Module 0: Megafauna Scaling Laws

The intellectual thread of biological scaling begins with Galileo’s Two New Sciences (1638). In the “Second Day,” Galileo reasoned that if two animals are geometrically similar but one has linear dimension \(L\) twice the other, then its weight (scaling as \(L^3\)) grows eight-fold while the cross-sectional area of its limb bones (scaling as \(L^2\)) grows only four-fold. The stress on those bones must therefore double. He concluded “it would be impossible to build up the bony structures of men, horses, or other animals so as to hold together and perform their normal functions if these animals were to be increased enormously in height.”

Nearly three centuries later, J. B. S. Haldane (1926) — “On Being the Right Size” — distilled Galileo into accessible prose: “you can drop a mouse down a thousand-yard mine shaft; and, on arriving at the bottom, it gets a slight shock and walks away. A rat is killed, a man is broken, a horse splashes.” He pointed out that drag scales as area \(L^2\) while gravitational energy dissipates per unit time as \(L^3\), so terminal velocity rises as \(L^{1/2}\). The mouse’s terminal velocity is only a few metres per second.

The third founding text is Archibald Vivian Hill’s (1950) “The dimensions of animals and their muscular dynamics.” Hill derived the dynamic similarity principle: geometrically similar animals should run at the same Froude number \(\mathrm{Fr}=v^2/(gL)\), implying maximum running speed scales as \(v_{\max}\propto L^{1/2}\). This turns out to be almost right for small mammals and badly wrong for elephants, a discrepancy we resolve in the elastic-similarity and constant-stress discussion below.

In 1932 the Swiss agricultural chemist Max Kleiber compiled basal metabolic rate measurements for 13 species of mammal spanning from a 150 g rat to a 679 kg steer and plotted \(\log B\)against \(\log M\). The slope was \(0.74\pm0.02\), decisively ruling out the \(2/3\) exponent that naive surface-to-volume arguments (radiative heat loss) predicted. The canonical form of the law is:

For an African bull elephant \(M = 6000\) kg, this predicts \(B \approx 2.3\) kW basal, or a daily energy requirement of roughly 200 MJ — consistent with the 150–300 kg of forage consumed per day. For a 200 kg lion, \(B \approx 180\) W; for a 4 kg serval, \(B \approx 9.6\) W; for a 1.8 g Etruscan shrew,\(B \approx 27\) mW. Mass-specific metabolic rate \(B/M \propto M^{-1/4}\) means the shrew is consuming oxygen ~300 times faster per gram than the elephant.

2.1 West–Brown–Enquist Fractal Derivation (1997)

For 65 years Kleiber’s exponent was an empirical mystery. In 1997 Geoffrey West, James Brown, and Brian Enquist (Science 276, 122) proposed a derivation from three assumptions about the vascular network:

1. Space-filling. The network must reach every cell, so the sum of cylindrical volumes of the terminal capillaries fills the body volume: \(N_c\, l_c\, r_c^2 \propto V \propto M\).
2. Invariant terminal units. Capillaries have species-invariant size and flow (a mouse capillary and an elephant capillary are essentially identical).
3. Minimum dissipation. Natural selection minimises hydrodynamic power loss. For pulsatile blood flow in an elastic tree, this is achieved by a self-similar network with constant area-preservingbranching for large vessels (pulse propagation without reflection) and area-increasing branching near the capillaries (viscous-dominated Poiseuille flow).

Under these constraints, the radius ratio \(\beta_k = r_{k+1}/r_k = n^{-1/2}\) in the large-vessel region (where\(n\) is the branching number) and the length ratio \(\gamma_k = l_{k+1}/l_k = n^{-1/3}\) throughout (space-filling). The total number of levels \(N\) from aorta to capillary therefore satisfies \(n^N = N_c\). Combining:

The \(3/4\) exponent emerges from the interplay of 3D space filling (exponent 1/3) and the area-preserving hydrodynamic constraint (exponent 1/2). The same argument recovers \(4/3\) scaling of aorta radius, a\(-1/12\) scaling of blood velocity, and a \(1/4\) scaling of circulation time — all observed.

