Satellite observed global vegetation dynamics and its relations with biosphere-atmosphere carbon exchange

Satellite observed global vegetation dynamics and its relations with biosphere-atmosphere carbon exchange

Adv. Space Res. Vol. 9, No. 7, pp. (7)229-(7)237, 1989 Printed in Great Britain. 0273-1177/89 $0.00 +.50 1989 COSPAR SATELLITE O B S E R V E D G L O...

525KB Sizes 1 Downloads 13 Views

Adv. Space Res. Vol. 9, No. 7, pp. (7)229-(7)237, 1989 Printed in Great Britain.

0273-1177/89 $0.00 +.50 1989 COSPAR

SATELLITE O B S E R V E D G L O B A L V E G E T A T I O N D Y N A M I C S A N D ITS R E L A T I O N S WITH B I O S P H E R E ATMOSPHERE CARBON EXCHANGE B. J. Choudhury* and I. Y. Fung** *Hydrological Sciences Branch~624; **Goddard Institute for Space Studies, NASA/Goddard Space Flight Center, Greenbelt, MD 20771, U.S.A.

ABSTRACT Satellite observations of visible and near-infrared reflectances and brightness temperatures at 37 GHz frequency are studied to quantify spatial and temporal variations of land-surface vegetation. These satellite data are further correlated with the temporal variations of the atmospheric CO 2 concentration and the terrestrial primary productivity. INTRODUCTION The biosphere plays a significant role in the geographic and temporal variations of the atmospheric CO^ concentration. The biosphere decreases the atmospheric CO^ via • z . . . . z. photosynthesls and increases the concentratlon vla resplratlon and decomposltion. The net carbon incorporated into the biosphere as a result of this CO 2 exchange, which could be quantified through net primary productivity (NPP), is recognized to represent the single most important function of the biosphere. Thus, a global, multitemporal data for vegetation would be expected to play a key role in quantitative understanding of the atmospheric CO concentration and NPP. This paper discusses satellite observations of 2 spatial and temporal variation of vegetation as quantified through shortwave reflectances and microwave emission, and empirical relations between these observations and the atmospheric CO 2 concentration and NPP. SATELLITE OBSERVATIONS The satellite data analyzed in this paper is summarized in Table i. The vegetation indices are based upon shortwave reflectances and microwave emission. The advanced TABLE

I

Summary of Satellite Data for Monitoring Vegetation Analyzed

VEGETATION Source and Other Description

Normalized Difference

in This Paper

INDEX

Polarization

Difference

Satellite Sensor Equator Crossing Wavelength (approximate)

NOAA- 7 AVHRR 1430 0.6 - 0.7 and 0.6 - i.i #m

Nimbus -7 SMMR 1200 8 mm

Spatial Resolution

4 km, resampled 15 km

25 km

Temporal Resolution

Daily, composited monthly

Weekly, Composited monthly

Temporal

April 1982 January 1985

January 1979 December 1985

Global

Global

Span

Spatial Domain

very high resolution radiometer (AVHRR) on board NOAA satellites provide reflectances the visible (Ch i) and near-infrared (Ch 2) spectral bands from which the normalized difference (ND) vegetation index is computed as

(7)229

in

(7)230

B.J. Choudhury and I. Y. Fung

Ch 2 - Ch i ND . . . . . . . . . . . . . . Ch 2 + Ch i

(1)

The ND values are based upon the daily global area coverage (GAC) data, which has a nominal spatial resolution of about 4 km. For a pre-selected spatial grid of nominal resolution 15 to 20 km, one GAC pixel is selected to be representative of the entire grid area based upon a temporal compositing which attempts to minimize the effects of cloud and aerosols. It should be noted that (i) the effect of cloud and aerosols are not completely eliminated, (2) changes in the solar illumination, time of observation, and background (e.g., soil and snow) reflectance affect the ND values, and (3) the ND values are based upon channel counts rather than reflectances. The ND values are generally negative for water and clouds, range between 0.0 - 0.I for deserts or very sparsely vegetated soils, and between 0.i ° 0.5 for varying amounts of green vegetation /1,2/. Analysis of field-observed ND values for uniform stands of agricultural crops with varying green leaf area index (L) and the radiative transfer equation give a semi-empirical relation between ND and L as /3/. ND = NDma x

