Species distribution model (SDM)

A species distribution model predicts the geographic range where a species is likely to occur by statistically relating places it has actually been recorded to environmental predictors such as temperature, rainfall, elevation, soil type and land cover. The output is usually a continuous suitability surface across a map, which can be thresholded into a binary predicted-range boundary.

Models are fitted from occurrence records, often paired with background or pseudo-absence points, using methods ranging from logistic regression to maximum entropy (MaxEnt) and random forests. This differs from an expert-drawn range map, which encodes field knowledge rather than a fitted statistical relationship, and from habitat fragmentation analysis, which measures how broken up existing habitat is rather than predicting where a species could live. Results are commonly displayed and queried inside a geographic information system.

SDMs support conservation prioritization, invasive-species risk screening, and biodiversity net gain assessments, and increasingly combine with Environmental DNA survey data to validate predicted ranges against direct evidence. The main pitfalls are sampling bias, occurrence records cluster near roads and research stations, and extrapolation, projecting a model into environmental conditions well outside the range it was trained on produces unreliable output even though the map looks confident.