2.2 Controversies: Dodds 2001, White & Seymour 2003

Not everyone accepts \(3/4\). Dodds, Rothman & Weitz (2001)reanalysed Kleiber’s and McNab’s compilations using modern phylogenetic correction and obtained\(b = 0.686\pm0.014\) — statistically indistinguishable from the surface-area value 2/3. They argued that Kleiber’s 3/4 was an artefact of including large domesticated mammals with inflated BMR. White & Seymour (2003) split endotherms from ectotherms and found\(b_{\text{endo}} \approx 0.67\) and \(b_{\text{ecto}} \approx 0.76\), the opposite of what WBE predict. The debate is alive: Savage et al. (2004) reasserted 3/4 for the “true basal” subset, and Kolokotrones et al. (2010) argued the relationship is actually slightly curved in log–log space.

For the purposes of this course, we adopt \(b = 3/4\) as the default with the understanding that exponents for specific physiological quantities may deviate by \(\pm 0.1\). The derivation given here is less important than the approximate constancy of the underlying invariant: the terminal exchange unit.

Heat production scales approximately with \(M\) (cellular respiration per unit mass), but heat loss through the integument scales with surface area \(S \propto V^{2/3} \propto M^{2/3}\). The ratio is:

A shrew, all skin and no interior, must eat constantly just to keep warm; Etruscan shrews consume their own body mass in insects every day. A blue whale, with \(M^{1/3}\) some 400 times larger than a shrew’s, has the opposite problem: it must dump heat, which is why large cetaceans have counter-current vascular bundles in their flukes and why elephants evolved their enormous ears (a 6000 kg African bull has ~5 m² of ear surface, ~20% of total body surface). Swetaing is inefficient for mega-herbivores because sweat glands are a 2D phenomenon on a 3D creature. Elephants have very few sweat glands and instead cool by ear-flapping (forced convection) and dust-bathing (evaporative cooling through spray).

Galileo’s problem in quantitative form. The peak bending stress \(\sigma\) in a cylindrical long bone of radius \(r\) loaded at a moment arm \(d\) by a ground-reaction force \(F\) is

Under pure geometric similarity (\(L \propto M^{1/3}\), \(r \propto M^{1/3}\), \(\theta\) constant) we obtain\(\sigma \propto M\cdot M^{1/3}/M = M^{1/3}\), doubling every time mass grows eight-fold. A 6-tonne elephant with mouse-like crouched posture would have bone stresses of several hundred MPa — well above the 180 MPa yield of cortical bone. Biewener (1989, 1990) showed how three mechanisms conspire to keep peak stress roughly constant at \(\sigma_\text{peak} \approx 40\text{--}80\) MPa across 5 orders of magnitude of mass:

1. Posture shift. Large mammals align the limb more vertically during stance (\(\theta \propto M^{-0.09}\)), reducing the moment arm \(d\). The elephant stands with\(\theta < 10^\circ\) — a “columnar” posture like a four-legged piano.
2. Elastic similarity. McMahon (1973) showed bones should resist buckling at constant safety factor under their own weight: \(L \propto d^{3/2}\), giving \(r \propto M^{3/8}\) and\(L \propto M^{1/4}\). Alexander’s 1985 survey of femur dimensions in 37 mammal species confirmed\(r \propto M^{0.36\pm0.04}\), very close to 3/8.
3. Gait restriction. Elephants cannot leave the ground with all four feet simultaneously; they do not run in the aerodynamic sense, only “amble” at up to 25 km/h (Hutchinson 2003).

