FSR_NDVI.DOC


                                  1.  TITLE

1.1  Data Set Identification

     Fourier-Adjustment, Solar zenith angle corrected, Interpolated 
     Reconstructed (FASIR) Normalized Difference Vegetation Index (NDVI).

     FASIR NDVI (Monthly ; NASA/GSFC)

1.2  Data Base Table Name.

     Not applicable.

1.3  CD-ROM File Name. 

     \DATA\VEGETATN\FSR_NDVI\YyyMmm.sfx

     Note: capital letters indicate fixed values that appear on the CD-ROM 
     exactly as shown here, lower case indicates characters (values) that 
     change for each path and file.

     The format used for the filenames is: YyyMmm.sfx, where yy is the last 
     two digits of the year (e.g., Y87=1987), and mm is the month of the year 
     (e.g., M12=December).  The filename extension (.sfx), identifies the data 
     set content for the file (see Section 8.2) and is equal to .FSR for this 
     data set.

1.4  Revision Date Of This Document

     April 5, 1995.

                             2.  INVESTIGATOR(S)

2.1  Investigator(s) Name And Title.

     Sietse O. Los, Compton J. Tucker, Chris O. Justice
     Biospheric Sciences Branch
     NASA/Goddard Space Flight Center
     Greenbelt MD

2.2  Title Of Investigation.

     Earth Observing System - Inter-Disciplinary Science
     project (Sellers - Mooney)

2.3  Contacts (For Data Production Information)

________________________________________________
              |             Contact 1           |
______________|_________________________________|
2.3.1 Name    |Sietse O. Los                    |
2.3.2 Address |Biospheric Sciences Branch       |
              |code 923                         |
              |NASA Goddard Space Flight Center |
      City/St.|Greenbelt MD                     |
      Zip Code|20771                            |
2.3.3 Fax     |(301) 286-1757                   |
2.3.4 Email   |Sietse@Jello.gsfc.nasa.gov       |
______________|_________________________________|
NOTE: Providing information on these data is not part of my daily routine; 
      Please read literature and descriptions first. Allow for some delay in 
      my response.

**********************************************************
* FOR GENERAL QUESTIONS REGARDING THE DATA CONTACT THE   *
*             GODDARD DAAC (SEE SECTION 13).             *
**********************************************************

2.4  Requested Form of Acknowledgment

     Please cite the following publications when ever these data are used:

      Sellers, P.J., S.O. Los, C.J. Tucker, C.O. Justice, D.A. Dazlich, G.J. 
         Collatz, and D.A. Randall, 1994. A global 1*1 degree NDVI data set 
         for climate studies. Part 2: The generation of global fields of 
         terrestrial biophysical parameters from the NDVI.  International 
         Journal of Remote Sensing, 15(17):3519-3545.
      Sellers, P.J., S.O. Los, C.J. Tucker, C.O. Justice, D.A. Dazlich, G.J. 
         Collatz, and D.A. Randall, 1995b. A revised land surface 
         parameterization (SiB2) for atmospheric GCMs. Part 2: The generation 
         of global fields of terrestrial biophysical parameters from satellite 
         data. Submitted to Journal of Climate.
      Los, S.O.,  C.O. Justice, C.J. Tucker, 1994. A global 1 by 1 degree NDVI 
         data set for climate studies derived from the GIMMS continental NDVI 
         data. International Journal of Remote Sensing, 15(17):3493-3518.

     ACKNOWLEDGMENTS

     This work was funded by a NASA Earth Observing System - Interdisciplinary 
     Science (EOS-IDS) grant (Sellers-Mooney). David Strauss of COLA, Dept. of 
     Meteorology of the University of Maryland is thanked for his advice on 
     Fourier series. The members of the GIMMS group at NASA Goddard Space 
     Flight Center, in particular Wayne Newcomb, are thanked for their 
     assistance. Computer code that aided in processing of the FASIR data set 
     was developed by Hong Zhang, Laboratory for Global Remote Sensing, 
     University of Maryland.

                                 3.  INTRODUCTION

3.1  Objective/Purpose

     The 1 X 1 degree FASIR - NDVI data set was compiled to correct the 1 X 1
     degree NDVI data (see the NDVI.DOC on this CD-ROM) for:

     - Data drop-outs occurring fewer than three months in a row.
       These drop-outs may be caused by cloud cover, atmospheric constituents
       (aerosols, molecules, gases, water vapor), or could be missing data.
     - Changes in the NDVI as a result of variations in solar zenith angle.
     - Missing data (>= 3 months) for evergreen needle bearing vegetation
       types, occurring for the most part in the Northern Hemisphere during 
       winter.
     - Persistent cloudiness (>= 3 months) over tropical forests. 

