Vous êtes sur la page 1sur 14

Journal of Hydrology, 117 (1990) 241 254

Elsevier Science Publishers B.V., Amsterdam - - Printed in The Netherlands

241

111

GLOBAL RIVER R U N O F F C A L C U L A T E D FROM A GLOBAL A T M O S P H E R I C G E N E R A L CIRCULATION MODEL

GARY L. RUSSELL and JAMES R. MILLER

~ NASA/Goddard Space Flight Center, Institute for Space Studies, 2880 Broadway, New Yortz, N Y 10025 (U.S.A.) '-'Department of Meteorology and Physical Oceanography, Cook College, Rutgers University, New Brunswick, NJ 08903 (U.S.A.)
(Received June 23, 1989; accepted after revision October 24, 1989)

ABSTRACT Russell, G.L. and Miller, J.R., 1990. Global river runoff calculated from a global atmospheric general circulation model. J. Hydrol., 117:241 254. The purpose of this paper is to show that an atmospheric general circulation model (GCM) can be used to calculate runoff for the world's major rivers, that river runoff provides an important diagnostic for climate modelers, and that the model runoff provides useful information for hydrologists. The global atmospheric model of Hansen et al. has been used to calculate the annual river runoff for the world's major rivers. The model has a horizontal resolution of 4 x 5, but the runoff from each grid box within a particular river's drainage basin is summed on a resolution of 2 2.5 to obtain the runoff at the river mouth. The mean annual runoff is calculated and compared with observations for 33 of the world's largest rivers. The runoff depends on the model's precipitation and parameterizations of groundwater storage and evapotranspiration, which are affected by soil type and vegetation.

INTRODUCTION

The global hydrologic cycle is one of the principal components of our climate system. Across the global air-sea interface there is a net flux of water out of the ocean because evaporation exceeds precipitation. Over land, evaporation is less than precipitation. The long-term global water budgets for both continents and oceans are balanced by the continental runoff of water back to the ocean. The distribution of global runoff has been discussed in several papers (Baumgartner and Reichel, 1975; Korzoun et al., 1977; Milliman and Meade, 1983). Atmospheric general circulation models (GCM) have been used to simulate the present climate. In the past few years, global climate modelers have begun to examine the effects of more complex parameterizations of land-atmosphere biosphere interactions. The biosphere atmosphere transfer scheme (BATS) described in Dickinson (1984), Dickinson et al. (1986), and Wilson et al. (1987) and the simple biosphere (SiB) of Sellers et al. (1986) provide a model

0022-1694/90/$03.50

((; 1990 Elsevier Science Publishers B.V.

242

G,L. RUSSELL AND J R , MILLER

framework for a more accurate representation of surface processes in GCM, particularly the interactions with vegetation and vegetation canopies. Such studies are needed if we are fully to understand the physical processes that affect the global hydrologic cycle. An important component of the Earth's hydrologic cycle is river runoff. The world's 20 largest rivers account for ~40% of the total continental runoff (Baumgartner and Reichel, 1975). The river runoff for a particular drainage basin depends on the precipitation and evaporation budgets within the basin and on the ability of the ground to store water, which depends on the soil type and the vegetative cover. Flow rates depend on the topography. The comparison of model-generated river runoff with observations provides a useful diagnostic for climate modelers to obtain a better understanding of the parameterizations which affect the hydrologic cycle in their models. An important reason for hydrologists to study model-generated river runoff is to understand and to predict future changes in river runoff that may accompany global climatic changes. Since the prediction of future changes depends on some type of model, it is essential for hydrologists and climate modelers to develop the best possible surface parameterizations that affect model-generated river runoff. The primary purpose of this paper is to demonstrate that the calculation of river runoff in an atmospheric GCM provides a useful diagnostic for climate modelers and a useful tool for hydrologists. To accomplish this, we use the runoff from each grid box of the global atmospheric model of Hansen et al. (1983) to calculate the model-generated runoff from the world's rivers which have a mean annual discharge greater than 100 km3year 1 or a drainage basin area greater than 5 105km 2. The mean annual runoff for each river is calculated from a four-year model simulation and is compared with observations. The effect of the model precipitation, evaporation, and groundwater storage on the runoff is examined. THE ATMOSPHERICMODEL A four-year simulation with the Climate Model II of Hansen et al. (1983) was run with a horizontal resolution of 4 latitude by 5 longitude and nine vertical layers. Arakawa's B grid scheme is used for the dynamics. The source terms include a comprehensive radiation calculation and parameterizations of condensation and surface interaction. At the surface, grid boxes are divided into land and ocean fractions. The sea surface temperature (SST) and ocean ice distribution is specified from monthly climatologies interpolated on a daily basis. The SST is from Robinson and Bauer (1982). The ocean ice distribution is from Walsh and Johnson (1979) in the Northern Hemisphere and Alexander and Mobley (1976) in the Southern Hemisphere. The land distribution and continental topography is from a corrected version of Gates and Nelson (1975). The runoff in each grid box depends on the precipitation, evaporation, and

