NOAA/NESDIS's Office of Research and Applications (ORA) Climate Research and Applications Division (CRAD) Land Surface Team have been routinely producing and transferring a Third Generation version of the GVI for data from April 1985 to the present. This development was funded by NOAA's Climate and Global Change Program. The Third Generation is a further improvement over the Second Generation, with enhancements which include consistent post-launch calibration of AVHRR Channels 1 and 2, non-linearity corrections for Channels 4 and 5; and the generation of quality/cloud data for each map cell (Gutman et al., 1994 and Gutman et al., 1995). There are three products created by the Third Generation: weekly, monthly and climatology products which correspond to B-, C-, and D-levels, respectively. See Gutman et al., 1994 and Gutman et al., 1995 for details of procedures and a description of product level nomenclature. Appendix M describes the composition of the tapes for the three data levels, Appendix N supplies software for reading the image data and Appendix O contains the reprint of the Gutman et al. 1995 paper.

5.1 THIRD GENERATION GVI WEEKLY COMPOSITE PRODUCT (B-LEVEL)

NESDIS has been operationally producing the Second Generation of the weekly composite dataset since April 1985. Details about the Second Generation composite dataset can be found in Section 4. While the Second Generation product is an enhancement over the First Generation product, there are still some inherent problems, such as spurious trends, discontinuities, and noise due to satellite orbit and sensor instabilities, satellite replacement, atmospheric effects (clouds, stratospheric and tropospheric aerosols, water vapor, Rayleigh and ozone) and surface anisotropy effects.

The issues of cloud contamination and calibration have been addressed in the new, enhanced Third Generation GVI product. Although weekly composites contain fewer cloudy pixels than the original daily images, cloud contamination persists in global composite imagery (Gutman et al., 1994). Research during the last several years revealed the instability of AVHRR solar channels and their impact on the NDVI time series (Kaufman and Holben, 1993). It has also been found that the thermal IR channels needed a non-linearity correction to their in-flight calibration (Weinreb et al., 1990).

The enhancements of the Third Generation GVI product (the NOAA-9 and NOAA-11 observational periods) include consistent post-launch calibration of channels 1 and 2, non-linearity corrections for channels 4 and 5, and generation of quality/cloud (QC) files (masks), which allow researchers to have an indication of the quality of each map cell.

5.1.1 PROCESSING PROCEDURE

5.1.1.1 Channels 1 and 2

Channels 1 and 2 have been calibrated using the updated Pathfinder coefficients (Rao and Chen, 1994), and then normalized by the cosine of the Solar Zenith Angle (SZA) and corrected for sun-earth distance. The resulting bi-directional reflectances are in percent reflectance units, which are packed into 8-bit values in the output files. The Pathfinder coefficients used are a function of satellite and date from launch. The following equations were used for calibrating the NOAA-9, NOAA-11 and NOAA-14 data:

NOAA-9:

NOAA-11:

NOAA-14:

5.1.1.2 Channels 4 and 5

Channels 4 and 5 were converted from scaled GOES counts to brightness temperatures (in Kelvin), Ch4 and Ch5, using the following equations:

these equations were then corrected for calibration non-linearity by using tables given by Weinreb et al., 1990 (see also Rao, 1993). This correction depends upon the internal target temperature, which is not available in the GVI dataset. Table 5.1.12-1 contains the non-linearity coefficients (A0, A1, A2, A3) that were used for Channels 4 and 5 for NOAA-9, NOAA-11, NOAA-14.

The non-linearity coefficients were applied to the Ch4 and Ch5 brightness temperatures by using the following equations:

The AVHRR IR channels have shown problems with temperature values that exceed 326 K. Therefore, the maximum temperature used for scaling is 326 K. The user is cautioned about using GVI values over a hot desert.

5.1.1.3 Solar Zenith and Viewing Geometry

The AVHRR Solar Zenith Angle (SZA) and Scan Angle (SCA) files remained unchanged from the Second Generation product with the same scaling:

where C8 represents the corresponding 8-bit count values.