       The corrected NDVI is referred to as FASIR-NDVI (see section 3.2).  The 
       FPAR, LAI, Greenness, Roughness Length and Albedo data sets were 
       calculated (see following documents FPAR.DOC, LAI.DOC, GREENNSS.DOC, 
       ROUGHNSS.DOC and SUR_ALBD.DOC) using the FASIR-NDVI data set.  These 
       five data sets were used in the simple biosphere model (SiB2) of 
       Sellers et al (1994, 1995a), to calculate the energy and water exchange 
       between the biosphere and the lower boundary layer of the atmosphere.

3.2  Summary of Parameters

     FASIR:    Fourier based Adjustment, Solar zenith angle correction, 
               Interpolation of missing data and Reconstruction of 
               evergreen broadleaf land cover types (tropics). FASIR is 
               calculated from NDVI and VEG_CLSS.VGC data.

     Inputs to FASIR-NDVI:

     NDVI:         Normalized Difference Vegetation Index calculated from 
                   AVHRR channel 1 and 2 normalized radiance data with NDVI = 
                   (L2 - L1)/(L2 + L1)
     VEG_CLSS.VGC: Land cover classification data (this CD-ROM).
                   Solar Zenith Angle: calculated with NOAA/NASA AVHRR Land 
                   Pathfinder navigation routines (James and Kalluri, 1994).

3.3  Discussion.

     FASIR NDVI was developed to account for errors in the NDVI data.
     In order to fix some of the problems in the NDVI data set, approximate
     corrections were developed to obtain a more complete and consistent
     coverage of the globe. Some of the corrections to the NDVI data are made,
     specifically with the use of SiB2 in mind, e.g., SiB2 uses NDVI values
     to calculate FPAR and LAI values, which in turn are used to calculate 
     (amongst others) photosynthetic rate, albedo and roughness length.  
     Missing NDVI data (which can be substituted by zero NDVI values) for 
     needle bearing evergreen vegetation during winter, are not a problem for 
     the calculation of photosynthetic rates, because these values will be 
     close to zero. However, for the calculation of roughness length, a zero
     NDVI value is not acceptable, because this value would translate to a 
     zero LAI, and by the way SiB2 is formulated, to a zero roughness length.
     This would result in erroneous estimates for the aerodynamic resistance.
     (see also NDVI document, and Sellers et al 1994).

                          4.  THEORY OF MEASUREMENTS

     Note: the description in the NDVI document provides general information
     on the purpose of this data set.

     The derivation of the FASIR corrections depends heavily on assumptions
     made with respect to the distribution of errors in the NDVI. FASIR is
     divided into the following steps:

     FA: Fourier Adjustment

     The Fourier adjustment requires an input of 12 monthly values at a time. 
     In general, it is assumed that the effects of atmosphere, clouds, off-
     nadir viewing etc. will lower the value of the NDVI. It is further 
     assumed that the development of NDVI-time series is smooth, i.e., sudden 
     changes in the NDVI are most likely to be related to distortions of the 
     time-series. Using these two assumptions, a technique was developed which 
     fits a smooth series through an NDVI sequence for a year. The smoothed 
     NDVI curve is compared with the values of the original series. Points 
     below the fitted curve are likely to be unreliable (errors tended to 
     decrease the NDVI). Based on the distance to the fitted curve weights 
     were calculated and a new curve was fitted with strong emphasize on the 
     reliable points (points above the first fitted curve).

     S: Solar zenith angle adjustment:

     We currently do not have sufficient data to do a full solar zenith angle
     correction over the entire range of NDVI. By comparing the tails (i.e., 
     the high and low end) of NDVI distributions taken for a number of 
     vegetation types and for different solar zenith angle intervals, and by 
     assuming that the the NDVI values at these tails are from equally green 
     (or dead) targets, (in other words we say that changes at the tails of 
     distributions taken for different solar zenith angle intervals are the 
     result of changes in solar zenith angle) we estimate the result of solar 
     zenith angle effect for the high and low ends of the NDVI. Lacking 
     further information, we assume the solar zenith angle effect to be linear 
     over the range of NDVI, even though we know from ground measurements that 
     quite often this is not the case.  This procedure requires an assumption 
     about vegetation type.  In this version, we made use of the satellite-
     based vegetation classification of DeFries and Townshend (1994a, b) 
     instead of the in situ-based product described in Sellers et al (1994).  
     The DeFries and Townshend (1994a, b) classification data are on the 
     ISLSCP Initiative I CD-ROM Volume I.

     I: Interpolation

     The interpolation of NDVI data for evergreen needle leaf ground cover 
     types during the winter months is necessary to estimate values for leaf 
     area index, roughness length, and albedo using SiB2.

