Skip to main content
Back to Biology

ECOLOGY · SAMPLING AND BIODIVERSITY

Biodiversity & Sampling Lab

File ▾

CHOOSE YOUR INVESTIGATION LEVEL

How do you count what you can’t count?

Both levels use the same four stations: a 64-plot meadow or stream bed to survey with quadrats and transects, a field of beetles for mark and recapture, a before-and-after disturbance census, and repeated surveys to measure sampling error.

About this model

What kind of model. Codex’s science core (model.js), with every site, survey and catch made by a seeded random-number generator: the same site or population number always gives the same result. The four taxa (A–D) and their counts are synthetic: a meadow or stream bed of 64 plots of 1 m², not a real survey. They are drawn as local-looking organisms (dandelion, clover, daisy and grass; stonefly and mayfly nymphs, caddisfly larva, pond snail) only so you can tell them apart.

① Patterns and surveys. Uniform: every plot has the same counts. Random: each taxon is scattered independently. Clumped: each taxon is concentrated within two plots of a random centre. These are simplified count generators. You count taxon B: density = count ÷ area sampled, estimate = density × 64 m². Census error = estimate − the full count; it is evidence about sampling, and a correct calculation can still have a large error.

② Mark and recapture. A closed population of 80–250 beetles (set by the population number). Lincoln–Petersen: N ≈ M × n ÷ m. It assumes no births, deaths or movement in or out, lasting marks, complete mixing and equal catchability, all true in the model. With few recaptures the estimate is unstable (and tends high); with zero it is not estimable. No Chapman correction or confidence interval is used. The beetles’ wandering is drawn; which beetles are caught comes from the model.

③ Diversity. Richness = number of taxa present. Simpson complement = 1 − Σn(n−1)/(N(N−1)) (the finite-sample form; undefined for N < 2). Evenness = Shannon H ÷ ln S (Pielou), undefined for S < 2. The disturbance is a supplied rule (each plot keeps ⌊n × (1 − severity × k/3)⌋ of taxon k = 0…3: A untouched, D removed most), not a measured ecological response; because it rounds down in every plot, it removes more than the nominal share where counts per plot are small.

④ Survey effort. The same random survey of n plots is repeated 20 times; RMSE = √(mean of the squared errors). Larger samples tend to have smaller errors, but each 20-survey run is one realised experiment, and error need not fall with every step in n. A full census (64 plots) has zero error.

What am I looking at?

How the lab works. Pick Explore or one lab at the top. Each lab has five tabs: Pre-lab, Bench, Data & graph, Claim · Evidence · Reasoning, and Test. On the Bench, the coach lists your steps; they tick themselves as you work. Press What do I do now? when you’re stuck.

① Survey. A 64 m² site from above, cut into 8 × 8 plots of 1 m² (row numbers on the left). Every organism is drawn. ▶ Survey lays quadrat frames (white and orange squares) on the sampled plots and counts taxon B in each; a transect is a yellow tape along one row. Tap plots to choose your own quadrats. The field sheet on the right does the arithmetic. The graph: census error against area sampled (■ uniform, ● random, ▲ clumped, ◇ transect).

② Mark & recapture. A fenced field of beetles. Catch 1 circles the beetles that get a yellow paint dot; then they are released and mix; catch 2 rings the beetles caught again: ○ unmarked, ☆ marked. The notebook works out the estimate.

③ Diversity. The whole site before and after a disturbance. Bars: each taxon’s count now (filled) and before (dashed outline), with the organism under each bar.

④ Survey effort. The same-size random survey 20 times: each ◆ is one survey’s estimate, the dashed line is the census, the shaded band is census ± RMSE.

Controls. ▶ runs the station; Step › moves one beat (one quadrat, one phase of the capture, a quarter of the disturbance, one survey) and pauses; the speed menu slows it to 0.1× or 0.05×. Keys: Space play / pause · → step · R reset · Z / X zoom. Drag to move a zoomed-in scene.