(NDma x - NDmin) exp (-~L)

(2)

where ND . and ND are, respectively, the minimum and the maximum values of ND . mln .max achleved as L varles and the empirical constant ~ depends mostly upon the foliage orientation and the soil reflectance. The other vegetation index discussed in this paper is based upon the difference of vertically and horizontally polarized brightness temperatures (AT) observed at 37 GHz frequency of the scanning multichannel microwave radiometer (SMMR) on board the Nimbus-7 satellite. The spatial resolution of the AT data is about 25 km, and the global data has been gridded on a Mercator projection in 0.25 ° x 0.25 ° (latitude x longitude) cells. At any location about four AT values are available per month from which the second lowest AT value is kept so as to minimize the effects of cloud and soil wetness variations on AT /4,5/. The AT values range between 15-30 deg K for deserts or very sparsely vegetated soils, and between 4-15 deg K for varying amounts of vegetation /4,5/. Analysis of field observations over agricultural crops and the radiative transfer equation give the relation between AT and L as /5,6/: AT = ATmi n + (ATma x - ATmin) exp(-~L)

(3)

where ~ is an empirical constant depending primarily upon the canopy structure, and AT and AT . are, respectively, the maximum and the minimum values of AT achieved as L max increases, m1~t should be noted that apart from green leaves the AT values are also affected by surface roughness, woody structure, exposed water, snow and dead vegetation. It is clear from the semi-empirical eqns. (2) and (3) that a close relation should exist between ND and AT for uniform stand of agricultural crops. With increasing vegetation density the ND increases, while the AT decreases. Combining eqns. (2) and (3) one can obtain the direct relation between ND and AT which might be expected for agricultural crops as

AT = ATmi n + (ATma x - ATmin)

NDma x - ND [ ............... ]~ ND ND . max mln

(4)

where ? = ~/~, which is expected to be a soil- and crop-specific parameter. An extrapolation of eqn. (4) to satellite data must be tentative because many factors other than green leaf dynamics are expected to affect both ND and AT. Neither eqn. (2) nor eqn. (3) has been validated for satellite observations. Concurrent satellite-observed time series of ND and AT from April 1982 to January 1985 for three locations, contrasting in vegetation density, are shown in Fig. I. The ND values over the temperate deciduous forest in U.S.A. range between 0.15 and 0.4, while the AT values range between 5 and 9 deg K. The ND and AT values show a pronounced seasonality for this forest. Over the Sahara the ND values range between 0.03 and 0.05, while the AT values range between 25 and 28 deg K. The ND values do not show any noticeable seasonality, while the AT values show a weak seasonality (lower AT values during August - September). Both ND and AT values show a fair degree of seasonality over the shrubland of Kalahari; the ND values range between 0.I and 0.2, while the AT values range between 8 and 16 deg K. The general patterns of these time series are in qualitative agreement with eqns. (2) and (3), although vegetation characteristics other than leaf area index need to be considered in the quantitative interpretation of these data. The annually averaged values of ND and AT for 1983 and 1984 are shown in Fig. 2 for 14 globally distributed locations. For this annually averaged values a highly significant inverse correlation could be established statistically, as has been annotated within the

Global Vegetation Dynamics

(7)231

Fig. 2. These data could also be fitted by eqn. (4) /6/. One should note that the slope of AT vs. ND relation (either calculated from equation 4 or the statistically derived equation give in Figure 2) progressively increases as ND decreases. From this slope variation one may conclude that AT provides a more sensitive indicator of vegetation than ND for arid and semi-arid regions /6/,

0.4 0.3 zo 0.2 0.1

.2

0.1



.0 I0

APRIL

DEC

1982

1983

DEC

1984

DEC

TIME (MONTH)

Fig. I.

