Database Reference
In-Depth Information
To generate these data products, some standard processing steps are applied
including calibration, navigation, and registration. These data products are
then provided to other scientists, who use it for their own process.
83
One of the key observations of Frew and Bose
83
is that scientists may
use many different command line applications (e.g., Perl scripts, MATLAB,
TerraScan tools) to generate new satellite data products. The Earth System
Science Workbench (ESSW) system developed at the University of California,
Santa Barbara, therefore keeps track of which command line applications are
used to modify the data products and records all intermediate data products.
Using ESSW, scientists can determine which satellites were used in generating
a data product, what version of the calibration algorithm was used, and the
number of scan samples used in the original sensor data.
Provenance is also critical in regenerating large-scale data products derived
from satellite imagery. An example is how imagery from the Moderate Resolu-
tion Imaging Spectroradiometer Satellite is processed and stored at NASA's
Goddard Space Flight Center using the MODIS Adaptive Data Processing
System (MODAPS). The imagery obtained directly from the satellite is known
as Level 0 data. This raw data is irreplaceable and must be archived. However,
the initial transformation of this raw data into more useful data called Level
1B data is too large to be archived. Level 1B data includes calibrated data and
geo-located radiances for the 36 channels of the satellite.
*
Level 1B data is then
processed and consolidated to create Level 2 data products, which are finally
translated into visual images. The Level 2 and final images are the most use-
ful for scientists. Because of its size, Level 1B data is typically kept for 30-60
days and then discarded. To enable this data to be reproduced the MODAPS
systems maintains enough process documentation to reproduce the Level 1B
data from the raw satellite data. The process documentation includes the
algorithms used, their versions, the original source code, a complete descrip-
tion of the processing environment, and even the algorithm design documents
themselves. Essentially, provenance enables a
virtual archive
of satellite data,
which otherwise would be lost or dicult to recreate.
84
Provenance for geospatial data is also extremely important for merging data
from multiple sources. This aspect was discussed in detail in Chapter 10 in
the context of interoperability and data integration in geosciences.
12.6.3 Provenance for Oceanographic Applications
When studying the oceans, scientists require data from multiple data
sources—shipboard instruments, buoy-based sensors, underwater vehicles,
and permanent stations. These data sources measure everything from the tem-
perature of the ocean to its salinity and the amount of chlorophyll present.
*
http://modis.gsfc.nasa.gov/data/dataprod/dataproducts.php?MOD NUMBER=02.
Accessed
July 20, 2009.
