presenceonly

Presence-only refers to datasets that record only locations where a species has been observed, with no records of where the species was searched for and not found. Such data are common in natural history collections, biodiversity surveys, and citizen science platforms, and are often used to infer species distributions when true absence data are unavailable.

In presence-only species distribution modeling, the goal is to estimate the environmental conditions associated with occurrences and to predict where suitable habitat exists. Because absence data are missing, models rely on background or pseudo-absence samples drawn from the study region to contrast with presence records. The resulting outputs typically indicate relative suitability rather than an absolute probability of presence.

Common methods include maximum entropy models (MaxEnt), Poisson point process approaches, and other algorithms adapted to presence-only data. These models link presence to environmental covariates such as climate, topography, and land cover, often with regularization to avoid overfitting.

Interpretation and limitations: presence-only models are sensitive to sampling bias and geographic bias in the presence data, and the choice of background data can strongly influence results. They cannot directly infer true absences and should be validated with independent data or using cross-validation; results are best treated as proxies for habitat suitability rather than confirmed occupancy.

Data sources commonly used include global biodiversity databases and citizen-science platforms such as GBIF and iNaturalist, as well as museum collections. Presence-only modeling is a core tool in ecology, biogeography, and conservation planning, informing habitat protection and niche characterization.