UC Irvine
UC Irvine Previously Published Works
Title
Primary production of the biosphere: integrating terrestrial and oceanic components
Permalink
https://escholarship.org/uc/item/9gm7074q
Journal
Science (New York, N.Y.), 281(5374)
ISSN
0036-8075
Authors
Field, CB
Behrenfeld, MJ
Randerson, JT
et al.
Publication Date
1998-07-01
DOI
10.1126/science.281.5374.237
Copyright Information
This work is made available under the terms of a Creative Commons Attribution License,
availalbe at https://creativecommons.org/licenses/by/4.0/
Peer reviewed
eScholarship.org Powered by the California Digital Library
University of California
Primary Production of the
Biosphere: Integrating
Terrestrial and Oceanic
Components
Christopher B. Field,
*
Michael J. Behrenfeld, James T. Randerson,
Paul Falkowski
Integrating conceptually similar models of the growth of marine and terrestrial
primary producers yielded an estimated global net primary production (NPP)
of 104.9 petagrams of carbon per year, with roughly equal contributions from
land and oceans. Approaches based on satellite indices of absorbed solar ra-
diation indicate marked heterogeneity in NPP for both land and oceans, re-
flecting the influence of physical and ecological processes. The spatial and
temporal distributions of ocean NPP are consistent with primary limitation by
light, nutrients, and temperature. On land, water limitation imposes additional
constraints. On land and ocean, progressive changes in NPP can result in altered
carbon storage, although contrasts in mechanisms of carbon storage and rates
of organic matter turnover result in a range of relations between carbon storage
and changes in NPP.
Biological processes on land and in the
oceans strongly affect the global carbon cycle
on all time scales (1–4). In both components
of the biosphere, oxygenic photosynthesis is
responsible for virtually all of the biochemi-
cal production of organic matter. Mecha-
nisms of and constraints on photosynthesis on
land and in the oceans are similar in many
respects, but past syntheses of primary pro-
duction from photosynthesis have focused on
the terrestrial or ocean components individu-
ally. Consequently, models of the global car-
bon cycle are compartmentalized, with limit-
ed opportunities for comprehensive or com-
parative analyses. Here, we present integrated
estimates of primary production based on
satellite measurements for both oceanic and
terrestrial ecosystems. This integrated ap-
proach builds from parallel data sets and
model formulations toward a truly biospheric
view.
The biologically mediated parts of the
carbon cycle in terrestrial and ocean biomes
involve both production and turnover of or-
ganic matter. At the biochemical level, pho-
tosynthesis and the biosynthesis of organic
compounds, the processes that result in net
primary production (NPP), are very similar.
NPP, originally defined as the amount of
photosynthetically fixed carbon available to
the first heterotrophic level in an ecosystem
(5), is also the difference between autotrophic
photosynthesis and respiration (6). NPP is a
major determinant of carbon sinks on land
and in the ocean (7, 8) and a key regulator of
ecological processes, including interactions
among trophic levels (9, 10). Because ocean
NPP is dominated by phytoplankton, nearly
all of the plant biomass is photosynthetic.
Therefore, relatively short-term measure-
ments (24 hours) can account for both pho-
tosynthesis and respiration. In contrast, the
major components of terrestrial plant biomass
are roots and stems, which respire but do not
generally photosynthesize. In terrestrial eco-
systems, it is relatively straightforward, in
principle, to determine NPP from incremental
increases in biomass plus litter fall over
weeks, months, or years. Below-ground pro-
cesses, however, add numerous challenges to
these conceptually simple measurements.
NPP on land and in the oceans has been
modeled with a variety of approaches with a
range of fundamental mechanisms, specific
details, and levels of integration (11, 12). A
common contemporary approach, developed
independently for land and ocean models,
calculates NPP as a function of the driving
energy for photosynthesis, the absorbed pho-
tosynthetically active (400 to 700 nm) solar
radiation (APAR), and an average light utili-
zation efficiency ()(13, 14)
NPP ⫽ APAR ⫻ (1)
Models based on this approach are diverse in
terms of mechanistic detail, but they are all
strongly connected to global-scale observa-
tions. For the oceans, APAR can be related to
satellite-derived measurements of surface chlo-
rophyll (C
sat
)(14), and for terrestrial systems, it
can be determined from satellite-based esti-
mates of vegetation greenness, often the nor-
malized difference vegetation index (NDVI)
(15). APAR depends on the amount and distri-
bution of photosynthetic biomass (the primary
source of variability in C
sat
and NDVI), as well
as the amount of downwelling solar radiation
and the fraction that is in the visible (photosyn-
thetically active) wavelengths. is an effective
photon yield for growth that converts the bio-
mass-dependent variable (APAR) into a flux of
organic compounds (NPP). For both terrestrial
and oceanic models, cannot be directly mea-
sured from space and must be parameterized
with field measurements.