     R: reconstruction:

     The reconstruction over tropical forests (broad leaf evergreen vegetation
     type, class 1 in VEG_CLSS.VGC) is done to remove the effects of cloud, 
     not removed by the Fourier Adjustment. We take the maximum NDVI over the 
     year to represent the value for each month. This is not a very desirable 
     procedure, since it removes the seasonality in the data.  However, 
     failing to adjust the data, will result in underestimates for the NDVI, 
     which translates into low estimates of FPAR.  This would result in severe 
     underestimation of LAI, because the FPAR-LAI relationship is exponential.

                                5.  EQUIPMENT

5.1  Instrument Description.

     See the NDVI document.

     5.1.1  Platform (Satellite, Aircraft, Ground, Person...).

            See the NDVI document.

     5.1.2  Mission Objectives.

            See the NDVI document.

     5.1.3  Key Variables.

            See the NDVI document.

     5.1.4  Principles of Operation.

            See the NDVI document.

     5.1.5  Instrument Measurement Geometry.

            See the NDVI document.

     5.1.6  Manufacturer of Instrument.

            See the NDVI document.

5.2  Calibration.

     See the NDVI document.

     5.2.1  Specifications.

            See the NDVI document.

            5.2.1.1  Tolerance.

                     See the NDVI document.

     5.2.2  Frequency of Calibration.

            See the NDVI document.

     5.2.3  Other Calibration Information.

            See the NDVI document.

                                   6.  PROCEDURE

6.1  Data Acquisition Methods.

     The 1 X 1 degree FASIR-NDVI data set is derived from the 1 X 1 degree 
     global NDVI data set (1982 - 1990), and the 1 X 1 degree land cover
     data (VEG_CLSS.VGC). Two years from the full 8 years data set are 
     extracted for this CD-ROM.

6.2  Spatial Characteristics. 

     6.2.1  Spatial Coverage. 

            The coverage is global.  Data in file are ordered from North to 
            south and from West to East beginning at 180 degrees West and 90 
            degrees North.  Point (1,1) represents the grid cell centered at 
            89.5 N and 179.5 W (see section 8.4).

     6.2.2  Spatial Resolution. 

            The data are given in an equal-angle lat/long grid that has a 
            spatial resolution of 1 * 1 degree lat/long.

6.3  Temporal Characteristics. 

     6.3.1  Temporal Coverage. 

            January 1987 through December 1988.

     6.3.2  Temporal Resolution. 

            Monthly.

                                 7.  OBSERVATIONS

7.1  Field Notes.

     Not applicable.

                               8.  DATA DESCRIPTION

8.1  Table Definition With Comments.

     Not applicable.

8.2  Type of Data 

--------------------------------------------------------------------------------
|                 8.2.1                |             |              |          |
|Parameter/Variable Name               |             |              |          |
--------------------------------------------------------------------------------
|    |               8.2.2             |     8.2.3   |    8.2.4     |  8.2.5   |
|    |Parameter/Variable Description   |Range        |Units         |Source    |
--------------------------------------------------------------------------------
|FASIR_NDVI (Sellers et al., 1994)     |             |              |          |
|    |Normalized Difference Vegetation |Min = 0.05 $ |[Unitless]*   |Los et al.|
|    |Index, which has been corrected  |Max = 0.65 $ |              |(1994);   |
|    |for distortions caused by        |             |              |Sellers et|
|    |persistent cloud cover,          |             |              |al.       |
|    |atmospheric aerosols and water   |             |              |(1994a)   |
|    |vapor, solar zenith angle        |             |              |          |
|    |effects, missing data in the     |             |              |          |
|    |Northern Hemisphere during       |             |              |          |
|    |winter, and persistent           |             |              |          |
|    |cloudiness over tropical forests |             |              |          |
|    |                                 |             |              |          |
--------------------------------------------------------------------------------
$The minimum and maximum NDVI are for land surfaces, desert and dense 
 vegetation respectively. Both values are approximate.  A minimum value of 
 0.033 has been inserted into the FASIR_NDVI data sets for regions (Antarctic, 
 Greenland...) that are classified as missing data in the NDVI data set (see 
 NDVI.DOC).
*FASIR-NDVI Units are non-dimensional, a fraction with a potential range 
 between -1 and 1.

8.3  Sample Data Base Data Record.

     Not applicable.

8.4  Data Format

     The CD-ROM file format is ASCII, and consists of numerical fields of 
     varying length, which are space delimited and arranged in columns and 
     rows.  Each column contains 180 numerical values and each row contain 360 
     numerical values.  