5.1.1.4 NDVI and Precipitable Water Index

In addition to the original channel and geometry files, the Third Generation GVI weekly composite product contains NDVI and Precipitable Water Index (PWI) files. The NDVI and PWI values are calculated from the calibrated/corrected reflectances and brightness temperatures:

5.1.1.5 Quality/Cloud Files

The Third Generation GVI weekly composite product has a quality/cloud (QC) file which allows one to: 1) identify residual cloud contaminated map cells and 2) test the quality of the data for missing or out of range values. The generation of QC flags is based on 6-year (1985-1991) means and standard deviations, , derived from clear-sky statistics of Ch1, Ch4 and PWI on a 2 x 2 degree monthly basis (Gutman et al, 1994). Although the QC file consists of 8 bits for each box, only Ch4 screening capability has been investigated, with bit 2 presently recommended to users. All bits are set to zero unless the criteria listed in Table 5.1.1.5-1 are met.

Table 5.1.1.5-1. Criteria for QC file.
Bit #	Contents
1	Ch1>Ch1+; bit 1 is set to 1
2	Ch4; bit 2 is set to 1
3	PWI>PWI+; bit 3 is set to 1
4	Ch1>Ch1+2; bit 4 is set to 1
5	Ch4; bit 5 is set to 1
6	PWI>PWI+2 ; bit 6 is set to 1
7	ChN<0 (N=1,2) or ChN<200K (N=4,5) or PWI>20K or NDVI>0.7 or SZA>90 degrees ; bit 7 is set to 1
8	Missing data (count of 255); bit 8 is set to 1

5.1.1.6 Scaling

All the enhanced GVI weekly composite floating point results are scaled back to the original standard 8-bit format. To obtain the equivalent floating point values, use the following equations:

Amended March 25, 1998

5.1.2 DIGITAL FORMAT

The Third Generation GVI weekly composite product tapes contain four weeks of data in the Plate Carrée projection only. Each week contains nine files. All Third Generation GVI weekly products use the same mapping projection (Plate Carrée) as the Second Generation weekly products, i.e. 904 x 2500 latitude/longitude arrays with about 0.15 degree resolution and with the first map cell at the northwest corner of the geographic domain (75N - 55S, 180W - 180E). The content of the data files are arranged as shown in Table 5.1.2-1.

All records are 2500 bytes in length. The first 512 bytes of each file contains a header which includes information about the satellite, navigation and display, along with ancillary and image data characteristics. The remaining 2,260,000 bytes contain the 904 latitude x 2500 longitude digital image. The format of the Third Generation weekly composite GVI header record is contained in Table 5.1.2-2. In this table, the field containing the number of bytes also includes the type of field; whether integer (I) or real (R). Real numbers (or floating point) were generated using C language and are not easily converted to useable numbers. However, Appendix K contains two software modules written in C language and FORTRAN which read the Third Generation header and output its contents. In addition, Appendix K contains IBM Job Control Language (header.jcl) for running the FORTRAN program in a batch mode.

The relative azimuth angle is not supplied with the Third Generation weekly composite GVI dataset but can be calculated using the FORTRAN read program supplied in Appendix L.

In the header record, bytes 19-20 (number of scan lines/orbit) and bytes 21-22 (number of scans/scan line) are set to zero since it is a weekly composite Plate Carrée projection. Bytes 23-24 can have values ranging from 0 to 9; where 0 = Ch. 1, 1 = Ch. 2, 3 = Ch. 4, 4 = Ch. 5, 5 =SZA, 6 = SCA, 7 = NDVI, 8 = PWI, and 9 = Quality/Cloud. The equator crossing time (bytes 33-36) is given in terms of UTC, e.g., a time of "14.75", would correspond to 2:45 PM.

Bytes 47-48 describe the type of storage interleaf. There are three possible methods of storage: scan interleaf, pixel interleaf and channel sequential. However, for the Third Generation GVI weekly composite dataset, scan interleaf was chosen. Scan interleaf implies that the data are stored by scan line. For the Third Generation GVI dataset there are 2500 bytes per scan line.

In the navigation portion of the header record, for a Plate Carrée image, bytes 51-52 (Polar Stereo) are set to zero, and bytes 53-56 (Mercator) are also set to zero. Bytes 57-72 contain the lower and upper latitude/longitudes for the Plate Carrée projection for the ranges shown in Table 5.1.2-2.

