Spotlight #63 – Applied Science: Engineering, Medicine & Agriculture in the Browser

I. Retaining Wall — From Earth Pressure to a Factor of Safety

Soil is not a solid block. Pile it up and it wants to slump to its natural angle of repose; a retaining wall exists to hold that mass in place. The first question an engineer asks is: how hard does the soil push? Rankine’s earth-pressure theory, from 1857, answers it with a single coefficient derived from the soil’s internal friction angle.

The active coefficient K_a falls as the friction angle rises — loose, slippery soil pushes harder than dense, gravelly soil. Because the pressure increases linearly with depth, integrating it over the wall height gives a triangular distribution whose resultant thrust acts at one-third of the height from the base. That single force, and where it acts, drives everything else.

K_a = tan^2(45 - phi/2)            (active earth-pressure coefficient)
P_a = 0.5 * K_a * gamma * H^2      (resultant thrust per metre of wall)
     acting at height H/3 above the base

Here is the design philosophy that surprises newcomers: you do not compute one safety margin, you compute several, because a wall can fail in genuinely different ways. It can rotate about its toe (overturning), it can slide forward bodily (sliding), or it can crush the soil beneath one edge (bearing). Each failure mode has its own free-body diagram and its own factor of safety, and they are checked separately because nothing about passing one guarantees passing another. A wall heavy enough not to slide may still be too narrow to resist overturning.

Convention sets different thresholds for each: typically a factor of at least 2.0 against overturning and at least 1.5 against sliding. The simulation also tracks the eccentricity of the resultant on the base — if it strays outside the middle third, part of the base lifts into tension, which masonry and soil cannot sustain.

FS_overturning = M_resisting / M_overturning   >= 2.0
FS_sliding     = (mu * N) / P_a                >= 1.5
eccentricity e = B/2 - M_net/N;  keep e <= B/6  (middle third)

II. Haemodialysis — How a Membrane Replaces a Kidney

A healthy kidney filters the entire blood volume many times a day, removing urea and other nitrogenous wastes. When kidneys fail, a dialyser does the job with a bundle of hollow semipermeable fibres: blood flows inside the fibres, a clean dialysate fluid flows outside, and urea diffuses across the membrane down its concentration gradient.

The single most important design choice is to run the two fluids countercurrent — in opposite directions. If they flowed the same way, the concentrations would equalise partway along and diffusion would stall. Flowing in opposition, the blood always meets dialysate that is slightly cleaner than itself, so a gradient persists along the whole length of the fibre and clearance is maximised. It is the same trick fish gills and mammalian kidneys evolved.

Clinically, waste removal is described by single-pool kinetics: treat urea as if it were dissolved in one well-mixed volume, and its concentration decays exponentially as the dialyser clears it.

C(t) = C0 * exp(-K * t / V)
  K = dialyser clearance (mL/min)
  t = session duration
  V = urea distribution volume (≈ total body water)

Two numbers tell the clinician whether a session was good enough. The dimensionless Kt/V bundles clearance, time, and patient size into one adequacy index; a per-session target of at least 1.2 is widely used. The urea reduction ratio, the simpler bedside measure, is just the fractional drop in blood urea from start to finish. The simulation links the three knobs — clearance, time, and volume — to both outputs, so you can see why a larger patient needs either a faster dialyser or a longer session to reach the same adequacy.

Kt/V = -ln(C_end/C0)   (single-pool, simplified);  target >= 1.2
URR  = (C0 - C_end) / C0 * 100%

III. Photosynthesis Light Response — The Law of Limiting Factors

Shine more light on a leaf and it fixes more carbon — but only up to a point. The light-response curve has three regimes. In dim light the rate rises almost linearly, limited purely by how many photons arrive; its slope is the quantum yield. As light strengthens the curve bends over and saturates, now limited by the downstream biochemistry rather than by photons. And at the very bottom, below the compensation point, respiration releases more CO₂ than photosynthesis fixes, so the leaf is a net source.

The simulation models net assimilation as a non-rectangular hyperbola minus a constant dark-respiration term, marking both the compensation point (net rate zero) and the saturation point (the curve has flattened).