GLOBAL RIVER RUNOFF

243

w a t e r storage within the land p o r t i o n of the grid box. The m e a n a n n u a l distribution of p r e c i p i t a t i o n of the model is shown in Fig. la. F o r comparison, o b s e r v a t i o n s of p r e c i p i t a t i o n by Shea (1986) are s h o w n ifi Fig. lb and a difference plot is shown in Fig. lc. The e v a p o r a t i o n ( k g m 2s 1) is c a l c u l a t e d as the product:
E = flpCV(q~ qa)

(1)

where fl is a dimensionless efficiency f a c t o r for e v a p o r a t i o n or evapotranspiration, p ( k g m a) is the surface air density, C is a dimensionless drag coefficient t h a t depends on stability, V (m s- 1) is the surface wind speed, qs is the surface s a t u r a t i o n specific h u m i d i t y t h a t depends on the g r o u n d t e m p e r a t u r e and the surface pressure, and qA is the surface air specific h u m i d i t y at 10m above the surface. The a n n u a l c h a n g e in g r o u n d w a t e r storage over a four-year r u n is insignific a n t c o m p a r e d with a n n u a l precipitation, e v a p o r a t i o n and runoff. Nevertheless, the c u r r e n t stored g r o u n d w a t e r directly affects both fl and the runoff. E a c h grid box has two layers of g r o u n d w a t e r storage. The upper l a y e r responds immediately to e v a p o r a t i o n and p r e c i p i t a t i o n and the lower l a y e r acts as a seasonal reservoir. T h e r e is a two-day time c o n s t a n t for diffusion of w a t e r between the two layers, except d u r i n g the growing season when the u p w a r d diffusion o c c u r s i n s t a n t l y over v e g e t a t e d areas. The w a t e r field capacities of the two layers depend on the v e g e t a t i o n c h a r a c t e r i s t i c s of each grid box and are given in Table 1 (from H a n s e n et al., 1983). The choice of v e g e t a t i o n types is described more fully in M a t t h e w s (1983). If the soil is u n s a t u r a t e d , runoff, R ( k g m 2 s 1) is c a l c u l a t e d as:
R = -~ P W 1

(2)

where P (kg m-2 s - l ) is the p r e c i p i t a t i o n and W~is the r a t i o of w a t e r in the first l a y e r divided by the w a t e r field capacity. The f a c t o r fl in eqn. (1) is equal to W~ unless the g r o u n d is snow covered in which case fl = 1. F o r s a t u r a t e d soil
R=P.

CALCULATION OF RIVER RUNOFF The d r a i n a g e basins for the rivers in this study were defined on a h o r i z o n t a l r e s o l u t i o n of 2 2.5 and were e x t r a c t e d from K o r z o u n et al. (1977) and the TABLE 1 Water field capacity (kgm -2) as a function of vegetation characteristics in Model II. Layer 1 2 Desert 10 10 Tundra 30 200 Grass 30 200 Shrub 30 300 Woodland 30 300 Deciduous 30 450 Evergreen 30 450 Rainforest 200 450

244

G.L. RUSSELL ANDJR. MILI,ER

PRECIPITATION (mm/day)
1312llI09876-

Model

II

(4 0 S )

Annual

4~

:t
(a)

PRECIPITATIOH (mm/day)
1312' 11. 10" 9. 8" 7" 6' S~

Shea Observations

Annual

4~

3i

i
(b) Fig. 1. Mean annual distribution of precipitation from (a) the atmospheric model and (b) the observed data of Shea (1986). The model precipitation minus the observed precipitation is given in (c).

(HA)BAI, RIVER RUNOFF

245
Model I I - Shea Annual

,4 PRECIPITATION (mmlday)

3I

2-

1-

0 _~

II

Fig. l(c).