          Grid arrangement

             ARRAY(I,J)
             I  = 1 IS CENTERED AT 179.5W
             I INCREASES EASTWARD BY 1 DEGREE
             J  = 1 IS CENTERED AT 89.5N
             J INCREASES SOUTHWARD BY 1 DEGREE

             90N - | - - - | - - - | - - - | - -
                   | (1,1) | (2,1) | (3,1) |
             89N - | - - - | - - - | - - - | - -
                   | (1,2) | (2,2) | (3,2) |
             88N - | - - - | - - - | - - - | - -
                   | (1,3) | (2,3) | (3,3) |
             87N - | - - - | - - - | - - - |
                  180W   179W    178W   177W

            ARRAY(360,180)

8.5  Related Data Sets 

     1 X 1 degree Normalized Difference Vegetation Index (NDVI) global data 
           (on this CD-ROM).
     1 X 1 degree global land cover classification image (on this CD-ROM).
     1 X 1 degree fraction of photosynthetic active radiation absorbed by the 
           vegetation canopy (FPAR) monthly global data (on this CD-ROM).
     1 X 1 degree leaf area index (LAI) global data (on this CD-ROM).
     1 X 1 degree Greenness global data (on this CD-ROM).
     1 X 1 degree roughness length monthly global data (on this CD-ROM).
     1 X 1 degree snow free Surface Albedo monthly global data (on this 
           CD-ROM).
     1 X 1 degree vegetation classification map (Dorman and Sellers 1989, 
           Nobre et al. 1991), on this CD-ROM.

     Also see section 8.5 (Related Data Sets) in the NDVI document.

                            9.  DATA MANIPULATIONS

9.1  Formulas.

     Data have been adjusted using both formulas and rules, the description of 
     the complete procedure is provided below in section 9.2.

     9.1.1  Derivation Techniques/Algorithms.

            See section 9.2 below.

9.2  Data Processing Sequence 

     See Sellers et al., 1994, 1995b for a more elaborate description.

     9.2.1  Processing Steps and Data Sets.

            The following data sets are used for the calculation of FASIR-
            NDVI.

            1. 1 X 1 degree NDVI data set (1982 - 1990, two years were 
               extracted for this CD-ROM) for FA, S, I and R corrections.
            2. Land cover classification (VEG_CLSS.VGC, this CD-ROM) for S, I 
               and R corrections.
            3. Estimated solar zenith angles at time of data collection by the
               satellite for S correction.
            4. Solar zenith angles for noon, local time (15th of the month)
               for S correction.


            Processing steps

            1. Fourier adjustment (uses NDVI data only, data are converted to
               NDVI units, different results will be obtained when using 
               counts).  Fill up and count missing data of temporal NDVI data 
               series Y

            IF Y_i .EQ. missing value THEN Y_i = 0

            IF number of missing values .LT. 3/4 * 12 THEN BEGIN (else do 
               nothing)

                      For each 1 X 1 degree cell a Fourier series of 3 
                      harmonics is fitted with least squares through a time-
                      series of 12 months using the matrix equation, solve [c] 
                      with

                      ( [F]^T [F] ) [c] = [F]^T [Y]

            with

                 | 1, cos[phi_01], sin[phi_01], cos[2 phi_01], sin[2 phi_01] |
                 | .,      .     ,      .     ,      .       ,       .       |
           [F] = | .,      .     ,      .     ,      .       ,       .       |
                 | .,      .     ,      .     ,      .       ,       .       |
                 | 1, cos[phi_12], sin[phi_12], cos[2 phi_12], sin[2 phi_12] |

                      phi_i = 2 * pi *(i - 1) / 12; i refers to month i

            [F]^T = transposed [F]

                    |a_0|
                    |a_1|
            [c]   = |b_1|       Fourier coefficients to be solved
                    |a_2|
                    |b_2|

                    |Y_1 |
                    | .  |
            [Y]   = | .  |      the NDVI time-series for a particular 1*1 cell
                    | .  |
                    |Y_12|

            The Fourier series Y' is calculated by

            Yi'   = Sum[a_i * cos[(j-1) phi_i] + b_i * sin[(j-1) phi_i]; 
                    {j=1,3}]

            b_0 can not be determined because sin[0] = 0;
            A weight is calculated based on the difference between Y and Y'

            m = Median[ Abs[ Y - Y'] ]

            U = [ Y - Y'] / m

            Four cases are distinguished to calculate the weight

                 Wi = 1                        -R <=  U_i <= R
                 Wi = (1 + (U_i - R)/K)^2             U_i >  R
                 Wi = (1 + (U_i + R)/K)^4             U_i <  R
                 Wi = 0                        otherwise
            and
                 0 <= W1  <= 1
                 0 <= W12 <= 1

            A second Fourier series is fitted through the data with each point
            weighted

                          ( [Fw]^T [Fw] ) [c] = [Fw] [Yw]

            with

            [Fw]<i>  =  [W]<i> . [F]<i>
            [Yw]<i>  =  [W]<i> . [Y]<i>
           (each column is multiplied by the weights)

