Logistic Growth and Carrying Capacity

Exponential growth, dN/dt = rN, assumes the per-capita growth rate r is constant no matter how many individuals are already present — an assumption that only holds when resources are effectively unlimited relative to population size. The logistic model, first written down by Pierre-François Verhulst in 1838 and rediscovered for biology by Raymond Pearl and Lowell Reed in the 1920s, relaxes that assumption. It proposes that per-capita growth rate declines as a linear function of population size N, reaching zero exactly when N equals the environment's carrying capacity, K — the maximum population size that the environment's resources can sustain indefinitely. The logistic differential equation is:

dN/dt = rN(1 − N/K) Here r is the intrinsic (maximum, density-independent) per-capita growth rate — the same r as in the exponential model — and the added factor (1 − N/K) is the density-dependent term. When N is small relative to K, (1 − N/K) is close to 1, so growth is nearly exponential. As N climbs toward K, (1 − N/K) shrinks toward 0, throttling growth. If N ever exceeds K (an "overshoot"), the term goes negative and the population declines back toward K. K is therefore a stable equilibrium: population size that starts above or below K is pulled toward it, which is why K is often called an attractor. The biological content hidden inside that one algebraic term (1 − N/K) is density dependence: the idea that birth rates fall and/or death rates rise as population density increases, because individuals increasingly compete for the same finite pool of food, nest sites, light, or territory (this foreshadows the fuller treatment of Density-Dependent Factors). The logistic model is the simplest possible mathematical statement of that idea — a single linear negative feedback — and its elegance, more than its precision, explains why it remains the starting point for population regulation theory.

Step 1 — Start from exponential growth. Begin with dN/dt = rN, unconstrained growth. This is the density-independent baseline: every individual, on average, contributes the same net number of offspring per unit time regardless of how many neighbors it has.

Step 2 — Add a resource-limitation term. Nature does not actually supply unlimited per-capita resources. As N grows, each individual's share of food, breeding sites, or space shrinks. Verhulst modeled this by multiplying r by the fraction of carrying capacity still "unused": (K − N)/K, algebraically identical to (1 − N/K). This single move converts a constant per-capita rate into one that falls linearly with N.

Step 3 — Locate the equilibria. Growth rate dN/dt = 0 under two conditions: N = 0 (no population, trivially stable-ish — extinction) and N = K (the population exactly fills the available capacity). Between them, for 0 < N < K, dN/dt > 0, so the population grows; above K, dN/dt < 0, so it shrinks. K is therefore a stable equilibrium and N = 0 is unstable (any tiny positive population moves away from zero, toward K).

Step 4 — Watch per-capita growth rate fall. Divide both sides by N: (1/N)(dN/dt) = r(1 − N/K). This is the per-capita growth rate — the quantity biologists actually measure by tracking birth and death rates in the field. It is a straight line when plotted against N: it starts at r when N is near zero and declines linearly to zero at N = K. This is the direct empirical signature of logistic dynamics, and it is what field ecologists look for when testing whether a real population is density-regulated.

Step 5 — Integrate to get the S-curve. Solving the differential equation analytically gives: N(t) = K / (1 + [(K − N₀)/N₀] · e^(−rt)) At t = 0 this returns N₀; as t → ∞, the exponential term vanishes and N(t) → K. The curve has an inflection point — where growth rate is fastest — exactly at N = K/2, found by taking the second derivative of N(t) and setting it to zero. Before the inflection, growth accelerates (the curve looks convex, exponential-like); after it, growth decelerates (concave, approaching a plateau). This inflection point matters practically: maximum sustainable yield in fisheries and wildlife management is harvested at N = K/2, because that is where the population regenerates individuals fastest.

Step 6 — Recognize the biological triggers of density dependence. The mathematics doesn't specify why growth slows — that requires biology. Classic mechanisms include:

Raymond Pearl and Lowell Reed's Drosophila bottles (1920). Working at Johns Hopkins, Pearl and Reed reared fruit flies in fixed-volume culture bottles with a constant, limited food supply and counted the population daily. The population grew nearly exponentially at first, then decelerated, then leveled off at a stable number determined by the bottle's food and space — the first quantitative demonstration that a real population follows the sigmoidal shape Verhulst had predicted decades earlier on purely mathematical grounds. Pearl went on to fit the same curve to U.S. and other human census data (a use now regarded as historically important but scientifically overreaching, since human K is not fixed).

