Mathematical Modeling

⚡ Key Takeaways

Mathematical modeling is the art and science of translating a real-world situation into mathematical language (equations, functions, graphs, or simulations), analyzing that representation, and interpreting the results back in the real world. Key result: a good model captures the essential features of a system precisely enough to explain or predict its behavior while remaining simple enough to be solved and understood. Where it applies: physics, biology, economics, engineering, medicine, climate science — anywhere quantities change or interact. Who/when: developed continuously from Newton (1687) through Lotka–Volterra (1920s) to modern computational and statistical modeling.

💡 The Core Insight

Reality is infinitely complex; a model deliberately throws away most of that complexity and keeps only what matters for the question being asked. The central skill is choosing good assumptions — the model is only as trustworthy as the simplifications behind it. "All models are wrong, but some are useful" (George Box). Modeling is therefore a cycle: build, solve, test, refine — not a one-shot calculation.

Formal Statement

DefinitionThe Modeling Cycle

A mathematical model is a triple consisting of (1) a set of variables representing measurable quantities, (2) a set of assumptions restricting how those variables may behave, and (3) a set of relationships (equations, inequalities, or rules) linking the variables. The modeling cycle is the iterative procedure: 1. Identify the real-world problem and the question to answer. 2. Assume & simplify — list what is ignored and what is kept. 3. Formulate — write mathematical relationships. 4. Solve / analyze — apply mathematics to extract predictions. 5. Interpret — translate the math back into the real world. 6. Validate — compare with data; if poor, return to step 2.

A classic worked relationship is the exponential growth model, in which the rate of change of a quantity is proportional to the quantity itself:

1
dPdt=kP\quad \frac{dP}{dt} = kP

Derivation hint: assume each individual contributes a constant per-capita rate kk; total rate = rate per individual × number of individuals PP. Here P=P(t)P = P(t) is the population (or amount) at time tt, and kk is the growth constant (units of inverse time). Solving this gives:

2
P(t)=P0 ekt\quad P(t) = P_0\, e^{kt}

Derivation hint: separate variables, ∫dPP=∫k dt\int \tfrac{dP}{P} = \int k\,dt, integrate to ln⁡P=kt+C\ln P = kt + C, exponentiate and set P(0)=P0P(0)=P_0. When growth is limited by resources, the logistic model replaces the constant rate with one that shrinks as PP approaches a carrying capacity KK:

3
dPdt=kP(1−PK)\quad \frac{dP}{dt} = kP\left(1 - \frac{P}{K}\right)

Derivation hint: multiply the exponential rate kPkP by a "braking factor" (1−P/K)(1 - P/K) that equals 1 when P≈0P\approx 0 and 0 when P=KP = K. Its solution is the logistic curve:

4
P(t)=K1+Ae−kt,A=K−P0P0\quad P(t) = \frac{K}{1 + A e^{-kt}}, \qquad A = \frac{K - P_0}{P_0}

Derivation hint: separate variables, use partial fractions 1P(1−P/K)=1P+1/K1−P/K\tfrac{1}{P(1-P/K)} = \tfrac{1}{P} + \tfrac{1/K}{1-P/K}, integrate and solve for PP.

TheoremBehavior of the Logistic Model

For the logistic equation [3] with 0<P0<K0 < P_0 < K and k>0k > 0, the solution P(t)P(t) is strictly increasing, bounded above by KK, and satisfies lim⁡t→∞P(t)=K\lim_{t\to\infty} P(t) = K. The curve has a single inflection point where P=K/2P = K/2, at which the growth rate dP/dtdP/dt is maximized.

Why It Holds / Derivation

ProofSolving the Logistic Equation

Step 1. Start from [3]: dPdt=kP(1−PK)\dfrac{dP}{dt} = kP\left(1 - \dfrac{P}{K}\right).

Step 2. Separate variables:

dPP(1−P/K)=k dt.\frac{dP}{P\left(1 - P/K\right)} = k\,dt.

Step 3. Decompose the left side by partial fractions:

1P(1−P/K)=1P+1/K1−P/K.\frac{1}{P(1 - P/K)} = \frac{1}{P} + \frac{1/K}{1 - P/K}.