For marine systems, can be parameter-
ized from thousands of
14
C-based field mea-
surements of NPP (16–18). Terrestrial values
are less abundant, largely because depends
on time-consuming determinations of NPP
and APAR (19, 20). Uncertainty in is a
primary source of error in land and ocean
NPP estimates. With few exceptions, ocean
NPP models estimate solely as a function of
sea-surface temperature (11, 16, 21–23). In
terrestrial ecosystems, varies with ecosys-
tem type and with stresses from unfavorable
levels of temperature, nutrients, and water
(20, 24, 25).
In this study, we combined results from
conceptually similar land and ocean NPP
models, the Carnegie-Ames-Stanford ap-
proach (CASA) (26) for land and the Verti-
cally Generalized Production Model (VGPM)
(16) for the oceans (27). Both of these models
are simple formulations designed with an em-
phasis on integrating spatially extensive sat-
ellite observations rather than describing the
mechanistic details of NPP. In essence, both
models use versions of Eq. 1, expanded to
provide an effective interface with observed
variables. The fundamental relation in the
CASA model is
NPP ⫽ f (NDVI) ⫻ PAR ⫻ ε* ⫻ g共T兲 ⫻ h共W兲
(2)
where APAR (in megajoules per square
meter per month) is a function of NDVI and
downwelling photosynthetically active solar
radiation (PAR) and (in grams of C per
megajoule) is a function of the maximum
achievable light utilization efficiency * ad-
justed by functions that account for effects of
temperature g(T ) and water h(W ) stress (26).
For the VGPM, the fundamental equation is
NPP ⫽ C
sat
⫻ Z
eu
⫻ f共PAR兲 ⫻ P
opt
b
(T) (3)
where C
sat
is the satellite-derived, near-sur-
face phytoplankton chlorophyll concentration
(in milligrams per cubic meter), Z
eu
is the
depth (in meters) to which light is sufficient
C. B. Field, Department of Plant Biology, Carnegie
Institution of Washington, Stanford, CA 94305, USA.
M. J. Behrenfeld and P. Falkowski, Institute of Marine
and Coastal Sciences, Rutgers University, New Bruns-
wick, NJ 08901– 8521, USA. J. T. Randerson, De-
partment of Plant Biology, Carnegie Institution of
Washington, Stanford, CA 94305, USA, and Depart-
ment of Biological Sciences, Stanford University,
Stanford, CA 94305, USA.
*To whom correspondence should be addressed.
R EPORTS
www.sciencemag.org SCIENCE VOL 281 10 JULY 1998
2
to support positive NPP, f (PAR) describes
the fraction of the water column from the
surface to Z
eu
in which photosynthesis is light
saturated, and P
opt
b
(T ) is the maximum, chlo-
rophyll-specific carbon fixation rate (in mil-
ligrams of C per milligram of chlorophyll per
day), estimated as a function of sea-surface
temperature (11, 16). For the VGPM, varia-
tion in the fraction of absorbed PAR is a
function of depth-integrated phytoplankton
biomass (that is, C
sat
⫻ Z
eu
). The product of
P
opt
b
and f (PAR) yields an average water
column light utilization efficiency, making it
the corollary of in Eq. 1. The VGPM op-
erates with a daily time step, whereas CASA
has a monthly time step.
Biospheric NPP was calculated from Eqs.
