Automated Classification:
The automated portion of the classification methodology utilized the Feature Analyst extension for ArcGIS. Feature extractions were performed by providing the alogorithm with examples of impervious features to extract from the Quickbird imagery. Once the extractions were completed an internal review was performed by C-CAP analysts at NOAA CSC.
Classes
0 - Pervious 1 - Impervious
Errors or areas that could not be consistently extracted from the imagery were resolved through manual edits and using ancillary data sets. In some areas of the islands, impervious surface features are obscured by clouds, shadows or tree canopy. For these types of issues NOAA CSC relied on acillary imagery and vector data provided by the CNMI Coastal Zone Management Program to fill in missing features. Due to spectral confusion between impervious surfaces and other features such as bare land manual editing was performed in Erdas Imagine to correct comission errors within the map. The same technique was used to capture missing impervious features that could not be derived from ancillary data sources.