Major River Dra ina0e Bas ins


%

~J

Fig. 2. D r a i n a g e b a s i n s of t h e rivers in t h i s study. T h e letter located at t h e m o u t h of e a c h river c a n be used to identify t h e r i v e r a c c o r d i n g to t h e key in Table 2. T h e l a r g e s t r i v e r in e a c h c o n t i n e n t is s h o w n in blue, o t h e r rivers w i t h r u n o f f > 2 0 0 k m ~ y e a r -~ are s h o w n in g r e e n a n d o r a n g e , a n d interior d r a i n a g e b a s i n s are s h o w n in yellow. Some of t h e i d e n t i f y i n g l e t t e r s are u s e d m o r e t h a n
once.

246 Times Atlas of the World observed exceeding Meade mean annual (1967). T h e c r i t e r i a exceeding

G.L. RUSSELLAND JR. MILLER for selecting a river were an basin and

runoff

1 0 0 k m ~ y e a r -1 o r d r a i n a g e to the list of rivers in Milliman

5 10 ~ k m ~ i n a r e a a c c o r d i n g

(1983). basin areas are listed by summing t h e a r e a s o f t h e 2 2.5

The rivers considered in this study and their drainage i n T a b l e 2. M o d e l a r e a s a r e c a l c u l a t e d

TABLE 2 Model and observed drainage basin areas (101 m 2), annual runoff (km 3year- 1), and annual precipitation (km 3 year 1) for the world's major rivers River Area Model A B C D F H I K L M N O P Q S T U V W Y Z A C F I K M N O R S Y Z Amazon Brahma-Ganges Columbia Danube Fraser Hsi Chiang Irrawady McKenzie Lena Mississippi Niger Ob LaPlata Orinoco St. Lawrence Tigris-Euphrates Yukon Mekong Yellow Yangtze Congo Amur Colorado San Francisco Indus Kolyma Magdalena Nile Orange Murray Severnay Dvina Yenesei Zambesi 611 155 69 85 24 45 40 169 231 327 121 266 286 111 117 100 88 82 113 197 382 190 65 66 95 62 24 282 104 110 33 267 126 Obs. a 615 148 67 81 22 44 43 181 250 327 121 250 283 99 103 105 84 79 74 194 382 185 64 64 97 64 24 296 106 106 35 258 120 Runoff Model 2332 1229 303 298 150 400 769 562 544 517 351 504 404 474 462 79 492 712 515 1304 2165 316 83 210 300 338 314 586 123 117 117 501 253 Obs. a 6300 971 251 206 112 302 428 306 514 580 192 385 470 1100 447 46 195 470 49 900 1250 325 20 97 238 71 237 83 11 22 106 560 223 Precipitation Model 12267 1874 754 996 289 942 1345 1176 1414 2645 1922 1250 2941 2173 1206 499 780 1997 1407 3442 8829 1369 417 1155 654 470 863 3509 972 727 224 1484 1589 Obs. b 11639 1794 438 596 203 647 771 654 770 2439 1217 1117 3297 1720 1033 396 385 1126 547 1976 5596 939 165 898 521 162 380 1915 414 596 168 989 1275

aFrom Milliman and Meade (1983). bFrom Shea (1986); accumulated over the model's drainage basin.