2 and 3, on the basis of observations averaged
over several years. Because the satellite data
necessary for estimating APAR cover differ-
ent time periods for the oceans and land, the
averaging periods are different: 1978 to 1983
for the oceans and 1982 to 1990 for land. The
input data include C
sat
from the Coastal Zone
Color Scanner (CZCS) (28), NDVI from the
Advanced Very High-Resolution Radiometer
(AVHRR) (29–31), cloud-corrected surface
solar radiation (32), sea-surface temperature
(33), terrestrial surface temperature (34), pre-
cipitation (35), soils (36), and vegetation
(37), plus field-based parameterizations of
(16, 21, 26). Our results based on time-aver-
aged data are likely to characterize typical
NPP from this time period but certainly miss
key anomalies such as El Nin˜o–Southern Os-
cillation, as well as progressive global chang-
es. The contribution of models like the one
used here to quantifying these changes will
depend on continuous, high-quality data, over
extended periods.
Using the integrated CASA-VGPM bio-
sphere model, we obtained an annual global
NPP of 104.9 Pg of C (Table 1), with similar
contributions from the terrestrial [56.4 Pg of
C (53.8%)] and oceanic [48.5 Pg of C
(46.2%)] components (38). This estimate for
ocean productivity is nearly two times greater
than estimates made before satellite data (39,
40). Average NPP on land without permanent
ice cover is 426 g of C m
⫺2
year
⫺1
, whereas
that for oceans is 140 g of C m
⫺2
year
⫺1
. The
lower NPP per unit area of the ocean largely
results from competition for light between
phytoplankton and their strongly absorbing
medium. For the average ocean C
sat
of 0.19
mg m
⫺3
(16, 41), only 7% of the PAR inci-
dent on the ocean surface is absorbed by the
phytoplankton (14), with the remainder ab-
sorbed by water and dissolved organics. In
contrast, leaves of terrestrial plants absorb
about 31% of the PAR incident on land with-
out permanent ice cover. Although primary
producers in the ocean are responsible for
nearly half of the biospheric NPP, they rep-
resent only 0.2% of global primary producer
biomass (3, 16, 21). This uncoupling between
NPP and biomass is a consequence of the
more than three orders of magnitude faster
turnover time of plant organic matter in the
oceans (average 2 to 6 days) (1) than on land
(average 19 years) (42).
On land and in the oceans, spatial hetero-
geneity in NPP is comparable, with both
systems exhibiting large regions of low pro-
duction and smaller areas of high production.
In general, the extreme deserts are even less
productive than the vast mid-ocean gyres
(Fig. 1). Maximal NPP is similar in both
systems (1000 to 1500 g of C m
⫺2
year
⫺1
),
but regions of high NPP are spatially more
restricted in the oceans (essentially limited to
estuarine and upwelling regions) than in ter-
restrial systems (for example, humid tropics)
(Fig. 1). On land, 25.0% of the surface area
without permanent ice (3.3 ⫻ 10
7
km
2
) sup-
ports an NPP greater than 500 g of C m
⫺2
year
⫺1
, whereas in the oceans, that figure is
only 1.7% (5.0 ⫻ 10
6
km
2
). Highly produc-
tive (that is, eutrophic) regions in the oceans
contribute less than 18% to total ocean NPP
(Table 1).
Globally, NPP reaches maxima in three
distinct latitudinal bands (Fig. 2). The largest
peak (⬃1.6 Pg of C per degree of latitude)
near the equator and the secondary peak at
midtemperate latitudes of the Northern Hemi-
sphere are driven primarily by regional max-
ima in terrestrial NPP. The smaller peak at
midtemperate latitudes in the Southern Hemi-
sphere (Fig. 2) results from a belt of enhanced
oceanic productivity corresponding to en-
hanced nutrient availability in the Southern
Subtropical Convergence (43). At mid and
low latitudes, ocean NPP is remarkably uni-
form, consistent with the predominant influ-
ence of large-scale ocean circulation patterns.
Seasonal fluctuations in ocean NPP are
modest globally, even though regional season-
ality can be very important (44). Ocean NPP
ranges from 10.9 Pg of C in the Northern
Hemisphere spring (April to June) to 13.0 Pg of
C in the Northern Hemisphere summer (July to
September) (Table 1). The July to September
maximum in ocean NPP is largely a result of
SP
-60
-30
EQ
30
60
NP
180 120 W 60 W 0 60 E 120 E 180
0 100 200 300 400 500 600 700 800
Fig. 1. Global annual
NPP (in grams of C per
square meter per year)
for the biosphere, cal-
culated from the inte-
grated CASA-VGPM
model. The spatial res-
olution of the calcula-
tions is 1° ⫻ 1° for
land and 1/6° ⫻ 1/6°
for the oceans. Input
data for ocean color
from the CZCS sensor
are averages from
1978 to 1983. The
land vegetation index
from the AVHRR sen-
sors is the average
from 1982 to 1990.
