Digital Vegetation Delineation on Scanned Orthorectified Aerial Photography of Petersburg National Battlefield

Abstract

I developed a new methodology to produce an orthorectified mosaic and a vegetation database of Petersburg National Battlefield using mostly digital methods. Both the mosaic and the database meet National Map Accuracy Standards and proved considerably faster than traditional aerial photograph interpretation methods. I classified vegetation polygons to the formation level using the Nature Conservancy's National Vegetation Classification System. Urban areas were classified using Mitchell's Classification Scheme for Urban Forest Mapping with Small-Scale Aerial Photographs. This method reduced the production time by 2/3, compared to traditional methods. It also reduced the chance of user error because re-tracing of the linework is not required.

My method started with scanning 75 aerial color IR photos, provided by Petersburg National Battlefield, at 600 dpi. Erdas Imagine was used to rectify the images using United States Geological Service (USGS) Digital Elevation Models (DEM) and black and white USGS Digital Orthophoto Quarter Quadrangles (DOQQ) as reference. The images were then mosaiced to create a seamless color infrared orthorectified basemap of the park. The vegetation polygons were drawn onscreen using ArcMap from Environmental Systems Research Institute, Inc. (ESRI) with the color, orthorectified mosaic as a background image. Stereo pairs of the aerial photos were referenced as needed for clarification of the vegetation. I used a minimum mapping unit (mmu) of 0.2 hectares, which exceeds guidelines defined by the United States Geological Survey — National Park Service Vegetation Mapping Program. This methodology is easily learned quickly and has already been applied to several other studies.

The production of an orthorectified mosaic, created during the process, from the aerial photographs greatly increases the value of the photographs at little additional cost to the user. The orthorectified basemap can then be used as a backdrop for existing data layers or it can be used to create new GIS data layers. I used a minimum mapping unit (mmu) of 0.2 hectare, which exceeds guidelines defined by the United States Geological Survey-National Park Service Vegetation Mapping Program.

Traditionally, vegetation polygons are delineated on acetate for each photograph. The linework on the acetates is then transferred to a basemap using a zoom transfer scope or other transfer instrument. The linework is traced again to digitize it for use in a GIS program. This process is time consuming, and the linework is drawn three times. The redundant tracing increases the chance of user error. My new methodology requires that polygons be delineated only once. I wanted to avoid using the zoom transfer scope and to avoid the redundant linework.

A total of 228 polygons were delineated over 20 separate vegetation and land cover classes with an overall thematic accuracy of 87.42% and a Kappa of .8545. Positional accuracy was very good with a RMSE of 1.62 meters in the x direction and 2.81 meters in the y direction. The Kappa and RMSE values compare favorably with accuracies obtained using traditional vegetation mapping methods.