Combining \(r \propto M^{3/8}\), \(L \propto M^{1/4}\), and \(\theta \propto M^{-0.09}\):

Schmidt-Nielsen’s Scaling: Why Is Animal Size So Important? (1984) compiled decades of data showing:

Heart mass is a species-invariant \(\approx 0.6\%\) of body mass — an elephant’s heart weighs ~28 kg, a shrew’s ~12 mg. Stroke volume follows heart mass, so cardiac output\(\dot{Q} = V_\text{stroke}\cdot f_\text{HR}\) scales as \(M\cdot M^{-1/4} = M^{3/4}\), exactly tracking metabolic demand. Heart rate drops from ~1000 beats/min in a shrew to 25–35 beats/min in a resting elephant, to 8–10 in a blue whale.

The total number of heartbeats in a mammalian lifetime is approximately invariant:\(N_\text{beats} \sim f_\text{HR}\cdot \tau \sim M^{-1/4}\cdot M^{1/4} \sim 10^9\). A shrew and an elephant get the same ~one billion heartbeats — just compressed or stretched in time (Schmidt-Nielsen 1984). Humans, with our anomalously long lifespan, get 2–3 billion.

One consequence: vascular transit time \(\tau_\text{circ} = V_\text{blood}/\dot{Q} \propto M^{1/4}\). A drop of blood completes a loop in 15 s in a shrew, 2 min in a human, 10 min in an elephant, and ~15 min in a blue whale. This sets the timescale for endocrine signalling — one reason large animals have slower “clocks.”

Demment & Van Soest (1985) observed that gut capacity scales as\(V_\text{gut}\propto M^{1.0}\) while metabolic demand scales as \(B\propto M^{3/4}\). The ratio — the mean retention time — therefore grows with body mass:

Empirically: a rabbit digests in ~20 h, a domestic cow in ~80 h, an African elephant in ~45–50 h (elephants are hindgut fermenters with less efficient digestion than ruminants despite their size). The extra retention time allows large herbivores to break down fibrous, low-quality forage that a ruminant can ferment more efficiently, so Jarman (1974) and Bell (1971) predicted and observed that megaherbivores can subsist on forage that smaller species cannot touch — the “Jarman–Bell principle.”

Numerical application. An elephant consuming 1.5% of body mass in dry matter per day processes \(\sim 90\) kg of cellulose-rich vegetation with ~30% digestive efficiency. The hindgut acts as a giant anaerobic bioreactor, with methanogenic archaea contributing a planet-scale 1–2 Mt CH₄/yr of emissions from the ~400 000 African savanna elephants alive today.

Eric Charnov (1993) compiled life-history data across 14 mammalian orders and found remarkable invariants when rescaled:

The ratio of age-at-first-reproduction to lifespan is approximately a species-invariant “Charnov’s constant” \(\alpha/\tau_\text{life}\approx 0.3\). An elephant with \(\tau_\text{life}\approx 65\) yr and a gestation of 22 months reaches first reproduction at about 14 yr, and the whole “life movie” — gestate, wean, sexually mature, die — plays out ~15 times slower than that of a 1 kg mouse-deer. Large mammals live in proportionally the same life, only on a slower clock.

For the modules that follow, the critical consequence is that elephants, rhinos, and hippos are all K-selected species with low intrinsic rates of increase (\(r_\text{max}\approx 0.06\) yr for African elephants). Poaching that removes even a few per cent of adults per year drives rapid demographic collapse — a biophysics-mediated conservation disaster.

The Froude number \(\mathrm{Fr}=v^2/(g L)\) is the dimensionless ratio of inertial to gravitational force for a pendulum of length \(L\). Alexander (1976, 1989) showed that animals of very different sizes switch gait at nearly the same Froude number:

With leg length \(L\propto M^{1/4}\) (elastic similarity), top speed at fixed Fr scales as\(v\propto L^{1/2}\propto M^{1/8}\). Garland (1983) and McMahon (1975) confirmed this up to ~120 kg. Above that, bone-stress and heat-dissipation limits force a break: maximum speed rises up to a ~300 kg optimum (cheetahs, greyhounds, pronghorn) and declines for larger animals. The elephant never achieves Fr > 1, meaning it never has an aerial phase. Hutchinson (2003) timed racing elephants at 25 km/h but found all four feet never left the ground simultaneously — biomechanically a fast amble rather than a true gallop.

