Figure 1.
Study area with soils surveys (in black) in the Laurentian Mixed Forest (black and grey) of Minnesota.
Table 1.
Environmental variables for modeling.
Table 2.
The AUC values using reserved polygons without present cases as pseudoabsences, for models with random pseudoabsences (rand), random pseudoabsences followed by pseudoabsences with probabilities <25% (rand_2), pseudoabsences from surveyed plots (surv), surveyed pseudoabsences followed by pseudoabsences with probabilities <25% (surv_2).
Figure 2.
The AUC values compared to area (fraction of total area) predicted as present by species.
Random pseudoabsences (rand) had high AUC values and little difference in area by species albeit extremely parsimonious areas, whereas both random pseudoabsences followed by pseudoabsences with probabilities <25% (rand_2) and surveyed pseudoabsences followed by pseudoabsences with probabilities <25% (surv_2) had a range of AUC values and area predicted as present although all predicted areas were large, and pseudoabsences from surveyed plots (surv) had a range of AUC values and area predicted as present.
Table 3.
Count of FIA surveys (2004–2008), mean predicted probabilities, and area (fraction of total area) of polygons with predicted probabilities ≥75% for models with random pseudoabsences (rand), random pseudoabsences followed by pseudoabsences with probabilities <25% (rand_2), pseudoabsences from surveyed plots (surv), surveyed pseudoabsences followed by pseudoabsences with probabilities <25% (surv_2).
Table 4.
Correlation (all species combined) among predicted probabilities for models with random pseudoabsences (rand), random pseudoabsences followed by pseudoabsences with probabilities <25% (rand_2), pseudoabsences from surveyed plots (surv), surveyed pseudoabsences followed by pseudoabsences with probabilities <25% (surv_2).
Figure 3.
Predicted probabilities (black indicates areas of ≥75% predicted probability) for uncommon yellow birch (black circles indicates approximate location) by models with a) random pseudoabsences, b) random pseudoabsences followed by pseudoabsences with probabilities <25%, c) pseudoabsences from surveyed plots, d) surveyed pseudoabsences followed by pseudoabsences with probabilities <25%.
Figure 4.
Predicted probabilities (black indicates areas of ≥75% predicted probability) for widespread aspens by models with a) random pseudoabsences, b) random pseudoabsences followed by pseudoabsences with probabilities <25%, c) pseudoabsences from surveyed plots, d) surveyed pseudoabsences followed by pseudoabsences with probabilities <25%.
Figure 5.
Predicted probabilities (black indicates areas of ≥75% predicted probability) for jack pine by models with a) random pseudoabsences followed by pseudoabsences with probabilities <25% and b) pseudoabsences from surveyed plots.
Figure 6.
Predicted probabilities (black indicates areas of ≥75% predicted probability) for white cedar by models with a) random pseudoabsences followed by pseudoabsences with probabilities <25% and b) pseudoabsences from surveyed plots.
Figure 7.
Predicted probabilities (black indicates areas of ≥75% predicted probability) for balsam fir by models with a) random pseudoabsences followed by pseudoabsences with probabilities <25% and b) pseudoabsences from surveyed plots.