Table 3. Logistic regression of GKR presence in relation to NDVI.
| Model | AIC | ΔAIC | wi |
| −1.26+6.11*MaxNDVIT1+4.15*MaxNDVIT0 − 23.79*MinNDVIT0 − 3.19*SlopeNDVIT0 −0.25*year2006 − 0.55*year2010+0.44*year2011 | 2,425.8 | 0 | 1 |
| MaxNDVIT1+MaxNDVIT0+MinNDVIT0+year | 2,439.3 | 13.50 | 0 |
| MaxNDVIT1+MinNDVIT0+SlopeNDVIT0+year | 2,452.7 | 26.91 | 0 |
| MaxNDVIT1+MinNDVIT0+year | 2,461.5 | 35.72 | 0 |
| MaxNDVIT0+MinNDVIT0+SlopeNDVIT0+year | 2,502.9 | 77.05 | 0 |
| MaxNDVIT0+MinNDVIT0+year | 2,514.7 | 88.94 | 0 |
| MaxNDVIT1+year | 2,623.9 | 198.15 | 0 |
| MaxNDVIT1+SlopeNDVIT0+year | 2,624.8 | 199.03 | 0 |
| MaxNDVIT1+MaxNDVIT0+year | 2,625.2 | 199.35 | 0 |
| MaxNDVIT1+MaxNDVIT0+SlopeNDVIT0+year | 2,626.1 | 200.34 | 0 |
| MaxNDVIT0+year | 2,702.4 | 276.56 | 0 |
| MaxNDVIT0+SlopeNDVIT0+year | 2,703.6 | 277.82 | 0 |
The full model performed the strongest. In this model, GKR presence is positively correlated with peak primary productivity in the current year and previous year, and negatively correlated with minimum NDVI and NDVI slope in the previous year. This suggests that the best predictor of GKR presence in a given year is a positive correlation with resource abundance over two years, and presence in the area the previous year.