Mass-specific cost of transport scales as \(C\propto M^{-0.32}\) (Taylor 1982): large animals travel more cheaply per kilogram–kilometre, but their absolute energy budget for migration (\(\propto M^{0.68}\)) is still large. Migrating elephants spend ~10% of daily energy budget on travel alone over 30 km/day dry-season movement.

McMahon (1973) argued that biological tissues scale to preserve elastic rather than geometric similarity. For a column loaded by its own weight, the critical buckling load is \(P_\text{cr}\propto r^4/L^2\). Setting \(P_\text{cr}\)equal to the gravitational load \(\rho g L r^2\) and demanding a constant safety factor yields:

Real mammals are intermediate between geometric (\(r\propto M^{1/3}\)) and elastic (\(r\propto M^{3/8}\)) similarity: Alexander 1985 measured an average exponent of \(0.36\) across 37 species. In graviportal mammals (rhinos, elephants, sauropods) the exponent approaches and occasionally exceeds 3/8 — they are over-engineered in bone diameter, which is why their long bones look so thick and tree-trunk-like.

McMahon extended elastic similarity to tree trunks, bamboo stems, and dinosaur femora. The prediction for sauropods is particularly striking: a 35-tonne Apatosaurus by elastic similarity should have femoral radius\(r\approx 0.015\cdot(35{,}000)^{3/8}\approx 0.9\) m, close to the measured 0.85 m. Even the 70-tonnePatagotitan fits the scaling if we allow a slight exponent inflation to 0.40 characteristic of graviportal forms. The continuity of the allometric law from shrew to sauropod across 10 orders of magnitude in mass suggests that bone-stress invariance is a fundamental constraint of mechanical design, not a contingent feature of any particular clade.

Gillooly, Brown, West, Savage & Charnov (2001) extended the WBE framework to unify size and temperature effects on metabolism through the Arrhenius factor for enzymatic rate-limiting steps:

For endotherms, \(T\) is body temperature (relatively constant at 37–38 °C), so the temperature factor drops out and we recover pure Kleiber. For ectotherms, body temperature tracks ambient, giving\(Q_{10}\approx 2.5\) sensitivity per 10 K. A Savanna grass-rat faces thermal stress above 40 °C; an elephant faces the opposite problem, dumping excess heat even at a moderate 30 °C because of its low surface-to-volume ratio and poor sweating capacity. This drives the very different thermoregulatory strategies we will examine in Module 2.

An under-appreciated consequence: because metabolic rate sets the timescale of every physiological process (wound healing, growth, neural-firing cost), endotherms in the tropics effectively “live faster” than those in the arctic. The African elephant’s calf-growth rate exceeds its Asian cousin’s, matching the slightly warmer mean ambient temperature. Temperature corrections sharpen every scaling law presented earlier.

Example 1: Daily food requirement of a 4 t white rhino. Applying Kleiber with \(B_0=3.4\) W/kg^(3/4): \(B = 3.4\cdot 4000^{0.75}\approx 1{,}700\) W. Multiplying by 86 400 s/day gives \(E_\text{day}\approx 147\) MJ. Field-metabolic rate (FMR, active above basal) is \(\sim 2.5\times\) BMR, so \(E_\text{FMR}\approx 368\) MJ/day. At 8 MJ/kg gross energy in fresh savanna grass and a digestive efficiency of 40%, required intake is\(\approx 368/(8\cdot 0.4)\approx 115\) kg/day of fresh forage — close to observed field values.

