U1: Oral Presentation C Friday, 10:30 – 11:00 am C Gooseberry Falls Room

Impervious Surface Area Classification and Mapping Using Satellite Remote Sensing

Jean Doyle, Marvin Bauer
University of Minnesota
Department of Forest Resources
Remote Sensing and Geospatial Analysis Laboratory
115 Green Hall
1530 Cleveland Avenue North
St. Paul, MN  55108
doyl0037@tc.umn.edu

The amount of impervious surface in a landscape directly affects the amount of runoff to streams and lakes, and is also related to water quality in receiving waters. The distribution of impervious areas can be important when planning for runoff control in an urban setting. This study evaluated whether Landsat Thematic Mapper (TM) satellite imagery can be used to estimate the percentage of impervious area across a large area. The objective was to determine the percentage of imperviousness at the scale of a single pixel, 30 meters or approximately 1/4-acre. 

The percentages of impervious surfaces in the seven-county Twin Cities metropolitan area were classified from 1991 and 1998 Landsat TM images. The process consisted of four steps. First, the feasibility of using a principle components transformation of the Landsat images, known as “tasseled cap”, was evaluated. The second component of tasseled cap, greenness, is strongly related to the amount of green vegetation and conversely to the amount of bare or impervious surface area. Greenness values for selected sites were compared to measured pervious/impervious percentages obtained from DOQs. The second step consisted of selecting training sites of various impervious percentages throughout the area. Third, models were developed based on the linear regression of the training sites and applied to the greenness data. The resulting classification provided a continuous range of impervious area from 0 -100% for each date. Classification accuracy was estimated to be between 82% and 88%. Finally, the impervious area classification image was combined with previously classified Landsat TM images that identified 10 land cover classes, two of which were urban. The new impervious area image replaced the two urban classes with continuous values. The model developed for this project can be easily modified for use with images from other dates or geographic areas.