Tateishi and Kojiwara Monthly Maximum Global
Vegetation Index and Land Cover Classifications from NOAA-9
(JAN 1986 - DEC 1989)
Principal Investigators:
Ryutaro Tateishi and Koji Kajiwara
Remote Sensing and Image Research Center
Chiba University, Japan
Summary:
Dataset consits of four thematic coverages; a) Scaled Normalized Difference
Vegetation Indices, b) Land cover Classification, c) Source Example: Scaled
normalized Difference Gegetation Indices and d) Source Example: Land Cover
Classification. Data are gridded at a resolution of 10 and 5 minutes.
Primary References:
Tateishi, R. and Kajiwara, K. 1991. Global Land-Cover Monitoring
by NOAA GVI Data. International Geoscience and Remote Sensing Symposium
Vol. II, No. 1 , pp 1637-1639.
Tateishi, R. and Kajiwara, K. 1991.Land
Cover Monitoring in Asia by NOAA GVI Data. Geocarto International
Vol.
6, No. 4, pp 53-64.
Tateishi, R., Kajiwara, K., and Odajima,
T.1991. Global Land-Cover Classification by Phenological Methods Using
NOAA GVI Data. Asian-Pacific Remote Sensing Journal Vol. 4, No. 1 ,
pp 41-50.
Odajima, T., Kajiwara, K., and Tateishi, R. 1990. Global Land-Cover
Classification by NOAA AVHRR Data. Proceedings of the 11th Asian Conference
on Remote Sensing Vol. II, , pp S-3-1 to S-3-6.
Tateishi and Kojiwara GVI and Land
Cover Classification
DATASET DESCRIPTION
Dataset Description
INTEGRATED DATASET
DataSet Citation:
Ryutaro Tateishi and Koji Kajiwara. 1993. Monthly Maximum Global Vegetation
Index and Land Cover Classifications from NOAA-9 (JAN 1986 - DEC 1989).
Digital Raster Data on a 10-minute Cartesian Orthonormal Geodetic (lat/long)
1080x2160 grid. In: Global Ecosystems Database Version 2.0. Boulder, CO:
NOAA National Geophysical Data Center. Forty-eight independent and one
dependent single-attribute spatial layers. 121,361,591 bytes in 105 fiiles.
Projection:
Global Cartesian Geodetic (Platte Carree
- latitude/longitude)
Spatial Representation:
(1) (SNDVI) 10-minute grid produced from source grid using a "cell overlap"
resampling technique
(2) (Land Cover Classification) 5-minute grid produced from source
grid using a nearest neighbor cell overlap resampling technique
Temporal Representation:
Monthly maximum of weekly values
Data Representation:
(1) one byte unsigned integers representing Scaled Normalized Difference
Vegetation Indices
(2) one byte unsigned integers representing land cover classes
Layers and Attributes:
Raster: Time series of 48 independent single-attribute spatial layers and
one derived single-attribute spatial layer.
Dataset Description
DESIGN
Variables:
(1) scaled normalized difference vegetation index (SNDVI)
(2) land cover classification derived from 1987 SNDVI data (see (1)
above)
Origin:
NOAA-9 & NOAA-11 Polar Orbiting Satellites with Advanced Very High
Resolution Radiometer sensors
Geographic Reference:
Cartesian Geodetic (Plate Carreé
-- latitude/longitude)
Geographic Coverage: Global
Global
Maximum Latitude: +75 degrees (N)
Minimum Latitude: -55 degrees (S)
Maximum Longitude: +180 degrees (E)
Minimum Longitude: -180 degrees (W)
Geographic Sampling:
Last ("random") element of each 4x4 array of Global Area Coverage (GAC)
4 km values, mapped onto a 904x2500 global Plate Carreé (lat/long)
grid. GAC values are 1x4 km averages (along scan-line) of sampled values
within each 4x4 array of 1 km cells. Look-angle varies between pixels due
to temporal compositing.
Time Period:
January 1986 to December 1989
Temporal Sampling:
(1) (SNDVI) monthly maximum of weekly values. Weekly values are maximum
of 7 daily values. Time of day varies between pixels.
(2) (Land Cover Classification) static composite derived from 1987
SNDVI
Dataset Description
SOURCE
Source Data Citation:
Ryutaro Tateishi and Koji Kajiwara. 1990. Monthly Maximum Global Vegetation
Index and Land Cover Classifications from NOAA-9 (JAN 1986 - DEC 1989).
Digital Raster Data on a 8.6-minute Plate Carreé (lat/long) 904x2500
grid. Chiba, Japan: Remote Sensing and Image Research Center, Chiba University.