            <i> column i
            The second Fourier series Y" is calculated with coefficients c
            The restored data Yf are calculated from Y and Y" according to
            the following rules

            Yf_i = Max[Y_i, Min [Y"_i, 1.02 * Max [Y_i-2, .. , Y_i+2] ] ]

            If three or more missing values in a row in the original values 
            then

            If (Y_j .. Y_j+k .EQ. missing value) .AND. (k .GE. 2) 

            Yf_i .. Yf_i+k = missing value


            The Fourier adjustment is applied to series of 12 months of NDVI
            values at a time with 6 month between consecutive series. The 
            middle 6 months of each series were selected to obtain a 9 year 
            Fourier adjusted series according to the following scheme

            input (month) 1..12,  7..18, 13..24, ...,  97..108
            selected      1.. 9, 10..15, 16..21, ..., 100..108

            2. Solar zenith angle adjustment 

               Input a. Fourier adjusted NDVI, years 1985 - 1990, 1982 - 1984
                        was not incorporated because of a volcanic eruption in 
                        1982 in Chiapas, Mexico, which affected the NDVI data 
                        around the equator. For the interpolation, the entire 
                        data set was used, because no effect of the volcanic 
                        eruption was found at high latitudes where most of the 
                        needle bearing evergreen forests occur.

                     b. VEG_CLSS.VGC, see Defires and Townshend (1994a, b)

                     c. Mask with 2 degree intervals of solar zenith angle.
                        A guess had to be made about the value of the solar 
                        zenith at the time of overpass of the NOAA satellite, 
                        because this information was discarded during 
                        processing of the GIMMS continental NDVI data set. The 
                        estimation of solar zenith angles was done as follows:

                        The solar zenith angle was calculated for 10 degrees 
                        in the forward scatter direction (=average angle from 
                        which data are selected by compositing), then an 
                        assumption was made for the average day of the month 
                        from which the composite tends to select the data. It 
                        was decided that the minimum solar zenith angle of 5, 
                        15 and 25th day on average, would be likely to concur 
                        with maximum NDVI conditions for the month. 
                        (especially, in the case of growing or senescing 
                        vegetation, high NDVI is concurrent with low solar 
                        zenith angle). For the calculation of solar zenith 
                        angle, the navigation subroutines of the NOAA/NASA 
                        Pathfinder AVHRR Land project (see James and Kalluri, 
                        1994) were used.

                     d. Mask with solar zenith angles at noon for the middle 
                        of the month. Two effects are concurrent with solar 
                        zenith angle changes, 1: a change in the NDVI as a 
                        result of illumination geometry, this is the effect we 
                        want to account for, and 2: a decrease in the NDVI as 
                        a result of lower ambient temperatures caused by high 
                        solar zenith angles during late autumn, winter and 
                        early spring. Inclusion of this second effect in the 
                        estimation of solar zenith angle effect is clearly 
                        not desirable. To eliminate this second effect from 
                        our data, we calculated a mask with solar zenith  
                        angles at noon, and only used those values in itme c 
                        above.  for which the solar zenith angle at noon was 
                        smaller than 35 degrees.

            Procedure

              I. Create 6 X 12 monthly masks with stratification according to 
                 land  cover class (b) and solar zenith angle interval (c. 
                 masked with d.) for the entire period from which data were 
                 collected

             II. For each month calculate Fourier adjusted NDVI-distributions 
                 for classes of equal solar zenith angle interval and land 
                 cover class.

            III. Combine monthly distributions of equal solar zenith angle 
                 interval  and vegetation class for nine years.

             IV. Calculate 98 % value of histogram for each solar zenith angle 
                 interval (for 30 < sza < 60) for the following vegetation 
                 classes.

                 1 + 6          evergreen broadleaf (used for classes 1,6 
                                and 8)
                 2 + 3 + 4 + 5  mix deciduous/evergreen broad/needle leaf
                                (also used for 10)
                 12             agriculture (used for 7, 9, 11, 12)

                 for classes with bare soil and bare soil and shrubs (9 and 
                 11) combined calculate NDVI at 2%

              V. Plot NDVI 98% against solar zenith angle and fit data with 
                 a non-linear least squares method (routine nls from S-Plus
                 language, see Bates and Chambers, 1992) for the range of 30
                 to 60 degrees solar zenith angle.

