A presence‑only species distribution modelling method that estimates a relative suitability (or probability distribution under specified sampling assumptions) across a study area by finding the probability distribution of maximum entropy subject to constraints that empirical averages of selected environmental features match those observed at occurrence locations.

Principle

MaxEnt finds the least‑committal (maximum‑entropy) distribution consistent with feature constraints derived from occurrence samples; under certain conditions it is mathematically equivalent to a Poisson point process model and yields relative habitat suitability scores when background sampling is appropriately defined.

Demonstration

Illustrative scenario — Situation: Presence‑only records of a mosquito vector and raster environmental layers are available. Recognition: Absence data are unreliable. Action: Select environmental features, sample background locations to represent available environment, fit MaxEnt, and produce a raster of relative suitability and response curves. Consequence: The output ranks landscape cells by relative suitability; with careful background choice and bias correction it supports targeted surveillance planning, but outputs require calibration before interpreting as absolute probabilities.

Misapplication

Interpreting raw MaxEnt outputs as absolute probabilities of presence without accounting for sampling bias, background selection and prevalence; using an inappropriate background sample that reflects uneven survey effort; or overfitting by including many correlated features without regularisation.

Consequence

MaxEnt is effective for presence‑only data and can deliver robust relative suitability maps with small samples when assumptions are addressed; misapplication yields overconfident or biased suitability maps that can misguide surveillance and control priorities.

Reversal

When reliable presence‑absence or repeated‑survey occupancy data exist, methods that estimate prevalence or detection (e.g., occupancy models, GLMs with known denominator) may be preferable. Strong sampling bias, unsuitable background definition, or violation of feature expectation assumptions require bias correction or alternative modelling approaches.

Boundary

Clearly within: presence‑only occurrence data, environmental covariates and a defensible background sample for available environment. Boundary case: small sample sizes or heavily biased sampling requiring explicit bias correction. Clearly outside: presence‑absence or abundance datasets better handled by methods estimating true probability of presence or density models.

Semantic Tension

Mathematical generality and performance with limited data (maximum‑entropy convenience) ↔ ecological interpretability and dependence on sampling design (background choice and bias).

Synthesis

MaxEnt is a principled presence‑only estimator that produces relative suitability surfaces by maximising entropy under empirical feature constraints; its practical value depends critically on background selection, bias correction and cautious interpretation of outputs as relative rather than absolute measures without calibration.