Table 1. Summary of variable importance in conditional inference random forest models.
| Variable name | Rank | Models | Variable description |
| ndvi_jun4 | 1.4 | 5 | NDVI, June, 4 years previous |
| ndvi_jun1 | 2.3 | 11 | NDVI, June, previous year |
| ndvi_juna1 | 2.4 | 5 | NDVI, first half of June, previous year |
| ndvi_maya3 | 3.7 | 3 | NDVI, first half of May, 3 years previous |
| ndvi_sepa1 | 4.6 | 5 | NDVI, first half of September, previous year |
| ndvi_esummer4 | 4.7 | 7 | NDVI, early summer, 4 years previous |
| ndvi_esummer1 | 5.2 | 16 | NDVI, early summer, previous year |
| ndvi_jun0 | 5.3 | 12 | NDVI, June, previous year |
| ndvi_juna4 | 5.5 | 2 | NDVI, first half of June, 4 years previous |
| ndvi_junb4 | 5.5 | 2 | NDVI, second half of June, 4 years previous |
| ndvi_juna5 | 7.0 | 1 | NDVI, first half of June, 5 years previous |
| ndvi_may3 | 8.1 | 7 | NDVI, May, 3 years previous |
| ndvi_apr3 | 9.0 | 1 | NDVI, April, 3 years previous |
| ndvi_auga2 | 9.3 | 4 | NDVI, first half of August, 2 years previous |
| ndvi_lsummer5 | 9.3 | 4 | NDVI, late summer, 5 years previous |
Only the top 15 variables (out of 270 total potential predictors) are shown. Variables are ordered by the mean rank (from node purity) computed by the random forest algorithm; the third column gives number of models across which this mean was computed.