Global NPP is 104.9
Pg of C year
⫺1
(104.9 ⫻ 10
15
gofC
year
⫺1
), with 46.2%
contributed by the
oceans and 53.8%
contributed by the
land. Seasonal ver-
sions of this map are
available at www.
sciencemag.org/feature/data/982246.shl. NP, North Pole; EQ, equator; Sp, South Pole.
R EPORTS
10 JULY 1998 VOL 281 SCIENCE www.sciencemag.org238
open-ocean blooms north of 30°N (Fig. 1).
Despite the greater ocean area in the Southern
Hemisphere, a similar bloom-induced increase
in NPP does not occur during the Austral sum-
mer (Fig. 2), perhaps reflecting the more com-
mon occurrence of iron limitation in the high-
latitude southern oceans (45, 46). Seasonal fluc-
tuations in terrestrial NPP are much greater,
with global production during the Northern
Hemisphere summer (July to September) about
60% greater than that in January through March
(Table 1).
Spatial variation in NPP in both the ter-
restrial and ocean components of our bio-
sphere model is driven mostly through vari-
ation in light capture by photosynthetic bio-
mass or APAR and secondarily through vari-
ation in (12, 16). Spatial and seasonal
variation in photosynthetic biomass is, in
turn, largely controlled by the availability of
other resources. Nitrogen, iron, and light are
critical in the oceans. On land, water stress,
temperature, and other nutrients such as
phosphorus also play a role (47). Conse-
quently, regional and seasonal distributions
of NPP reflect the interface between physical
(for example, precipitation, PAR, ocean cir-
culation, and water-column stratification) and
biological processes (for example, species
composition, microbial activity, and interac-
tions among organisms).
Over most of the ocean, as on land, nutri-
ents required to support NPP are primarily
supplied through local decomposition, rather
than from sources of new nutrients. Biologi-
cally mediated carbon sinks are, however,
largely dependent on inputs of new nutrients,
supplied from processes such as upwelling,
biological N fixation, deposition from the
atmosphere, and cultural eutrophication (48,
49). In both systems, progressive changes in
NPP over periods of decades to centuries can
have a range of impacts on the global carbon
cycle, depending on the turnover rates of the
pools that receive the NPP (7 ). In terrestrial
systems, where plant and soil pools typically
have turnover times in the range from years
to decades (50), even modest increases in
NPP potentially result in substantial carbon
storage in plants and soils (7). Because of the
rapid turnover of oceanic plant biomass, even
large increases in ocean NPP will not result in
substantial carbon storage through changes in
phytoplankton standing stock. They do, how-
ever, impact ocean carbon storage through
effects on fluxes of inorganic and refractory
organic carbon to the ocean interior (3).
The future development and application
of whole biosphere models, such as that de-
scribed here, can play a major role in the
emergence of integrated, comprehensive per-
spectives on the function of the Earth system.
NPP is critical for these efforts, because it is
central to carbon and nutrient dynamics and it
links biogeochemical and ecological process-
es. The global carbon cycle and the ecologi-
cal processes that contribute to it are not in
steady state but are highly dynamic (51–53).
Current capabilities to interpret these dynam-
ics and their implications for the future of the
biosphere are constrained by gaps in the data
record, limitations on data quality, and in-
complete understanding of some of the mech-
anisms. The successful launch of the Sea-
viewing Wide-Field-of-view Sensor (Sea
WiFS) in September 1997 plus other forth-
coming remote-sensing missions will provide
marked improvements in the quality of APAR
measurements for both land and ocean. These
programs need to be paralleled by efforts to
improve the characterization of spatial and tem-
poral variation in and the fate of carbon after
it is fixed in photosynthesis.
References and Notes
1. P. G. Falkowski and J. A. Raven, Aquatic Photosynthe-
sis (Blackwell, Oxford, UK, 1997).
2. R. F. Keeling and S. Shertz, Nature 358, 723 (1992).
3. P. G. Falkowski, R. T. Barber, V. Smetacek, Science
281, 200 (1998).
4. C. B. Osmond, in Photosynthesis, W. R. Briggs, Ed.
(Liss, New York, 1989), pp. 5–17.