GLOBAL RIVER RUNOFF

247

grid boxes t h a t lie in the a p p r o p r i a t e d r a i n a g e basins. O b s e r v e d a r e a s and r u n o f f are from M i l l i m a n a n d M e a d e (1983). E a c h of t h e G C M 4 x 5 grid boxes overlies four 2 x 2.5grid boxes. T h e G C M m e a n a n n u a l r u n o f f f r o m e a c h grid box is d i s t r i b u t e d into the 2 x 2.5 land grid boxes. T h e G C M m e a n a n n u a l p r e c i p i t a t i o n is d i s t r i b u t e d into all 2 x 2.5 grid boxes r e g a r d l e s s of s u r f a c e type. F o r the model d a t a in T a b l e 2, the m e a n a n n u a l r u n o f f or p r e c i p i t a t i o n of a r i v e r ' s d r a i n a g e b a s i n is the s u m m a t i o n of the r u n o f f or p r e c i p i t a t i o n from e a c h 2 x 2.5 grid box w i t h i n the basin. O b s e r v e d p r e c i p i t a t i o n (Shea, 1986) is s u m m e d o v e r the s a m e grid boxes. In the GCM, r u n o f f is c a l c u l a t e d but t h e n d i s a p p e a r s and is not used in s u b s e q u e n t c a l c u l a t i o n s r e l a t e d to the h y d r o l o g i c cycle. T h e r u n o f f a n d p r e c i p i t a t i o n in T a b l e 2 are given in a n n u a l w a t e r m a s s for e a c h r i v e r basin. In T a b l e 3, t h e y are given per u n i t area. T h e model r u n o f f a n d p r e c i p i t a t i o n and the o b s e r v e d p r e c i p i t a t i o n from T a b l e 2 are divided by the model d r a i n a g e area, w h e r e a s the o b s e r v e d r u n o f f is divided by the o b s e r v e d area. The e v a p o r a t i o n is c a l c u l a t e d as p r e c i p i t a t i o n m i n u s runoff. The d r a i n a g e basins of the rivers are s h o w n in Fig. 2. W h i t e a r e a s are o c e a n grid boxes on the 2 x 2.5 resolution. B l a c k a r e a s d r a i n to the o c e a n b u t are not included in a n y of the r i v e r basins in this study. The r i v e r s with the l a r g e s t d i s c h a r g e s on e a c h c o n t i n e n t are s h o w n in blue. Yellow a r e a s are i n t e r i o r d r a i n a g e basins t h a t do not r e a c h the ocean. T h e l e t t e r at the m o u t h of e a c h r i v e r identifies the r i v e r s a c c o r d i n g to the k e y in T a b l e 2. Some letters are used m o r e t h a n once. One p r o b l e m a s s o c i a t e d w i t h the model grid r e s o l u t i o n is t h a t a p a r t i c u l a r grid box m a y c o n t a i n o n l y a p o r t i o n of a r i v e r d r a i n a g e basin. We h a v e a c c o u n t e d for this s o m e w h a t by dividing the m o d e l ' s grid boxes into four 2 2.5 boxes for the r u n o f f c a l c u l a t i o n s . If a p a r t i c u l a r grid box is a s s i g n e d to a r i v e r ' s d r a i n a g e basin, all the c o n t i n e n t a l r u n o f f in t h a t box is assigned to the r i v e r flow e v e n t h o u g h the a c t u a l d r a i n a g e b a s i n does not c o v e r the w h o l e grid box. E v e n w i t h this finer r e s o l u t i o n t h e r e are some rivers for w h i c h the d r a i n a g e b a s i n does n o t e x t e n d far e n o u g h b e c a u s e grid boxes closer to the m o u t h c o n t a i n too m u c h area. This p r o b l e m is g r e a t e s t for the s m a l l e r and n a r r o w e r d r a i n a g e basins, p a r t i c u l a r l y in m o u n t a i n o u s r e g i o n s in Asia. COMPARISON WITH OBSERVATIONS T h e d r a i n a g e basins in Fig. 2 were used to c a l c u l a t e the m e a n a n n u a l r i v e r r u n o f f from the model. F i g u r e 3 shows the m e a n a n n u a l r u n o f f for the five l a r g e s t rivers and c o m p a r i s o n with the o b s e r v e d v a l u e s of M i l l i m a n and M e a d e (1983). The m o d e l - g e n e r a t e d r u n o f f from the A m a z o n a n d O r i n o c o Rivers in S o u t h A m e r i c a is less t h a n h a l f the o b s e r v e d values. H o w e v e r , the model r u n o f f for the C o n g o is a b o u t twice the o b s e r v e d value. The model o v e r p r e d i c t s the r u n o f f for the B r a h m a p u t r a - G a n g e s and the Y a n g t z e R i v e r s by a b o u t 30%. A l t h o u g h the five l a r g e s t r i v e r s h a v e o b s e r v e d m e a n a n n u a l r u n o f f of a b o u t 1000 k m 3 y e a r -1, except for t h e A m a z o n w h i c h is six times larger, m o s t of the

