Each sample was analysed for
To examine reliability of the vegetation mapping, 236 sampling points were randomly assigned. The number of validation points assigned to each basin was proportional to its size. Care was taken to ensure that the areas validated were representative of PFT coverage across the basin as a whole. Each point was uploaded onto a Garmin eTrex10 GPS and validated in the field during summer 2013. The pixel resolution of the CIR imagery (0.5 m) was less than the GPS position accuracy (~ 5 m) and to account for this, the PF-04880594 vegetation within a 5 m radius of the given point was recorded. Subsequently, in ArcGIS 10, a 5 m buffer was placed around each validation point and the percent coverage of each vegetation class falling within progesterone extracted. Using these data, a confusion matrix was generated, which cross correlates remote sensing classifications with points validated in the field, this enables the quantification of total error and misclassification rates for each PFT class (Campbell and Whynne, 2011). Validation points were deemed accurate if the majority class noted in the field matched the majority class extracted from the vegetation map.