work underway on a water-discharge structure we designed for poor soil conditions
project filter
downloadable resources
back to projects

vegetation classification with GIS and remotely sensed data

vegetation classification with GIS and remotely sensed data
confidential pipeline client
upper Midwest

processing high-resolution satellite images enables detailed vegetation analysis and classification

For a client proposing a crude-oil pipeline that would cross three Midwestern states, Barr collaborated with the University of Minnesota’s Remote Sensing and Geospatial Analysis Laboratory to provide specialized GIS services for quantifying forested vegetation communities along the pipeline’s corridor. The final deliverable was a summary of vegetation-community classes that could be used by our client and the U.S. Fish and Wildlife Service in establishing the amount of “habitat equivalency compensation” that would be required for the pipeline if it were approved.

Vegetation-classification data available from the U.S. Geological Survey’s Gap Analysis Program was developed from 30-meter-resolution Landsat images collected between 1999 and 2001, and is not intended to be used at scales larger than 1:100,000. For this pipeline project, Barr recommended using images taken by a five-satellite network in 2012 and 2013. With resolution down to 5 meters, the new images provide exponentially more data than the Landsat photos, allowing the identification of small variations in vegetation communities, as shown in the images above (the one on the left was produced with older GAP data, the one on the right with recent, higher-res data).   

Using the high-res data to assess and quantify vegetation-community classes involved acquiring satellite imagery; developing “training” data for the computer model that would analyze the images; developing and classifying areas of vegetation (known as polygons); analyzing vegetation heights and age; and assessing the accuracy of vegetation classifications.

The Barr team combined purchased satellite imagery with data available through public GIS sources, data that had been collected in the field, and, where it existed, lidar (light detection and ranging) data. We used that information to develop data that was then employed to train the image-analysis program.

The actual analysis involved processing the satellite images in multiple GIS applications to increase their quality, group them into polygons corresponding to different vegetation-community types, and produce shapefiles for mapping.

Following two rounds of manual inspection to catch misclassifications or other errors, Barr’s team used a GIS tool to assign heights to trees, and then created a database that estimated tree age based on height. The final step was to use a GIS application that incorporated field-collected data for grassland and wetland areas.

Our team then summarized the data by county, vegetation class, age class, and impact type; calculated the vegetated acres for each category; and presented the information in one comprehensive, easy-to-read table.