Describing and measuring biodiversity · Lesson 4

Detection Limits and Non-detection

5 min read

Lesson objectives

  • Distinguish confirmed detection, effort-bounded non-detection and uncertain identity in a visit ledger.
  • Design a comparable revisit before generalising a species presence or absence judgment.
Reading tools

“Not detected” describes a visit, method and effort. It is not the same as “absent from the site.” An organism can be missed because it is seasonal, hidden, scarce, or searched for at an unsuitable time or place. Define the sampling unit first: a plot, route, water point or camera image. Then record its date, duration, method, weather and identification quality. A zero from a daytime bird survey cannot be combined with a zero from a plant inventory to assert that a third organism is absent. One confirmed occurrence establishes an observation at that unit and time; it estimates neither population size nor occupation of the entire habitat. Species-occurrence data cleaning principles help distinguish confirmed identity and valid location and date from records requiring review. A cleaned occurrence record is still not an exhaustive census.

Before comparing sites, ask whether identification followed the same method, whether the organism could have been visible on each visit, and whether effort, season and area were comparable. Record three practical states: confirmed detection, no detection under stated effort, and uncertain identity or location. Do not silently recode the third state as either a detection or a zero. Preserve zeros and ambiguous photographs in the visit table with separate quality flags. Dropping zeros biases the record, while the fraction of visits with a detection is not the true probability of presence. For a consequential decision, repeat visits during a suitable detection window and settle identity before drawing a habitat boundary or claiming absence.

Lesson summary

This lesson shows how detection window, effort and identification quality limit species occurrence records. Learners keep confirmed detection, effort-bounded non-detection and unresolved identity as distinct visit states. They compare plots with unequal seasons and effort, then design a valid revisit instead of turning a zero into categorical absence or a detection fraction into presence probability. The outputs are a detection ledger and a resurvey memo with worked solutions and stated limits.