8
WATER RESOURCES RESEARCH, VOL. 32, NO. 11, PAGES 3367-3374, NOVEMBER 1996 30 $ OriljH-ack's law Riccardo Rigon,1-2 Ignacio Rodriguez-Iturbe,1 Amos Maritan,3 Achille Giacometti,4 David G. Tarboton,5 and Andrea Rinaldo6 Abstract. Hack's law is reviewed, emphasizing its implications for the elongation of river basins as well as its connections with their fractal characteristics. The relation between Hack's law and the internal structure of river basins is investigated experimentally through digital elevation models. It is found that Hack's exponent, elongation, and some relevant fractal characters are closely related. The self-affine character of basin boundaries is shown to be connected to the power law decay of the probability of total contributing areas at any link and to Hack's law. An explanation for Hack's law is derived from scaling arguments. From the results we suggest that a statistical framework referring to the scaling invariance of the entire basin structure should be used in the interpretation of Hack's law. 1. Introduction Hack [1957] demonstrated the applicability of a power func tion relating length and area for streams of the Shenandoah Valley and adjacent mountains in Virginia. He found the equa tion L = 1.440'6 (1) where L is the length of the longest stream in miles from the outlet to the divide and A is the corresponding area in square miles. Hack also corroborated his equation through the mea surements oi Langbein [1947], who had measured L and ,4 for nearly 400 sites in the northeastern United States. Gray [1961] later refined the analysis, finding a relationship L « /I0'568. Many other researchers have corroborated Hack's original study, and, although the exponent in the power law may slightly vary from region to region, it is generally accepted to be slightly below 0.6. Equation (1) rewritten as L t*Ah with h > 0.5 is usually termed "Hack's law." Midler [1973], on the basis of extensive data analysis of several thousand basins, found that the exponent in Hack's equation was not constant but that it changed from 0.6 for basins less than 8,000 square miles (20,720 km2) to 0.5 for basins between 8,000 and 105 square miles (20,720-259,000 km2), and to 0.47 for basins larger than 105 square miles (259,000 km2). As Mesa and Gupta [1987] point out, Midler's empirical observations are not consistent with the implications of the random topology model of channel network structure first in- 'Department of Civil Engineering, Texas A&M University, College Station. Permanently at Dipartimcnto di Ingcgneria Civile e Ambientale, Universita di Trento, Trento, Italy. 3Istituto Nazionalc di Fisica della Materia, Scuola Internazionale Superiore di Studi Avanzati and Istituto Nazionale di Fisica Nucleare, Trieste, Italy. 4Dipartimento di Scienze Ambientali, Universita di Venezia, Ven- ezia, Italy>^^ 5Departmenf of Civil and Environmental Engineering, Utah State University, Logan. "^ 6Istituto di Idraulica "G. Poleni," Universita di Padova, Padova, Italy. Copyright 1996 by the American Geophysical Union. Paper number 96WR02397. 0043-1397/96/96WR-O2397$O9.O0 traduced in the classic paper of Shreve [1966]. In fact, they theoretically derived the value of. Hack's exponent, h, for the random topology model of channel networks as M») = 2 1 /ir+ (irlnY rr - 1/n (2) where n is the basin's magnitude. Equation (2) implies a con tinuously decreasing h(n) with an increasing n. For n = 10,100, and 500 the exponent h(n) is 0.68, 0.530, and 0.513, respectively. When n tends to infinity, h tends to the asymp totic value of 0.5. This result makes clear the importance of the magnitude of the network in the exponent h under the pre mises of the random topology model. Further and more gen eral resultS/Ojt__andom trees can also be found in work by Durret et K [1991}/ The classit^rexplanation for the exponent h being larger than 0.5 was to conjecture that basins have anisotropic shapes and tend to become narrower as they enlarge or elongate. The hypothesis of basin elongation was verified by Ijjasz-Vasquez et al. [1993] under the framework of optimal channel networks (OCNs), which are the result of the search of fluvial systems for a drainage configuration whose total energy expenditure is minimized [Rodriguez-Iturbe et al., 1992a; Rinaldo et al., 1992]. Thus Hack's relationship may result from the competition and minimization of energy in river basins. Mandelbrot [1983] suggested that an exponent larger than 0.5 in L « Ah could arise from the fractal characters of river channels which cause the measured length to vary with the spatial scale of the object. Thus (1) would be a reflection of a fractal dimension of river channels close to d, = 2 X 0.6 =1.2 [e.g., Feder, 1988]. However, as explained in detail by Feder [1988], to derive the fractal dimension of a river from the previous argument, one needs to assume that the shapes of the river basins are self-similar. This assumption does not seem to be verified in nature [Ijjasz-Vasquez et al. 1994; Nikora and Sapozhnikov, 1993]. Peckham [1995] derived a relationship between Hack's ex ponent h and a fractal dimension, say DSST, of a self-similar tree as h = 1/DSst- In tne caSG of a space-filling tree, DSST = 2 and h = 0.5, while for DSST ^ 2 Hack's exponent h can be higher. Similar arguments connecting Hack's exponent to the global fractal characters of the network have been proposed by Nikora and Sapozhnikov [1993]. One goal of this paper is to suggest a new statistical uiter- 3367