Time series of monthly data (April 1982 to January 1985) for normalized difference (ND) vegetation index from AVHRR and the 37 GHz polarization difference (AT) from SMM_R for three locations contrasting in vegetation density: temperate deciduous forest of eastern U.S.A., the Sahara desert, and shrubland of Kalahari.

~

z w

30

,

I

'

I

]

ANNUAL AVERAGE DATA 1983 ANO 1984

25

20

Y

Z

o_ J\6\

N

'

=

3.2

+

r =

0.97

5

N =

t.05/X 28

15

< a-

Fig. 2.

10

~

5

m

,[ C 0.0

rc-

LEGEND O SAHARA 1 --MALl 2 -MAURITANIA 3 - SENEGAL 4 - G T VICTORIA DESERT 5 - G T . SANDY DESERT 6 -- 8URKINA FASSO 7 -BOTSWANA 8 - N . W . TEXAS (U.S.) 9 - C . INDIA X - S . W , TEXAS (U,S.) Z~ GEORGIA (U,S.) • -CENTRAL AFRICAN REPUBLIC • - CAMEROON

,

I

0.1 NORMALIZED

l

I

0.2 DIFFERENCE

,

I

0.3 (AVHRR)

The annually averaged values of ND and AT for 1983 and 1984 from 14 globally distributed locations.

As a further illustration regions, Fig. 3 shows the for 42 locations in Africa recognized to be the major arid regions /7,8/. Thus,

of the sensitivity of AT to vegetation in arid and semi-arid relation between the annually averaged AT and annual rainfall and Australia. Here, one should note that rainfall has been determinent of vegetation density and primary productivity in the implicit variable providing the correlation between AT

(7)232

B.J.

C h o u d h u r y a n d I. Y . F u n g

and rainfall should be considered to be the vegetation density. A linear relationship between AT and rainfall could explain 86 percent of the variance. The AT value decreases by about 0.5 deg K per i0 mm increase in rainfall, according to the linear regression analysis of this data. A better appraisal of AT - rainfall relation for arid regions needs to be established by increasing the number of locations and geographic diversity. By concurrent time series and by annual averages we illustrated above two satellite observed vegetation indices, namely ND derived from AVHRR data and AT derived from SMMR data. The physical mechanisms governing the behavior of these two vegetation indices are different; the ND index exploits differential absorption and scattering of visible and near-infrared radiation by leaves, while AT index exploits the attenuation and scattering of microwave radiation emitted from the soil and within the vegetation as it passes through the canopy. Water within the foliage and green stem absorbs microwave radiation, and dead litter and woody structures scatter this radiation. The atmosphere intervening the land surface and satellite affects the ND index more than the AT index. For the ND data one 4 km GAC pixel is assumed to be representative of a much larger cell area (nominally 15 km). The spatial resolution of the AT pixels are about 25 km and it

30

z

i

t,'~ 25 i; ~

n

p





~

I

l

HOT ARID AND SEMI-ARID REGIONS n =42 "Z = - 0.93 Y = 27-0.048X

,

~ •

u_ N 20 z _o I--< N 25

I

~

o• •

o•

o~



°





o o

2 N "r" (.~

~

10

s 0

L 0

Fig. 3.

i 100

i

i

200 RAINFALL(mmy

i

i 300

i 400

1)

The seven year (1979-1985) average AT values for 42 locations in Africa and Australia are plotted against climatologic annual rainfall. The climatologic rainfall is chosen because it might be considered to be areally representative and thus commensurate with the spatial resolution of the AT data (25 km).

as been gridded in 0.25 ° x 0.25 ° (latitude x longitude) cells. Thus, there are basic differences in the spatial areas represented by the ND and the AT data. These data are also not temporally in exact coincidence; the maximum ND value or the second lowest AT value have been chosen among all observations during any month at each location. The temporal and spatial patterns of these two data may not match exactly due to the data processing itself. In the following we will discuss relations between these two vegetation indices and the atmospheric CO 2 concentration and the net primary productivity. BIOSPHERE-ATMOSPHERE

