Detecting Large Scale Change Using LandSat Satellite Imagery and GIS

Oaks ā€˜nā€™ Folks ā€“ Volume 12, Issue 1 ā€“ February, 1997

Change detection is the process of identifying changes in land cover over time. As human and natural forces continue to alter the landscape,various public agencies are finding it increasingly important to develop monitoring methods to assess these changes. Changes in vegetation result in changes in wildlife habitat, fire frequency and effects, aesthetic and historical values, ambient air quality, and other resource values, which in turn influence policy decisions. Currently, the US Forest Service (USFS) is interested in assessing drought- and insect-caused mortality within coniferous forests, while State and local agencies are concerned with the loss of oak woodlands to fuel wood harvest and urban development.

Methods for monitoring vegetation change at a landscape scale range from fieldwork-intensive plot inventories to utilization of remotely sensed data that includes aerial photography and satellite imagery. The USFS, in collaboration with the California Department of Forestry and Fire Protection (CDF), is conducting a large-area change detection project using LandSat ThematicMapper satellite imagery in an effort to minimize monitoring costs and improve the quality of monitoring data. The goal of the project is to identify vegetation change that has occurred over a five-year period for the entire state.

The state has been divided into five areas; each area encompasses a distinct region and will take one year to complete. The first area encompasses approximately 16 million acres in the southern Sierra Nevada Mountain Range and includes four national forests (Stanislaus, Inyo, Sierra, and Sequoia), two national parks (Yosemite, and Sequoia-Kings Canyon), private forestland, and hardwood rangeland. Work on the first area began in October of 1995. The northern Sacramento Valley and northern Sierra Nevada mountain range were selected for change detection work in 1996-1997.

A two-phased approach was used to describe broad, landscape changes between 1990 and 1995 for the southern Sierra Nevada and to determine causes of those changes. One of the most important, albeit time-consuming, steps is preparing the imagery for processing. Image processing requires registering one image to another and correcting for atmospheric differences between the two dates. Image registration ensures that the location of pixels in one image exactly match the location of pixels in the second image. (Pixels are the smallest land area, 30 meters by 30 meters, where differences in reflectance can be measured.) If pixels do not correspond, then artificial changes due to pixel misregistration will occur. Differences in atmospheric conditions at the time each satellite image was acquired also affect the pixel values in each image and must be accounted for. Image processing techniques were then used to derive a change map consisting of data representing changes in greenness, brightness, and wetness in the imagery. Decreases in greenness and wetness and increases in brightness all indicate decreases in vegetation.A classification algorithm was then applied to the data, which resulted in a gradient of change- classes from a large decrease in vegetation to a large increase in vegetation.

Existing vegetation maps were used in a Geographic Information System(GIS) to identify specific areas of concern: productive coniferous forestland and hardwood rangeland. Fieldwork was necessary to accurately identify and quantify changes in the change map. Because conducting fieldwork on all change areas throughout the entire project area is not economical, a sampling technique was used to choose areas representative of the entire area.

In this project, areas of mortality were identified on the four national forests and sample plots were delineated within these areas. Fieldwork in these plots resulted in identifying and quantifying vegetation change as a result of conifer mortality. Analyses of this fieldwork will result in similar actions across the entire area.

Similarly, fieldwork conducted by UC Cooperative Extension in hardwood rangelands has identified decline in hardwood canopy cover due to fire,thinning, harvest, urban development, mortality and other factors. Although quantitative measurements were not taken, a strong correlation is evident between areas of change on the change map and decreases and increases in hardwood canopy cover.

The final products for the Southern Sierras are (1) a change map identifying large decreases and large increases in vegetation, and (2) a GIS database identifying the locations of known vegetation change for both coniferous,public forestland, and hardwood rangeland. The change maps and the database are available to anyone interested in monitoring vegetation change. For additional information on this project, please contact Lisa Levien, USFSat (916) 454-0803 or Cindy Bell, CDF-Fire and Resource Assessment Program at (916) 227-1338.


Lisa Levien and Cindy Bell

prepared and edited by Justin Vreeland, Bill Tietje, and Pam Tinnin

Lisa Levien
USDA Forest Service,
Region 5, Sacramento, CA

Cindy Bell
California Department of Forestry and Fire Protection, Sacramento,CA

Barbara Maurizi
Pacific Meridian Resources, Sacramento, CA