Einzelbeispiel satellitengestützter Erderkundung
dazu eine Animation 'Frühlingseinzug'
Acquired : 02-OCT-95 00:00:00
Updated on : 11-OCT-95 Product Type : Week_Composite
Satellite Number : noaa14 Projection Code : xx_STER_0
Channels : (ch2-ch1)/(ch1+ch2) Interpollation : NN
Radio Res : .003155 Upper Left :(-35 26'24",64 13'48")
Min NDVI : -.099685 Upper Right:(65 26'24",64 13'48")
Water Flag : 0 Lower Left :(-7 08'59",27 15'36")
Cloud Flag : 255 Lower Right:(37 08'59",27 15'36")
Receiving Station : Oberpfaffenhofen
NOAA AVHRR Normalized Difference Vegetation Index Maps (NDVI)
Source, Image Characteristics, and Processing
This product is part of DLR's AVHRR land pathfinder activities. The goal of the product is to provide the user daily, weekly and monthly NDVI synthesis maps in a defined format easy to access with the highest possible reliability on the thematic quality. After a phase of tests between March and June 1994, the operational production chain was launched on July 1st, 1994. Since then, both weekly and monthly NDVI maps covering the European continent were available until September 13th, 1994, when the AVHRR on board the NOAA-11 spacecraft failed. The production of daily, weekly and monthly NDVI maps was resumed on February 20, 1995, based on NOAA-14 AVHRR
Special emphasis is given to a precise image registration and a reasonable cloud screening procedure to ensure that only cloudfree pixels are taken for the later process of composition. Due to the importance of these two tasks, the production of the NDVI maps is a mixture of unsupervised pre-processing steps and a supervised parametrization of the cloud tests and an image navigation control. Thus, before an image was sent to the data archive and becomes available to the user, it was manually controlled regarding the navigation quality and the cloud tests.
All NDVI products consist of one NDVI channel. The major processing steps are descibed below:
* Automatic pre-navigation / Interactive supervision
Firstly, a process of auto-navigation is performed using WDB-II coastline and river data to improve the accuracy of the geo-referenzation given by the daily updated "twoline" elements. This process is done over the entire pass before the remapping procedure is applied. Firstly, appropriate coastline areas with significant features are selected in 1deg*1deg boxes and tested regarding cloudiness. For the remaining cloudfree boxes a cross correlation algorithm between the "real" coastline in the satellite image and the coastline of the reference data set is performed. Based on the yielding vector array the satellite's yaw, pitch, tilt, and roll angles are corrected. The complete procedure is firstly done unsupervised, then the results of the navigation process are checked and if necessary corrected. Also, if no valid boxes for the correlation process could be found, the navigation procedure in done manually. Therefore, best possible quality is guaranteed.
The solar channels 1 and 2 are calibrated into % technical albedo as described by NOAA. For the NOAA-11 AVHRR pre-launch calibration coefficients were applied for all products between July 1994 and September 1994. Time-adjusted coefficients provided by TEILLET & HOLBEN will be applied for all past NOAA-11 data if required. For the NOAA-14 AVHRR, pre-launch coefficients are used as provided by NOAA. The thermal channels 3,4, and 5 are calibrated into radiances
and then into brightness temperatures by inverting the PLANCK function. The non-linearity between counts and radiances is considered. Reference data from the internal Black Body are taken every 100 lines. The solar channels are used to calculate the NDVI, the thermal data are used in conjuction with the solar channels to perform the cloud/water detection algorithm.
* Cloud/water detection
To ensure that only NDVI values over cloudfree land surfaces are derived, a couple of cloud/water tests are performed. They are based on the principal spectral characteristics of land, water, and cloud surfaces. The applied thresholds may vary from case to case with regard to the specific characteristics of every single pass. The procedure is performed in two steps:
(1) Arid surface test: the first four combined tests are used to calculate valid NDVI values over arid areas where cloudfree areas are assumed to be warm and bright. Because some arid areas are brighter than clouds, these would be flagged as cloudy using an adequate maximum channel 2 brightness threshold for cloudfree vegetated areas. If all conditions are true, NDVI values will be calculated.
(2) The second set of tests determines wether a pixel is cloud/water contaminated or not. If one test is true, the pixel is flagged as cloud or water and excluded from further NDVI processing. For all remaining pixels the NDVI is calculated.
The whole cloud/water detection scheme is listed below. It was developed by Dr Stefan Dech at DFD's Value Adding & Visualization group (VAV) in 1994 (Copyright DLR 1995).
