in corresponding natural ecosystems.
It in-
Human Appropriation of Renewable Fresh Water
Sandra L. Postel, Gretchen C. Daily, Paul R. Ehrlich*
Humanity now uses 26 percent of total terrestrial evapotranspiration and 54 percent of
runoff that is geographically and temporally accessible. Increased use of evapotranspi-
ration will confer minimal benefits globally because most land suitable for rain-fed ag-
riculture is already in production. New dam construction could increase accessible runoff
by about 10 percent over the next 30 years, whereas population is projected to increase
by more than 45 percent during that period.
U n l i k e other important commodities such
as oil, copper, o r wheat, fresh water has n o
substitutes for most of its uses. It is also
impractical to transport the large quantities
of water needed in agriculture and industry
more t h a n several hundred kilometers ( I ) .
Fresh water is now scarce in lnanv, r e ~ i o ~of
ls L>
t h e world, resulting in severe ecological
degradation, limits o n agricultural and in-
dustrial production, threats to human
health, and increased potential for interna-
tional co~lflict12,, 3).
, estimate transpiration separately from evap-
I n this report, we estimate how much of
Earth's renewable fresh water is realistically
accessible t o humanity; what portion of this
accessible supply humarlity now uses direct-
lv, diverts into human-dominated svstems.
dr appropriates; and by how much h u m a n
access to fresh water is likely to expand over
t h e next 3 0 vears. O n that basis, we derive
a n indicator of Earth's carrying capacity, as
well as a measure of the sustainability of
current water trends.
Fresh water co~lstitutesonly -2.5% of
t h e total volulne of water o n Earth, and
two-thirds of this fresh water is locked in
glaciers and ice caps (4). Just 0.77% of all
water (-10,665,000 km3) is held in aqui-
fers, soil pores, lakes, swamps, rivers, plant
life, and the atmosphere ( 4 ) .
Only fresh water flowing through t h e
solar-powered hydrological cycle is renew-
able (Fig. I ) . No~lreple~lishable(fossil)
ground water can be tapped, but such ex-
traction deuletes reserves i n much t h e same
way as extractions from oil wells do. T h e
terrestrial renewable fresh water supply
(RFWSIc,,,cI)equals precipitation o n land
(P,~,,,),which t h e n subdivides into two ma-
jor segments: evapotranspiration from t h e
land (ETI,,,cI)and ru11off t o t h e sea ( R ) .
Because ground water and surface water are
often hvdraulicallv connected, we i11clude
soil infiltration ground-water replenish-
,+
m e n t as Dart of this runoff comDonent.
Thus, RFWS~,, = Pi ,,,' = ETI',,, R.
Global water balance estunates are de-
S. L Postel, Global Water Policy Project, 17 Msgr,
O'Brien Hahwav. Cambrdae. MA 02141-1817. USA.
G. C ~ a i l q a n d ' ~R.. ~ h r l E h Department
, of ~ i o o g c a l
Scences, Stanford University, Stanford, CA94305, USA.
'To whom correspondence should be addressed.
rived from interpolation of climatic, vege-
tation. and soil information for different
geographic zones. T h e methoits are inher-
ently imprecise; estimates of annual runoff
range from 33,500 klni to 47,000 klni (5).
W e use t h e esti~natesof L'Vovich et al. (6),
which yield runoff values near t h e ~ n i d d l eof
this range (Fig. 2 ) .
Transniration is t h e u ~ t a k eof moisture
by plants and its release back into t h e at-
mosphere. O n a large scale, it is difficult to
oration; hence the joint term. Evapotrans-
piration represents the water supply for all
11011irrigated vegetation, including forests
and woodlands, and rain-fed
crons. Runoff is t h e source for all h u m a n
diversions or withdrawals of water for irri-
gated agriculture, industry, and municipali-
ties, as well as for a wide varietv of instream
water uses, i ~ l c l u d i ~ lthe
g maintenance of
aquatic life (for example, fisheries), naviga-
tion, t h e dilution of pollutants, and the
generation of hydroelectric power.
T o estimate the share of ET appropriated
by human activity, we started with the Vi-
tousek et al. (7) calculation of the fraction of
terrestrial net primary production (NPP)
that hulna~lity11ow co-opts. Co-opted N P P
is material used directlv, bv, humans or useit in
human-dominated ecosystems by cornmun1-
ties of orgallisms that are different from those
Land
Fig. 1. A simplified depiction of the global hydro-
logical cycle, adapted from Geick (5). Flows are
approximate, f a within ranges of estimates in (5),
and are in cubic kilometers ,Der vear.
