PI method | IC methods | ||||||
---|---|---|---|---|---|---|---|
High-resolution imagery (1 m) | Low-resolution imagery (30 m) | ||||||
Study area tree canopyb | Study area water | Study area impervious surface | Study area non-tree vegetation | Study area tree canopy | Census block tree canopyc | Study area tree canopyd | |
Land cover estimate | 14.7% | 1.1% | 31.1% | 45.6% | 16.1% | 11.3% | 6.7% |
Producer’s accuracya | - | 100.0% | 100.0% | 90.2% | 96.9% | - | - |
User’s accuracya | - | 100.0% | 89.8% | 97.4% | 96.9% | - | - |
Overall accuracy | - | 96.0% | - |
↵a Producer’s accuracy indicates the probability that a pixel is correctly classified by the computer algorithm while user’s accuracy measures the probability that a classified pixel actually matches what it is on the ground (Campbell and Wynne 2011).
↵bWeighted mean of ten independently replicated runs of i-Tree Canopy at varying point sample sizes (see Table 1).
↵cTree canopy cover estimate derived from i-Tree Landscape using three Census Block Groups that overlapped the study area (4.51 km2).
↵dTree canopy cover estimate for the study area (3.58 km2) manually extracted from the 2011 NLCD USFS Tree Canopy data that comprises i-Tree Landscape.