eMODIS ALASKA
Calli B. Jenkerson, Senior Scientist1
Gail L. Schmidt, Senior Software Engineer2
1
ADNET Systems, Inc., Contractor to U.S. Geological Survey
Work performed under contract 08HQCN0005
2
SAIC, Contractor to U.S. Geological Survey
Work performed under contract 03CRCN0001
Earth Resources Observation Science Center
Sioux Falls, SD 57198
jenkerson@usgs.gov
gschmidt@usgs.gov
ABSTRACT
Operational monitoring of Alaskan vegetation using remote sensing data is a challenging project due to a
general lack of consistency and coverage from historical and existing platforms. With a more frequent repeat cycle
than Landsat and higher spatial resolutions than the Advanced Very High Resolution Spectroradiometer (AVHRR),
the National Aeronautics and Space Administration (NASA) Earth Observing System (EOS) Moderate Resolution
Imaging Spectroradiometer (MODIS) is a desirable data source for Alaska. For operational monitoring however, the
benefits of MODIS are counteracted by usability issues with the standard map projection, file format, composite
interval, high-latitude “bow-tie” effects, and production latency. eMODIS Alaska responds to a community-specific
need for alternatively packaged EOS-Terra MODIS data, addressing each of these factors for historical trend
analysis. Produced at the U.S. Geological Survey (USGS) Earth Resources Observation and Science (EROS)
Center, eMODIS delivers Normalized Difference Vegetation Index (NDVI) and surface reflectance over Alaska,
extending into Canada to cover the Yukon River Basin (YRB). eMODIS products are composited over 7-day weeks
during the growing season (February – October). Each 250-meter (m), 500-m, and 1,000-m composite represents
the entire geographic region on the Alaska Albers Equal Area mapping grid and is distributed in Geostationary Earth
Orbit Tagged Image File Format (GeoTIFF). eMODIS enhancements include the implementation of a sun angle
threshold (83º) and use of a snow filter to tag snowy pixels when no others are available. A simple File Transfer
Protocol (FTP) distribution site is currently enabled on the Internet for direct download of eMODIS products
(ftp://elpdl02.cr.usgs.gov/eMODIS).
eMODIS ALASKA CHARACTERISTICS
The eMODIS suite includes a 7-day data set over Alaska, extending into Canada to cover the YRB. The
historically produced data are composited within 7-day intervals during the Alaskan growing season (FebruaryOctober) based on a January 1 start date. Each data set includes NDVI, reflectance, acquisition, and quality
information at 250-m, 500-m, and 1,000-m resolutions. The Terra MODIS-based products are projected to the
Alaska Albers Equal Area mapping grid, and delivered in GeoTIFF format.





NDVI (Band 2 – Band 1 / Band 2 + Band 1)
Atmospherically corrected surface reflectance
o Band 1 (620-670 nanometer (nm)) Red
o Band 2 (841-876 nm) Near Infrared
o Band 3 (459-479 nm)
o Band 4 (545-565 nm)
o Band 5 (1230-1250 nm)
o Band 6 (1628-1652 nm)
o Band 7 (2105-2155 nm)
Quality of NDVI and reflectance
Day of Acquisition
Metadata
ASPRS 2009 Annual Conference
Baltimore, Maryland ♦ March 9-13, 2009
1A
1B
1C
1D
Figure 1. Samples of the eMODIS Alaska product suite. (1A). RGB image created with Bands 1, 4, 3 with 500-m
data acquired in August 2006. (1B). Grey-scaled 250-m NDVI calculated with data from the same time period.
(1C) depicts the per-pixel day of acquisition during the August interval. (1D) is an example of per-pixel quality
taken from a June 2003 interval.
Native spatial resolution for the reflectance bands is 250-m in Bands 1 and 2 and 500-m in Bands 3-7. All
bands are aggregated accordingly for the eMODIS 500-m and 1,000-m products, ie., Bands 1 and 2 are resampled to
500-m in the 500-m product, and all bands are resampled to 1,000-m in the 1,000-m product. The 250-m product,
however, includes only Bands 1 and 2 as original 250-m pixels. Table 1 contains the image dimensions for each
product, and the geographic dimensions of the Alaska composites are defined in Table 2 with the output map
projection parameters.