248 TABLE 3

G.L. RUSSELl,AND J.R MILI,ER

Model and observed runoff, precipitation and evaporation per unit area (m year ~) for the world's major rivers. River Runoff Model A B C D F H I K L M N O P Q S T U V W Y Z A C F I K M N O R S Y Z Amazon Brahma Ganges Columbia Danube Fraser Hsi Chiang Irrawady McKenzie Lena Mississippi Niger Ob LaPlata Orinoco St. Lawrence Tigris Euphrates Yukon Mekong Yellow Yangtze Congo Amur Colorado San Francisco Indus Kolyma Magdalena Nile Orange Murray Severnay Dvina Yenesei Zambesi 0.38 0.80 0.44 0.35 0.64 0.88 1.92 0.33 0.24 0.16 0.29 0.19 0.14 0.43 0.40 0.08 0.56 0.87 0.45 0.66 0.57 0.17 0.13 0.32 0.32 0.55 1.28 0.21 0.12 0.11 0.35 0.19 0.20 Obs. ~ 1.02 0.66 0.37 0.25 0.51 0.69 1.00 0.17 0.21 0.18 0.16 0.15 0.17 1.11 0.43 0.04 0.23 0.59 0.07 0.46 0.33 0.18 0.03 0.15 0.25 0.11 0.99 0.03 0.01 0.02 0.30 0.22 0.19 Precipitation Model 2.01 1.2l 1.09 1.17 1.23 2.08 3.36 0.70 0.61 0.81 1.59 0.47 1.03 1.96 1.03 0.50 0.88 2.45 1.24 1.75 2.31 0.72 0.64 1.75 0.69 0.76 3,52 1,24 0.93 0,66 0,68 0,55 1.26 Obs. h 1.90 1.16 0.63 0.70 0.86 1.43 1.92 0.39 0.33 0.75 1.00 0.42 1.15 1.55 0.88 0.39 0.44 1.38 0.48 1.00 1.47 0.50 0.25 1.36 0.55 0.26 1.55 0.68 0.40 0.54 0.51 0.37 1.01 Evaporation' Model 1.62 0.42 0.65 0.82 0.59 1.20 1.44 0.36 0.38 0.65 1.30 0.28 0.89 1.53 0.64 0.42 0.32 1.57 0.79 1.08 1.75 0.56 0.52 1.43 0.37 0.22 2.24 1.04 0,82 0.55 0.32 0.37 1.06 Obs. 0.88 0.50 0.26 0,44 0.35 0.74 0.93 0.22 0.13 0.57 0.85 0.27 0.99 0.44 0.45 0.35 0.20 0.78 0.42 0.54 1.14 0.32 0.22 1.21 0.30 0.15 0.56 0.65 0.39 0.52 0.21 0.15 0.83

aFrom Milliman and Meade (1983). bFrom Shea (1986); accumulated over the model's drainage basin. "Precipitation minus runoff. other major river drainage basins have mean annual runoff between 200 a n d

600 k m 3 y e a r 1. F i g u r e 4 s h o w s t h e m o d e l a n d o b s e r v e d r u n o f f f o r t h e s e o t h e r m a j o r r i v e r s . T h e m o d e l r u n o f f is w i t h i n ~ 2 0 % o f t h e o b s e r v e d r u n o f f f o r h a l f o f t h e 16 r i v e r s s h o w n i n F i g . 4, a n d w i t h i n 3 ( ~ 4 0 % f o r t h e r e s t e x c e p t f o r t h e Mekong, Irrawaddy and McKenzie, which are all overpredicted by the model. Runoff was also calculated for other rivers with relatively large drainage b a s i n s ( a r e a s > 5 1 0 5 k m 2) b u t l o w e r r u n o f f . T h e s e a r e s h o w n i n F i g . 5. T h e

GLOBAL RIVER RUNOFF

249

ANNUAL RIUER RUIiOFF


6500
oa

6000
(@

5500
5000

i E c-

0 N E

0 0 (_>

0 u o ~0

C@

"2
"~

4500

4000

_%
v

3500

9qo 3000 c~ 2500 L


> t',v" !5 0 0 2000

500 0

EO0bserved

~Model

(4x5 )

Fig. 3. Mean annual runoff for the world's five largest rivers. Comparison between model runoff and the observed runoff of Milliman and Meade (1983) Fraser River is also included b e c a u s e its runoff is > 100 km 3 year 1. The model runoff is overpredicted for all of these rivers, and is more than s e v e n times too large for the N i l e and Y e l l o w Rivers. M a n y of the river basins in this figure h a v e m e a n a n n u a l observed precipitation < 0 . 5 m y e a r 1 (Table 3) and the m o d e l overpredicts the r u n o f f in all of these basins. The precipitation field over c o n t i n e n t s is o n e of the major deficiencies of the model since the model overpredicts the m e a n a n n u a l precipitation over c o n t i n e n t s by ~ 25%. For the runoff c a l c u l a t i o n s this b e c o m e s more critical for basins w i t h l o w e r m e a n a n n u a l runoff as s h o w n in Fig. 5. A n o t h e r problem a s s o c i a t e d w i t h the model r u n o f f in Fig. 5 is that the model does n o t a l l o w runoff to e v a p o r a t e as it m o v e s from o n e grid box to another. Since m a n y of these rivers are in dry regions, the n e g l e c t of this e v a p o r a t i o n is likely to lead to o v e r p r e d i c t i o n of the runoff. As an e x a m p l e of w a t e r loss in the N i l e River, C h a n and E a g l e s o n (1980) h a v e s h o w n that ~ 60% of the flow of the W h i t e N i l e disappears in s w a m p s before r e a c h i n g the j u n c t i o n w i t h the Blue Nile at K h a r t o u m . If this were i n c o r p o r a t e d into the model, the predicted runoff for the N i l e w o u l d be significantly reduced. Figure l a s h o w s that the model precipitation in n o r t h e r n Asia is g e n e r a l l y low, a l t h o u g h Fig. l c s h o w s that the model precipitation is s o m e w h a t larger