Example 2: Expected elephant lifespan. Charnov gives\(\tau_\text{life}\propto M^{0.20}\) with prefactor ~0.5 yr for a 1 kg mammal. For 6000 kg:\(\tau_\text{life}\approx 0.5\cdot 6000^{0.20}\approx 3.6\) — wait, that’s too small. The correct intercept for placental mammals is \(\sim 9\) yr at 1 kg, giving\(\tau_\text{life}\approx 9\cdot 6000^{0.20}\approx 64\) yr, matching observed ~65 yr for wild African elephants. The lesson is that allometric intercepts are clade-specific; only exponents are universal.

Example 3: Heart rate of a newborn elephant calf. Birth mass\(\approx 120\) kg gives \(f_\text{HR}= f_0\cdot M^{-1/4}\) with\(f_0\approx 240\) min for a 1 kg mammal: \(f_\text{HR}\approx 240\cdot 120^{-0.25}\approx 72\) bpm. Measured neonatal elephant heart rates are 70–90 bpm, in good agreement.

Example 4: Surface-to-volume heat ratio. For geometric similarity,\(S/V \propto M^{-1/3}\). An elephant (6000 kg) compared to a 0.003 kg shrew:\((6000/0.003)^{-1/3}\approx 0.008\) times the shrew’s per-unit-volume skin area. The elephant’s sweating area is \(\sim 125\) times smaller per unit mass than the shrew’s, explaining why elephants evolved radiator ears instead of sweat.

The African savanna hosts an unusually wide spread of body-mass classes in a single biome, from 12 g Karoo hairy-footed gerbils to 6000 kg elephants. Plotting their physiological quantities on log–log axes traces the universal allometric exponents within a single ecosystem. The following visual summarises six of the key scaling relationships simultaneously, using colour-coded predictions from WBE/Biewener/McMahon against observed savanna-species data.

Notice the vertical ordering near the elephant: metabolism is highest (top), lifespan is high, bone radius sits where WBE predicts, and heart rate has fallen to the bottom. Every quantity in a grown animal’s body can be positioned within a couple of tick marks of its theoretical curve.

Species share traits through common descent. A naive OLS regression on species means violates the independence assumption of statistical inference — a concern first formalised by Felsenstein (1985). The remedy is the method of independent contrasts: for every pair of sister taxa on a dated phylogeny, compute the difference in log(B) and log(M) scaled by \(\sqrt{t_i + t_j}\) branch length. The contrasts are statistically independent under a Brownian-motion model of evolution.

When Symonds & Elgar (2002) applied phylogenetic contrasts to the Kleiber data, the estimated exponent flattened from 0.72 to 0.67, closer to the 2/3 value but with broader confidence intervals. Savage et al. (2004) countered that phylogenetic correction may over-correct if metabolic-rate variation is not Brownian. The current consensus is that any exponent in the range 0.67–0.75 is empirically defensible, and that the cleanest test of the WBE prediction is not on metabolic rate but on vascular geometry itself — where modern MRI angiography of 50 mammalian species (Huo & Kassab 2012) confirms the predicted fractal branching ratios to within experimental error.

For the practising biophysicist, the takeaway is: use the exponent that the underlying physics predicts, and bound the intercept by species-specific measurement where possible. An elephant Kleiber estimate differs from measurement by roughly ±10% at most, small enough that feeding budgets, drug doses, and veterinary anaesthesia protocols can be derived directly from the allometric formula.

Hibernators. Marmots and ground squirrels drop body temperature to near 0 °C during torpor; their metabolic rates collapse by a factor of 30, far below the Kleiber prediction. Recovery follows \(Q_{10}\) exponentially. The measurement date therefore matters: BMR should be collected during the active season at thermoneutrality.

Marine mammals. Cetaceans and pinnipeds sit ~30% above the Kleiber line at rest, but during dives their metabolism drops 20–40% below. The elevated resting BMR is a thermoregulatory premium for life in a cold conductive medium; the dive reduction is oxygen-conservation bradycardia. Elephant seals hold their breath for 90 min by dialling BMR to 10% of resting while the heart drops from 90 to 4 bpm — a factor of 22 reduction.