Georgy Gause's Paramecium microcosms (1934). This is the most cited direct test. Gause grew Paramecium caudatum in test tubes with a fixed daily ration of bacterial food, counting individuals under a microscope every day for about a week to ten days. Results:

• Population grew rapidly for roughly the first 2–3 days (near-exponential phase).
• Growth then decelerated sharply.
• The population stabilized at a roughly constant density — an experimentally observed K — that Gause could shift up or down by changing the daily food ration, directly confirming that K is set by resource supply, not by some innate property of the organism.
• When Gause added a competing species (P. aurelia) to the same tube, one species was driven toward exclusion — a result central to his separate, equally famous competitive exclusion work, but built on the same logistic scaffolding.

The Gause experiments were decisive because they isolated the mechanism: by controlling food input directly, Gause showed that per-capita division rate fell as density rose (matching the linear (1 − N/K) prediction) and that the plateau population size tracked resource supply, not some fixed ceiling — exactly what the logistic model requires and exponential growth cannot explain.

H. G. Andrewartha and L. C. Birch, and later field tests. Subsequent field ecologists tested whether logistic dynamics appear in wild populations, with mixed results. Song sparrow populations on Mandarte Island, and controlled experiments with flour beetles (Tribolium) by Thomas Park, showed density dependence in birth and death rates but also revealed departures from the simple logistic curve — overshoots, oscillations, and lagged responses — because real populations respond to density with a time delay rather than instantaneously. This is why the pure logistic equation is now understood as a foundational idealization, not a literal description of most wild populations; time-lag and discrete-generation versions (e.g., the Ricker model) were developed to capture the oscillatory and chaotic dynamics that appear once realistic delays are included.

Why these experiments matter as evidence, not just illustration: they show cause and effect, not mere correlation with an S-shaped curve. Because Gause could manipulate food ration independently and watch K move accordingly, and because per-capita growth rate could be measured directly and shown to fall linearly with density, the experiments established density dependence as a causal mechanism regulating population size — not simply fit a curve after the fact.

Example 1 — A fishery approaching maximum sustainable yield. A fish stock has intrinsic growth rate r = 0.4 per year and carrying capacity K = 100,000 tonnes of biomass. The stock is currently at N = 25,000 tonnes. Step 1: Growth rate right now. dN/dt = rN(1 − N/K) = 0.4 × 25,000 × (1 − 25,000/100,000) = 0.4 × 25,000 × 0.75 = 7,500 tonnes/year. Step 2: Where is growth fastest? At N = K/2 = 50,000 tonnes. dN/dt at N = 50,000: 0.4 × 50,000 × (1 − 0.5) = 0.4 × 50,000 × 0.5 = 10,000 tonnes/year — this is the maximum sustainable yield (MSY), the largest harvest that can be removed indefinitely without depleting the stock. Step 3: Management implication. A manager wanting to maximize long-term catch should aim to keep the stock near 50,000 tonnes, not at K. Harvesting the stock down to K/2 and holding it there (removing exactly 10,000 tonnes/year, matching regrowth) yields the highest sustainable catch — the theoretical basis (with important caveats about time lags and stochasticity) for MSY targets in fisheries policy. Step 4: Contrast with unmanaged exploitation. If the fleet instead depletes the stock to N = 5,000 tonnes (5% of K), growth rate = 0.4 × 5,000 × (1 − 0.05) = 1,900 tonnes/year — far below MSY, illustrating why overfishing past the inflection point is doubly costly: it removes fish and collapses the population's own regenerative capacity.

