A New Look at an Old Data Set

  Page 1Page 3
 

Sietse Los received his PhD in Earth Sciences from the Free University in Amsterdam. His research focus is on land vegetation and the roles that terrestrial plants play in the global water and carbon cycles. When he came to NASA’s Goddard Space Flight Center in 1989, he was looking for ways to analyze multi-spectral data on vegetation in three dimensions—both through time and space. While attending a summer conference at Goddard organized for visiting scientists, Los saw a presentation by a colleague who showed a movie of sea surface temperature together with precipitation patterns. Clearly, the movie showed that temperature anomalies like El Niño exert a profound influence on rainfall patterns around the globe.

But why stop there? Why not go a step farther and see if there is a relationship between sea surface temperature and plant growth? Los decided to do just that.

Back at Goddard, Los approached colleagues James Collatz, Compton Tucker, Lahouari Bounoua, and Piers Sellers to help him reprocess and analyze remote-sensing data collected by NOAA’s Advanced Very High Resolution Radiometer (AVHRR). This would seem to be the perfect data set since NOAA AVHRR sensors have been collecting global-scale data since the early 1980s. But, according to Los, there were problems with the data.

"AVHRR was never designed for vegetation monitoring," Los states. "It was designed mainly for monitoring the atmosphere (i.e., weather patterns, cloud cover, and air temperature) as well as measuring sea surface temperature and observing the extent of snow and ice on the surface."

Los explains that the AVHRR missions did not have the stringent calibration and orbital requirements that today’s Earth observing satellite missions have. Some of the NOAA AVHRR missions are known to have drifted by as much as 4 hours in their orbits, which changed the times of their equatorial crossings. (An extreme example is NOAA-9, which started as an afternoon satellite. It was replaced by NOAA-11, but was turned on again when NOAA-11 failed. By that time NOAA-9 had become a morning satellite.) In turn, this means that the relative angle of the sun would differ at the time data were collected over a given location, which has an impact on the data. "What you observe on the surface changes as a function of the sun’s angle of illumination," Los says.
 

 

NOAA POES Satellite
The image above shows one of the NOAA Polar Orbiting Environmental Satellites that carries the Advanced Very High Resolution Radiometer (AVHRR) instruments. The AVHRR has been collecting the data used to measure vegetation and sea surface temperature since 1978. Long-term monitoring is an essential part of studying climate trends. New instruments such as the Moderate-Resolution Imaging Spectroradiometer (MODIS) and Global Land Imager (GLI) will extend and improve on AVHRR’s heritage.

Image courtesy NASA GSFC

 

Solar Zenith Angle Comparison

Moreover, the performance of all satellite sensors degrades over time. They experience extreme temperature shifts, bombardment from cosmic and microwave radiation, possible impacts from micrometeorites, and corrosive outgassing from the satellite itself. Without adequate onboard calibrators, it is difficult to quantify how sensors change over time. "You sometimes see the sensitivity of the sensor changes over time," Los observes. "The AVHRR aboard NOAA-9, for instance, measured different amounts of vegetation on the surface over time. And its measurements differed from observations made by the AVHRR aboard NOAA-11."

James Collatz adds that AVHRR does not have the spectral sensitivity of more modern remote sensors like the Moderate-resolution Imaging Spectroradiometer (MODIS) flying aboard NASA’s Terra spacecraft. Whereas, AVHRR is sensitive in only 5 channels of the spectrum, MODIS has 36 channels, thus providing much finer multi-spectral detail. Consequently, in reprocessing the AVHRR data set, the team had to painstakingly fine-tune the data to correct for atmospheric effects that could interfere with its measurements, such as aerosol particles in the atmosphere that scatter sunlight and thus weaken the signal from the surface. Such atmospheric corrections are far easier with sensors like MODIS.
 

 

The angle between the sun and the Earth’s surface, called solar-zenith angle, changes reflected sunlight in several ways. At low solar zenith angles, such as local noon, top, the sunlight passes through relatively little atmosphere, minimizing scattering of light by the atmosphere and any effects of pollution, haze, or water vapor. The sunlight is also perpendicular to the Earth’s surface, so it is scattered directly back towards a sensor.

At high solar-zenith angles, like the bottom image (55°), atmospheric scattering is increased, decreasing the amount of shorter wavelength light (blue) in incident sunlight. Some surfaces reflect light differently at high angles than low ones. An additional effect is the increase in the apparent depth of forest canopy at high angles.

Images by Robert Simmon, NASA GSFC

 

Difference in Calibrated and Uncalibrated NDVI Data

The team took all of the above factors into consideration as they reprocessed the AVHRR data set. Their objective was to "correct" the data so as to remove any artifacts that may have erroneously been introduced into the data so they could be sure that what they were seeing was the signal of reflected sunlight and emitted thermal infrared energy from the surface. Once they confirmed the reprocessed data set gave them a clearer picture of global plant growth, they were ready to begin their analysis.

"Using various analysis techniques, we can now extract signals from the vegetation data that relate to the climate system," Los states. "And we can now correlate vegetative response to climate change in three dimensions—through time and space."

next When Plants are Thriving
next Watching Plants Dance to the Rhythms of the Ocean

 

The two images at left show the difference between uncalibrated (top) and calibrated AVHRR vegetation data in West Africa. Notice that the density of green foliage (defined as Normalized Difference Vegetation Index, or NDVI) is greater in the calibrated NDVI data. Errors in the data can either raise or lower measured vegetation index values.

Images courtesy Dan Slayback, NASA GSFC Laboratory for Terrestrial Physics.