A_net = (phi*I * A_max) / (phi*I + A_max) - R_d
  phi   = quantum yield (mol CO2 / mol photons)
  I     = incident photon flux
  A_max = light-saturated assimilation rate
  R_d   = dark respiration
Compensation point: A_net = 0;   Saturation: dA/dI -> 0

This is a clean illustration of Blackman’s 1905 law of limiting factors: when a process depends on several inputs, the rate is set by whichever input is in shortest supply. At dawn, light limits; at midday, light is plentiful and CO₂ or temperature takes over as the bottleneck. Pushing the abundant factor harder does nothing — you must lift the scarce one.

The C3 and C4 presets show why this matters agronomically. C4 plants — maize, sugarcane, many tropical grasses — concentrate CO₂ around their fixing enzyme, suppressing the wasteful photorespiration that plagues C3 plants in bright, hot conditions. The result is a higher saturation plateau and better water-use efficiency exactly where C3 crops struggle.

IV. Irrigation & Soil Water Balance — The Bucket Model

To an irrigation engineer, the crop root zone is a bucket. Water in — rainfall and irrigation; water out — evapotranspiration, deep drainage past the roots, and surface runoff. The change in stored soil water over any interval is just the difference. Scheduling irrigation is the art of refilling the bucket before the crop runs dry, without over-filling it and wasting water to drainage.

delta_S = P + I - ET - D - R          (soil-water balance)
  P,I = precipitation, irrigation;   D,R = drainage, runoff
ET    = ET0 * Kc                      (reference ET x crop coefficient)

The bucket has two important marks. Field capacity is how much water the soil holds a day or two after a soaking, once gravity has pulled the excess away. The wilting point is the level below which roots can no longer overcome the soil’s grip on the remaining water. The difference between them, multiplied by root depth, is the total available water the crop can actually use.

You do not wait until the wilting point to irrigate — the crop would be stressed long before. Instead a management-allowed depletion fraction sets a comfortable trigger, commonly around half of the available water, at which the next irrigation is scheduled. Crop demand itself is the weather-driven reference evapotranspiration scaled by a crop coefficient that rises and falls through the growing season.

TAW = (theta_FC - theta_WP) * root_depth   (total available water)
Trigger irrigation when depletion >= MAD * TAW   (MAD ≈ 0.5)
Deficit irrigation: apply < full demand to lift water-use efficiency

The simulation also explores deficit irrigation, where a grower deliberately supplies less than full demand. Yield falls, but often far less than the water saved — so the water-use efficiency, the crop produced per unit of water, can rise. In water-scarce regions that trade-off is the whole game.

V. El Niño (ENSO) — A Quasi-Periodic Climate Swing

El Niño is not a weather event but a slow rearrangement of the tropical Pacific that ripples through climate worldwide. It emerges from a tight coupling between ocean and atmosphere first described by Jacob Bjerknes: warm sea-surface temperatures in the east weaken the easterly trade winds, the weaker winds let the thermocline flatten, and the flatter thermocline brings still warmer water to the surface. Left alone, that positive feedback would run away.

What stops it — and turns a runaway into an oscillation — is delay. Slow ocean waves, set off by the same wind changes, cross the Pacific basin, reflect off its western boundary, and return months later carrying the opposite signal. The delayed-oscillator model captures this with a single delay-differential equation: a local growth term, a delayed negative feedback, and a cubic term that caps the amplitude.

dT/dt = a*T(t) - b*T(t - tau) - eps*T(t)^3
  a   = Bjerknes coupled growth (positive feedback)
  b   = delayed wave feedback (negative), arriving after lag tau
  eps = cubic saturation, bounding the amplitude

The interplay of growth, delay, and saturation produces a self-sustaining swing with a period of roughly two to seven years. But it is quasi-periodic, not clockwork: the equation sits close to chaotic regimes, and small perturbations from weather noise nudge each cycle’s timing and strength. That sensitivity is precisely why ENSO forecasts beyond a year remain hard, and why the “spring predictability barrier” frustrates seasonal forecasters every year.

Operationally, events are declared using the Oceanic Niño Index — a running mean of sea-surface temperature anomalies in a defined Pacific region. Cross +0.5°C for long enough and it is an El Niño; cross −0.5°C and it is a La Niña. The simulation tilts the thermocline in real time and tracks the index across that threshold.

The Common Thread

Five very different fields — geotechnics, nephrology, plant physiology, agronomy, and climate dynamics — and the same intellectual move in each: take a physical mechanism, write it as an equation, and then read a decision off the result. A factor of safety, a Kt/V, a limiting factor, a depletion trigger, an index crossing a threshold. The equations are the easy part; knowing which number to watch is the engineering. Open any of the five and the answer changes as you turn the dials.