Identifying Large Areas of Change

Submitted by Anonymous (not verified) on

The ability to quickly identify where change on the ground occurs on a national level can enable focused mapping efforts and help identify trends in landscapes over time. The USGS National Geospatial Technical Operations Center undertook preliminary research into methods to evaluate change. The shift from areas of low or no amounts of vegetation to high, or vice versa, can indicate where larger land cover/ land use changes are occurring due to e.g., landslides or urbanization. Normalized Difference Vegetation Index (NDVI) datasets created from Moderate Resolution Imaging Spectroradiometer (MODIS) raster images were obtained from NASA as the MOD13Q1 Version 6 product, with a spatial resolution of 250 m per pixel and the best available pixels from the acquisitions during a 16-day period incorporated into a single dataset. Two datasets were obtained for each area of interest representing the same season. These datasets were incorporated into automated toolsets to highlight areas of change between two dates of interest for large regions.

To highlight areas of significant vegetation change, the MODIS NDVI images were subtracted, agricultural areas masked out using the National Land Cover Database as a guide, and any pixel with a value within 2.5 standard deviations from the mean value of each image was extracted. These pixels were converted to area and summed within the overlapping USGS 1:100,000 quadrangle boundaries to determine the percentage of changed area for each polygon. Shortcomings to this method do exist. Because growing seasons vary from year to year, the direct comparison of the same time period may show changes in NDVI that do not directly reflect actual land cover change, but rather changes in the growing season. Masking out agricultural areas may mitigate this somewhat, but the risk will remain for non-agricultural areas. In addition, the results from this method will not illustrate what types of land cover change are occurring—only that changes in vegetation have occurred.

Initial results are promising. This method allows for rapid identification of  significant vegetation change over large areas and can serve to focus efforts and highlight areas where mapping updates may be needed.

Author Name
Kristina H Yamamoto
Author Email