166
D. Koslowsky
even if not completely filled by data, is defined by the corner coordinates:
55CNIlOOW
55 0 N/42°E
27°N/lOOW
27°N/42°E.
in latitude/longitude representation (geographic projection) with 111 00° resolution.
The total set of data contains 2800 lines and 5200 columns. It is subdivided for
faster access by regional users into four overlapping subframes of identical 1536
pixels line length defined by the coordinates of the upper left corners at 10 0 W, 5°E,
l5°E,and 27°E. This covers respectively the Iberian Peninsula, Italy, the Balkans,
and the eastern Mediterranean.
With the calibrated A VHRR data of its channels I to 5 the following
supplemental information is delivered: observation geometry (local zenith distance
and azimuth of satellite and sun, scattering angle), an origin indicator (link to date
and pass number), bitmaps (land-water, cloud mask), local time, NDVI, broad band
reflectance, and split-window land/sea surface temperatures as well as cloud
classification (currently in preparation). With theses data a congruent 16 channel
data set is build up. Daily as well as ten days composite data sets according to the
maximum local NDVI's are generated. For the years 1989 to 1994 only the western
and the central part so far are processed while the whole Mediterranean is covered
for the period 1995 until now amounting to about 2 Tbyte of data.
Data handling and access was found to be one of the most important prerequisite
for a successful evaluation of such large data sets. As long as the data were stored
only on about 500 magnetic tape cartridges it was a very time consuming job to run
evaluations throughout the full archives. Thus efforts were made to built up a large
array of magnetic disks in order to provide the data set in fast random access. The
actual capacity is now at 2.5 Tbyte capable to store the actual state of the daily 12
years MEDOKADS data set and will be expanded to about 4 Tbyte till the end of the
year 2001 to include the missing data for the eastern part of the years 1989 to 1994.
1.2 Inter-Calibration of NOAA 11 and NOAA 14
A critical issue is the necessary absolute calibration of the measured signals. No
calibration information of the two short wave channels of the A VHRR are supported
by the instrument. Preflight calibrations were found to be not applicable and thus
targets at the earth' surface have to be used for this task as well as to control the
degradation of the sensor response.
Reflectances, albedo and the NDVI are very sensitive to calibration errors and
changes of the environment can be detected if artificial trends like sensor
degradation are removed with great reliability. Even more than 20 years after the
first mission of an A VHRR instrument these problems are not settled finally. A
detailed study was performed (Koslowsky, 1996, 1997) to quantify the influences of
sensor degradation, atmospheric attenuation by changing local observation time and
influence of the bidirectional reflectance distribution function of the desert
D. Koslowsky
even if not completely filled by data, is defined by the corner coordinates:
55CNIlOOW
55 0 N/42°E
27°N/lOOW
27°N/42°E.
in latitude/longitude representation (geographic projection) with 111 00° resolution.
The total set of data contains 2800 lines and 5200 columns. It is subdivided for
faster access by regional users into four overlapping subframes of identical 1536
pixels line length defined by the coordinates of the upper left corners at 10 0 W, 5°E,
l5°E,and 27°E. This covers respectively the Iberian Peninsula, Italy, the Balkans,
and the eastern Mediterranean.
With the calibrated A VHRR data of its channels I to 5 the following
supplemental information is delivered: observation geometry (local zenith distance
and azimuth of satellite and sun, scattering angle), an origin indicator (link to date
and pass number), bitmaps (land-water, cloud mask), local time, NDVI, broad band
reflectance, and split-window land/sea surface temperatures as well as cloud
classification (currently in preparation). With theses data a congruent 16 channel
data set is build up. Daily as well as ten days composite data sets according to the
maximum local NDVI's are generated. For the years 1989 to 1994 only the western
and the central part so far are processed while the whole Mediterranean is covered
for the period 1995 until now amounting to about 2 Tbyte of data.
Data handling and access was found to be one of the most important prerequisite
for a successful evaluation of such large data sets. As long as the data were stored
only on about 500 magnetic tape cartridges it was a very time consuming job to run
evaluations throughout the full archives. Thus efforts were made to built up a large
array of magnetic disks in order to provide the data set in fast random access. The
actual capacity is now at 2.5 Tbyte capable to store the actual state of the daily 12
years MEDOKADS data set and will be expanded to about 4 Tbyte till the end of the
year 2001 to include the missing data for the eastern part of the years 1989 to 1994.
1.2 Inter-Calibration of NOAA 11 and NOAA 14
A critical issue is the necessary absolute calibration of the measured signals. No
calibration information of the two short wave channels of the A VHRR are supported
by the instrument. Preflight calibrations were found to be not applicable and thus
targets at the earth' surface have to be used for this task as well as to control the
degradation of the sensor response.
Reflectances, albedo and the NDVI are very sensitive to calibration errors and
changes of the environment can be detected if artificial trends like sensor
degradation are removed with great reliability. Even more than 20 years after the
first mission of an A VHRR instrument these problems are not settled finally. A
detailed study was performed (Koslowsky, 1996, 1997) to quantify the influences of
sensor degradation, atmospheric attenuation by changing local observation time and
influence of the bidirectional reflectance distribution function of the desert