112 MB in 52 files on one quarter-inch cartridge tape.
Contributor:
Ryutaro Tateish
Distributor:
NGDC, Global Ecosystems Database Project
National Geophysical Data Center
325 S. Broadway, E/GC1,
Boulder, CO 80303
Date of Production:
1990
Lineage & Contacts:
-
Dr. Ryutaro Tateishi and Koji Kajiwara
GIS/Remote Sensing and Land Cover Monitoring
Chiba University
1-33 Yayoi-cho, Inage-ku, Chiba 263
Japan
-
John J. Kineman and David C. Schoolcraft
Global Ecosystems Database Project
National Geophysical Data Center
325 S. Broadway, E/GC1,)
Boulder, CO 80303
Email: dschoolcraft@ngdc.noaa.gov
fax: (303) 497-6513
Web: http://www.ngdc.noaa.gov/seg/eco
Dataset Description
ADDITIONAL REFERENCES
See references cited in Kineman, J.J., and Ohrenschall, M.A. et al. 1992.
Global Ecosystems Database Version 1.0: Disc A, Documentation Manual. Key
to Geophysical Records Documentation No. 27. USDOC/NOAA National Geophysical
Data Center, Boulder, CO. 240p.
Dataset Description
FILE LISTS
Tateishi and Kojiwara GVI and Land
Cover Classification
DATASET ELEMENT DESCRIPTIONS
Scaled Normalized Difference Vegetation Indices
Description:
Normalized Difference Vegetation Index (NDVI) derived from the NOAA AVHRR
Satellite with calibration corrections.
Structure:
10-minute Cartesian Geodetic
(latitude / longitude) 1080x2160 grid
Series:
48 month time-series
System Files:
Notes:
(1) The minimum and maximum values for all files are set to the time series
minimum and maximum of 0 (zero) and 255.
Land Cover Classification
Description:
Thirteen land cover classes derived from the Scaled Normalized Difference
Vegetation Indices
Structure:
5-minute Cartesian Geodetic
(latitude / longitude) 2160x4320 grid
Series:
None
System Files:
Notes:
(1) The original resolution of the source grid was 0.144, which was resampled
by nearest neighbor to a .0833 cell size for comparability with the other
datasets in the GED. The original resolution is indicated in the metadata
file (TKCL13.doc), although it actually varies across the image due to
the fractional difference in cell size between original and resampled grids.
Source Example: Scaled Normalized Difference
Vegetation Indices
Description:
Source data example (before resampling for GED integration): Scaled Normalized
Difference Vegetation Indices.
Structure:
8.6-minute Cartesian Geodetic (latitude
/ longitude) 904x2500 grid
Series:
None
System Files:
Notes:
None
Source Example: Land Cover Classification
Description:
Source data example (before resampling for GED integration): Land Cover
Classification
Structure:
8.6-minute Cartesian Geodetic
(latitude / longitude) 904x2500 grid
Series:
None
System Files:
Notes:
None
Tateishi and Kojiwara GVI and Land Cover
Classification
TECHNICAL REPORT
DATA INTEGRATION
John J. Kineman and David C. Schoolcraft
National Geophysical Data Center
Boulder, CO 80303 USA
The data were received as 8.6 minute raster grids
in Plate Carreé (lat/long) projection. To achieve integration with
the GED grid structure, the Vegetation Index data were resampled to a 10-minute
grid using an area-weighted cell overlap method (similar to distance weighted
linear interpolation). The land cover classification data were resampled
to a 5-minute grid using a nearest neighbor method to preserve the class
identities. Nearest neighbor resampling to a coarser grid (e.g. 10-minutes)
was shown to produce unacceptable artifacts owing to the loss of rows and
columns inherent in this method. Source examples are provided for comparison
and determination of error. Also, the original data are available separately
from NGDC.
Registration accuracy was assured by first determining
the precise origin of the original grids, then resampling to the output
grid using corresponding offsets in latitude and longitude. The origin
was determined by a best fit of Micro World Data Bank II coastline data
with coastlines derived from the Vegetation Index data by taking the first
derivative (slope). This tends to be greatest at the transition between
water (low GVI) and land, thus revealing coastlines at approximately one
pixel resolution. Best fit was determined by maximizing the calculated
slope values along the entire coastline, extracted globally along the MWDB-II
coastline. A second method of determining the coordinates for small islands
visible in the GVI images was also used, however this was found not to
be a precise as the coastline comparisons. We then confirmed that the time-series
was uniform in origin by determining the relative position of clearly visible
features (e.g., the Nile river). These results were later confirmed with
Dr. Tateishi.
The original dataset was provided with three color
separates for the classed image, to provide a preferred color scheme. These
were converted into a palette file that accompanies the dataset.