5. R. L. Lindeman, Ecology 23, 399 (1942).
Fig. 2. Latitudinal distribution of
the global NPP in Fig. 1. (A) The
global total (land plus ocean)
NPP (solid line), land total NPP
(dotted line), and ocean total
NPP (dashed line). (B) Land NPP:
April to June (solid line), July
to September (dotted line), Oc-
tober to December (short dashed
line), and January to March (long
dashed line). (C) Ocean NPP:
The four seasonal periods are as
in (B). The seasonal information
is available as maps at www.
sciencemag.org/feature/data/
982246.shl
Table 1. Annual and seasonal NPP of the major units of the biosphere, from CASA-VGPM. Ocean color
data are averages from 1978 to 1983. The land vegetation index is from 1982 to 1990. All values are in
petagrams of carbon (1 Pg ⫽ 10
15
g). Ocean NPP estimates are binned into three biogeographic
categories on the basis of annual average C
sat
for each satellite pixel, such that oligotrophic ⫽ C
sat
⬍ 0.1
mg m
⫺3
, mesotrophic ⫽ 0.1 ⬍ C
sat
⬍ 1mgm
⫺3
, and eutrophic ⫽ C
sat
⬎ 1mgm
⫺3
(21). The macrophyte
contribution to ocean production from (38) is not included in the seasonal totals. The vegetation classes
are those defined by (37).
Ocean NPP Land NPP
Seasonal
April to June 10.9 15.7
July to September 13.0 18.0
October to December 12.3 11.5
January to March 11.3 11.2
Biogeographic
Oligotrophic 11.0 Tropical rainforests 17.8
Mesotrophic 27.4 Broadleaf deciduous forests 1.5
Eutrophic 9.1 Broadleaf and needleleaf forests 3.1
Macrophytes 1.0 Needleleaf evergreen forests 3.1
Needleleaf deciduous forest 1.4
Savannas 16.8
Perennial grasslands 2.4
Broadleaf shrubs with bare soil 1.0
Tundra 0.8
Desert 0.5
Cultivation 8.0
Total 48.5 56.4
R EPORTS
www.sciencemag.org SCIENCE VOL 281 10 JULY 1998 239
6. W. H. Schlesinger, Biogeochemistry: An Analysis of Glob-
al Change (Academic Press, San Diego, CA, 1991).
7. M. V. Thompson, J. T. Randerson, C. M. Malmstro¨m,
C. B. Field, Global Biogeochem. Cycles 10, 711
(1996).
8. N. Gruber, J. L. Sarmiento, T. F. Stocker, ibid., p. 797.
9. C. T. Tynan, Nature 392, 708 (1998).
10. S. J. McNaughton, ibid. 345, 613 (1990).
11. M. Behrenfeld and P. Falkowski, Limnol. Oceanogr.
42, 1479 (1997).
12. C. B. Field, J. T. Randerson, C. M. Malmstro¨m, Remote
Sens. Environ. 51, 74 (1995).
13. J. L. Monteith, J. Appl. Ecol. 9, 747 (1972).
14. A. Morel, Prog. Oceanogr. 26, 263 (1991).
15. P. J. Sellers, Remote Sens. Environ. 21, 143 (1987).
16. M. J. Behrenfeld and P. G. Falkowski, Limnol. Ocean-
ogr. 42, 1 (1997).
17. W. M. Balch and C. F. Byrne, J. Geophys. Res. 99, 7555
(1994).
18. J.-F. Berthon and A. Morel, Limnol. Oceanogr. 37, 781
(1992).
19. G. Russell, P. G. Jarvis, J. L. Monteith, in Plant Cano-
pies: Their Growth, Form and Function, G. Russell, B.
Marshall, P. G. Jarvis, Eds. (Cambridge Univ. Press,
Cambridge, 1989), pp. 21–39.
20. A. Ruimy, G. Dedieu, B. Saugier, J. Geophys. Res. 99,
5263 (1994).
21. D. Antoine, J. M. Andre, A. Morel, Global Biogeochem.
Cycles 10, 57 (1996).
22. W. Balch et al., J. Geophys. Res. 97, 2279 (1992).
23. R. R. Bidigare, B. B. Prezelin, R. C. Smith, in Primary
Productivity and Biogeochemical Cycles in the Sea,
P. G. Falkowski, Ed. (Plenum, New York, 1992), pp.
175–212.