Table 1. 7-Day Alaska Composite Image Dimensions
Product
X (samples)
Y (lines)
250-m
500-m
1,000-m
9,322
4,661
2,330
7,064
3,532
1,766
Table 2. 7-Day Alaska Composite Mapping Grid Parameters
Projection
Datum
Central Meridian (DMS)
Latitude of Origin (DMS)
Standard Parallel 1 (DMS)
Standard Parallel 2 (DMS)
Alaska Albers Equal Area
WGS84
-154 00 00.00
50 00 00.00
55 00 00.00
65 00 00.00
ASPRS 2009 Annual Conference
Baltimore, Maryland ♦ March 9-13, 2009
Semimajor Axis (m)
Semiminor Axis (m)
Upper Left Lat/Lon DD (center of pixel)
Lower Right Lat/Lon DD (center of pixel)
Upper Left Output X/Y (center of pixel)
Lower Right Output X/Y (center of pixel)
6378137
6356752.31
72 00 00.00, -173 00 00.00
54 00 10.64, -128 00 04.26
2538877.50, -666333.75
773127.50, 1663916.25
In contrast to the NASA-EOS MODIS Vegetation Indices (VI) which layer VI, reflectance, and quality
information in a single Hierarchical Data Format for EOS (HDF-EOS) file, eMODIS provides separate GeoTIFF
files for each product in a 7-day interval. Every reflectance band and the NDVI have five associated files: data,
quality, metadata, acquisitions image, and acquisitions table (see eMODIS ALASKA PROCESSING,
Acquisitions Files below). Consequently, 40 files can be expected per 500-m and 1,000-m data set, while the 250m product will have only 15. The GeoTIFFs, “.met,” and “.txt” are packaged into a “.zip” file. There is one zip for
all files associated with NDVI and one for reflectance. The average volume for each zip is shown in Table 3.
Table 3. 7-Day Alaska Composite Zip File Volumes
Resolution
250-meters
500-meters
1,000-meters
NDVI ZIP Volume Megabytes (MB)
112.4
28.9
6.6
REFL. ZIP Volume (MB)
237.1
211.1
53.9
A typical data set includes the products listed below compressed into the two zip files. File specifications are
detailed in Table 4.
NDVI
composite 250m_TERRA_NDVI_2005_35.zip
250m_composite_ndvi.tif
250m_composite_ndvi_bq.tif
250m_composite_ndvi.met
250m_composite_ndvi_acq.tif
250m_composite_ndvi_acq_table.txt
NDVI data
NDVI data quality
NDVI data metadata
NDVI data Acquisitions Image
NDVI data Acquisitions Table
Reflectance
composite250m_TERRA_REFL_2005_35.zip
250m_composite_b*.tif
250m_composite_b*_bq.tif
250m_composite_b*.met
250m_composite_b*_acq.tif
250m_composite_b*_acq_table.txt
*Repeated per band (1-7)
reflectance data
reflectance data quality
reflectance data metadata
reflectance data Acquisitions Image
reflectance data Acquisitions Table
all of the above
Table 4. 7-Day Alaska Composite Output File Specifications
File
Band 1-7
250-m
Avg. Vol.
(MB)
126
500-m
Avg. Vol.
(MB)
32
1,000-m
Avg. Vol.