OriljH-ack's law - The University of Vermontpdodds/files/papers/others/1996/rigon1996a.pdf · OriljH-ack's law Riccardo Rigon,1-2 Ignacio Rodriguez-Iturbe,1 Amos Maritan,3 ... surements

  • Upload
    doandat

  • View
    221

  • Download
    0

Embed Size (px)

Citation preview

Page 1: OriljH-ack's law - The University of Vermontpdodds/files/papers/others/1996/rigon1996a.pdf · OriljH-ack's law Riccardo Rigon,1-2 Ignacio Rodriguez-Iturbe,1 Amos Maritan,3 ... surements

WATER RESOURCES RESEARCH, VOL. 32, NO. 11, PAGES 3367-3374, NOVEMBER 1996

30$

OriljH-ack's lawRiccardo Rigon,1-2 Ignacio Rodriguez-Iturbe,1 Amos Maritan,3Achille Giacometti,4 David G. Tarboton,5 and Andrea Rinaldo6

Abstract. Hack's law is reviewed, emphasizing its implications for the elongation of riverbasins as well as its connections with their fractal characteristics. The relation betweenHack's law and the internal structure of river basins is investigated experimentally throughdigital elevation models. It is found that Hack's exponent, elongation, and some relevantfractal characters are closely related. The self-affine character of basin boundaries isshown to be connected to the power law decay of the probability of total contributingareas at any link and to Hack's law. An explanation for Hack's law is derived from scalingarguments. From the results we suggest that a statistical framework referring to the scalinginvariance of the entire basin structure should be used in the interpretation of Hack's law.

1. IntroductionHack [1957] demonstrated the applicability of a power func

tion relating length and area for streams of the ShenandoahValley and adjacent mountains in Virginia. He found the equation

L = 1 . 4 4 0 ' 6 ( 1 )

where L is the length of the longest stream in miles from theoutlet to the divide and A is the corresponding area in squaremiles. Hack also corroborated his equation through the measurements oi Langbein [1947], who had measured L and ,4 fornearly 400 sites in the northeastern United States. Gray [1961]later refined the analysis, finding a relationship L « /I0'568.Many other researchers have corroborated Hack's originalstudy, and, although the exponent in the power law may slightlyvary from region to region, it is generally accepted to beslightly below 0.6. Equation (1) rewritten as L t*Ah with h >0.5 is usually termed "Hack's law."

Midler [1973], on the basis of extensive data analysis ofseveral thousand basins, found that the exponent in Hack'sequation was not constant but that it changed from 0.6 forbasins less than 8,000 square miles (20,720 km2) to 0.5 forbasins between 8,000 and 105 square miles (20,720-259,000km2), and to 0.47 for basins larger than 105 square miles(259,000 km2).

As Mesa and Gupta [1987] point out, Midler's empiricalobservations are not consistent with the implications of therandom topology model of channel network structure first in-

'Department of Civil Engineering, Texas A&M University, CollegeStation.Permanently at Dipartimcnto di Ingcgneria Civile e Ambientale,Universita di Trento, Trento, Italy.3Istituto Nazionalc di Fisica della Materia, Scuola Internazionale

Superiore di Studi Avanzati and Istituto Nazionale di Fisica Nucleare,Trieste, Italy.

