Table 2. Selected NPMR models for each ponderosa pine (Pinus ponderosa) variety and haplotype, with final predictors, response variable sensitivity to predictors (Q), evaluation metrics (LogB, xR², AUC), and number of presence / absence points (see S1 Table for additional model evaluation metric results).
| Model | Predictors (Q)1 | LogB | xR² | AUC | Presence/Absence |
|---|---|---|---|---|---|
| var. ponderosa | SHMI (0.62) ELEV (0.28) PRATIO (0.21) WINP (0.21) | 1165.3 | 0.78 | 0.99 | 1059/8941 |
| var. scopulorum 1 | PRATIO(0.56) MTWM (0.42) SUMP (0.41) ELEV (0.11) | 827.2 | 0.62 | 0.98 | 861/9139 |
| var. scopulorum 2 | MTWM (0.49) SUMP (0.34) TRIX (0.09) ELEV (0.09) | 830.1 | 0.62 | 0.98 | 861/9139 |
| Haplotype 1 | SDAY(0.11) SHMI (0.10) PRATIO (0.06) WINP (0.03) | 47.6 | 0.25 | 0.93 | 45/3125 |
| Haplotype 2 | SMRPB (0.04) TDIFF/LogGSP (0.04) | 17.5 | 0.34 | 0.78 | 14/3098 |
| Haplotype 3 | LogMAP (0.10) ELEV (0.08) SUMP (0.04) | 43.2 | 0.21 | 0.95 | 45/3098 |
| Haplotype 4 | PRATIO (0.04) SMRPB (0.04) | 14.3 | 0.14 | 0.99 | 9/3098 |
| Haplotype 5 | PRATIO (0.06) LogGSP (0.06) | 34.1 | 0.26 | 0.99 | 22/3125 |
| Haplotype 6 | MTCM (0.12) TDIFF/LogGSP (0.07) WINP (0.02) | 23.7 | 0.06 | 0.93 | 34/3098 |
| Haplotype 7 | SHMI (0.10) PRATIO (0.07) SMRPB (0.07) | 26.3 | 0.13 | 0.99 | 23/3098 |
| Haplotype 8 | MTWM (0.08) ELEV (0.06) WINP (0.02) | 26.2 | 0.14 | 0.96 | 23/3125 |
| Haplotype 9 | ELEV (0.02) PRATIO (0.01) | 7.0 | 0.04 | 0.83 | 5/3125 |
| Haplotype 10 | PRATIO (0.01) ELEV (0.01) | 4.1 | 0.06 | 0.99 | 3/3125 |
1Models with elevation (ELEV) or topographic roughness index (TRIX) as predictors are only shown when LogB improved by greater than 5% over the climate-only model. Refer to Table 1 for predictor abbreviation descriptions.
Q = sensitivity of response variable after adjusting predictor values (by +/- 5% of their ranges), expressed as the proportion of the range of response variable, and generated by evaluating original “un-tuned” models. LogB = log10(likelihood ratio), indicating improvement of a new model over the naive model, presented for un-tuned haplotype models and aggressively tuned variety models. xR² = cross-validated R² (the cross-validated coefficient of determination, with a maximum of 1.0, and no theoretical minimum), representing evaluation of the un-tuned haplotype models and the aggressively tuned variety models. AUC = Area under Receiver Operating Characteristic, calculated for each haplotype model and for the aggressively tuned variety models.