(MB)
8
NDVI
126
32
8
Quality
63
16
4
Data Type
Valid Range
Fill
Value
Scale
Factor
16-bit
signed
integer
16-bit
signed
integer
8-bit
unsigned
-100 – 16,000
-28,672
* 0.0001
-1,999 – 10,000
-2,000
* 0.0001
0 - good
1 - cloudy
10
na
ASPRS 2009 Annual Conference
Baltimore, Maryland ♦ March 9-13, 2009
integer
Acquisitions
Image
126
32
8
Acquisitions
Table
Metadata
0.02
0.02
0.02
3
kilobytes
(KB)
3 KB
3 KB
16-bit
unsigned
integer
Delimited
text
ASCII
2 - bad band
quality
3 - negative
reflectance
4 – snow
10 – fill
101 - 36699
na
na
101 – 36699
na
na
na
na
na
The NDVI data is specified with a valid range “-1,999 - 10,000.” The value “-1,999” is assigned to any VI
computation between “-1,998” and “-10,000.” VI computations between “-1” and “-1,997” are assigned face value.
When surface reflectance input pixels contain negative or fill values, the pixel will have a “-2,000” for NDVI and a
corresponding “10” in the band quality layer.
eMODIS ALASKA VALIDATION
A full-scale validation of eMODIS products is currently underway, and with intentions to publish in 2009. A
general characterization of data consistency in terms of processing artifacts, geolocation, data quality, and data
accuracy accompanies a “readme” document on the eMODIS FTP site (ftp://elpdl02.cr.usgs.gov/eMODIS/Alaska/).
eMODIS ALASKA PROCESSING
The eMODIS 7-day Alaska composites are produced using a combination of the software used to create
standard NASA-EOS MODIS products and software from the former USGS EROS MODIS Direct Broadcast
System (DBS). Collection 5 MODIS code acquired from MODIS Adaptive Processing System (MODAPS)
Services Software Distribution (http://modaps.nascom.nasa.gov:9500/production/) is integrated as the eMODIS
“house” algorithm. It is used to generate atmospherically corrected level-2 surface reflectance swaths from which
USGS EROS MODIS DBS software processes the final composites. The input level-1B (L1B) data are downloaded
from the Level 1 and Atmospheres Archive and Distribution System (LAADS) into the historical processing stream.
The compositing process begins with all level-2 surface reflectance swaths relevant to the Alaska extent. The
first step strips the outer 150,000-m of data from the east and west borders of the level-2 surface reflectance data to
remove the pixels whose distance from nadir are likely to cause a “bow-tie effect.” Next, the algorithm grids the
needed swath products (stripped surface reflectance, geolocation, and cloud mask) within the Alaska Albers
coordinates listed in Table 2.
A simple NDVI calculation using Band 1 (red) and Band 2 (near infrared) is then made for each gridded surface
reflectance file. The resulting gridded NDVI files, representing all available coverage for Alaska are fed into an
enhanced maximum value composite (MVC) algorithm. A traditional MVC would be populated using the highest
NDVI chosen from all available values for a pixel in the 7-day period. For the eMODIS 7-day Alaska composites,
the MVC algorithm is modified to incorporate band quality, negative surface reflectance, solar angle, cloud mask,
snow cover, and view angle as described in Composite Selection Process below.
The MVC outputs are finally stitched together by an EROS DBS compositing process to create the eMODIS 7day Alaska NDVI products. The surface reflectance and acquisitions composites are put together using the pixels
selected by MVC for NDVI.
Composite Selection Process
The pixels populating the final eMODIS product are selected by an enhanced MVC algorithm using a six-step
process to filter through input surface reflectance with bad quality, negative values, low sun angles, clouds, snow
cover, or low view angles.
ASPRS 2009 Annual Conference
Baltimore, Maryland ♦ March 9-13, 2009
1.
The band quality information in the level-2 surface reflectance data is used to determine if a certain pixel is
of “bad quality.” The standard MODIS Land Science Team (MODLAND) Quality Assessment (QA) bits
are carried into the production of the 250-m, 500-m, and 1,000-m surface reflectance data. Bad quality
surface reflectance pixels are not used in eMODIS production, leaving pixels with QA values of “00” as the
only input candidates.
MODLAND QA bits 0-1
corrected product produced at
corrected product not produced due to
00 -- ideal quality all bands
01 -- less than ideal quality some or all bands
10 -- cloud effects all bands
11 -- other reasons some or all bands may be fill value
[Note that a value of (11) overrides a value of (01)]
2.
