| Pseudo-absence data are locations of assumed species absence that researchers use in modeling species habitat in lieu of confirmed species absence. The use of pseudo-absence data could make presence-only datasets such as the California Natural Diversity Database (CNDDB) valuable resources for modeling endangered species habitat. However, the use of pseudo-absence data can adversely affect logistic regression habitat models. These effects were investigated by developing a model of Foothill pine in Mount Diablo California State Park using absence data derived from remotely sensed imagery. In a series of model runs, the observed absence data were gradually removed from the model and replaced with random pseudo-absence data. Models with as much as 27% of their absence sample coming from randomly generated pseudo-absence did not show greatly compromised performance. However, models with greater than 27% pseudo-absence under-predicted presence and over-predicted absence. These results suggest using pseudo-absence data to supplement observed absence data might improve the results of a logistic regression habitat model when observed absence data is limited. However, excessive reliance on pseudo-absence data may invalidate model results in a logistic regression habitat model. |