, Downward
arrows signify precipitation; upward arrows signify
eva~otrans~iration. The horizontal arrow repre-
senis the transfer of atmospheric moisture from
sea to land and the arrow beow it represents
runoff from land to sea
SCIENCE VOL. 271 9 FEBRUARY 1996
cludes, for example, cropland, grazing lanit,
and trees harvested for f~lelwoodand timber.
Vitousek et al. estimate that co-opted terres-
trial N P P is 40.6 billio11 metric tons, or more
than 30% of total N P P (Table 1).
T o arrive a t a global estimate of t h e
average volume of E T required to produce a
unit of biomass, we divided total terrestrial
N P P of 132 billion metric tons per year (8)
by the global terrestrial annual E T estimate
above, yielding 1.9 kg of biomass per t o n of
ET, or about 2 g of biomass per kilogram (or
liter) of water (9). W e t h e n applied this
global average to t h e calculated co-opted
N P P (Table I ) , maki11g two aitjustments.
Approsilnately 16% of t h e world's cropla11d
is irrigated to supplemellt i n situ rainfall
(10). T o avoid double counting, we sub-
tracted from our estimate of E T o n cultivat-
ed land (Table 1 ) t h e share provided by
irrigation water, -2000 k ~ n ~ / ~ eW a re . also
assume that half of t h e ET associated with
lawns, parks, alld other h u m a ~ ~ - o c c u ~ i e d
areas is supplied by irrigation; thus t h e total
E T co-opted is -18,200 k1n3. This repre-
sents 26% (18,200 km3/69,600km') of total
terrestrial ET. T h e remaining 74% must
meet the water needs of all other land-based
species and natural communities.
W e adjusted total runoff (40,700 k m 3 )
for geographic and temporal i ~ ~ a c c e s s i b i l i t ~
t o estimate t h e portion that is realistically
available for h u m a n use; we call this acces-
sible runoff ( A R ) . T h e distribution of global
runoff among t h e continents is highly LIII-
even and correspo~ldspoorly to t h e distri-
bution of world population (Table 2). Asia,
with 60% of world population, contains
36% of global runoff. South America, with
-5% of world population, contains 25% of
runoff. Moreover, much of t h e ru~loffin t h e
tropics and high norther11 latitudes is virtu-
ally inaccessible to the h u m a n economy
and is likely to remain so for t h e foreseeable
future.
Table 1. Estimatesof ET appropriated for human-
dominated land uses. A total of 26.2% of terres-
trial ET is appropriated (18,200 km3/69,600 km3).
NPP
co-opted3 ET
Land type (10 V o o - o p t e d - ; -
metric (km3)
tons)
Cultivated land 15.0 5,500t
Grazing land 11.6 5,800
Forest land 13.6 6,800
Human-occupied areas 0.4 1OO$
(lawns, parks, golf
courses, and so forth)
Total appropriated 40.6 18,200
"NPP from intermedate calculation of Vitousek et a/.
(7). ;-Assumes 2 g of bomass produced for each liter
of water evapotranspred :;Adjusts for share of ET
requirement met through irrigation.
 T h e Amazon River accounts for 15% of
global runoff (1 1 ) . It is currently accessible,
however, to -25 ~llillion people (12)-
0.4% of world population-and n o masslve
expansion of irrigation is likely that would
warrailt major ~ ~ l v e r s i o nfrom
s it. W e thus
co~lsider95% of its flow inaccessible. T h e
Zalre-Congo ranks second in global runoff
(3.5% of the total) (11) and supports
-1.3?h of worlit (12). W e judge
half of its flow to be inaccessible for pur-
poses of irrigation anit initi~strialand mu-
nlcipal use over t h e next 3 0 years.
T h e final subtraction is for t h e relnote
rivers of North America and Eurasia, 55 of
which have n o d a ~ n so n their lnain chan-
nels ( 1 3). Most of this river flow is in tundra
and taiga biomes that are relnote from pop-
ulation centers. T h e combined average an-
nual flow of these northern untapped rivers
is 1815 k ~ n ' / ~ e a and
r , we subtract 95% of it.