Pixels with negative surface reflectance values are immediately flagged to exclude them as potential input
to the gridded NDVI files unless there are no positive values available. In the rare case that a negative
surface reflectance value is retained, the resulting gridded NDVI value is flagged as “-3,000” so that the
pixel for that day in the period will be ignored in the MVC process.
3.
The solar zenith angle taken from the 1,000-m level-2 geolocation product is used to determine sun angle
and is resampled for the 250-m and 500-m products. Any pixels acquired at sun angles greater than 83º are
eliminated as composite candidates.
4.
The level-2 (L2) cloud mask product (MOD35_L2) is used to minimize the overall presence of cloudy
pixels in the production stream. Data from the first byte (band) in the cloud mask is used for cloud
assessment. If bit 0 (see below) is “not determined,” then the associated pixel is treated as cloudy. If bit 0
is “determined” and bits 1-2 indicate a “cloudy” pixel, the pixel is excluded unless there are no suitable
pixels available in the 7-day record and the pixel’s data were from the most recent acquisition (see Quality
Assignment below).
If bit 0 is “determined” and bits 1-2 indicate “probably clear” or “confident clear,” then the corresponding
surface reflectance pixel is eligible in the MVC process.
5.
Bit field
0
Description
Cloud Mask Flag
2, 1
Unobstructed FOV Quality Flag
Key
0 = Not determined
1 = Determined
00 = Cloudy
01 = Uncertain
10 = Probably Clear
11 = Confident Clear
At this point in the filtering process, the two highest NDVI values in the remaining ideal-quality, nonnegative, non-cloudy pixels are checked for snow cover, which is read from the first byte in the level-2
cloud mask product.
Bit field
5
Description
Snow Flag
Key
0 = Snow/Ice
1 = No Snow/Ice
If one of the two highest NDVI values is flagged as snowy, it is eliminated from the selection process, and
the single remaining pixel is used in the final composite. If neither or both values are snowy, the pixels are
examined for their position from nadir.
6.
The view angle (distance from nadir) finally determines which gridded NDVI pixel is used in the Alaska
composite. The sensor zenith angle taken from the 1,000-m level-2 geolocation product is used to
determine the view angle and is resampled for the 250-m and 500-m products. The two highest NDVI
values in the remaining ideal-quality, non-negative, non-cloudy, non-snowy pixels are assessed for their
ASPRS 2009 Annual Conference
Baltimore, Maryland ♦ March 9-13, 2009
position from nadir using the geolocation information. Whichever of these two pixels is closest to nadir
will be used to populate the final composite.
NOTE: It is possible for off-nadir pixels to be selected for the Alaska composite since in some cases the
off-nadir pixels may have a higher NDVI than those closer to nadir.
The composite resulting from the MVC process is then stitched inside Alaska space to provide the eMODIS 7day composites.
Quality Assignment
A quality file is produced by the USGS EROS MODIS DBS software to track each pixel in the composite.
Most pixels will have good quality data. However there can be situations when there are no good quality surface
reflectance values for a particular pixel. The best available pixel is always preferred for the final composite but in
cases when the “best” is snowy, cloudy, or negative, the accompanying QA product will describe its condition. The
worst case, in which all available pixels have negative surface reflectance, are cloudy, snow-covered, and at very
low sun and view angles, MVC will use the pixel most recently acquired during the 7 day interval that does not
carry an original fill value (-28672).
The following values are used in the QA band:
0 = good quality
1 = cloudy pixel
2 = bad band quality
3 = negative surface reflectance
4 = snow
10 = fill
Acquisition Files
Because the products can originate from well over 100 acquisitions per 7-day interval, the eMODIS Alaska
composites include Acquisitions files to identify which of the possible inputs were used to populate the final
composite. As described above (eMODIS ALASKA CHARACTERISTICS), eMODIS products include both a
composite image of dates and an ASCII matrix listing the dates for each pixel. The Acquisitions Image and
Acquisitions Table contain the same information in different formats. The original acquisition used in composite
production is identified in an integer formatted as “DOY AQ.” “DOY” is the day of year and “AQ” is the acquisition
number which for eMODIS represents not the time of acquisition but the order of capture and is one-based. For
example, an Acquisitions value of “11702” identifies the second capture (02) from April 27 (117) as the input
selected for the composite. Likewise, an acquisition value of “1001” represents day 10 (January 10), first capture.
