How Do You Count a Whole Meadow? Count One Square and Multiply

The device is simply a square frame of known area, commonly half a metre or a metre on each side, sometimes divided into a grid of smaller squares by strings. Placing it on the ground defines a boundary, and everything inside that boundary is recorded, whether by counting individuals, by estimating the percentage of ground each species covers, or by recording presence and absence. Repeating that at many positions and averaging gives an estimate for the whole area, together with a measure of how variable it is, which is the point of taking many samples rather than one large one.

Several measures are used and each answers a different question:

• •Density, meaning the number of individuals per unit area
• •Percentage cover, which suits plants that cannot be counted individually
• •Frequency, meaning the proportion of frames a species appears in
• •Presence and absence only, which is fastest and least informative
• •Abundance on a graded scale, which trades precision for speed
• •Height or biomass, where structure matters as much as identity

The single largest source of error is where the frames are put, and the temptation to place them is what the method exists to resist. A person choosing positions will unconsciously favour interesting patches, or avoid nettles, or keep near the path, and the resulting estimate is biased in a direction nobody can quantify afterwards. Genuine random placement requires generating coordinates and going to them regardless of what is there. Where a site has recognisable zones, stratified sampling divides it and samples each in proportion. Where a gradient is the subject, frames are placed at intervals along a line across it, which answers a different question and should not be averaged as though it were random.

Deciding how many samples to take is a statistical question with a practical answer that fieldworkers use routinely. Recording the running total of species found as frames accumulate produces a curve that rises steeply at first and then flattens, and the point at which it flattens indicates that further frames are adding little, which is a species accumulation curve and is drawn in the field as the work proceeds. For abundance rather than species lists, the running mean is plotted instead and sampling continues until it stabilises. Both are rough, and a formal calculation from the variability of an initial sample gives a defensible number where the result has to withstand scrutiny.

Several difficulties are easy to overlook and each biases results systematically. Deciding whether a plant rooted outside the frame but overhanging it counts requires a rule applied consistently, since different observers choose differently. Estimating percentage cover by eye is notoriously unreliable, with the same patch assessed differently by different people and by the same person on different days, and training with reference images reduces but does not remove that. Mobile animals move in and out while being counted. And the frame size itself determines what can be detected, since a small frame misses rare species and a large one is impractical to work through carefully.

A frame of known area defines a boundary within which everything is counted, and many such samples averaged give an estimate for a whole site along with its variability. Where the frames go is the largest source of error, since anybody choosing positions favours interesting ground unconsciously, so coordinates are generated and followed. Percentage cover estimated by eye varies between observers and between days.