Hive pallets sit in an orchard block. Bees fly out with an exponentially / Gaussian decaying distance; each tree's fruit set saturates with visits; the contract pays only if the extra crop beats the rent.
The science (equations and honest simplifications)
1. Where bees go. A foraging bee lands at distance r from its hive with a 2-D probability density per m^2 that integrates to 1: exponential K(r)=exp(-r/λ)/(2πλ^2) with mean distance 2λ, or Gaussian K(r)=exp(-r^2/2σ^2)/(2πσ^2) with mean σ√(π/2). The fraction of visits inside radius R has closed forms 1-e^(-R/λ)(1+R/λ) and 1-e^(-R^2/2σ^2). Honey bees can fly 1-3 km but, with abundant bloom, most visits are within a few hundred metres; the slider sets the mean distance.
2. Visit field. With np hives at pallet p: v(x)=κ·s·fT·104·Σp np K(|x-xp|) effective visits per flower per bloom. A uniform density of h hives/ha gives v=κ s fT h. With no hives nearby the block is isolated and bees that fly outside it are lost, so edges are under-visited. With neighbours stocked alike (default, a large orchard district) the pattern is treated as periodic (3x3 image blocks, rescaled so the mean visit rate is exactly N/L2): bees leaving one block are replaced by bees arriving from the next, and only the clumping of pallets matters.
3. Colony strength and weather. Foragers scale with frames of bees: s=frames/10. Flight needs warmth: fT=smoothstep(9–19 °C)·(1-0.8·smoothstep(32–42 °C)). This is a smooth stand-in for hours-per-day; rain and wind are not modelled.
4. Fruit set. Pollination-limited saturation: u=1-exp(-v/v0), fruit set S=Y0+(1-Y0)u, where Y0 is the share set without managed bees (wind, wild insects, self-pollination). Because the curve is concave, uneven coverage loses yield (Jensen).
5. Calibration (an assumption, flagged). κ is set per crop so that the recommended stocking density, with a 10-frame colony in warm weather and perfectly even spread, gives u=90%. So v0, the recommended densities and Y0 are ballpark values for teaching, not field data or agronomic advice. Prices and rents are illustrative.
6. Money. Grower: gain=A·Ypot·P·(S̄-Y0), net=gain-R·N. Beekeeper: margin=N(R-c), with cost c = $60 colony upkeep/loss allowance + trucking $0.012/km (illustrative). Break-even rent for the beekeeper is c; the grower's break-even rent is gain/N.
7. Optimal stocking. Under even spread, net(h)=G(1-e^(-ah))-Rh with a=ln10·s·fT/hrec and G=YpotP(1-Y0). Marginal gain equals marginal cost at h*=ln(G a/R)/a (0 if G a<R). The graph's optimum for the real hive layout is found numerically; it differs from h* because of edge leakage and clumping.
Not modelled: competing bloom, hive health/disease, nectar flow, neighbouring blocks, bumblebees, pollenizer varieties, multi-year effects, contract risk. Trees and hives are drawn larger than life; 1 world unit is shown in the results as metres. Hive and orchard meshes/textures were made in Blender (Cycles-baked) and rendered with Three.js.
| Crop (ballpark) | hives/ha | v0 | Y0 | t/ha | $/t |
|---|