Viewing Vegetation in a New Light

  Page 2Page 4

Several years ago, Imhoff and a group of researchers at NASA’s Goddard Space Flight Center set out to examine just how all this urbanization affects vegetation and, consequently, carbon dioxide levels. Their first obstacle lay in simply measuring urbanization’s extent. Cities and suburbs sprout up fast and their boundaries are uneven, often spreading over the land in a seemingly organic fashion like mold on fruit. Getting an accurate up-to-date map of just one urban area can be exceedingly difficult even with the help of planes or conventional remote sensing satellites such as Landsat.

Imhoff says he came across a solution when he discovered satellite images displaying the illumination that cities and towns generate at night. The images were taken by Defense Meteorological Satellite Program’s Operational Linescan System (OLS). The satellite network was originally designed to aid in aircraft navigation by detecting the lunar illumination off of nighttime clouds. What the Air Force realized is that on evenings when there was a new moon, the satellite was sensitive enough to record the illumination from city lights. Over a period of several new moons, the data the satellite retrieved could be pieced together to produce a global image of city lights.

Employing computer algorithms and additional data, Imhoff figured out a way to create maps of population density across an entire country or continent from the images. “We essentially scaled back on the brightness levels of the imaging data,” says Imhoff. The first full map of urban areas he constructed was of the United States. Using statistics taken from the U.S. Census Bureau, the Goddard team was able to place all land area in the United States into three classes urban, peri-urban, and non-urban—and assign population densities to those classes.


Map of Sacremento, California

Urban regions, Imhoff explains, constitute any area that has 1, 000 people or more per square mile. These are regions where the ecosystem has been significantly transformed into a human-devised habitat filled with office buildings, housing developments, and strip malls. Peri-urban areas, on the other hand, have only been lightly populated. They usually consist of farmland, light suburban development, or small towns and are classified as having an average of 100 people per square mile. Finally, non-urban regions, such as central Montana and western Maine, harbor only ten people or less per square mile.

Now that Imhoff and his team had the outlines of the urban and peri-urban areas, the next step was to calculate how urbanization has transformed the local vegetation. In short, he would have to compare the present day vegetation in urban areas to the vegetation that existed before the cities were built.

Measuring vegetation density for the present day cities was fairly straightforward. For years scientists have gathered global readings of vegetation using satellites. Remote sensing satellite instruments such as the Advanced Very High Resolution Radiometer (AVHRR) aboard NOAA’s polar satellites have been observing the Earth’s surface on a regular basis since the 1970s. The instrument records light of different bands (colors) reflected and emitted from the Earth, including red, green, and near infrared. To extract the information about the amount and extent of vegetation from the satellite imagery, scientists typically use what is known as the normalized difference vegetation index (NDVI). NDVI is produced by calculating the difference in reflectance between the visible and near-infrared light in a satellite image. It is a relative measure of the vegetation density and an indicator of how much photosynthesis is taking place on a given plot of ground. To obtain current readings of urban vegetation density, Imhoff simply acquired the most recent NDVI readings of the United States from NOAA. He then laid the vegetation density data on top of his city lights map and retrieved the values of vegetation density for regions designated as urban and peri-urban.

  The map at left shows the urban areas of Sacramento, California derived from city lights data and superimposed on census tracts. Each census tract has a similar total population, so smaller tracts correspond to denser populations. Light blue represents urban areas, dark blue peri-urban areas, and light grey rural areas. (Image courtesy Marc Imhoff, NASA GSFC)

Map of Land Cover Classification

Obtaining vegetation density readings for these areas from before the cities were constructed presented a more complex problem. After all, many major cities in the United States were founded well before people ever took flight or even knew what photosynthesis meant. “We looked at what the NDVI is inside urban areas and compared it to what the NDVI is immediately outside urbanized areas in the same land cover class,” says Imhoff. The team consulted a map created by the United States Geological Survey (USGS) displaying natural ecosystems such as savanna, mixed forest, and shrub land across the United States. Since these maps were general in nature and ignored the presence of urban areas, the scientists could pinpoint the natural ecosystems that predated present day cities and suburbs. They then retrieved satellite-derived vegetation density readings in untouched areas around the city with the same ecosystem. In Miami, for instance, the USGS maps indicated that most urban areas lay on savanna. The team located relatively untouched savanna nearby and obtained NDVI values. They swapped these readings for the vegetation density values in their present day urban area maps to get a picture of what vegetation was like before the city was built.

next Urban Sprawl’s Impact on Vegetation
back Our Changing Landscape

  Imhoff and his colleagues used USGS land cover classification data to determine the type of vegetation that existed before a city was built. The change in productivity (which is proportional to the amount of carbon consumed by vegetation) due to urbanization differs depending on the original type of land cover. (Image by Robert Simmon, based on data from the USGS.)
Print this entire article