Abstract: Characterization of weed size using remotely sensed data could provide a useful means of scheduling post-plant weed control in plantation forests. Data from four forestry stands, which included field measurements of weed competition and multispectral data collected from an unmanned aerial vehicle (UAV) and satellite platform (WorldView3), were used to (i) develop models to predict overall weed percentage cover (C), height (H), and a competition index, CI, (C × H/100) derived from these two metrics; and (ii) determine the optimal spatial resolution for imagery from these two platforms for characterizing these measures of weed competition.
Bivariate analysis of the three weed metrics showed the optimal spatial resolution to be 5 m for the WorldView3 data and 1 m for data obtained from the UAV. The best multiple regression models developed to predict C, H, and CI had coefficient of determinations (R2) of, respectively, 0.33, 0.68, and 0.56 for data derived from WorldView3 and, respectively, 0.56, 0.79, and 0.71 for data derived from the UAV. Vegetation indices, derived from the multispectral UAV data and satellite imagery, were strong predictors of weed metrics for all but one of the six models developed. These results suggest that it may be possible for forest managers to use satellite imagery and UAV data to schedule weed control operations.
Authors: Michael S. Watt, Marie Heaphy, Andrew Dunningham, Carol Rolando
Associations: Scion, New Zealand