250

G.L, RUSSELL AND J.R. MILLER

ARMUAL RIVER RUMOFF


800 "M~SA~ES

~oo

oo

+~

~
~0 E

500

,Ik
E]Observed l~Model (4x5

300

Fig. 4. Model and observed runoff for rivers with observed mean annual runoff > 200km~year 1 that are not shown in Fig. 3. Observed runoff from Milliman and Meade (1983). t h a n the o b s e r v a t i o n s . This is an e x a m p l e of a case w h e r e the r u n o f f m a y be modeled well e v e n t h o u g h the p r e c i p i t a t i o n is too large. This i n d i c a t e s t h a t t h e r e is too m u c h e v a p o r a t i o n , p a r t l y b e c a u s e of the p a r a m e t e r i z a t i o n of g r o u n d w a t e r s t o r a g e a n d p a r t l y b e c a u s e of the influence of v e g e t a t i o n . A l t h o u g h we h a v e c o n s i d e r e d only the m e a n a n n u a l runoff, the s e a s o n a l v a r i a t i o n of b o t h p r e c i p i t a t i o n and r u n o f f should be e x a m i n e d f u r t h e r in f u t u r e studies. A l t h o u g h t h e l a r g e h i g h l a t i t u d e rivers in Asia are in good a g r e e m e n t with o b s e r v a t i o n s , t h a t is not t r u e of the M c K e n z i e and Y u k o n Rivers in N o r t h A m e r i c a . T h e c o m p a r i s o n of the p r e c i p i t a t i o n fields in Fig. 1 shows t h a t the model g e n e r a t e s too m u c h p r e c i p i t a t i o n in t h e s e two r i v e r b a s i n s and t h a t the excess model r u n o f f is d i r e c t l y r e l a t e d to the excess p r e c i p i t a t i o n . T h e modeled r u n o f f and p r e c i p i t a t i o n for the Mississippi a n d St. L a w r e n c e Rivers are in good a g r e e m e n t w i t h the o b s e r v e d values. S o u t h A m e r i c a h a s two of the w o r l d ' s five l a r g e s t r i v e r s and the model u n d e r p r e d i c t s the r u n o f f for b o t h the A m a z o n and O r i n o c o Rivers by > 50%. A g a i n it is useful to e x a m i n e the p r e c i p i t a t i o n field in Fig. 1 and T a b l e s 2 and 3. The model p r e c i p i t a t i o n for b o t h basins is a c t u a l l y too large, w h i c h seems to

GLOBAL RIVER RUNOFF

251

AHHUAL RIVER RUHOFF


800 to

700

-~.

(J to ~ ~
>~ ~

i. cW o -c]
~0

(o
600
0 tO

~-

tO (II
",~ 500

~o

400

Cx:
300 L

200

I00

Eli Observed

~ Model (4g )

Fig. 5. Model and observed runoff for other rivers with mean annual runoff > 100 km3year i or drainage basin area > 5 105km2. Observed runoff from Milliman and Meade (1983). be i n c o n s i s t e n t with our low runoff. The problem may lie in the parameterization of evapotranspiration or in the parameterization of groundwater storage. Abramopoulos et al. (1988) have developed a new h y d r o l o g y scheme to use in GCM to better model evapotranspiration and soil water movement. D i c k i n s o n and Henderson-Sellers (1988) and Henderson-Sellers et al. (1988) have s h o w n the importance of using improved parameterizations of micrometeorological processes in the forest c a n o p y for a better simulation of the hydrologic cycle in tropical rainforests. If the groundwater storage capacity is too large, water will remain in the grid box to evaporate back into the atmosphere rather than leave the grid box as runoff. If the capacity were reduced, the model's runoff in the Amazon basin might improve. This could produce increased runoff and also reduce the a m o u n t of water that evaporates from the ground to be recycled as precipitation. Delworth and M a n a b e (1988) have discussed the significance of soil wetness parameterizations, potential evaporation and precipitation on the hydrologic balance and runoff in atmospheric models. In s o u t h e a s t Asia, the model's high runoff in the Mekong, Yangtze and Irrawaddy Rivers is primarily o w i n g to the model's excess rainfall as s h o w n in

