Table 3. Model selection results for linear mixed-effects models optimized on pairwise genetic differentiation.
| Surface | K | Avg. rank | Ω̄ | π̂ |
|---|---|---|---|---|
| Village | 4 | 1.865 | 0.375 | 0.592 |
| Fire density | 4 | 2.393 | 0.243 | 0.288 |
| TPI | 4 | 3.537 | 0.109 | 0.089 |
| Distance | 2 | 4.470 | 0.048 | 0.016 |
| Rugged | 4 | 4.611 | 0.085 | 0.014 |
| Historical forest | 7 | 5.428 | 0.047 | 0.001 |
| Railroad | 3 | 6.280 | 0.047 | 0.000 |
| Current forest | 4 | 7.417 | 0.046 | 0.000 |
Abbreviations: Avg. rank, average model ranking from 10 000 bootstrap iterations; K, number of parameters in the transformation of continuous surfaces or number of land cover classes in categorical surfaces; TPI, topographic position index; Ω̄, Akaike weight representing the probability that a model is the best in the model set, averaged over 10 000 bootstrap iterations; π̂, proportion of 10 000 bootstrap iterations that a model was the top model. Only surfaces that were selected as the top model more frequently than distance alone were used to develop composite models.