer-term epidemic models may need demographic turnover, disease mortality, vaccination, loss of immunity, or repeated infection.
For example, an SIRS structure allows
\[ S\rightarrow I\rightarrow R\rightarrow S. \]The correct compartment structure depends on the biological mechanism being represented.
Vaccination Models
Vaccination can be represented in several ways depending on whether the model needs to distinguish reduced susceptibility, reduced infectiousness, reduced severe outcomes, delayed protection, waning protection, or multiple doses.
A model should therefore avoid treating "vaccinated" as a single universal state unless that simplification is adequate for the intended analysis.
Heterogeneous Mixing
The basic SIR model often assumes homogeneous mixing. Real populations are structured by age, household, workplace, geography, behavior, and contact networks.
A structured transmission model may replace one transmission parameter with a contact matrix \(C\), where \(C_{ij}\) represents contact from group \(i\) to group \(j\).
Network Epidemiology
When individual contact structure matters, people or groups may be represented as nodes in a network and potentially infectious contacts as edges.
Degree distribution, clustering, centrality, community structure, and dynamic contact patterns can substantially alter epidemic behavior compared with homogeneous-mixing models.
Stochastic Epidemic Models
Deterministic differential equations describe expected or average behavior. When populations are small or events are rare, randomness may be central.
A stochastic model can assign probabilities or rates to individual infection, recovery, birth, death, or migration events.
Extinction Probability
Even when \(R_0>1\), a newly introduced infection can disappear by chance before a large outbreak develops.
This distinction between expected growth and stochastic extinction is especially important during early outbreak phases or in small populations.
Spatial Population Models
A population density may vary across space:
\[ u=u(\mathbf{x},t). \]Spatial models can include movement, local growth, habitat variation, and boundaries.
Diffusion
Random movement can be approximated by diffusion:
\[ \boxed{ \frac{\partial u}{\partial t} =D\nabla^2u }. \]The diffusion coefficient \(D\) controls the rate of spatial spreading.
Reaction-Diffusion Models
Combining local population change with diffusion gives
\[ \boxed{ \frac{\partial u}{\partial t} =D\nabla^2u+f(u) }. \]Such equations can model invasion fronts, ecological spread, population dispersal, and spatial epidemic dynamics.
Advection and Directed Movement
When movement has a preferred direction, an advection term may be included:
\[ \boxed{ \frac{\partial u}{\partial t} +\mathbf{v}\cdot\nabla u =D\nabla^2u+f(u) }. \]This can represent migration, river transport, prevailing wind, host movement, or directional habitat flow.
Environmental Forcing
Biological rates can depend on temperature, rainfall, season, resource supply, habitat quality, or policy conditions.
A parameter such as \(r\) or \(\beta\) may therefore be modeled as time dependent:
\[ r=r(t), \qquad \beta=\beta(t). \]Time-varying parameters can improve realism but also make inference more difficult.
Seasonality
A simple periodic transmission rate might be written
\[ \boxed{ \beta(t) =\beta_0 \left[ 1+a\cos\left(\frac{2\pi t}{T}\right) \right] }. \]Here \(a\) controls seasonal amplitude and \(T\) the seasonal period.
Delay Models
Biological systems may respond to past rather than current conditions. Birth delays, maturation, incubation, reporting delays, and immune response can introduce time lags.
A delay differential equation may include a term such as
\[ N(t-\tau), \]where \(\tau\) is a delay time.
Discrete-Time Models
Some populations reproduce in distinct seasons or generations. A discrete model may be more natural:
\[ \boxed{N_{t+1}=F(N_t)}. \]The discrete logistic map
\[ \boxed{x_{t+1}=rx_t(1-x_t)} \]can display fixed points, cycles, bifurcations, and chaotic behavior depending on \(r\).
Observation Models
The true biological state is often not observed directly. Reported cases, survey counts, trap counts, test positivity, hospital admissions, or wildlife sightings are measurements generated from an underlying process.
A useful modeling framework separates
\[ \boxed{\text{process model}} \]from
\[ \boxed{\text{observation model}}. \]This separation is critical when detection is incomplete or biased.
Underreporting
Suppose only a fraction \(p\) of true events is observed. A simple expected observation model is
\[ \boxed{Y_t=pX_t}, \]where \(X_t\) is the true count and \(Y_t\) is the expected reported count.
If \(p\) changes through time, interpreting reported trends becomes more difficult.