24. G. Joel, J. A. Gamon, C. B. Field, Remote Sens. Environ.
62, 176 (1997).
25. J. Runyon, R. H. Waring, S. N. Goward, J. M. Welles,
Ecol. Appl. 4, 226 (1994).
26. C. S. Potter et al., Global Biogeochem. Cycles 7, 811
(1993).
27. Confidence in NPP estimates from these and similar
models can be derived from several sources, including
ability to reproduce small-scale NPP measurements,
model intercomparisons (11), fidelity in tracking in-
terannual variations in agricultural NPP over large
regions [C. M. Malmstro¨m et al., Global Biogeochem.
Cycles 11, 367 (1997)], and ability to reproduce the
annual cycle of atmospheric CO
2
[W. Knorr and M.
Heimann, Tellus Ser. B 47, 471 (1995; J. T. Randerson,
M. V. Thompson, I. Y. Fung, T. Conway, C. B. Field,
Global Biogeochem. Cycles 11, 535 (1997)], when
coupled with appropriate atmospheric and heterotro-
phic models. Even with these approaches, the actual
values still include uncertainties related to both the
data and the structure of the models.
28. G. Feldman et al., Eos 70, 634 (1989).
29. J. D. Tarpley, S. R. Schneider, R. L. Money, J. Clim.
Appl. Meteorol. 23, 491 (1984).
30. The processed NDVI data set used for this report is
from S. O. Los [thesis, Vrije University, Amsterdam
(1998)].
31. Until the launch of SeaWiFS in September 1997,
there was no single instrument appropriate for esti-
mating APAR in the oceans and on land at the global
scale. For the oceans, there has been no color sensor
at all since the end of the CZCS mission in 1986. For
land, the AVHRR instruments aboard the NOAA sat-
ellites have been invaluable, even though their design
was not optimized for vegetation monitoring.
32. J. K. B. Bishop and W. B. Rossow, J. Geophys. Res. 96,
16839 (1991).
33. Sea-surface temperatures are climatological averages
from the U.S. Navy Marine Climatic Atlas (CD-ROM)
(Naval Command Oceanographic Detachment Center,
Asheville, NC, 1995), with processing described in (16).
34. Terrestrial temperatures were calculated for 1982 to
1990 from long-term averages (54 ) plus anomalies
from the updated data set of J. Hansen and S. Leb-
edeff [ J. Geophys. Res. 92, 13345 (1987)]. These
temperatures were then averaged by month, for the
9-year period.
35. Terrestrial precipitation was calculated for 1982 to
1990 from long-term averages (54) plus anomalies
from C. B. Baker, J. K. Eischeid, T. R. Karl, and H. F. Diaz
(paper presented at Ninth Conference on Applied
Climatology, American Meteorological Society, Dal-
las, TX, 1995). These values were then averaged by
month, over the 9-year period.
36. L. Zobler, A world soil map for global climate mod-
eling (NASA Technical Memorandum 87802, NASA,
New York, 1986).
37. R. S. DeFries and J. R. G. Townshend, Int. J. Remote
Sens. 15, 3567 (1994).
38. This estimate includesa1PgofCcontribution from
macroalgae [S. V. Smith, Science 211, 838 (1981)].
Differences in ocean NPP estimates in (16) and those
in Fig. 1 and Table 1 result from (i) addition of Arctic
and Antarctic monthly ice masks, (ii) correction of a
rounding error in previous calculations of pixel area,
and (iii) changes in the designation of the seasons to
correspond with (3).
39. O. J. Koblentz-Mishke, V. V. Volkovinsky, J. G. Ka-
banova, in Scientific Exploration of the South Pacific,
W. S. Wooster, Ed. (National Academy of Science,
Washington, DC, 1970), pp. 183–193.
40. G. M. Woodwell et al., Science 199, 141 (1978).
41. This value corresponds to an integrated phytoplank-
ton from the surface to
Z
eu
of about 15 mg m
⫺2
.
42. M. V. Thompson and J. T. Randerson, Global Change
Biol., in press.
43. G. L. Pickard and W. J. Emery, Descriptive Physical
Oceanography: An Introduction (Pergamon, Oxford,
UK, 1982).