Example 2 — Reading a bacterial growth curve for K and r. A microbiologist inoculates a flask with N₀ = 1 × 10³ cells/mL. After 6 hours the population has reached N = 5 × 10⁶ cells/mL and is still climbing steeply; by 24 hours it has plateaued at 8 × 10⁷ cells/mL and stays there. Step 1: Identify K. The plateau value is the experimental estimate of carrying capacity: K ≈ 8 × 10⁷ cells/mL (set by nutrient depletion and, in dense cultures, by accumulating metabolic waste and falling pH — real biological mechanisms behind the abstract "1 − N/K" term). Step 2: Locate the inflection (fastest growth). Theory predicts the steepest part of the curve at N = K/2 = 4 × 10⁷ cells/mL. If the recorded data show the doubling rate peaking around hour 10–12 near that density, this is direct empirical support that the culture is following logistic, not pure exponential, dynamics. Step 3: Estimate r from the early phase. Between t = 0 and t = 6 h, growth is still close to exponential because N ≪ K (10³ vs 8×10⁷, so N/K ≈ 0.0000125, negligible). Using N(t) = N₀e^(rt): 5×10⁶ = 1×10³ × e^(6r) → e^(6r) = 5000 → 6r = ln(5000) ≈ 8.517 → r ≈ 1.42 per hour. Step 4: Predict N at hour 10 using the full logistic solution. N(t) = K / (1 + [(K−N₀)/N₀] e^(−rt)) At t = 10: exponent term = e^(−1.42×10) = e^(−14.2) ≈ 6.8×10⁻⁷. [(K−N₀)/N₀] ≈ 8×10⁷/10³ = 8×10⁴. Denominator ≈ 1 + (8×10⁴)(6.8×10⁻⁷) ≈ 1 + 0.0544 ≈ 1.054. N(10) ≈ 8×10⁷ / 1.054 ≈ 7.6 × 10⁷ cells/mL — already close to K, consistent with the observed near-plateau by hour 24, and demonstrating how quickly the logistic curve saturates once N/K becomes non-negligible.

Pierre-François Verhulst was a Belgian mathematician working in the 1830s who was reacting directly against Thomas Malthus's grim prediction that unchecked human population growth would always outrun food supply. Verhulst reasoned that growth must be self-limiting and, in 1838, published the differential equation that we now call the logistic — he coined the term "logistique" himself. His work was mathematically elegant but had almost no immediate empirical following; it languished for nearly ninety years without experimental confirmation, in part because 19th-century biology lacked the controlled culture techniques needed to test it. It was rediscovered and popularized in the 1920s by the American biometrician Raymond Pearl, who fitted the curve to fruit fly cultures and (more controversially) to human census growth, and it was given rigorous experimental grounding by the Soviet ecologist Georgy Gause in the 1930s, whose Paramecium microcosms remain the textbook demonstration to this day.

Fisheries and wildlife management. Maximum sustainable yield theory, still used (with heavy caveats) by agencies such as NOAA Fisheries and the International Whaling Commission, is built directly on the logistic model's prediction that yield is maximized at N = K/2. Post-whaling-ban recovery targets for species like humpback and gray whales are modeled with logistic or logistic-derived curves to estimate how populations should rebound toward K.

Invasive species and biosecurity. Predicting how fast an invader (zebra mussels, Burmese pythons, cane toads) will saturate a new habitat, and at what density it will plateau, uses logistic curve-fitting to time intervention — early exponential phase is the cheapest and most effective window for control, before density-dependent saturation sets in.

Epidemiology and public health. The logistic curve is mathematically identical in shape to the early, resource-analogous phase of an SIR epidemic curve, where "susceptible individuals" play the role of unused carrying capacity; case-count curves for outbreaks (including COVID-19 case curves in closed populations) are routinely logistic-fitted to estimate peak timing and eventual plateau size.

Conservation biology and endangered species recovery. Estimating a species' K in a given reserve — and recognizing that habitat restoration raises K while fragmentation lowers it — underlies decisions about minimum viable population sizes and reserve design, feeding directly into models of Population Regulation.

Agriculture and bioprocessing. Industrial fermentation (biofuels, antibiotics, vaccine production in bioreactors) is optimized using logistic growth curves to time nutrient feeding and harvest before cultures hit their carrying-capacity plateau, maximizing yield per batch.

Human demography. Even though human carrying capacity is notoriously hard to pin down (it depends on technology, diet, and consumption patterns, not just fixed biological limits), logistic-style S-curve reasoning still frames debates about long-term population stabilization and the concept of sustainable global population size.

Ask the tutor: How does adding a time lag to the logistic equation produce population overshoot and oscillation around K?