Together, t h e inaccessible remote flows
of the Amazon, Zaire-Congo, and northern-
tier undeveloped rivers anlount to 7774 km'
per year (Table 3 ) , or 19% of total annual
runoff. This leaves -32,900 km' geograph-
ically accessible. O u r esti~nateis conserva-
tive hecause we made n o subtractions for
Total
evapotranspiration
on land
(69,600 km3/year)
many northern) rivers that
have verv, large " flows relative to t h e h u m a n
population size and water needs of their
geographic areas (14).
W e next adlusted for telnuoral inacces-
sibility. Irrigated agriculture, industry, and
h o ~ ~ s e h o l drequire
s that water be supplied
when alld where it is needed. This degree of
control over runoff is n o t easy t o achieve.
Approxilnately 11,100 klni of global runoff
(-27% of t h e total) is renewable
water and base river flow (6). As long as
extraction does not exceed renlenishment.
these sources c a n provide a reliable renew-
able supply. T h e remaining runoff, -29,600
km3, is much harder to capture, because
most of it is flood water. In Asia, for in-
stance, 80% of runoff occurs from May to
October ( 4 , 15). Capturing flood runoff
generally requires the construction of dams.
T h e present storage capacity of large dams
collectively totals 5500 km', of which 3500
k m 3 is actively used in t h e regulation of
river runoff ( 6 , 16).
Adding together the base flow and t h e
surface runoff controlled bv danls gives a n
estimate of t h e total stable'flow. i s s u n l i n g
that t h e geographically accessible runoff is
Remote flow
(7774 km3/year)
Uncaptured
floodwater
(20,426 km3/year)
f
Geographicallyand
temporally access~ble
runoff (AR)
(12,500 km3/year)
divided between base and flood flow i n t h e
same proportion that total runoff IS, we
t h e n reduced t h e esti~nateof total base flow
by t h e share of it containeit in t h e relnote
rivers, 2100 km' (0.27 x 7774 km'), lead-
ing to a n accessible base flow of 9000 km'
(11,100 - 2100). Addition to this of t h e
estimated 3500 k1n3 of runoff regulated by
existi~lg reservoirs yields a n estimate of
present total A R of 12,500 km3lyear.
W e next estinlated what portion of AR
humanity now uses. Three categories of wa-
ter use are (i) withdrawals or abstractions,
which represent water r e ~ n o v e dfrom rivers,
lakes, and aquifers for hulnan activities
(also k n o ~ v nas water demand or water use);
(ii) c o n s ~ ~ m p t i owhich
n, refers to withdraw-
als that are not available for a second or
third use; and (iii) h u m a n instreall1 flow
needs. Together, withdrawals and instream
uses provide a measure of h u m a n appropri-
ation of runoff, and we estimate them sep-
arately here.
Agriculture uses by far t h e [nost A R
worldwide. W e estimated agricultural water
withdrawals by multiplying a n average wa-
ter application rate of 12,000 m3/ha (17) by
t h e 1990 esti~nateof 240 ~llillionhectares of
world irrigated area (10). This yields a total
agricultural water demand of -2580 km'
(Table 4 ) . T h e ratio of consumption to
withdrawals 17arieswith cli~naticfactors, t h e
crops grown, and irrigation efficiency, and
typically ranges between 5 0 and 80% (4).
W e assume that -65% of agricultural water
~vithdrawalsare consumed, for a global total
of 157G km'.
Industrial water use has leveled off or
d e c l ~ n e dIn many wealthier countries, but is
growlng rap~dlyIn much of the develop~ng
world (2). Shiklomanov ( 4 ) estinlated that
industrial use is -975 klni globally, incluit-
ing the ther~noelectricpower in~lustry. In
contrast to agriculture, only a s~llallshare of
Lvater used in initustry is consumed; i no st of
it is ~ t i s c h a r ~ eback
d t o t h e environment,

lnstream uses
[2350 km3/year(19%)] Table 2. Share of global runoff and population by
I
continent.