Parameter Estimation
Parameters can be estimated by comparing model predictions with observed data. For example, one may estimate \(r\), \(K\), \(\beta\), \(\gamma\), or interaction coefficients by minimizing a discrepancy function.
A least-squares objective may be
\[ \boxed{ J(\theta) =\sum_{i=1}^{n} \left[y_i-\hat{y}(t_i;\theta)\right]^2 }. \]Likelihood-Based Estimation
When a probability model for observations is available, parameters may be estimated by maximizing likelihood:
\[ \boxed{ \hat{\theta} =\arg\max_\theta L(\theta\mid y) }. \]Count data may require Poisson, negative-binomial, binomial, or other observation models rather than ordinary least squares.
Bayesian Inference
Bayesian methods combine prior information with observed data:
\[ \boxed{ p(\theta\mid y) \propto p(y\mid\theta)p(\theta) }. \]The posterior distribution represents uncertainty about parameters after incorporating the data and model assumptions.
Identifiability
A model may contain parameters that cannot be uniquely estimated from the available observations.
If different parameter combinations produce nearly identical outputs, the model may be practically non-identifiable even when numerical optimization returns a specific parameter estimate.
Calibration Is Not Validation
Calibration adjusts parameters so the model matches selected data. Validation tests whether the calibrated model performs adequately on relevant observations or conditions not used for fitting.
A model that reproduces historical data can still fail when conditions change.
Out-of-Sample Validation
One validation strategy divides data into calibration and testing periods. Parameters are fitted on the calibration data, and predictions are then compared with observations not used during fitting.
This helps expose overfitting and structural weaknesses.
Sensitivity Analysis
Sensitivity analysis asks how outputs change when parameters or assumptions are varied.
Important outputs might include
- equilibrium population;
- time to extinction;
- epidemic peak size;
- peak timing;
- total infections;
- probability of invasion;
- spatial spread rate.
Uncertainty Analysis
Biological models may contain uncertainty in initial conditions, parameters, contact behavior, environmental forcing, observation processes, and model structure.
A useful model report should distinguish
- measurement uncertainty;
- parameter uncertainty;
- stochastic variability;
- scenario uncertainty;
- model-form uncertainty.
Scenario Analysis
A scenario is not necessarily a forecast. It asks what the model predicts under a specified set of assumptions.
Ecological and epidemiological models are frequently more reliable for comparing scenarios than for predicting a single exact future trajectory.
Intervention Models
Management or policy actions can enter the model as changes in parameters, state transitions, or control variables.
Examples include
- harvest limits;
- habitat restoration;
- vaccination;
- testing and isolation;
- vector control;
- contact reduction;
- targeted treatment.
Control Variables
A time-dependent intervention can be represented by a control variable \(u(t)\). For example,
\[ \beta(t)=\beta_0[1-u(t)]. \]This structure allows optimization questions such as how to reduce total infections or ecological damage while limiting intervention cost.
Optimization in Biological Systems
A management objective might minimize
\[ \boxed{ J = \int_0^T \left[ w_1 I(t)+w_2u(t)^2 \right]dt }. \]The first term penalizes disease burden, while the second penalizes intervention intensity. Changing the weights changes the policy tradeoff.
Dimensionless Variables
Scaling variables can reveal the essential structure of a model. For logistic growth, define
\[ x=\frac{N}{K}, \qquad \tau=rt. \]Then
\[ \boxed{ \frac{dx}{d\tau}=x(1-x) }. \]The dimensional parameters disappear, showing that all logistic systems share a common normalized structure under this scaling.
Time Scales
Characteristic time scales help determine appropriate numerical resolution and interpretation. Examples include
\[ \frac{1}{r}, \qquad \frac{1}{\gamma}, \qquad \frac{1}{\sigma}, \]corresponding to population-growth, infectious, and latent-period scales in suitable models.
Numerical Time-Stepping
When no closed-form solution exists, differential equations can be solved numerically. Forward Euler approximates
\[ \boxed{ x_{n+1} =x_n+h f(t_n,x_n) }. \]The step size \(h\) must be small enough to control numerical error and preserve important qualitative behavior.
Numerical Stability and Positivity
Biological state variables such as populations and compartment counts should usually remain nonnegative.
A poor numerical method or excessively large time step can create negative populations even when the continuous model does not. This is an implementation failure, not a biological prediction.
Agent-Based Models
Agent-based models represent individual organisms, people, households, or other entities explicitly. Each agent may have attributes, location, behavior, and interaction rules.
These models can represent heterogeneity that compartment equations average away, but they require more data, computation, and validation.