CARBON EXCHANGE

The theoretical basis for quantifying the dry matter accumulation or net primary productivity in terms spectral reflectances in visible and near-infrared regions was provided by Kumar and Monteith /9/. The dry matter accumulation occurs as a result of photosynthesis, and because canopy photosynthesis is directly proportional to solar radiation, the dry matter accumulation is, to a large extent, determined by the amount of solar radiation absorbed by the canopy. Thus, one can write /9/ T NPP = e f o

S(t) f(t) dt

(5)

where the time period (O,T) is the duration for which NPP is being determined (generally one year), S(t) is the incident solar (or photosynthetically active) radiation, f(t) is the fractional absorption of incident radiation by the canopy at time t and e is the efficiency of the canopy for converting the absorbed radiation into dry matter.

Global Vegetation Dynamics

(7)233

An extensive sensitivity analysis of the relationships between f(t) and vegetation indices derived from visible and near-infrared reflectances for varied soil and canopy characteristics representing uniform stands of agricultural crops has been performed by Choudhury /i0/. Although the relationship between f(t) and ND was shown to be generally curvilinear, the following linear equation provides a first approximation for extensive homogeneous canopies:

(6)

f(t) = 1.15 [ND(t) - 0.15]

From eqns. (5) and (6) one would expect that a temporal integration of ND would be a good indicator of NPP for uniform stands of a crop provided that the solar radiation is a moderately conservative parameter. Gall® et al. /ii/ have experimentally demonstrated that a temporal integration of ND does provide a good measure of the yield of maize. Although a direct link between NPP and ND has been established for uniform stands of agricultural crops, an extrapolation of the above results to ND derived from satellite observations to plant communities in general must be considered tentative due to spatial heterogenieties and atmospheric effects, and also because of data processing itself. Figure 4 shows the relation between published NPP values for different biomes in North and South America and the temporally integrated values of ND derived from the AVHRR data /12/. The monthly composite ND values from April 1982 to March 1983 were used to compute the annual integral. Such an annual integral of ND was computed for an average of 12 locations within each biome in North America and an average of 14 locations for each biome in South America. The biome average of the integrated ND was then computed to correlate with the biome NPP values in Fig. 4o A linear regression between the integrated ND and NPP could explain 90 percent of the variance. With the exception of short grasses, all of the comparable biomes in South America were found to give a higher integrated ND values than their North American equivalents. Goward et al. /12/ suggested that real differences exist for NPP of equivalent biomes in North and South America. The correlation shown in Fig. 4 is quite encouraging but it needs a direct validation via observations and by a rigourous theory.

2~

®

I

I

E

®

@

>

®

1400

[]

[]

a.

B []

®

[] D® I ~ []Q

[]

®

%

<

z

®

©

2200

[] [] [] [] [] []

®

200

% o

[] i ®i

3O 4 O

@ ® © ® ® ® © ® ® Q ® ®

5O

6O

i

i

i

~

7O

8O

90

100

i I~

i

[]

SOUTH AMERICAN BIOMES TROPICAL RAIN FOREST LIGHT TROPICAL FOREST TROPICAL WOODLAND AND SHRUB CONIFEROUS FOREST MIXED CONIFEROUS AND BROADLEAF FOREST WOODLAND AND SHRUB (MEDITERRANEANI WOODED SAVANNA GRASS AND SHRUB SAVANNA TALL GRASS AND WOODED STEPPE SHORT GRASS DESERT AND DESERT SHRUB TUNDRA AND ALPINE NORTH AMERICAN BIDMES TUNDRA TUNDRA - CONIFEROUS ECOTONE BOREAL CONIFEROUS TEMPERATE MOIST CONIFEROUS CONIFEROUS DECIDUOUS ECOTONE DECIDUOUS OAK PINE SUBCLIMAX PINE CLIMAX GRASSLAND CULTIVATED WOODLAND AND SCRUB DESERT

120

INTEGRATED YEAR (APRIL1982-MARCH 1983)

NORMALIZEDDIFFERENCE*DAYS

Fig. 4.

Relation between net primary productivity of different biomes in North and South America and temporally integrated normalized difference from AVHRR data (redrawn from Goward et al./12/).