Total Share of Share of

Human appropriation Human appropriation river global global
Region runoff* river POPU-
[ I 8,200 km3/year(26%)] (6780 km3/year (54%)] (km3/ runoff lat~on?
year) (96) (96)

Human Europe 3,240

appropriation of Asla 14.550
accessible RFWS,, Afr~ca 11,320
North and 6.200
[24,980 km3/year (30%)]
Central
Human appropriation Amerca
of total RFWS,,, South America 10,420 25.6 5.5
Australia and 1.970 4.8 0.5
[24,980 km3/year (23%)]
Oceania
Fig. 2. Flow dagram of analysis of human appropriat~onof RFWS,,,,,. The f n a box shows human 40.700 100.0 100.0
appropr~at~on of est~matedaccessible RF\RIS,,,,, to be 30O'o (24,980 km3/82.100 km7 and human +RLI,O~~estimates from (6) i-popuat~onest~mates
appropriaton of total RFWSlai., to be 23% (24,980 km3/1 10.300 km"). from (32)

786 SCIENCE \ OL 271 9 FEIIRUrlRk 1996

Table 3. Estimates of inaccess~blerunoff of se- this rate t o t h e 1990 population yields a next 3 0 years. If average reservoir capacity
lected remote rlvers. d ~ l u t ~ oreiluirement
n of -4700 km3. If per dam rernai~lst h e same as in the period
50% of m ~ ~ n ~ c i panii a l iniiustrial waste from 1950 to 1985, as well as the proportion
Remote flow globally receives a t least secondary treat- that is dead storage or otherwise unavail-
R~verbasin or region
(kmyyear)
m e n t before il~scharge( 2 0 ) , t h e n t h e in- able for water supply, -1200 km3 would be
Amazon* (95% of total flow) 5387 stream flow requirement is 2350 k ~ n ~ / ~ e a r aiided . to the accessible supply circa (ca.)
Zaire-Congo* (50% of total) 662 I n actuality, some iiilution is accom- 2025 (28). Addition of this to existing ac-
Remote undammed northern plished by flooii flows rather t h a n hy A R , tlve storage capacity of 3500 km3 yields a
rivers-; (950f0of totals) anii some a i i d ~ t i o n a rollution
l comes from total of 4700 krn3 ca. 2025. Combining t h ~ s
North Amerca 979
disperseii (such as agricultural) sources, with the accessible base flow (9000 km3)
Eurasia 746
Total inaccessible remote 7774 but because we are using t h e dilution re- gives a n A R ca. 2025 of 13,700 km3iyear
runoff quirement as a proxy for i~lstreanl uses 129).
, ,

'Amazon and Zaire-Congo runoff from (11). -i North-
ern rivers from (73).
although it is often polluteid. S o ~ u e9%-or
-90 km3-of industrial water a~ithdraa.als
are consumed.
M ~ ~ n i c i p a use
l varies greatly among
countries anii regions. Shiklomanov ( 4 ) ac-
counteid separately for urban anil rural in-
h a h ~ t a n t su, s ~ n gcountry-level iiata o n idem-
ographic characteristics and water use. HIS
estimated w o r l d a ~ ~ dmunicipal
e use is 300
knli per year, of which -50 k m i (or -179j1)
is consumed.
In certain geograph~creglons, reservoir
generally, we made n o aiijustments for If average per capita water delnanii re-
these (21 ). nlaills the same in 2025 as a t present [ ~ v h ~ c h
Overall, we estinlate that -18% of A R is conservative, because withiirawals per
(2285 km3/12,500 km3) is consu~ned capita increased nearly 509'0 between 1950
directly for h u m a n purposes. Withdrawals anii 1990 ( 2 ) ] , global water dernand ca.
from rivers, streams, and aquifers combined 2025 ~ v o ~ total ~ l d -6400 km3/year. Further,
with lnstream flow reiluire~nentstotal 6780 ~f instream flow needs for pollution dilution
km', w h ~ c h suggests that a n aiiiiitio~~al increase 111 iiirect proportion to population,
36%-for a total of 54% of A R (6780 these woulii total -3430 km3/year ca. 2025,
km3/12,500 km3-1s c ~ ~ r r e n t appropriateii ly for a total h u m a n appropriation ca. 2025 of
for h u m a n purposes. W e estimate that hu- -9830 km3iyear, or >70% of e s t ~ m a t e dA R
man use of E T and runoff constitutes 30% ca. 2025.
of the total access~hleRFLYIS [(18,200 k m ' W e ignore the poss~bility that, during
+ 6780 km3)/(69,600km' + 12,500 km3)]. the next few decades, runoff patterns might
This is conservative, because it a s s ~ ~ m that es be altered substantially by temperature in-
all E T is accessible ( 2 2 ) . Comparison of creases and nrecinitation shifts associateii
A A

losses to evaporation c o n s t i t ~ ~ tae suhstan- 11~1manuse wit11 the total unadjusted RFWS with the buildup of greenhouse gases (30).