IF CHANNEL_2_ALBEDO > VALUE (land brightness) AND CHANNEL_4_BBT > VALUE (land temperature) AND CHANNEL_4-5_BBT_DIFF > VALUE (thin clouds) AND SATELLITE_ZENITH_ANGLE < VALUE (Geometry limit)
THEN Calculate NDVI
ELSEIF CHANNEL_2_ALBEDO < VALUE (max water brightness) OR CHANNEL_2_ALBEDO > VALUE (max land brightness) OR CHANNEL_4_BBT < VALUE (max cloud temp) OR CHANNEL_4-5_BBT_DIFF > VALUE (thin clouds) OR CHANNEL_3-4_BBT_DIFF > VALUE (water clouds) OR SATELLITE_ZENITH_ANGLE > VALUE (Geometry limit)
THEN CLOUD or WATER
ELSE Calculate NDVI
* Derivation of the Normalized Difference Vegetation Index:
The formula which is applied bases on the AVHRR channels
1 (RED) and 2 (NIR): NDVI = (NIR-RED)/(NIR+RED)
No atmospheric corrections are done so far. For the future, it is planned to calculate NDVI values after a correction of the two solar channels using a method developed by Dr Thomas Popp (DLR).
Preparations are in progress. Tests are foreseen for the first quarter of 1996
The data are remapped into a Stereographic projection with a given geometrical resolution of 1.1132 km at the center of the satellite map at 51.00 N / 15.00 E. The "nearest neighbor" technique is applied
to resample the pixels into the map. The total size of the maps is 4100 samples * 4300 lines.
* MV Compositing / synthesis
Via ISIS, daily, weekly and monthly maps are accessable.
Daily maps are composed based on the maximum NDVI value (MV) at every pixels' position which normally consist of three consecutive NOAA-14 passes (east, central, west). Daily composites are only available for NOAA-14 data. The weekly and monthly synthesis images are composed taking the daily maximum NDVI value at every pixels' position. However, there is a special user request (e.g. 10-day maximum NDVI, or averages) all single scenes are available in an off-line archive and can be composed with regard to the specific requirements. In that case, please contact the VAV group at
DFD (see below).
* Ocean masking
A WDB-II based land/sea mask in 1 km geometrical resolution is used to mask out remaining NDVI values over ocean areas (e.g. under sunglint conditions). The land/sea mask was created by Gerhard Gesell (DFD).
* Integer scaling
The NDVI data are scaled from their original 10-bit format into the user-friendly 8-bit integer format (values from 0-255). The NDVI values are stored as follows: Greyvalue "0" is referred to "WATER", greyvalue "255" is reserved for "CLOUD" and "NO DATA". The NDVI range starts with "-0.09968454" and is referred to greyvalue "1". The radiometric resolution is "0.0031546", greyvalue 254 is therefore referred to value "0.7" (maximum NDVI). Values below NDVI_MIN are set to
NDVI_MIN and values above NDVI_MAX are set to NDVI_MAX.
NOTE: White triangles in particular in Northern Africa can be seen in the daily NDVI composites (value 255 in the original data sets). They are the edges of the subsequent overpasses opening approching the equator direction where no pass overlapping appears. The triangle size may vary
depending on the daily pass geometry and the satellite zenith angle threshold used to limit NDVI calculation close the pass edges.
* Geographical coverage
The frame coordinates of the NDVI product are:
UL 64.229 N / 35.440 W UR 64.229 N / 65.440 E
LL 27.262 N / 07.152 W LR 27.262 N / 37.152 E
* Temporal Coverage
One daily NDVI map is composed using three consecutive NOAA-14 acquisitions at noon LT. Weekly and monthly maps are composed based on the daily maximum images using the maximum NDVI for every single pixels' position. Thus, a week composite normally consists of 21 AVHRR passes, a monthly synthesis map of about 90 passes.
* Data Format
One NDVI map consists of 1 layer in 8-bit resolution with the above given size. The used data format is VFF Raster Standard Format. Other formats (ERDAS LAN, SUNRASTER, etc.) are available. The Quicklooks are stored in JPEG Format. The Europe NDVI map consists of 17.63 Mbyte
(uncompressed). A very high compression rate can be achieved for the NDVI products.
Digital browse quicklooks can be accessed via network. Small quicklooks of the NDVI maps are showing the distribution of the NDVI using a color-code. Dark green = high NDVI's up to 0.7, going to brown = low NDVI's down to -0.1. A scalebar is imbeded for the NOAA-14 generation.
In the Q/L, ocean and lake areas are blue, and clouds are white.
Full resolution data are available via network.
* Remarks (14.03.1995)
Due to the moderate quality of the actual TBUS ("two-line") elements for the NOAA-14 spacecraft, the auto-navigation fails frequently. The interactive navigation process is very time-consuming and difficult to perform for the complete NOAA-14 AVHRR passes. In the case of synthesis images (e.g.
week composites) small pixel displacements along the edges between central and eastern passes could be detected in a order of up to 2 pixels in total.
For questions on AVHRR NDVI pathfinder products contact
Dr Stefan Dech, DLR, DFD-BS, D-82234 Oberpfaffenhofen (firstname.lastname@example.org)
For special requirements on NDVI synthesis images please contact
Andrea Holz, DLR, DFD-BS, D-82234 Oberpfaffenhofen (email@example.com) or
Tunga Padsuren, DLR, DFD-BS, D-82234 Oberpfaffenhofen (firstname.lastname@example.org)
For requirements regarding NDVI image animation on video please contact
Robert Meisner, DLR, DFD-BS, D-82234 Oberpfaffenhofen (email@example.com)
Quelle: Isis-Bilderdatenbank beim DFD