                 NDVI_{S,P98}= NDVI_{S=0, P98) - k1 * (S - PI()/6)^k2 

                 S     = solar zenith angle (in radians)
                 S=0   = overhead solar zenith angle
                 P98   = 98 % of NDVI distribution for a particular vegetation
                           class

                 The NDVI_{SZA, P98} is held constant for solar zenith angles 
                 larger than 60 degrees. The NDVI 2% (_02) of classes 9 and 11 
                 is  held constant throughout the range (the parameters k1 and 
                 k2 were found not to deviate significantly (at 95 % 
                 confidence level) from 0 and 1, respectively). We removed 
                 three outliers from the 98% data for the agriculture class 
                 (type 12 in VEG_CLSS.VGC). We used criterium residual < - 1.5 
                 median of absolute values of the residuals.

                 Data are adjusted according to

                 NDVI_{S=0} = (NDVI_{S} - NDVI_{P2}) * (NDVI_{S=0,P98} - 
                               NDVI_{P2}) (NDVI_{S,P98} - NDVI_{P2}) + 
                               NDVI_{P2}

                 with NDVI_{S,P98} dependent on class and NDVI_{P2} constant:

                 IV refers to land cover class, see documentation 
                 VEG_CLSS.VGC.

                 ================================================
                 vegetation parameters
                 ================================================
                  1    0.618          0.034     0.2403     2.8216
                  2    0.686          0.034     0.5529     2.5069
                  3    0.686          0.034     0.5529     2.5069 
                  4    0.686          0.034     0.5529     2.5069
                  5    0.686          0.034     0.5529     2.5069
                  6    0.618          0.034     0.2403     2.8216 
                  7    0.630          0.034     0.1702     1.6091 
                  8    0.618          0.034     0.2403     2.8216 
                  9    0.630          0.034     0.1702     1.6091
                  10   0.686          0.034     0.5529     2.5069 
                  11   0.630          0.034     0.1702     1.6091 
                  12   0.630          0.034     0.1702     1.6091 
                  IV NDVI_{S=0,P98}  NDVI_{P2}     k1         k2  

                  Table with NDVI_{S=0,P98} and NDVI_{P2} values for
                  each vegetation type

            3. Interpolation

               The missing NDVI data in evergreen needle bearing forests 
               during winter are a problem for the calculation of snow free 
               albedo and surface roughness within SiB2. We therefore have to 
               make an estimate of the "snow free" NDVI for these cases. This 
               is done by taking the average October value for the period of 
               1982 - 1990 to represent the NDVI of evergreen needle leaf 
               vegetation cover in the Northern Hemisphere. For the Southern 
               Hemisphere we use the average of the March values. We assume 
               that for these months near the end of the growing season, the 
               NDVI signal is primarily from evergreen vegetation because most 
               of the deciduous vegetation has no leaves.  We replace missing 
               values when we find three or more missing values in a row. 
               (Note: We can estimate up to two missing values with the 
               Fourier adjustment.) The rules used are:

               If (three or more missing data in a row) and (class 4)

               then for November until April

               NDVI = Max (NDVI, avg of all NDVI values for October for that
               particular pixel). Missing October values are excluded from
               averaging.

            4. Reconstruction of evergreen broadleaf classes

               Even though the Fourier adjustment is believed to have removed
               a large proportion of the cloud effects in the NDVI data, we
               still found fairly low NDVI values over the tropical forests,
               which, because of the exponential relationship between FPAR and
               leaf area index (LAI), resulted in fairly low estimates of LAI.
               This situation improved when we took the maximum NDVI of the
               year. This step is clearly a quick fix to the problem, we plan
               to use more advanced cloud screening procedures in future data 
               sets so that cloud problems are caught at an earlier stage.

               rule used for Reconstruction
               if (class .EQ. 1) then
               NDVI = Max (NDVI_jan .. NDVI_dec)

     9.2.2  Processing Changes.

            With this update of FASIR (release 2.0 some changes were made
            with respect to the description of FASIR in the Sellers et al
            (1994) paper. These changes are:
            - use of a different land cover classification (VEG_CLSS.VGC)
            - use of the multi-year NDVI data set to derive the solar zenith
              angle correction (1985 - 1990 instead of 1987)
            - different assumptions about the solar zenith angle at which data
              were collected by the satellite (10 degrees off-nadir in the
              forward scatter direction, minimum solar zenith angle of 5th,
              15th and 25th day of the month)
            - use of Pathfinder navigation software instead of the diagrams
              published in Kidwell (1988)
            - more selective procedure for selection of data (NDVI data with a
              solar zenith angle >35 degrees at noon for the 15th of the month
              are not used in the derivation of the technique)
            - direct estimation of parameters k1 and k2 (see section 9.2.1) 
              using a non-linear least squares technique (Bates and Chambers, 
              1992) instead of estimating k1 and k2 from the transformed data.

9.3  Calculations.

     Not available at this revision.

     9.3.1  Special Corrections/Adjustments.

            Not available at this revision.