t ~ a lshare of total runoff (18). W e assume ~ ~ l d i c a t ethat
s Homo snpiens is co-opting This, in turn, could alter iiam r e a ~ ~ l r e m e n t s
that a n average of 5% of the gross storage -23% of this life-support resource (18,200 and reservoir storage anii thus A R . G ~ v e n
capacity of reserl~oirsa ~ o r l d a ~ i d(5500
e km') kmi + 6780 km3/110,100 km'). the possible ~nonlineariries in the climatic
is lost to eva~orat1o1-1, or 275 km3/vear.
, , How much c a n A R be e x ~ ~ e c t eto d In- system, our ca. 2025 A R estimate may be
I n s t r e a ~ n f l o ~uses
~ incluiie mainte- crease during the next three idecaides? T h e optimistic.
n a n c e of navigation paths, water quality, principal means of expallding A R 1s to cap- T h e aquatic environment is alreaiiy
rlver deltas, fisheries, wildlife, riparian ture anii store more flood runoff or to ile- showing signs of degraiiation and decline,
\,egetation, other aquatic hiodivers~ty,and salinate seawater. Exotic o p t ~ o n s ,such as part~cularlybecause of dam construction,
r e c r e a t ~ o n a l opportun~ties. Because in- towing icebergs, are unl~kelyto yield appre- rlver diversions, heavy pollution loads, and
stream r e c l ~ ~ i r e ~ n e nvary
t s geographically ciable iluantities of water o n a global basis other habitat changes ( 2 7 , 31 ). Substant~al-
a n d seasonally, we useid pollution iiilution 111 the next 3 0 years. ly higher levels of human appropriation of
as a global proxy and assumeii t h a t t h e D e s a l ~ ~ l a t i o na ,~ h ~ csuppl~es
h -0.1%1 of A R could result In a severe falter~ng of
dilution reiluirement is sufficient to meet world water use (23), is a n expensive op- aquatic ecosystem services, including hroad
o t h e r Instream needs as well. A n often tlon, largely because ~t is energy-intensive. d e c ~ m a t i o nof fish populations and t h e ex-
~ ~ s edilution
d factor for assesslllg waste T h e theoretical m ~ n ~ m u energy m require- tinction of numerous beneficial species.
absorption capaclty is 28.3 l ~ t e r sper sec- lnent to relilove salt fro111 water IS 2.8 mil- Greater invest~nentsIn pollution preven-
o ~ per ~ d 1000 p o p u l a t ~ o n( 1 9 ) . Applying lion joules per cubic meter, but even the t1o11 a ~ o u l dfree up A R to meet rising hu-
best desalination plants now operating use man water needs while safeguarding ecolog-
30 times this alnount ( 2 4 ) . T e c h ~ ~ o l o g i c a l ical f ~ ~ n c t i o n sLikewise,
. greater e f f i c ~ e n c ~
Table 4. Estimated global water use and con
sumpton, by sector. ca. 1990. improve~nents 1n1g11t reduce energy neeids of water use, cha~lgesin agricultural crop-
to 10 times the theoretical minimum (24), p i ~ l gpatterns, anii t h e removal of marginal
Con- hut this IS still a s ~ l h s t a n t ~ energy
al require- laniis from irr~gationcoulii help slo~vt h e
Use ment. For the foreseeable future, ilesalina- g r o a ~ t hof h ~ l m a nappropriation of A R .
Sector (krn3/ SumPtlon tion is likelv to c o n t i n ~ ~toe he useLi L7runar-
(km3/
year)
year) ily to meet drinking water neeiis 111 water- REFERENCES AND NOTES
scarce, energy-rich nations.
Agr~culture* 2880 1870
Industly-i 975 90 T h e c r e a t i o ~ of
l new reservoirs will con- 1. A relatively small volume of fresh waters transported
tinue to expanii A R but at a slower rate. longer distances by tanker to supply drinking water
Municipalties-i 300 50 to water-scarce areas
Resenlo~rlosses:!: 275 275 Worliim.ide, a n a\.erage of 885 large dams 2. S. Postel, Lasi Oasis, Faclng i;l1aierScarc1iy (Norton,
Subtotal 4430 2285 (those a t least 15 m high) were constructeil New York, '992).
lnstream flow needs 2350 0 per year hetween 1950 anid the miid-1980s 3 P. H Gleick, Int. Secur. 18, 79 (summer 1993), N
Total 6780 2285 Myers, Dlinnate Secur1tY,.The Environmenial Bas6 of
( 2 5 ) . A t present, n o lnore than -500 large Pol~i~cal Stablllty, (Norton, New York, 1993).