4Dipartimento di Scienze Ambientali, Universita di Venezia, Ven-ezia, Italy>^^

5Departmenf of Civil and Environmental Engineering, Utah StateUniversity, Logan. "^6Istituto di Idraulica "G. Poleni," Universita di Padova, Padova,Italy.Copyright 1996 by the American Geophysical Union.Paper number 96WR02397.0043-1397/96/96WR-O2397$O9.O0

traduced in the classic paper of Shreve [1966]. In fact, theytheoretically derived the value of. Hack's exponent, h, for therandom topology model of channel networks as

M») = 21 /ir+ (irlnY

rr - 1/n (2)

where n is the basin's magnitude. Equation (2) implies a continuously decreasing h(n) with an increasing n. For n =10,100, and 500 the exponent h(n) is 0.68, 0.530, and 0.513,respectively. When n tends to infinity, h tends to the asymptotic value of 0.5. This result makes clear the importance of themagnitude of the network in the exponent h under the premises of the random topology model. Further and more general resultS/Ojt__andom trees can also be found in work byDurret et K [1991}/

The classit^rexplanation for the exponent h being largerthan 0.5 was to conjecture that basins have anisotropic shapesand tend to become narrower as they enlarge or elongate. Thehypothesis of basin elongation was verified by Ijjasz-Vasquez etal. [1993] under the framework of optimal channel networks(OCNs), which are the result of the search of fluvial systemsfor a drainage configuration whose total energy expenditure isminimized [Rodriguez-Iturbe et al., 1992a; Rinaldo et al., 1992].Thus Hack's relationship may result from the competition andminimization of energy in river basins.

Mandelbrot [1983] suggested that an exponent larger than0.5 in L « Ah could arise from the fractal characters of riverchannels which cause the measured length to vary with thespatial scale of the object. Thus (1) would be a reflection of afractal dimension of river channels close to d, = 2 X 0.6 =1.2[e.g., Feder, 1988]. However, as explained in detail by Feder[1988], to derive the fractal dimension of a river from theprevious argument, one needs to assume that the shapes of theriver basins are self-similar. This assumption does not seem tobe verified in nature [Ijjasz-Vasquez et al. 1994; Nikora andSapozhnikov, 1993].

Peckham [1995] derived a relationship between Hack's exponent h and a fractal dimension, say DSST, of a self-similartree as h = 1/DSst- In tne caSG of a space-filling tree, DSST =2 and h = 0.5, while for DSST ̂ 2 Hack's exponent h can behigher. Similar arguments connecting Hack's exponent to theglobal fractal characters of the network have been proposed byNikora and Sapozhnikov [1993].

One goal of this paper is to suggest a new statistical uiter-

3367

Page 2: OriljH-ack's law - The University of Vermontpdodds/files/papers/others/1996/rigon1996a.pdf · OriljH-ack's law Riccardo Rigon,1-2 Ignacio Rodriguez-Iturbe,1 Amos Maritan,3 ... surements

vrro

josanjBA sjmbsj 90 puB _c;o ussMjsq sjusuodxs s.jjobh snqx•ty pue 7</> ussMjsq sjuibjjsuoo sssoduii (^j) jsqj ssAissqo auo

,-(8) ssqduji 'puBq jsqjo sqj uo 'XjuBiyuis-jps -''<p juauodxaSuqEos aqj jo uoyonpojuyXs^ aqj joj [_66l 'AO^uifzodDg puowoyiN'1661 '7D(^"^5^*tJ§uojjBiaj snoiASjd uiojj sjsjjip(p\) JBqj aaijoM >-d> ='p = ^tp _\_q _m'4[966I] 7» 13 UDjuofujo uoijbjou aqj oj\aauajajaj qji^f ty 'juauodxa s.3p___ pu_ '7<J>'surea-js jo Xjjsonuj^rBjaBjj aqj '// 'XjiuyjB-jps ssjEpj qoiqM

t o i l \ - { y r < ? ) = H

[9661 l'l° 13 uvniDfl] aABq aM snqx

2-„j7 *> (»7)» _.

sppiX (9) qjiM psuiqiuoo qoiqM

a;.(€1)

(Zl)

•adBqs aqj uo spuadap „ puB x ajnSij ut UMoqs sspos qjpiMpuB qjSusj jo sajnsBaiu sjBudojddB aqj ajB "7 pus T 7 ajaqM

( £ ) l l 7 T 7 5 " = V

Xq pszusjoBJBqo aq Xbui suissq jbau jo uoipafojd jBusjd aqx•psuysp aq oj spaau jsjy aXjisonuis pj

objj,, pus ..uoijBSuop,, suusj aqj jo Suiuesui sqx -sujseq jsaijjo uo;jB§uop aqj puB 'spuuEqo uiBajjs jo Xjisonuis jbjobjj aqj'a\b[ s(3|____ ussMjsq uoposuuoo aqj sjspisuoo uoposs siqx