Flying vertebrates. Birds and bats lie systematically ~40% above the mammalian curve during flight. Their basal metabolic rate is comparable, but flight exceeds BMR by 30×. Mass-specific wing-beat power scales as \(M^{1/3}\), which is why 15 kg is close to the upper flight-mass limit for continuous powered flight (the Andean condor and kori bustard sit near this ceiling).

Sauropods and extinct megafauna. Scaling the elastic-similarity column prediction to a 70 t Patagotitan gives femoral radius ~1.0 m, actual ~0.85 m. The discrepancy suggests sauropods pushed their bones closer to yield stress than modern mammals — possibly enabled by the pneumatised skeleton (air sacs reducing effective density) and a more horizontal torso geometry that shortens the moment arms. Woolly mammoths at 10 t fit squarely on the mammalian curve.

Each exception illuminates the rule: when an animal deviates from Kleiber or Biewener, identify which invariant has changed (thermoregulation regime, locomotor medium, bone pneumatisation) and the new scaling prediction follows automatically.

1. Kleiber with uncertainty. Given a 6 t elephant with measured basal metabolic rate 2.1±0.3 kW, compute the intercept \(B_0\) and its uncertainty assuming exponent 3/4 exactly. Discuss whether the measurement is consistent with \(B_0=3.4\) W/kg^(3/4).
2. Heart-rate prediction. Derive the allometric heart-rate equation starting from (i) cardiac output \(\propto M^{3/4}\), (ii) heart mass \(\propto M\), and (iii) the assumption that stroke volume is proportional to heart mass. Predict the resting heart rate of a 100 kg cheetah.
3. Buckling safety factor. For a cylindrical bone modeled as an Euler column, compute the compressive load required to cause buckling of an elephant femur (length 1.3 m, radius 10 cm, Young’s modulus 17 GPa). Compare to the peak stance load of a 6000 kg elephant with safety factor.
4. Gait transition speed. Using Froude number\(\mathrm{Fr}=v^2/(gL)\) and leg length scaling \(L\propto M^{1/4}\), predict the walk–trot transition speed for (a) an impala (50 kg, L=0.6 m), (b) an African elephant (6000 kg, L=1.7 m). Why does the elephant never transition to a true gallop?
5. Gut-transit allometry in the field. An elephant eats 100 kg of grass per day and has gut capacity 600 L at 0.5 kg/L bulk density. Compute mean retention time and compare with the Demment–Van Soest prediction \(\tau\propto M^{0.25}\). Which parameter would you measure in situ to test the theory most stringently?
6. Metabolic ceiling. Compute the daily energy intake required to sustain a 40 t (hypothetical) graviportal mammal, assuming Kleiber scaling and a FMR:BMR ratio of 2.5. How much grass would it need to eat per day? Is such an animal ecologically feasible in African savanna?

Allometric scaling is not a separate discipline; it is the grammar by which every subsequent module of this course will be written. In M1–M2 we will see how Kleiber sets the elephant’s daily forage requirement and how the surface-area limit forces its radiator ears and dust baths. In M3 we will apply Biewener’s constant-stress rule to the giraffe’s absurdly long femur and cervical vertebrae. In M4 we will watch rhinos pushing the elastic similarity exponent past 3/8 at their 2.5-tonne mass. In M5–M6 we will derive the lion and cheetah’s running speed from Froude scaling at the sub-120 kg peak. And in M7 we will see how ungulate herds average their individual metabolic budgets into collective migration energetics scaling as\(N\cdot M^{3/4}\).

The one-line summary of this module: identify the physical invariant (bone stress, vascular dissipation, heat-flux density, buckling safety factor, gut-retention time), and every other physiological allometry follows by dimensional analysis. When experiment disagrees, a new invariant is waiting to be found.