9.4  Graphs and Plots.

     Not available at this revision.

                                   10.  ERRORS

10.1  Sources of Error.

      Some sources of error in the NDVI data set are not accounted for with 
      the FASIR corrections, these errors are caused by:

      - variations in off-nadir viewing
      - soil background reflectance
      - a bias as a result of compositing
      - long-term atmospheric effects. (e.g., after volcanic eruptions)

      The largest uncertainties in the FASIR correction are in steps 3 and 4, 
      since for these cases data are adjusted without sufficient information. 
      Errors (overestimates) may be introduced by the Fourier adjustment in 
      cases of rapid changes in the NDVI during the beginning and end of 
      season.  The solar zenith angle effect on NDVI is assumed to be linear. 
      This assumption was not always found to hold true in several ground 
      based and model studies (e.g., Deering et al., 1992).

10.2  Quality Assessment. 

      10.2.1  Data Validation by Source. 

              A rigorous verification of the FASIR-NDVI has still to be done 
              (Sellers et al 1995b).

      10.2.2  Confidence Level/Accuracy Judgment. 

              The FASIR-NDVI data set is believed to give large improvements 
              over the NDVI data set, especially for areas with persistent 
              cloud cover and for needle bearing evergreen vegetation during 
              winter.

      10.2.3  Measurement Error for Parameters and Variables.

              Errors are generally systematic and lower the NDVI value. For 
              dense vegetation cover (NDVI > 0.6), the errors approximately 
              are cloud contamination reduced by 70 - 90 % atmosphere reduced 
              if incidental (< 1 - 2 months) off-nadir viewing: +0.02 forward 
              -0.05 backscatter solar zenith angle: reduced missing data 
              (assume NDVI = October value for needle, otherwise 0), reduced 
              misregistration + 0.02 compositing (0 .. +0.05). For the low end 
              of NDVI, the soil can change the NDVI by ca +0.5 - + 1.5 units.

      10.2.4  Additional Quality Assessment Applied.

              Not available at this revision.

                                     11.  NOTES

11.1  Known Problems With The Data.

      Not available at this revision.

11.2  Usage Guidance 

      FASIR NDVI presents generalized patterns which may result in poor 
      representations of a specific locale, quantitative conclusions should be 
      drawn with caution. Nevertheless, FASIR-NDVI should provide a large
      improvement over previously used land cover schemes (e.g. Dorman and
      Sellers 1989) because the data are collected by one series of 
      instruments, and they give a more realistic representation of the 
      spatial and temporal variability of vegetation patterns over the globe.

11.3  Other Relevant Information.

      Not available at this revision.

                                 12.  REFERENCES. 

12.1  Satellite/Instrument/Data Processing Documentation. 

      Kidwell, K. B., 1988. NOAA polar orbiter data (TIROS-N, NOAA-6, NOAA-
         7, NOAA-8, NOAA-9, NOAA-10 and NOAA 11) users guide. (National
         Oceanic and Atmospheric Administration: Washington, DC 20233).


12.2  Journal Articles and Study Reports.

      DeFries, R. S. and J. R. G. Townshend, 1994a, NDVI-derived land 
           cover classification at global scales. International Journal of 
           Remote Sensing, 15:3567-3586. Special Issue on Global Data Sets.
      DeFries, R. S. and J. R. G. Townshend, 1994b. Global land cover:
           comparison of ground-based data sets to classifications with AVHRR 
           data. In Environmental Remote Sensing from Regional to Global 
           Scales, edited by G. Foody and P. Curran, Environmental Remote 
           Sensing from Regional to Global Scales. (U.K.: John Wiley and 
           Sons).
      Bates M, and Chambers, J.M., 1992, Nonlinear Models, In: Chambers, J.M.
         and Hastie, T.J., Statistical models in S, Wadsworth & Brooks, Ca. pp
         421-454.
      Deering, D.W., E.M. Middleton, J.R. Irons, B.L. Blad, E.A. Walter-
         Shea, C.J. Hays, C. Walthall, T.F. Eck, S.P. Ahmad and B.P. Banerjee. 
         1992. Prairie grassland bi-directional reflectances measured by 
         different instruments at the FIFE site. Journal of Geophysical 
         Research, 97:18,887-18,903.
      Dorman, J. L., and P.J. Sellers, 1989. A Global climatology of albedo,
         roughness length and stomatal resistance for atmospheric general
         circulation models as represented by the simple biosphere model 
         (SiB). Journal of Applied Meteorology, 28:833-855.
      James, M.E., S.N.V. Kalluri. 1994. The Pathfinder land data set: an 
         improved coarse resolution data set for terrestrial monitoring. 
         International Journal of Remote Sensing, 15(17):3347-3363.
      Los, S.O.,  C.O. Justice, C.J. Tucker, 1994. A global 1 by 1 degree NDVI 
         data set for climate studies derived from the GIMMS continental NDVI 
         data. International Journal of Remote Sensing, 15(17):3493-3518.
      Sellers, P.J., S.O. Los, C.J. Tucker, C.O. Justice, D.A. Dazlich, G.J. 
         Collatz, and D.A. Randall, 1994. A global 1*1 degree NDVI data set 
         for climate studies. Part 2: The generation of global fields of 
         terrestrial biophysical parameters from the NDVI.  International 
         Journal of Remote Sensing, 15(17):3519-3545.
      Sellers, P.J., D.A. Randall, C.J. Collatz, J.A. Berry, C.B. Field, D.A. 
          Dazlich, C. Zhang, and C.D. Collelo, 1995a. A revised land surface 
          parameterization (SiB2) for atmospheric GCMs. Part 1: Model 
          formulation. submitted to Journal of Climate.
      Sellers, P.J., S.O. Los, C.J. Tucker, C.O. Justice, D.A. Dazlich, G.J. 
         Collatz, and D.A. Randall, 1995b. A revised land surface 
         parameterization (SiB2) for atmospheric GCMs. Part 2: The generation 
         of global fields of terrestrial biophysical parameters from satellite 
         data. Submitted to Journal of Climate.