Step 4. Integrate both sides:

ln⁡P−ln⁡ ⁣(1−PK)=kt+C.\ln P - \ln\!\left(1 - \frac{P}{K}\right) = kt + C.

Step 5. Combine logarithms and exponentiate:

P1−P/K=ekt+C=Bekt.\frac{P}{1 - P/K} = e^{kt + C} = B e^{kt}.

Step 6. Solve algebraically for PP, then apply P(0)=P0P(0) = P_0 to fix the constant, yielding A=(K−P0)/P0A = (K - P_0)/P_0 and equation [4].

Step 7 (inflection point). Differentiate [3] and set d2P/dt2=0d^2P/dt^2 = 0; this occurs exactly when 1−2P/K=01 - 2P/K = 0, i.e. P=K/2P = K/2. ■\blacksquare

∎
mathematician · Alfred J. Lotka & Vito Volterra | 1880–1949 / 1860–1940

Lotka (a chemist) and Volterra (a mathematician) independently created the predator–prey equations in the 1920s — coupled differential equations modeling how the populations of, say, foxes and rabbits oscillate over time. Volterra was inspired by his son-in-law, a marine biologist puzzled by fish-catch fluctuations in the Adriatic. Their work showed that interaction between species, not external causes, can produce sustained cycles — a landmark demonstration that mathematics can reveal counterintuitive truths about living systems.

Building Intuition

🧠 Intuition

Think of modeling as drawing a map. A map of your city is not the city — it omits trees, traffic, and individual people. But a good map keeps exactly the features you need (streets, distances) and discards the rest, so you can navigate. A subway map distorts geography entirely yet is perfect for its purpose. Likewise, a model of a falling ball might ignore air resistance (a "frictionless map") — wonderful for a bowling ball, useless for a feather. The whole skill is matching the model's level of detail to the question. Ask: what can I safely ignore?

Figure
Exponential vs. logistic growth: the same starting slope, very different fates
time t population P carrying capacity K logistic exponential inflection (P=K/2)

Worked Examples

ExampleBacterial Growth (Exponential Model)

Given: A bacterial culture starts with P0=500P_0 = 500 cells and grows so that the population doubles every 3 hours. Predict the population after 12 hours.

Step 1 — find kk. Doubling in 3 h means P(3)=2P0P(3) = 2P_0, so from [2]:

2P0=P0e3k  ⇒  e3k=2  ⇒  k=ln⁡23≈0.231 h−1.2P_0 = P_0 e^{3k} \;\Rightarrow\; e^{3k} = 2 \;\Rightarrow\; k = \frac{\ln 2}{3} \approx 0.231\ \text{h}^{-1}.

Step 2 — evaluate at t=12t = 12 h. Note 12=4×312 = 4 \times 3 h = 4 doublings:

P(12)=500⋅e12k=500⋅24=500⋅16.P(12) = 500 \cdot e^{12k} = 500 \cdot 2^{4} = 500 \cdot 16.

Answer:

P(12)=8000 cells\boxed{P(12) = 8000\ \text{cells}}

(Check: model assumes unlimited nutrients — valid only for early growth.)

ExampleCooling Coffee (Newton's Law of Cooling)

Given: Coffee at 90∘C90^\circ\text{C} sits in a 20∘C20^\circ\text{C} room. Newton's law of cooling models the temperature T(t)T(t) by

dTdt=−r (T−Troom),\frac{dT}{dt} = -r\,(T - T_{\text{room}}),

whose solution is T(t)=Troom+(T0−Troom)e−rtT(t) = T_{\text{room}} + (T_0 - T_{\text{room}})e^{-rt}. After 5 minutes the coffee is 70∘C70^\circ\text{C}. Find rr, then the temperature at t=15t = 15 min.

Step 1 — set up. T0=90T_0 = 90, Troom=20T_{\text{room}} = 20, so T(t)=20+70 e−rtT(t) = 20 + 70\,e^{-rt}.