When Agent-Based Models Are Useful
Agent-based approaches are particularly useful when
- individual heterogeneity matters;
- contact networks are important;
- local spatial interactions matter;
- behavior changes with experience or policy;
- institutional rules influence outcomes.
Model Hierarchies
The same biological system may be studied with several model levels:
- a simple exponential or SIR model;
- a structured deterministic model;
- a stochastic compartment model;
- a spatial or network model;
- an agent-based simulation.
Complexity should be added only when it improves the model for its intended use.
Example: Exponential Growth
Suppose a population begins at
\[ N_0=500 \]and grows at \(r=0.08\) per year. After \(5\) years,
\[ N(5)=500e^{0.08(5)} \] \[ =500e^{0.4} \] \[ \boxed{N(5)\approx746}. \]Example: Logistic Growth Rate
Let
\[ r=0.20, \qquad K=1000, \qquad N=400. \]Then
\[ \frac{dN}{dt} =0.20(400)\left(1-\frac{400}{1000}\right) \] \[ =80(0.6) \] \[ \boxed{\frac{dN}{dt}=48} \]population units per unit time.
Example: SIR Threshold
Suppose
\[ \beta=0.30\ \text{day}^{-1}, \qquad \gamma=0.10\ \text{day}^{-1}. \]Then
\[ R_0=\frac{\beta}{\gamma} =\frac{0.30}{0.10} =\boxed{3}. \]In a fully susceptible population under the assumptions of this basic model, infection can initially grow because \(R_0>1\).
Example: Effective Reproduction Number
If \(R_0=3\) and the susceptible fraction falls to
\[ \frac{S}{N}=0.25, \]then
\[ R_{\text{eff}} =3(0.25) =\boxed{0.75}. \]In the simple SIR model, this indicates declining infection incidence.
Example: Mean Infectious Duration
If
\[ \gamma=0.20\ \text{day}^{-1}, \]then the model's mean infectious duration is
\[ D=\frac{1}{\gamma} =\frac{1}{0.20} =\boxed{5\ \text{days}}. \]Verification
Verification asks whether the mathematical and computational model has been implemented correctly.
Useful checks include
- population conservation where appropriate;
- nonnegative state variables;
- dimensional consistency;
- comparison with known analytical solutions;
- time-step convergence;
- equilibrium checks;
- independent recalculation of summary outputs.
Validation
Validation asks whether the model represents the real system well enough for its intended purpose.
A model might reproduce total population while failing to reproduce age structure, spatial distribution, epidemic peak timing, or response to an intervention. Validation should therefore focus on the outputs that matter for the decision.
Common Error: Treating Parameters as Universal Constants
Growth rates, carrying capacities, contact rates, transmission rates, and interaction coefficients can change with environment, behavior, season, policy, and population composition.
A fitted parameter may describe a particular setting rather than a permanent biological law.
Common Error: Equating Correlation with Mechanism
A statistical association can suggest a relationship, but a mechanistic model requires a defensible pathway linking variables.
Adding a term because it improves curve fit does not by itself establish that the corresponding biological mechanism is real.
Common Error: Ignoring Observation Bias
Reported cases, survey counts, and ecological observations may reflect changes in testing, reporting, sampling effort, detection probability, or surveillance practice.
A model that treats observed counts as perfect measurements of the hidden state can misattribute measurement changes to biological dynamics.
Common Error: Overfitting
Adding many parameters can improve fit to historical data while degrading interpretability and future performance.
Parameter count should be justified by information content and model purpose, not by fit alone.
Common Error: Confusing Scenarios with Forecasts
A conditional projection such as "if contact rates remain at this level" is a scenario. It should not be presented as an unconditional prediction when the future value of that assumption is unknown.
Common Error: Ignoring Stochasticity
Deterministic models can be misleading when populations are small, events are rare, or extinction and invasion depend strongly on chance.
In those settings, a stochastic model may be necessary to answer the actual question.
A Reliable Modeling Procedure
- Define the biological or public-health question.
- Choose state variables and the population structure to represent.
- Identify births, deaths, transitions, migration, and interactions.
- Translate those mechanisms into equations or simulation rules.
- Define parameters and their units.
- Specify initial conditions, boundaries, and observation processes.
- Estimate or calibrate parameters from data where needed.
- Verify numerical implementation and conservation properties.
- Validate against relevant observations not used for fitting when possible.
- Perform sensitivity, uncertainty, and scenario analysis before drawing decisions from the model.