Total as a percent 54% 18%
of AR (12,500 km;) clams are being comL~letedeach year ( 2 6 , 4. I. A Stiiklomanov, in Watei In Cilsis: A Guide to the
27), and we woiild expect this to drop fur- lV!/oi/d's Fresh ln,/atei Resouices, P. H Gleick, Ed.
'Assumes average applied water use of 12,000 m3iha (Oxford Univ. Press, New York, :993), pp. 13-24.
ther because of rising econo~llic.s o c i ~ l a11il .
and consumption equal to -65% of witiidrawals. 5. P. H. GleicK, Ed , Water in Cnss: A Guide to il7e
i Estimates are fioin (4). :Assumes evaporation loss e ~ l v ~ r o n m e n t acosts
l ( 2 ) . W e assunled a n lV!/oi/d'sFresh Watei Resouices (Oxford Unv. Press.
a l596 of gross reservoir storage capacity.
e q ~ ~to average of 350 new dams per year for t h e New York, 1993).

SCIENCE * VOL. 271 9 FEBRU.%RY 1996

 6. M. I. L'Vov~cheta1 , In The Earth as Transformed by,
Human Action, B. L. Turner et a/., Eds. (Cambr~dge
Univ. Press, Cambr~dge,1990), pp 235-252
7 P. M. Vitousek. P. R. Ehrlich, A. H. Ehrl~ch,P A
Matson, Bioscience 36, 368 (1986).
8 G. L. Ajtay, P Ketner, P. Duvgneaud, In m e Global
Caibon Cycle, B. Bolin. E. T. Degens, S Kempe, P.
Ketner, Eds ( W e y . New York. 1979). pp 129-1 82.
9. Our global estmate conforms well to values derived
from small-scale field studes w t h crops [B. A. Stew-
art, J. T Mus~ck,D, A. Dusek, Agron. J. 75. 629
(1983); Held Response to Water (U N. Food and
Agriculture Organization, Rome, 1979); Z. Z~xi,B A.
Stewart, F. Xiangjun, field Crops Res. 36. 175
(1994)l.
10. 1990 Pioduction Yearbook (UN. Food and Agrcul-
ture Organization, Rome, 1991), w ~ t hadjustments
for Unted States and Taiwan based on data from
U.S. Department of Agr~culture.
11. E. Czaya. Rivers of the World (Van Nostrand Ren-
hold, New York, 1981).
12. Populat~on est~mates from C. Haub and M.
Yanaglshta, Population Reference Bureau (personal
commun~cat~on, Wash~ngton,DC, January 1995).
13. M. Dynesius and C N~isson,Science 266, 753
(1994).
14. We do not Include In our est~mateof remote northern
river flows a large number of rlvers that have one or
two dams (typ~callyfor hydropower) on the~rm a n
channels but have flows vastly ~nexcess of water
supply needs ~nthe reglon, ncludlng, for example,
the Ob and Lena rlvers of Sberlan Russa, w ~ t ha
combined flow of 935 km". The amb~t~ous Sov~et
scheme to d~vertwater from the Ob to the Ara Sea
basln would n t a l l y have lnvolved 25 km3/year, just
6% of the Ob's annual average flow. L~kew~se, a G. E. Stout, Eds. (Oxford Unv. Press. Oxford, 1992),
proposal to s h ~ pwater vla undersea p~pellnefrom
southeast Alaska to C a f o r n a lnvolved 5 km3 annu-
ally, just under 5% of the combined average annual
flow of the Copper and S t k ~ n erlvers, leav~ng95% of
the~rflow st11remote [Alaskan Water for Cal~forn~a?
The Subsea P@eBne Option-Background Paper
(U.S. Off~ceof Technology Assessment, Washng-
ton, DC, 1992)l.
15. Uncaptured flood runoff provdes a varety of human
benefts, n c u d n g support of flood-recesston farmng,
fsheres, and generatlor- of hydroeectrlcty; however,
In these capaclt~es,t s use IS e~therlns~gn~f~cant glo-
bally or does not nvove actual approprlatlon.