252

G.L. RUSSEI,L AND ,JR, MILLER

Fig. lc. The same high runoff and precipitation occur in the model for the Congo and Niger Rivers. These results are generally consistent with our expectations th at overprediction of runoff will occur in regions where the precipitation is overpredicted. However, there are also regions, notably the Amazon and Orinoco Rivers, where the runoff is much too low even though the precipitation is too large. It is essential that the model includes a good parameterization of the groundwater storage and evapotranspiration.
DISCUSSION AND SUMMARY

River runoff is an important component of the global hydrologic cycle. It depends on the precipitation, evaporation and groundwater storage within the river basin. We have shown that global atmospheric models can be used to calculate the mean annual river runoff for the world's largest rivers. For about half of the rivers with runoff > 200km3/year 1, the model runoff is within 20% of the observed value. We will briefly discuss some of the limitations of the present model in predicting river runoff. The comparison of model river runoff whith observations is a good diagnostic for atmospheric models. Since the ratio of observed precipitation to observed river runoff varies from ~ 1.5 to 40, there is considerable variation in the amount of evaporation from different regions of the Earth's land surface. Because of this variability, it is difficult for a climate model to simulate river runoff accurately. In some cases, model river runoff may be accurate because of offsetting errors in precipitation and groundwater storage. One of the most important comparisons is the model precipitation with observations. Figure lc shows locations where the model precipitation is in best agreement with observations. By comparing the location of the river basins in Fig. 2 with the precipitation in Fig lc, or by using Tables 2 and 3, one can find the basins in which the model's precipitation is too large, although for some of these basins the runoff agrees with observations. This indicates the importance of good parameterizations of groundwater storage and evapotranspiration. The runoff in the Amazon and Orinoco basins shows that even when the model precipitation is too large, the runoff can be too low because there is too much storage of water in the ground or too much evaporation. Hence, the primary concerns of atmospheric modelers in predicting runoff is to obtain accurate surface precipitation, evapotranspiration and groundwater storage including the effects of soil type and vegetation. The grid resolution of the model can also affect the runoff. Although we did divide the model's 4 5 grid boxes into four smaller boxes to calculate the river runoff, whenever a particular 2 2.5 grid box was assigned to a river drainage basin, all the runoff in that grid box was assigned to the drainage basin. This assumption is likely to be most critical for some of the southeast Asian rivers where up to three major rivers pass through the same grid box in the Himalayan Mountains. One way to address this problem would be to refine the river drainage basins even more by assigning the appropriate percentage

GLOBALRIVER RUNOFF

253

of a grid box's runoff to the river drainage basin or by using a finer resolution model. Because the runoff in adjacent model grid boxes can vary by an order of magnitude or more, the resolution can significantly affect the total runoff in the drainage basin. There are other problems with the formulation of the model's runoff. The calculated runoff during a time step is assumed to disappear into an infinite ocean. It cannot evaporate during subsequent time steps or interact with adjacent grid boxes. Future model development should allow for such interactions and should incorporate runoff rates that depend on the surface topography. This would have a particularly large effect on monthly runoff since there are lags between river runoff at the mouth of a large river system and the precipitation near the headwaters. The seasonal variability of the model runoff should be compared with observations to determine whether the mean annual results, if predicted correctly, are the result of correctly predicting the seasonal variability. Another reason for differences between model-generated runoff and observations is the quality of the observed data. For several rivers there is a significant difference between the observed runoff given by Milliman and Meade (1983) and that given by Baumgartner and Reichel (1975). We used the more recent reference in our study. Milliman and Meade (1983) also included information on the quality of the data used to obtain the observed runoff in different basins. Overall, the model runoff agrees reasonably well with observations for rivers with runoff greater than 200 km 3year 1, but the model overpredicts the runoff for rivers with less than 200 km 3year '. Some of the inaccuracy occurs because of poor model precipitation fields, but the model's parameterizations of groundwater storage and evapotranspiration are also suspect. New parameterizations, including the effects of both soil type and vegetation, should be investigated. In addition to being a good diagnostic for atmospheric modelers, the runoff that may occur during future climate change can also be studied with these models.
ACKNOWLEDGEMENTS