Step 2 — use the data point T(5)=70T(5) = 70:

70=20+70e−5r  ⇒  e−5r=5070=0.714370 = 20 + 70 e^{-5r} \;\Rightarrow\; e^{-5r} = \frac{50}{70} = 0.7143
r=−15ln⁡(0.7143)≈0.0673 min−1.r = -\frac{1}{5}\ln(0.7143) \approx 0.0673\ \text{min}^{-1}.

Step 3 — evaluate at t=15t = 15:

T(15)=20+70 e−0.0673×15=20+70 e−1.010=20+70(0.364).T(15) = 20 + 70\,e^{-0.0673 \times 15} = 20 + 70\,e^{-1.010} = 20 + 70(0.364).

Answer:

T(15)≈45.5∘C\boxed{T(15) \approx 45.5^\circ\text{C}}

Common Misconceptions

⚠️ Common mistake

A model that fits the data is not automatically correct. With enough free parameters you can fit almost any curve (this is overfitting). A good model fits the data and makes accurate predictions about new situations and rests on plausible assumptions. Beware also of extrapolation: equation [2] predicts exponential growth forever, but no real population grows without bound — at some point food, space, or disease intervenes. A model is only valid within the range of conditions for which its assumptions hold.

counterextable

counterexample[Exponential Model Fails Long-Term] Suppose you fit bacteria data perfectly for the first 12 hours with P(t)=500 e0.231tP(t) = 500\,e^{0.231 t}. Extrapolating to t=48t = 48 h gives P≈500⋅e11.1≈3.3×107P \approx 500 \cdot e^{11.1} \approx 3.3 \times 10^{7}, and to t=120t = 120 h gives more bacteria than atoms in a petri dish could hold. The model is not wrong — it simply applied outside its valid domain. The logistic model [3], which the exponential model approximates only for small PP, corrects this by leveling off at KK.

Connections

connection

Mathematical modeling is the connective tissue of applied mathematics. The exponential and logistic equations are instances of Differential Equations, whose solution techniques (separation of variables, integrating factors) come from Integral Calculus. The growth constant kk and carrying capacity KK are estimated using Regression and Curve Fitting. Models with randomness (e.g. predicting dice, queues, or stock prices) draw on Probability Theory. When a model has no closed-form solution, it is solved approximately via Numerical Methods on a computer. The frequent reuse of one equation for many systems reflects the unifying power of Functions and Graphs and dimensional reasoning.

Real-World Applications

🔬 Where this appears

Historical Development

⏳ History
- 1687 — Isaac Newton's Principia introduces differential equations of motion, the first great predictive physical models.
- 1798 — Thomas Malthus publishes the exponential population principle, equation [1]–[2], warning of unchecked growth.
- 1838 — Pierre-François Verhulst introduces the logistic equation [3]–[4] to correct Malthus by adding a carrying capacity.
- 1840s — Newton's law of cooling formalized as an empirical exponential-decay model.
- 1920s — Lotka (1925) and Volterra (1926) independently develop predator–prey systems.
- 1927 — Kermack and McKendrick publish the SIR epidemic model.
- 1940s–50s — the advent of digital computers makes numerical solution of large models feasible; operations research is born in WWII.
- 1973 — Black, Scholes, and Merton publish the option-pricing model, revolutionizing finance.
- 1979 — George Box coins "all models are wrong, but some are useful," crystallizing the modeler's philosophy.

Subtleties & Limits

open-problem

How much detail is the right amount? There is no algorithm that tells you which features to keep — this remains a matter of judgment, testing, and experience. Modern challenges include modeling complex systems (ecosystems, economies, the brain) where many variables interact nonlinearly and small changes can cascade (chaos). Even when a model is "correct," its parameters may be unidentifiable from available data, and predictions of inherently chaotic systems (like weather beyond ~2 weeks) are fundamentally limited regardless of model quality. The frontier question — how do we model systems too complex to write equations for? — has spawned agent-based simulation and machine-learning models that trade interpretability for predictive power, raising the open issue of whether a black-box model that predicts well but cannot be understood truly "explains" anything.

See Also

Differential Equations · Exponential and Logarithmic Functions · Regression and Curve Fitting · Numerical Methods · Probability Theory · Dynamical Systems and Chaos · Functions and Graphs