16. Theoret~cay,a reservoir could be flied and empt~ed
more than once a year, creatng a greater effect~ve
capac~tyto regulate runoff than the storage capac~ty
alone would ndcate. We know of no estimates of
thls effectve storage capac~tyother than the state-
ment by K. Mahmood [Resenloir Sedimentat~on:lm-
pact. Extent, andMitigat1on (The World Bank, Wash-
ngton. DC. 1987)l that the usable reservoir storage
capacity "IS nearly used once every year." We there-
fore make no adjustments to the estmated 3500
km3 of capacity usable for runoff storage on an av-
erage annual bass.
17. This IS a somewhat higher rate than is implied by
Shikiomanov's estimates (4),which suggest rates of
10,700 to 11,000 m3/ha, We arrived at our figure
after examning data for California that suggest an
average water applcation rate on that state's irrigat-
ed area of -10,300 m3 ha [Californ~aWater Plan
Update (California Department of Water Resources,
Sacramento, CA, 1994), v o I], Because the aver-
age irrigation efficiency in C a ~ f o r n as reported to be
70%, whch IS substantially hgher than the worid-
wide average [S. Postel, in (5), pp. 56-66], we be-
lleve that 12,000 +/ha 1s closer to the actual global
average applicat~onrate. Moreover, the California
figures account only for on-farm water applcatons
and do not n c u d e the porton of d~versionslost to
seepage or evaporation between reservoirs and
farmers' f~elds.
18. Evaporative losses from Lake Nassar, for example,
have averaged 10 km3/year, w h ~ c his equal to 12%
of the Nlle's average annual flow [J. A. Allan, in The
Nile: SharingA Scarce Resource, P P. Howell and J
A. Alan. Eds. (Cambr~dgeUnv. Press, Cambridge,
1994). pp. 313-3201,
19. H. E. Schwarz. J. Emel. W. J. D~ckens,P. Rogers, J.
Thompson, In The Eariii as Transformed b y Human
Action, B. L Turner et a/., Eds. (Cambr~dgeUn~v
Press, Cambr~dge,1990), pp. 253-269.
20. Even in the countries of the Organzatlon for Eco-
nomic Cooperation and Development, domestic
wastewater treatment IS estimated to cover only
-60% of the population [A. K. Biswas. h'aterlnt. 17.
68 (February 1992)], nformatlon for developing
countries is sparse, but treatment coverage is cer-
t a n y far lower. Moreover, few regons control for
farm runoff and other dispersed pollut~onsources
that add substantal quant~tiesof sedment, pesti-
cides, and fertizers to water bodies.
21. Even f wastewater treatment coverage should be-
come nearly un~versal,substant~alnstream flows
would still be required to ma~ntainfsheries, support
recreat~onaldemands, and satisfy other instream
needs. For example, Cal~forn~a's Instream envron-
mental water requirements (after omlsslon of the
north coast hydrolog~cregon, w h c h conta~nssever-
al w ~ l dand scenc rlvers and thus may not be indc-
at~veof instream needs more narrowly def~ned)equal
22% of average annual runoff [Cal~forniaWater Plan
Update (Cal~forn~a Department of Water Resources,
Sacramento, CA, 1994).
22. We d d not cons~der~tfeas~bleto estmate accessbe
ET In a manner comparable to our estmate of AR. To
be conseivat~ve,we therefore assumed all terrestral
ET to be access~ble.
23 Wangnck Consult~ng,1990 IDA Worldln/~deDesalt-
ing Plants Inventory (Internatonal Desanaton Asso-
ciat~on.Englewood, NJ, 1990).
24. P. H. Gle~ck,Annu. Rev. Energy Envlron. 19, 267
(1994).
25, J. A. Veltrop, in Water for Sustainable Development
ln the Twenty-first Century, A K. B~swas,M. Jellall,
pp. 102-1 15.
26. Status of Dam Construction, 1991 (lnternat~onal
Commss~onon Large Dams, Parls, 1992), suggests
that -300 dams are now comm~ssonedeach year,
but these data Include only 64 countr~es.