;,uoiji.8uo|3[ Ajduii AiBq s_5pen S3oq

«wT

suiBjqo auo '(p) m (\) Suisfl 'uoncS-uop SuiXiduii 'sssesjoui "7 sb sssbsjosp (11j)d '1 > // joj

( I I ) , - . . 7 * > , 7 / T 7 - = ( 8 7 ) *

ssuioosq (i{j)n pun

( O l ) i + l l 7 s = V

uaqx -"7 > T7 uoijruysp Xq ssnsosq '. > 77 ajaqM

( 6 ) , 5 7 « T 7

:[£66I'Aoyuzoifdos Plw WORN -T66I '/» *w t/«.w/v "3-s] // juau-odxa jsanH aqj qSnojqj psypuBnb aq Xbui ssdEqs paqsjajBMjo XjiuyjB-jps "auyjB-jps ajB suiSBq puB sjejoejj 3jb suiBajjsJBqj si mb[ s(3p_H qjJM aiqijBduioo ssbo pjaus^S* ajoui aqx

( 8 ) H Z = 7 4 >

saqduii qoiqM '_/7<a,K » 7 snqj puB

( _ ) . _ > s n o - ~ 7 » / z 7 / K = J 7 / K

s_q auo '(jubjsuoo ~ ("7)0) adBqs ui jByims-jps aq oj paumssBajB suissq jbAjj jj -juauodxa 3urr_os [bjobjj b si ''tp ajaqM

(9) !7x 7sb Smpos si '7 lqj3uo| uiBajjs

ajojajaqj puB sjbjobjj ojb spuuEqo uiBajjs jBqj sf [_86l] fojq-ppuoffl Xq pajsa33ns mbj s^obj-j jo uoijBjajdjajui jaqjouy•K3uiSBajoui qjiM sjE§uop suiSBq 'mbj s^obh oj Suipjooos jBqj3uiX|dun 'sasBajoui y sb sasBajoap ("7)0 jsqj sjsaSSns siqx

(.) ro-K~K_,K» (U7)»

\ • _ . _ " ( ) = HJoj suiBjqo auo '(p) ui (j) 3uisn uaqx 'Jubjsuoo supiuaj s _iiq„\<B7 sb sapjos *7.'apiAip aqj oj jajjno aqj uiojj sjauuEqo 3uo[BpajnsBaui qj3uaj atjJTEqj si mbj s.jjoeh jo uoijEjajdiajui auo

•XjiurjjB-j|as saqduii y qjiM SuisBajo-ap ("7)0 ajiqM 'XjiJEjiuiis-jias saqduii uaqj ("7)11 jubjsuoo

t o ) ' y * " 7 / T 7 » | 7 / K = ( y 7 ) »

auyap mou 3m jubjsuooS M'lAA ' V m!M 3JB3UOJ3 pUB 3UIJJE-J|3S 3JB suissq 'JUBJSUOO3ui3q _ qjiM 'S3SB3J0UI ̂ SB S0SB3J33P "7/T7JI 'X|3AIJBUJ3J[V•jubjsuoo b si 's 'suiSBq sqj jo adsqs aqj ji pus 'jv 'se3je jje jojJubjsuoo si H7/T7 oijbj aqj ji jBiiuiis-jps parreo sq \\m ssdsqs

•UQijBnbs s(jjobh 3uijb3ijssauiusqM sjoqM b sb jy oj sjBnbspB jsoiu~squQu si_bsjb uiSBq uispnjiuSBiu jo sjspjo jo UBds s3je| b sousq puB 'uoijbisj s4>p&j-jui g-Q jo jusuodxs ub qjiM psjjy sq X^buosbsj ubo ' uuj Qtojdn zW OOI UJOJJ 'JSS BJBp sjisodiuoo B JBqj SMoqs BJBp JO UOIJ-osyoo [266ll ^lpupta Pu*> fowoSiuow 'sjouusqjjnj suisbq^jsau jo sjsjoBJBqo jusjjnosj psAissqo sqj 3uippiX 'sjoj jueu-luiopsjd b sXBjd uoijbziubSjo-j|ss 'ss[bos iieuis oj umipsui jbjBqj pbsjsui josdxs saa -ajEuiuiop sjojjuoo oi3o[oa3 s_j_os s3jbjXjsa jy -pajajjB aJB sdjqsuoijBjsj qj3usj-BSJB puB 'suuojpuB[SAB3UO0 JOJ SiqiSUOdsSJ SSSSSOOjd SAISOJS SAjJBJJUSQUOO qjiMjobjsjui sssssoojd OAisnjjip 'sspos jsmoj jv fsoioqo siqj Ajijsnfsuoijbaijoui [Bjusuiusdxs puB jBoijsjoaqx' '.uijj 000Z~0S i°sSubj sqj ui si uoisusjxs ssoqM suissq jo sjnjonjjs jbujojui sqjuo snooj sm 'XyBoypsds 'aouapiAa jBouiduia oqj jo uoijBjajd