44. Seasonal maps of global NPP are available at www.
sciencemag.org/feature/data/982246.shl
45. J. H. Martin, S. E. Fitzwater, R. M. Gordon, Global
Biogeochem. Cycles 4, 5 (1990).
46. J. H. Martin, R. M. Gordon, S. E. Fitzwater, Limnol.
Oceanogr. 36, 1793 (1991).
47. Over millennial time scales, water and temperature
are the dominant controlling factors for terrestrial
NPP, with effects of nutrient availability assuming a
major role on interannual to decadal time scales (51)
[D. S. Schimel et al., Global Biogeochem. Cycles 10,
677 (1996)].
48. P. M. Vitousek, H. A. Mooney, J. Lubchenco, J. M.
Melillo, Science 277, 494 (1997).
49. On land, but not in the oceans, elevated atmospheric
CO
2
could facilitate increased carbon storage per unit
of stored nutrients, through increased allocation to
recalcitrant tissues with high ratios of carbon to
nutrients [G. R. Shaver et al., Bioscience 42, 433
(1992)].
50. D. S. Schimel et al., Global Biogeochem. Cycles 8, 279
(1994).
51. R. H. Braswell, D. S. Schimel, E. Liner, B. Moore III,
Science 278, 870 (1997).
52. R. B. Myneni, C. D. Keeling, C. J. Tucker, G. Asrar, R. R.
Nemani, Nature 386, 698 (1997).
53. P. Ciais, P. Tans, M. Trolier, J. W. C. White, R. J.
Francey, Science 269, 1098 (1995).
54. D. J. Shea, Climatological Atlas: 1950 –1979 (TN-269,
National Center for Atmospheric Research, Boulder,
CO, 1986).
55. The CASA modeling activity has been supported
through NASA’s Earth Observing System program as
part of an Interdisciplinary Science grant to P. J.
Sellers and H. A. Mooney and a grant from the
Western Regional Center of the Department of En-
ergy National Institute for Global Environmental
Change to C.B.F. The VGPM activity has been sup-
ported through NASA grants to P.F. and M.J.B. J.T.R.
was supported by a NASA Earth System Science
Graduate Student Fellowship. Thanks to A. Lowry, D.
Kolber, Z. Kolber, M. Thompson, and C. Malmstro¨m
for assistance in developing and exercising the mod-
els. This is Carnegie Institution of Washington De-
partment of Plant Biology publication 1279.
28 March 1998; accepted 8 June 1998
Abrupt Shift in Subsurface
Temperatures in the Tropical
Pacific Associated with Changes
in El Nin˜o
Thomas P. Guilderson and Daniel P. Schrag
Radiocarbon (
14
C) content of surface waters inferred from a coral record from
the Gala´pagos Islands increased abruptly during the upwelling season (July
through September) after the El Nin˜o event of 1976. Sea-surface temperatures
(SSTs) associated with the upwelling season also shifted after 1976. The syn-
chroneity of the shift in both
14
C and SST implies that the vertical thermal
structure of the eastern tropical Pacific changed in 1976. This change may be
responsible for the increase in frequency and intensity of El Nin˜o events since
1976.
Several studies have noted that the pattern of
El Nin˜o–Southern Oscillation (ENSO) vari-
ability changed in 1976, with warm (El Nin˜o)
events becoming more frequent and more
intense (1). This “1976 Pacific climate shift”
has been characterized as a warming in SSTs
through much of the eastern tropical Pacific.
A recent study (2) proposed that this shift
originated when a subsurface warm water
anomaly in the North Pacific penetrated
through the subtropics and into the tropics.
This model is consistent with an association
of the shift in tropical temperatures with
changes in North Pacific sea-level pressures
(3). However, this interpretation is controver-
sial, and other mechanisms might be respon-
sible. Unfortunately, hydrographic observa-
tions have spatial and temporal biases that do
not allow for a definitive solution.
To examine changes in the origin of water
upwelling in the eastern Pacific during the
T. P. Guilderson, Department of Earth and Planetary
Sciences, Harvard University, Cambridge, MA 02138,
USA, and Center for Accelerator Mass Spectrometry,
Lawrence Livermore National Laboratory, Livermore,
CA 94550, USA. D. P. Schrag, Department of Earth and
Planetary Sciences, Harvard University, Cambridge,
MA 02138, USA.
R EPORTS
10 JULY 1998 VOL 281 SCIENCE www.sciencemag.org240