News & Events

Programs

Recycle

Donate

Volunteer

About KIB

Internships

Español

Tree Canopy Mapping Procedures

Data

New tree canopy cover estimates for Center Township, Indianapolis were developed using high resolution satellite imagery and digital aerial photography. Satellite imagery collected on April 25, 2005 from the QuickBird satellite (Digital Globe, Inc.) were provided through a grant from the Institute for Application of Geospatial Technology (IAGT). The QuickBird sensor collects multispectral imagery in the blue, green, red, and near infrared portions of the electromagnetic spectrum at a spatial resolution of 2.4 m (8 ft) and panchromatic (black and white) imagery at a spatial resolution of 61 cm. Spectral and spatial resolutions of the QuickBird bands are summarized in the table below.

 

Band	Spatial Resolution	Spectral Resolution
1	2.4 m (8ft)	450 - 520 nm (blue)
2	2.4 m (8ft)	520 - 600 nm (green)
3	2.4 m (8ft)	630 - 690 nm (red)
4	2.4 m (8ft)	760 - 900 nm (NIR)
Pan	61 cm (24 in)	445 - 900 nm (blue - NIR)

Large scale digital aerial photography was used to enhance the satellite image interpretation of tree canopy cover. Natural color air photos collected in 2004 at a spatial resolution of 15 cm (6 in) were provided by the Indianapolis Mapping and Geographic Infrastructure System (IMAGIS) through the Indiana Spatial Data Portal at Indiana University. The figures below provide an overview of these two image types at a greatly reduced scale in order to show the complete township.
 

Overview of color infrared QuickBird satellite image (2005).

Overview of color aerial photo mosaic (2004).

Methods

Tree canopy mapping was conducted in two phases. In phase one, digital image processing techniques were used to group pixels in the QuickBird satellite imagery into fifty categories based on similarity in spectral response patterns. Clustering was implemented using the Iterative Self Ordering Data Analysis Technique (ISODATA) unsupervised classification algorithm. The resulting spectral clusters were assigned to one of three categories based on visual interpretation of the imagery and spectral signatures: tree canopy, other, and mixed (tree canopy confused with other land cover types). Pixels in the mixed category were subjected to an additional ISODATA classification to further separate tree canopy from other land cover types. While the digital image processing techniques performed reasonably well, close inspection of the resulting map revealed areas were the algorithm incorrectly classified tree canopy as other land cover types (errors of omission) or other land cover types as tree canopy (errors of commission).

In phase two, the tree canopy map was refined using visual interpretation and on-screen digitizing in a geographic information system (GIS). The tree canopy map developed in phase one was overlaid on the 15 cm aerial photography. A remote sensing analyst visually compared the tree canopy map with the higher resolution aerial photography. Errors resulting from the digital classification process were manually corrected by changing the pixel values to the appropriate category (tree canopy or other).

The figures below illustrate the appearance of tree canopy cover in the QuickBird imagery, aerial photography, and the resulting tree canopy map for a small portion of the study area centered on Brookside Park, a few miles east of downtown Indianapolis. While the spatial resolution of the QuickBird imagery is coarser than the photography, QuickBird's infrared band provides enhanced discrimination between vegetation and other land cover types for digital image processing techniques. The improved spatial resolution of the aerial photography provides enhanced detail for visual interpretation. Thus, these two types of imagery compliment each other for tree canopy delineation, each providing important but different types of information.

Return to Previous Page