•'7 ajBsopis jsojjoqs aqx -uissq aqj ui juiod jsaqjjBj oqj oj japno aqjuiojj auq jq3iBJjs aqj sb pauyap '"7 jajauiBip aqj oj pyBjEdajB ijjoMjau aqj Suisopua s^ubjosj sqj jo sspis jss3uo| sqjuissqqns Xub joj -UMBjp osp sjb suiSBqqns siuos •T 7 'qjpiMsji puE J*7 'jsjsuiBip sji !uisBq jsau e jo qojs>is 1 ajnSij

AAVT S.-M3VH NO "TV 13 NODI _ 89ee

Page 3: OriljH-ack's law - The University of Vermontpdodds/files/papers/others/1996/rigon1996a.pdf · OriljH-ack's law Riccardo Rigon,1-2 Ignacio Rodriguez-Iturbe,1 Amos Maritan,3 ... surements

RIGON ET AL.: ON HACK'S LAW 3369

<f,t smaller than 1.14 and 1.2, respectively, for elongation tooccur. Higher values of <p,_ will imply contraction of riverbasins when increasing with size.

The relationships derived above were first tested in riverchannel networks extracted from digital elevations models(DEMs) using standard procedures [e.g., Jenson andDomingue, 1988; Tarboton et al., 1991]. The basins used in theanalysis are given in Table 1.

Figure 2a shows the result of Hack's analysis. While the bestestimate of the slope is close to the commonly found value of0.6, wejoticejhat the detailed study of the morphology of thefluvialb^sirisaUowed_byJDE_^^ the analysis- of thestatistjjal^flucluaJlQJ^-jiejent in Hack's type~of diagrams.These fluctuations were neither considered in the originalwork by Hack nor in the experimental studies following it.From Figure 2b we have

L.cc.4' (15)which gives a Hurst exponent H = 0.92. The relationshipbetween L and Ln is presented in Figure 2c:

L « Ll (16)

implying a fractal scaling exponent for channel sinuosity ofcbL = 1.15. With cbL = 1.15 and H = 0.92 in (14), one obtainsh = 0.60 that matches the value obtained by direct fitting inFigure 2a.

Thus assuming self-similar shapes and neglecting elongationwould have resulted in h = <pJ2 = 0.575, and taking streamsas nonfractal would have resulted instead in h = 1/(0.92 + 1)= 0.52. In both cases one underestimates the value of h. Themain contribution to the exponent h comes from fractal sinuosity, with a smaller contribution coming from elongation.

The above analysis gives standard results and also clarifiesthe conditions which define geometrical elongation. Nevertheless, the requirement that the shape coefficient s is a constantis too restrictive to be fulfilled exactly. To account for the

Table 1. Geometrical Data Relative to the River BasinsAnalyzed

River Basin LocationA,km2

L,km km

GuyandottcLittle CoalTug Dry ForkJohnsBig CoalRacoonPingeonMoshannonBrushyRock CastleSturgeonIslandWolfSextonBigIndianB i g ^Schoarie Creek HeadwaterPoundTipton Run.Blackberry Fork

W.Va.W.Va.W.Va.Ky.W.Va.Pa.W.Va.Pa.Ala.Ky.Ky.W.Va.Ky.Ky.Ky.W.Va.IdahoN . Y.Ky.Pa.

20889845864844494464053933253102952602121861521481461081056453

145.190.563.768.856.253.649.149.752.445.946.330.530.333.330.241.021.518.523.518.818.0

75.857.541.645.740.735.235.333.829.933.527.123.021.723.421.128.316.915.817.812.513.3

Data were obtained by processing U.S." Geological Survey 7.5"D E M s . \ ° *

TlrVy .

150. '

— 100. > T

E^ 7 0 . a ^ * r_:{? 50. Jr*""* ._ i a ^ / " ^ y .

% 30. a > " ^ * * y > ^ h = 0.6g>55 20. • J S ^ • 1

■15.

I

5 0 . 1 0 0 . 2 0 0 . 5 0 0 . 1 0 0 0 . 2 0 0 0 .Area [km2 ]

u_.

70. / •

50. •

30.

20.

15.