27. A P. Cov~ch[ ~ n(5), pp 4G55] lndcates that iarge
Rapid Collapse of Northern
Larsen Ice Shelf, Antarctica
Helmut Rott, Pedro Skvarca, Thomas Nagler
In January 1995, 4200 square kilometers of the northern Larsen Ice Shelf, Antarctic
Peninsula, broke away. Radar images from the ERS-I satellite, complemented by field
observations, showed that the two northernmost sections of the ice shelf fractured and
disintegrated almost completely within a few days. This breakup followed a period of
steady retreat that coincided with a regional trend of atmospheric warming. The obser-
vations imply that after an ice shelf retreats beyond a critical limit, it may collapse rapidly
as a result of perturbated mass balance.
I c e shelves cover 11% of the total area of
Antarctica ( 1 ) and play an i ~ n p o r t a ~role
lt
in the Inass budget and dynamics of the
Antarctic Ice Sheet. Most of the ice that
has accunlulated over the grounded parts of
Antarctica is discharged to ice shelves,
where it is lost as icehergs along the seaward
edges as well as by hasal melting (2). Be-
cause ice shel1:es are exposed to both atmo-
sphere and ocean, thei are sensitive tcj
chaln~esin the temnerature and circulation
H Rott and T. Nager, lnsttut fur Meteorologie und Geo-
phys~kder ~ n i v e r s ~ t aInnsbruck,
t lnnra~n5 2 , A-6020
Innsbruck. Austr~a.
P. Skvarca, nstituto Antartco Argentno. Cerr~to1248.
1010 Buenos A~res,Argerit~na.
SCIENCE VOL 271 9 FEBRUARY 1996
dams are currently beng completed at an average
rate of 500 per year, or 56% of the rate of the perod
from 1950 to 1986.
28. Because -85% of ex~stinglarge dams were bu~lt
snce mid-century (25), this calculation assumes that
8596 of total existing storage capacity was con-
structed since then, or 4675 km3 (5500 km3 x 0.85).
With the assumpton that 40% as many dams would
be constructed between 1990 and 2025 as between
1950 and 1985, and that capacity per dam remans
constant, 1870 km3 (4675 km3 x 0.40) of capacty
would be added by ca 2025, of w h c h 1190 km3
would be live storage for water supply
29. Even as dam construction IS adding to the total stable
runoff, other human activit~esare reducing it Defores-
taton and the paving over of aquifer recharge areas
often reduce ra~nwater~nfiltrat~on, thereby reducing
base flow and lncreasng surface flood runoff. More
Important giobaly, many reservoirs are loslng act~ve
storage capac~tyfaster than or~g~nally est~matedbe-
cause of rapd station from deforestat~on,so11ero-
s~on,and generally poor watershed management.
The Nzamsagar reservor In India, for Instance, lost
more than 60% of t s capaclty over 40 years [M. New-
son, Land, Water and Development: River Basin Sys-
tems and Their Sustainable Management (Routedge,
London, 19921. Lackng global estmates, we make
no subtract~onfor these losses.
30. P. E. Waggoner, Ed., Cl~mateChangeand U.S. Wa-
ter Resources (W~ley,New York, 1990).
31. Nat~onalResearch Councl, Restoration of Aquatic
Ecosystems (Nat~onaAcademy Press, Washington,
DC, 1992).
32. 1994 World Population Data Sheet (Popuaton Ref-
erence Bureau, Wash~ngton,DC, 1994).
33. We gratefully acknowledge comments from W Fa-
con, P. Gle~ck,R. Naylor, A. V~ckers,P. V~tousek,
and two anonymous reviewers. Supported by a
grant from Charles and Nancy Munger, the Wnslow
and Henz foundatons, and an anonymous donor.
28 September 1995; accepted 21 December 1995
of either (3).T h e 0C summer isotherm has
been taken as the clilnatic lilnit for the
existence of ice shelves along the \vest coast
of the Antarctic Peninsula (4). Between
1966 and 1989, the Wordle Ice Shelf (Fig.
1 ) decreased from -2000 to 700 krn', prob-
ably as a result of regional at~nospheric
\varming (5). Here, we report o n the recent
disintegration of the northern Larsen Ice
Shelf (LIS).
T h e LIS exte~ldsalong the eastern side
of the Antarctic Peninsula from latitude
64' to 74's (Fig. 1). The part of the LIS
llorth of ~ ~ b ~ ~& , ~~hasl retreated ~ ~ ~ d
" o w l ~but constalltl~since 1940s (67).
T h e retreat accelerated after 1975 and (a),