12.3  Archive/DBMS Usage Documentation.

      Contact the EOS Distributed Active Archive Center (DAAC) at NASA Goddard 
      Space Flight Center (GSFC), Greenbelt Maryland (see Section 13 below).
      Documentation about using the archive or information about access to the 
      on-line information system is available through the GSFC DAAC User 
      Services Office.

                             13.  DATA ACCESS

13.1  Contacts for Archive/Data Access Information.

      GSFC DAAC User Services
      NASA/Goddard Space Flight Center
      Code 902.2
      Greenbelt, MD 20771

      Phone:     (301) 286-3209
      Fax:       (301) 286-1775
      Internet:  daacuso@eosdata.gsfc.nasa.gov

13.2  Archive Identification.

      Goddard Distributed Active Archive Center
      NASA Goddard Space Flight Center
      Code 902.2
      Greenbelt, MD 20771

      Telephone:  (301) 286-3209
      FAX:        (301) 286-1775
      Internet:   daacuso@eosdata.gsfc.nasa.gov

13.3  Procedures for Obtaining Data.

      Users may place requests by accessing the on-line system, by sending 
      letters, electronic mail, FAX, telephone, or personal visit.

      Accessing the GSFC DAAC Online System:

      The GSFC DAAC Information Management System (IMS) allows users to 
      ordering data sets stored on-line.  The system is open to the public.

      Access Instructions:

      Node name:  daac.gsfc.nasa.gov
      Node number: 192.107.190.139
      Login example: telnet daac.gsfc.nasa.gov
      Username:  daacims
      password:  gsfcdaac

      You will be asked to register your name and address during your first
      session.

      Ordering CD-ROMs:

      To order CD-ROMs (available through the Goddard DAAC) users should 
      contact the Goddard DAAC User Support Office (see section 13.2).

13.4  GSFC DAAC Status/Plans.

      The ISLSCP Initiative I CD-ROM is available from the Goddard DAAC.

                        14.  OUTPUT PRODUCTS AND AVAILABILITY 

14.1  Tape Products.

      None.

14.2  Film Products. 

      None.

14.3  Other Products. 

      None.

                             15.  GLOSSARY OF ACRONYMS

AVHRR        Advanced Very High Resolution Radiometer
CD-ROM       Compact Disk (optical), Read Only Memory
DAAC         Destributed Active Archive Center
EOS          Earth Observing System
GAC          Global Area Coverage
GCM          General Circulation Model of the atmosphere
CIA          Central Intelligence Agency
FASIR (NDVI) Fourier Adjusted, Solar zenith angle correction,
             Interpolation (of missing data during winter), and
             Reconstruction of NDVI over tropical forests.
FPAR         Fraction of Photosynthetic Active Radiation absorbed by
             green vegetation
GIMMS        Global Inventory Monitoring and Modeling Studies at
             NASA GSFC
GSFC         Goddard Space Flight Center
GVI          Global Vegetation Index
IDS          Inter disciplinary Science
IFOV         Instantaneous Field Of View
ISLSCP       International Satellite Land Surface Climotology Project
LAC          Local Area Coverage
LAI          Leaf Area Index
NASA         National Aeronautics and Space Administration
NDVI         Normalized Difference Vegetation Index
NOAA         National Oceanographic and Atmospheric Administration
pixel        Picture element
SiB2         Simple Biosphere model (Sellers et al 1995a)
SR           Simple Ratio