Metadata Files
The metadata accompanying the data files summarize geographic bounds, projection parameters, and product
contact information.
eMODIS ALASKA DELIVERY
eMODIS NDVI and reflectance zip files are delivered immediately after production to a direct access FTP site.
Users navigate a simple directory tree to download eMODIS output.
ftp://elpdl02.cr.usgs.gov/eMODIS/
Alaska/
historical/
TERRA/
<year>/
comp_<endday>/
composite<resolution>m_TERRA_NDVI_<year>_<endday>.zip
composite<resolution>m_TERRA_REFL_<year>_<endday>.zip
ASPRS 2009 Annual Conference
Baltimore, Maryland ♦ March 9-13, 2009
where
<year>
<endday>
<resolution>
is acquisition year, eg., 2005
is the Julian date for the last day of a 7-day interval
is “250,” “500,” or “1000”
eg., ftp://elpdl02.cr.usgs.gov/eMODIS/Alaska/historical/TERRA/2006/comp_231/
7,796,087
composite1000m_TERRA_NDVI_2006_231.zip
53,638,740
composite1000m_TERRA_REFL_2006_231.zip
118,275,899
composite250m_TERRA_NDVI_2006_231.zip
245,779,548
composite250m_TERRA_REFL_2006_231.zip
29,802,197
composite500m_TERRA_NDVI_2006_231.zip
206,228,617
composite500m_TERRA_REFL_2006_231.zip
eMODIS ALASKA PLANS
eMODIS Alaska currently supports a specific effort researching historical trends in vegetation related to climate
change in the YRB. The capability exists to expand the system to include other geophysical parameters (such as
enhanced VI and land surface temperature) and to provide real-time products in Alaska by integrating an expedited
production stream similar to that used in the eMODIS conterminous U.S. (CONUS) system. Interested parties are
encouraged to contact the project.
CONTACT INFORMATION
Calli Jenkerson
jenkerson@usgs.gov
(605) 594-2638
ASPRS 2009 Annual Conference
Baltimore, Maryland ♦ March 9-13, 2009
ACRONYMS
ASPRS
AVHRR
CONUS
DBS
DD
DMS
DOY
EOS
EROS
EVI
FOV
FTP
GEOTIFF
HDF-EOS
KB
L1B
L2
LAADS
LRS
m
MB
MODAPS
MODIS
MODLAND
MVC
NASA
NASS
NDVI
NIR
NLCD
nm
NOAA
NRT
QA
RGB
USDA
USGS
VegDRI
VI
YRB
American Society for Photogrammetry and Remote Sensing
Advanced Very High Resolution Radiometer
Continental United States
Direct Broadcast System
Decimal Degrees
Degrees Minutes Seconds
Day of Year
Earth Observing System
Earth Resources Observations and Science
Enhanced Vegetation Index
Field of View
File Transfer Protocol
Georeferenced Tagged Image File Format
Hierarchical Data Format for EOS
kilobyte
Level-1B processing
Level-2 processing
L1 and Atmospheres Archive and Distribution System
Land Remote Sensing
meter
megabyte
MODIS Adaptive Processing System
Moderate Resolution Imaging Spectroradiometer
MODIS Land Discipline Team
Maximum Value Composite
National Aeronautics and Space Agency
National Agricultural Statistic Service
Normalized Difference Vegetation Index
Near Infrared
National Land Cover Database
nanometer
National Oceanic and Atmospheric Administration
Near Real Time
Quality Assessment
Red Green Blue
United States Department of Agriculture
United States Geological Survey
Vegetation Drought Response Index
Vegetation Index
Yukon River Basin
ASPRS 2009 Annual Conference
Baltimore, Maryland ♦ March 9-13, 2009