ChoudhurYo/13 / has studied the relationship between AT derived from the SMMR data and NPP for 5 latitude bands from 75°N to 55°S (Fig. 5). The AT data was averaged for seven years (1979-85) and formed into zonal (5 u latitude) means by weighting each 0.25 ° latitude band with the land surface area. The NPP values were obtained from Pearman and Hyson /14/. In the northern hemisphere the maximum and the minimum NPP values are found, respectively, within the latitude bands 50 ° - 55°N and 20 ° - 25°N, while the . . . . O minimum and the maximum AT values are found respectlvely wzthzn the latztude bands 50 • 'O ° ' ° • - 55°N and 15 ° - 20°N. The dzscrepancy of 5 latztude band for the mlnzmum NPP and the maximum AT might be due to the prevailing drought condition over the Sahel zone of Africa /15/. In the southern hemisphere a major discrepancy between AT and NPP is seen for the latitude band 50 ° - 55°S, which could be due to a rather large number of

(7)234

B . J . Choudhury and I. Y. Fung

80O

I

I

I

I

I

I

I

I

I

I

I

I

I

I

I

7OO

,z '?, E 6OO

24

a

>__ 5O0 I-.o

400

h-

-

20

Z

O__

\ li.t t f "--d\ NPP

/

16

'v,, ,+ /

12

N

5

O 0. ~9

8

b"°"od

rr 4

Z 100

09 i

t

60 °

i

=

40 °

=

i

20 °

t

i

i

0o

~

20 °

i

i

40 °

i

=

60 °

I

0

S

LATITUDE

Fig. 5.

The meridional patterns of AT derived from the SMMR data and NPP from Pearman and Hyson /14/. mixed (land-water) pixel values of AT. The latitudinal pattern of AT is seen to be in good agreement with that of NPP. While one could rationalize this agreement on the basis of the correlations shown in Figs. 2 and 4, a rigourous theoretical basis for these correlations and direct validation remain to be established. Associated with the NPP is the temporal and spatial variations of the atmospheric CO 2 concentration. There is a net depletion of atmospheric CO^ during the period of z vegetation growth due to photosynthesis. The biosphere then increases the atmospheric COo due to respiration and decomposition. In the northern hemisphere, the atmospheric CO~ concentration is maximum in the late spring, while minimum concentration occurs in o o z the autumn /16/. At Mauna Loa, Hawaii (156 W, 20 N) the CO2 cycle has a peak-to-peak amplitude of 5-6 ppm, with the maximum in May and minimum zn September. The globally averaged CO 9 cycle has a peak-to-peak amplitude of 4-5 ppm, and this amplitude is nearly in phase wi~h the cycle at Mauna Loa. Tucker et al. /17/ and Fung et al. /16/ have analyzed the relation between the CO 2 cycle and the monthly ND index derived from the AVHRR data for the period April 1982 to June 1984. The ND values were averaged in 5 ° latitude bands and weighted by the corresponding land surface area. To extract the seasonal variation of CO9, a 12-month running mean concentration was computed, which was then subtracted from tile observed actual concentration. When these relative C09 concentrations for any month at Mauna Loa were plotted against the corresponding monthl9 mean ND for I0 ° - 30°N latitude band an elliptical relation was found (Fig. 6a), which became an essentially linear relation when the CO 9 concentrations were lagged by one month (Fig. 6b). Similarly, with one month time lag, the globally averaged relative CO 2 concentration showed a linear relation with globally averaged ND (Fig. 7). A tzme lag between the seasonal ND and the CO 9 concentration would be expected due to boundary-layer mixing and atmospheric transport of CO 2 to the remote monitoring stations (like Mauna Loa). The time seri~s of SMMR observed AT from January 1979 to December 1985 for the northern • u o o hemzsphere (0 - 80 N) and the relative CO 9 concentratzon at Mauna Loa are shown in Fig, 8. (Note that the two time series in the figure have been offset by 2 months.) The maximum and the minimum AT for the northern hemisphere are observed, respectively, in February and July-August period. The relative COp concentration is maximum in May and minimum in September. Thus, the minimum AT and tile minimum CO 9 concentration lag by about 2 months, but the maximum AT and the maximum CO 2 concentration lag by about 3 months. Due to rising air temperature in the early sprzng, the biospheric release of

