At the lookout tower sites (where the understory GGS records were available), we extracted both of the AGDD and NDWI-values during the period of 2006 to 2008. To determine and validate the predictor-specific thresholds, we divided these datasets into two groups for calibration (consisting of all data acquired during 2006, which would be of the entire dataset) and validation (consisting of all data acquired during 2007 to 2008, which would be of the entire dataset). In the calibration phase (i.e., determining the understory GGS threshold), there were two steps. First, we computed an average and standard deviation for both of the AGDD and NDWI-values during the understory GGS period. Then we considered these average values as initial threshold values for understory GGS, where equal or greater amount of AGDD and NDWI would be the least requirement for understory GGS occurrence. Second, we varied the initial thresholds over a range of “ standard deviation” in increments of one-third standard deviation in order to determine the predictor-specific best thresholds for understory GGS. We selected the best thresholds as values providing the best agreement between predicted and ground-based observations of understory GGS. Note that we used “” and “” signs for premature and late predictions, respectively, compared to the ground-based understory GGS periods throughout the remaining article. For example, the 0, , and period of deviation meant that the: (1) MODIS-based prediction and ground-based observation fell in the same period, (2) MODIS-based prediction was 1 period earlier than that of ground-based observation, and (3) MODIS-based prediction was 1 period delayed from that of ground-based observation, respectively.