We would like to thank R. Reudy for helping to prepare the data files for this study. We would also like to thank A. Broccoli, W. Broecker, I. Fung, D. Rind, and C. Rosenzweig for helpful discussions on various aspects of the work. We are particularly grateful to B. Oppenheimer for helping to define the river basins.
REFERENCES Abramopoulos, F., Rosenzweig, C. and Choudhury, B. 1988. Improved ground hydrology calculations for global climate models (GCMs): Soil water movement and evapotranspiration. J. Climate, 1 (9): 921 941. Alexander, R.C. and Mobley, R.L. 1976. Monthly average sea-surface temperatures and ice-pack limits on a 1 global grid. Mon. W e a t h e r Rev., 104:143 148.

254

G.L. RUSSELL AND .1R. MILLER

Baumgartner, A. and Reichel, E., 1975. The World Water Balance. Elsevier, Amsterdam. Chan, S.-O. and Eagleson, P.S., 1980. Water balance studies of the Bahr El Ghazal Swamp. Rep. No. 161, Ralph M. Parsons Lab. Water Resourc. Hydrodyn., MIT, Cambridge, MA. Delworth, T. and Manabe, S., 1988. The influence of potential evaporation on the variabilities of simulated soil wetness and climate. J. Climate, 1 (5): 523 547. Dickinson, R.E., 1984. Modeling evapotranspiration for three-dimensional global climate models. In: Climate Processes and Climate Sensitivity. Geophys. Monogr. 29, Maurice Ewing Ser. AGU. Washington, D.C, 5, pp. 58-72. Dickinson, R.E., Henderson-Sellers, A., 1988. Modelling tropical deforestation: ,% study of GCM land-surface parameterizations. Q.J.R. Meteorol. Soc., 114:439 462. Dickinson, R.E., Henderson-Sellers, A., Kennedy, P.J. and Wilson, M.F., 1986. Biosphere atmosphere transfer scheme (BATS) for the NCAR Community Climate Model. NCAR, Boulder, CO, Tech. Note/TN-275 + STR. Gates, W.L., and Nelson, A.B., 1975. A new (revised) tabulation of the Scripps topography on a 1 global grid. Part I: Terrain heights. Rep. R-1276-1-APRA, Rand Corp., Santa Monica, 132 pp. Hansen, J., Russell, G., Rind, D., Stone, P., Lacis, A., Lebedeff, S., Reudy, R. and Travis, L., 1983. Efficient three-dimensional global models for climate studies: Models I and II. Mon. Weather Rev., 111(4): 609~62. Henderson-Sellers, A., Dickinson, R.E. and Wilson, M.F., 1988. Tropical deforestation: Important processes for climate models. Climatic Change, 13:43 67. Korzoun, V.I., Sokolov, A.A., Budyko, M.I., Voskresensky, K.P., Kalinin, G.P,, Konoplyantsev, A.A., Korotkevich, E.S., Lvovich, M.I. (Editors), 1977. Atlas of World Water Balance. USSR National Committee for the International Hydrological Decade, UNESCO, Paris. Matthews, E., 1983. Global vegetation and land use: New high-resolution data bases for climate studies. J. Climate Apph Meteorol., 22: 474--487. Milliman, J.D. and Meade, R.H., 1983. World-wide delivery of river sediment to the oceans. J. Geol., 91 (1): 1 2l. Robinson, M.K. and Bauer, R.A., 1982. Sea surface temperature climatology. In: S. Auer (Editor), NOAA Oceanographic Monthly Summary, April. Sellers, P.J., Mintz, Y., Sud, Y.C. and Dalcher, A., 1986. A simple biosphere model (SiB) for use within general circulation models. J. Atmos. Sci., 43 (6): 505-531. Shea. D., 1986. Climatological Atlas: 1950 1979, Surface Air Temperature, Precipitation, Sea-Level Pressure and Sea-Surface Temperature (45S-90 N). Nat. Center Atmos. Res./TN-269 ~ STR, Boulder, CO. Times Atlas of the World, 1967. The Times of London, J o h n Bartholomew, Edinburgh. Walsh, J. and Johnson, C., 1979. An analysis of Arctic sea ice fluctuations. J. Phys. Oceanogr. 9: 580 591. Wilson, M.F., Henderson-Sellers, A., Dickinson, R.E. and Kennedy, P.J., 1987. Sensitivitiy of the Biosphere Atmosphere Transfer Scheme (BATS) to the inclusion of variable soil characteristics. J. Climate Apph Meteorol., 26, 341 362.