Global Vegetation Dynamics

(7)235

4r~

583

3

,584

51,82 ~

• 6"B3 684 %,r82

° 4r~

100 30ON

=,82•"

10° 30°N

~ '82•• 4'83

483 • ,

_.. NO l I M E LAG

584 :

•3~3

SS35Sz • 3,a4

I

,2r~

• 783

7'84

, 2/83 =7182

1,84 682

0

0

i

1283

8'83 " 8'82

1

2

-1

• 12182

12,83

ONE MONTH LAG

6,.83

• • 3,'84 ,3/83

M A U N A LOA



~/83 ,12'82 11~3

7~2

2

• 11;82

7,83 ,1182

3i

11t83

•i0~2 ~o'e3

1883•

b 10~83 0'82 4

0.10

I

I

0.12

0.14

I

0.16

0.18

4 0 10

0.20

i 0.12

I 0,14

i 016

983

Sl82 •982 018

020

AVERAGE NORMALIZED DIFFERENCE

Fig. 6.

Relation between monthly relative~CO~ concentration and Mauna Loa and average ND from AVHRR data for i0 U -z30°N for the period April 1982 to June 1984. (a) no time lag, (b) one month time lag (redrawn from Tucker et al. /17/).

COp into the atmosphere (via decomposition) appears to be somewhat larger than the CO 2 extracted from the atmosphere via photosynthesis by the growing vegetation. The phasings of AT and the relative COp concentration are not simply related by a constant time offset /13/. Nevertheless, t~e dynamics of the atmospheric CO 9 concentration is seen to be highy correlated with the vegetation dynamics as quantified through ND and AT, while recognizing that both of these data are affected by factors other than vegetation (for example, snow).

3/84 •

• 2/84

2 / 8 ~ • 3 3/83

1/83

~

4/84

• 1/84 4/83 • • 12f83 • Z

• 4/82

• 12/82

11/B3

_o I<

• 11/82

~01 z

~0/83

• 5/83 5/84•

c) Z O f,D

•5/82 • 10/82 • 9/83

• 9/82

(J • 6J82

I,<

6/83 •

• 8183 • 882

7/83 •

-3 0,12

J

I 0.16 WEIGHTED

Fig. 7.

I AVERAGE

I 0,20 GLOBAL

• 7~82

I

I 0.24

NDVI

Relation between monthly globally averaged relative CO 2 concentration and the globally average ND with one month time lag (redrawn from Tucker et al.

i17/). SUMMARY AND CONCLUSIONS Spatial and temporal variations of vegetation as quantified through the normalized difference of visible and near-infrared reflectances (ND) and the polarization difference of 37 GHz brightness temperatures (AT) were discussed in this paper. The ND values are based upon NOAA-7/AVHRR data from April 1982 to January 1985, while the AT values are based upon Nimbus-7/SMMR data for January 1979 to December 1985. Analysis of the radiative transfer equation for uniform stand of agricultural crop suggested an inverse correlation to exist between AT and ND, which was confirmed by the satellite data. The theoretical analysis needs to be extended to natural vegetation under normal and drought conditions and complemented by field data to better understand the spatial and temporal patterns of ND and AT. The temporal variations of ND and AT were shown to be highly correlated with the temporal variations of the atmospheric CO 2 concentration. This correlation demonstrated that a direct link exists between the atmospheric and biospheric carbon dynamics. Additionally, on the annually integrated basis, both ND JASR 9:7-P

(7)236

B.J. Choudhury and I. Y. Fung

18,

14

I

o ,ULY

2

~o 8 2

L

1979

L

1980

I 1981

I 1982

L 1983

I 1984

I 1985

YEAR

Fig. 8.