5.2 MONTHLY PRODUCT (C-LEVEL)

The Third Generation GVI monthly products represent the C-level dataset as compared to the weekly B-level data (see B-product description in Section 5.1). The C-level products were derived from the B-level data by monthly averaging of cloud screened observations, spatial interpolation and smoothing. They consist of monthly mean values for each individual year (April 1985 - September 1994) of Channel 1 and 2 reflectances, Channel 4 and 5 brightness temperatures, normalized difference vegetation index (NDVI), and precipitable water index (PWI), as well as monthly mean solar zenith angles, scan angles and the number of cloud free weekly observations that went into the monthly averaging (NOBS).

5.2.1 PROCESSING PROCEDURE

The production flow for the Third Generation GVI monthly products consists of the following steps (also see Gutman et al., 1995):

1) Quality/Cloud (QC) flags are applied on a weekly basis, resulting in data gaps. Bits 2, 7 and 8 in the QC file were used to screen clouds and check for missing and spurious data (see description of B-level Quality/Cloud files in Section 5.1.1.5).

2) Each variable is averaged over five composites covering each month (because of frequent overlaps of weekly composites for a particular month with the adjacent months). This is done for each map cell to partially fill in the data gaps and reduce some of the angular variability. The number of cloud free weeks is stored as an additional map file (NOBS).

3) Bi-linear spatial interpolation is applied to the missing data areas with persistent cloudiness in monthly averaged images.

4) 3 x 3 map cell smoothing is done to partially account for the imperfection of cloud screening, to filter out atmospheric and angular variabilities, and to compensate for random spatial sampling from GAC data into the GVI map cells.

The above procedure yields mean top-of-atmosphere values (averaged over different viewing and solar zenith angles) of Channel 1 and 2 reflectances, NDVI, Channel 4 and 5 brightness temperatures and PWI at each map cell of each month of each year.

5.2.2 SCALING AND DIGITAL FORMAT

All Third Generation GVI Monthly products use the same mapping projection (Plate Carrée) as the Third Generation GVI weekly products, i.e. 904 x 2500 latitude/longitude arrays with about 0.15 degrees resolution and with the first map cell at the northwest corner of the geographic domain (75N - 55S, 180W - 180E). All values are scaled to 8-bit precision and can be converted to floating point values using:

The C-level products differ from the B-level products in that the files do not contain a header record. C-level products are available from April 1985 through September 13, 1994. (A gap in the C-level exists from October 1994 until data from the NOAA-14 satellite started being used in February 1995.) Each file contains 904 records. All records are 2500 bytes in length and contain data in the order shown in Table 5.2.2-1. Each tape contains three months of C-level data.

5.3 THIRD GENERATION CLIMATOLOGY PRODUCT (D-LEVEL)

The Third Generation GVI Climatology product consists of 5-year means and standard deviations for each month of Channel 1 and 2 reflectances, Channel 4 and 5 brightness temperatures values, scan angles, solar zenith angles, normalized difference vegetation index (NDVI), and precipitable water index (PWI). These D-level products represent statistics of the monthly C-level data (see description of C-level in Section 5.2).

5.3.1 PROCESSING PROCEDURE

The D-level products (means and standard deviations) were derived from the C-level data for each pixel for each month for each variable over a 5-year period before the Mount Pinatubo eruption for the period from April 1985 through March 1991 (excluding 1988). Data from 1988 were not used because of NOAA-9's late orbit which caused illumination problems (too low sun elevation angle).

5.3.2 SCALING AND DIGITAL FORMAT

All Third Generation GVI Climatology products use the same map projection (Plate Carrée) as the Third Generation GVI monthly products, i.e. 904 x 2500 latitude/longitude arrays with about 0.15 degree resolution and with the first map cell at the northwest corner of the geographic domain (75N - 55S, 180W - 180E). All values are scaled to 8-bit precision and can be converted to floating point values using the same equations as the C-level product (see Section 5.2.2).

The D-level products are arranged in six month segments on tape. There are two types of D-level products available: means and standard deviations. Each tape usually contains 48 files, consisting of the eight basic files (Ch1 reflectances, Ch2 reflectances, Ch4 brightness temperatures, Ch5 brightness temperatures, SCA, SZA, PWI and NDVI) for six months, for either the means or standard deviations. The months are arranged chronologically on each tape (January through June and July through December). The user must specify whether means or standard deviations (or both) are desired.

Each file contains 904 records of 2500 bytes. The files for each month of data on the tape are arranged as shown in Table 5.3.2-1. Each tape contains six months of D-level data (either means or standard deviations).

Source: http://www.ncdc.noaa.gov