Practice 1: Exponential Growth
A population begins at \(N_0=800\) and grows at \(r=0.05\) per year. Estimate the population after \(6\) years using the exponential model.
Practice 2: Doubling Time
Find the doubling time for an exponentially growing population with \(r=0.07\) per year.
Practice 3: Logistic Growth Rate
For
\[ r=0.30, \qquad K=1200, \qquad N=300, \]compute \(dN/dt\).
Practice 4: Maximum Logistic Growth
A logistic population has \(r=0.10\) and \(K=2000\). At what population is absolute growth maximal, and what is that maximum growth rate?
Practice 5: Basic Reproduction Number
In a simple SIR model,
\[ \beta=0.24\ \text{day}^{-1}, \qquad \gamma=0.08\ \text{day}^{-1}. \]Find \(R_0\).
Practice 6: Mean Infectious Duration
If \(\gamma=0.125\ \text{day}^{-1}\), find the mean infectious duration under the exponential-removal assumption of the basic model.
Practice 7: Effective Reproduction Number
If \(R_0=2.4\) and \(S/N=0.30\), find \(R_{\text{eff}}\).
Practice 8: Interaction Term
In a predator-prey model, explain why a term proportional to \(XY\) can represent predator-prey encounters.
Practice 9: Observation Bias
A surveillance system reports only \(40\%\) of true cases on average. If the expected true count is \(250\), what reported count does the simple observation model predict?
Practice 10: Choose the Better Model
A wildlife population is distributed across several isolated habitat patches, and migration between patches strongly affects persistence. Explain why a single logistic equation for total population may be inadequate and identify a more appropriate modeling direction.
Practice Answers
Practice 1
\[ N(6)=800e^{0.05(6)} =800e^{0.3} \] \[ \boxed{N(6)\approx1080}. \]Practice 2
\[ T_d=\frac{\ln2}{0.07} \] \[ \boxed{T_d\approx9.90\ \text{years}}. \]Practice 3
\[ \frac{dN}{dt} =0.30(300) \left(1-\frac{300}{1200}\right) \] \[ =90(0.75) \] \[ \boxed{\frac{dN}{dt}=67.5}. \]Practice 4
Maximum logistic growth occurs at
\[ N=\frac{K}{2} =\boxed{1000}. \]The maximum rate is
\[ \frac{rK}{4} =\frac{(0.10)(2000)}{4} =\boxed{50} \]population units per unit time.
Practice 5
\[ R_0=\frac{\beta}{\gamma} =\frac{0.24}{0.08} =\boxed{3}. \]Practice 6
\[ D=\frac{1}{\gamma} =\frac{1}{0.125} =\boxed{8\ \text{days}}. \]Practice 7
\[ R_{\text{eff}} =R_0\frac{S}{N} =(2.4)(0.30) =\boxed{0.72}. \]Practice 8
If predators and prey mix approximately at random, the number of possible predator-prey encounters increases with both populations. Doubling \(X\) while holding \(Y\) fixed approximately doubles encounters, and doubling \(Y\) while holding \(X\) fixed does the same. This motivates a rate proportional to the product \(XY\).
Practice 9
\[ Y=pX =(0.40)(250) =\boxed{100}. \]Practice 10
A single logistic equation ignores spatial structure and migration among habitat patches. A metapopulation model, coupled system of patch equations, network model, or spatial reaction-diffusion model would better represent persistence when connectivity matters.
What to Remember
- biological models are built from state variables, flows, interactions, and parameters;
- exponential growth assumes a constant per-capita growth rate;
- logistic growth introduces density dependence and a carrying capacity;
- multi-population systems can model predation, competition, and cooperation;
- compartment models represent transitions among epidemiological states;
- \(R_0\) and \(R_{\text{eff}}\) are threshold quantities whose meaning depends on model structure;
- spatial, network, stochastic, and agent-based models capture mechanisms omitted by simple aggregate equations;
- observation processes must be distinguished from the hidden biological process;
- calibration, validation, sensitivity, and uncertainty analysis are separate modeling tasks;
- model complexity should be driven by the decision and by the mechanisms that materially affect that decision.
Where We Go Next
Population, ecological, and epidemiological models show how interacting entities generate growth, thresholds, feedback, spread, and collective behavior. Many of the same mathematical structures appear in chemical systems, where reactants interact through concentration-dependent rates and conservation laws.
The next lesson, Chemistry and Reaction Systems, develops models of reaction rates, stoichiometry, equilibrium, reaction networks, transport, temperature dependence, and coupled chemical dynamics.