Time series of relative CO^ concentration at Mauna Loa from January 1979 to Z . O December 1985 and the SMMR observed AT for the northern hemlsphere (0 80°N). Note that the two time series have been plotted with 2 months offset.

and AT were found to be correlated with primary productivity, which needs to be understood on a rigourous theoretical basis. To what extent the interannual variation of ND or AT could be used as a quantitative indicator of the interannual variation of primary productivity of global ecosystems needs to be validated. ACKNOWLEDGEMENTS The financial support for this work was provided by the Land Processes Branch (Dr. R. E. Murphy, Chief) of NASA Headquarters. The data processing assistance was provided by Mr. R. E. Golus. The atmospheric CO 2 data were provided by NOAA/Geophysical Monitoring for Climatic Change Program. REFERENCES i.

B. N. Holben, Characteristics of maximum-value composite data. Int. J. Remote Sensing, 7: 1417-1434, 1986.

images from temporal AVHRR

2.

J. R. G. Townshend and C. O. Justice, Analysis using normalized difference vegetation index. 1445, 1986.

3.

G. Asrar, M. Fuchs, E. T. Kanemasu and J. L. Hatfield, Estimating absorbed photosynthetic radiation and leaf area index from spectral reflectance in wheat. Agron. J., 76: 300-306, 1984.

4.

B. J. Choudhury and C. J. Tucker, Monitoring global vegetation using Nimbus-7 37 GHz data: Some empirical relations. Int. J. Remote Sensing, 8: 1085-1090, 1987.

5.

B. J. Choudhury, Monitoring global land surface using Nimbus-7 and examples. Int. J. Remote Sensing, 1988.

6.

F. Becker and B. J. Choudhury, Relative sensitivity of normalized difference vegetation index (NDVI) and microwave polarization difference index (MPDI) for vegetation and desertification monitoring. Remote Sensing Environment, 24: 297311, 1988.

7.

H. N. LeHouerou, Rain-use efficiency Arid Environ 7: 3-45, 1984.

8.

H. Lieth, Modeling the primary productivity of the world, pp. 237-257. Primary Productivity of the Biosphere. In. H. Lieth and R. H. Whittaker (Editors), Springer-Verlag, N-Y, 1975.

9.

M. Kumar and J. L. Monteith, Remote sensing of crop growth, pp. 134-145. Plants and the Daylight Spectrum. In. H. Smith (Editor). Academic Press, NY, 1982.

of the dynamics of African vegetation Int. J. Remote Sensing, 7: 1435-

- A unifying concept

37 GHz data:

Theory

in arid land ecology.

J~

I0. B. J. Choudhury, Relationships between vegetation indices, radiation absorption, and net photosynthesis evaluated by a sensitivity analysis. Remote Sensing Environ., 22: 209-233, 1987. ii. K. P. Gallo, C. S. T. Daughtry and M. E. Bauer, Spectral estimation of absorbed photosynthetically active radiation in corn canopies. Remote Sensing Environ., 17: 221-232, 1985.

Global Vegetation Dynamics

(7)237

12. S. N. Goward, D. Dye, A. Kerber and V. Kalb, Comparison of North and South American biomes from AVHRR observations. Geocarto Int., i: 27-39, 1987. 13. B. J. Choudhury, Relating Nimbus-7 37 GHz data to global land-surface evaporation, primary productivity and the atmospheric CO 2 concentration. Int. J. Remote Sensing, 9: 169-176, 1988. 14. G. I. Pearman and P. Hyson, Global transport and inter-reservoir exchange of carbon dioxide with particular reference to stable isotopic distribution. J. Atmos. Chem., 4: 81-124, 1986. 15. S. E. Nicholson, 1393, 1985.

Sub-Saharan

rainfall

1981-84.

J. Clim. Appl. Meteorol.,

24:

1388-

16. I. Y. Fung, C. J. Tucker and K. C. Prentice, Application of advanced very high resolution radiometer vegetation index to study atmosphere-biosphere exchange of CO 2. J. Geophys. Res., 92: 2999-3015, 1987. 17. C. J. Tucker, I. Y. Fung, C. D. Keeling and R. H. Gammon, Relationship between atmospheric CO 9 variation and a satellite-derived vegetation index. Nature (Lond.), 319: 195-199,-1986.