• y r h' = 0.52

1

5 0 . 1 0 0 . 2 0 0 . 5 0 0 . 1 0 0 0 . 2 0 0 0 .

Area [km ]

150. - • / ■

_f 100.2£ 7 0 . y Co>

^ T •

__ 50. . • * * • * j <

E 1.15j / f_: 30.

CO 1

20.• j / .

15. 2 0 . 3 0 . 5 0 .

Diameter [km]

70. 100.

Figure 2. (a) Hack's law for the basins in Table 1. (b) Elongation for the same set of data as in Figure 2a. An exponentgreater than 0.5 means that basins elongate, (c) Stream lengthversus diameter for the basins in Table 1.

natural variability of river network structures, it is necessary torelax the deterministic definitions in favor of statistical ones.This will be pursued in the following sections, together with ageneralization of Hack's law that preserves the validity of (14).

3. Hack's Law and Scaling ArgumentsHack's original work dealt with nonoverlapping basins. In

this section we will start with the assumption that a Hack-typerelationship is valid for the different subbasins embedded in alarger basin. Thus a statistical framework is developed for the

Page 4: OriljH-ack's law - The University of Vermontpdodds/files/papers/others/1996/rigon1996a.pdf · OriljH-ack's law Riccardo Rigon,1-2 Ignacio Rodriguez-Iturbe,1 Amos Maritan,3 ... surements

aiuoojno aqj Pub '(5 - z)y «(w) m juaiuoui jsjg aqj jo juau-odxa aqj jo snrBA aqj uaaMjaq aouajajjip jusjsisuoo b SMoqssuissq aqj jo ouom -Z _iqu__ jo uumjoo pjiqj aqj ui payodajajB 1 ajqBx jo suissq jssSjbj aqj joj uoijnqujsip qjSusi aqjjo sjuauioui jsiij -sjjnsaj snoSojBUB qjiM t a|qBj_ ui payodajsuiseq jaqjo pajsaj os[B aABq a_\v 'XqdBjSodoj jo sdXj ado__n*iqb oj rBiAng b uiojj uoijisubjj aqj oj anp si sajsos irems Xisaaqj jb auq jqSiBJjs aqj uiojj uoijBiAafj -\q-q + 090 Pub '100+ WO 'lO'O + 9S'0 3Jb sadojs asoqM sSjsuis s__.11 jqSiBJjs jbojojsqj aAiasqo SM '_; puB lp '_ ssjnSiaj u; UMoqs bjb asaqx -j_ai_{suqof aqj puB jsai^j .fjoj Xj<_t 3nj_ aqj 'jaAry sjjopuBXnQ aqjJ0. ((93) uoijBnba '-g-a) b3jb snsjaA sjuauioui SAijnossuoo saijjsjq aqj jo oijbj aqj puB juauioui jsjg aqj ssjbos oiinqjuBSojs.qnop ui -uijjojd Xq >uomsujbjj aAoqB aqj psjssj aABq Syw

(93)

(SZ)

.» » X-uX)l"(ui;)

<i+_-_)¥0 » "<-£>

snqj puB

Aq usaiS si b B3JB Suijnqujuoo uoiuraoo b qj?M sqjSusjjo uonnqijjsip aqj jo juauioui Xub juomsujbjj siqj japunjBqj aapoN •„_• jojobj b Xq pspjossj 3jb sssqj usqM sqjSusijo uoijnqujsip sqj ui XjiugjE-jps oisBq b spsquis ji sb XjqsjoBJjssjBinjsod Xrjpqdxs (__) uoijEnbg -(33) jo snpA ubsui sqj ojSuijjsjsj JsjoBJBqo iboijsijbjs jpqdxs UB SBq mou ji "x = 5 uaqMtoz) uiojj pajaAooaj si diqsuoijEpj purSuo aqj uaqj 'uiSBq b uipappaqraa suisEqqns Xub joj pjiba pajapisuoo si mbj s.jjobh JI

toz) G-Z)HD ~ °<£> O S (%■) & f̂ o c->.aABq uaqj ____ "Z >:?

asnBoaq juBAapjji si uoijBjSajui aqj jo jiuni jaMoj aqj ajsqM

(hXq uaAiSsi v B3JB Suijnqujuoo uouiuioo b qjiM % jo anjBA ubsui aqx

(33)

(13)

( h O / X W - Z = ( » ' £ )■« •

samooaq (8l) snqx„_■ ~ (vn

Xaqo qiM asn aM qj3uaj oijsuaj-OBJBqo JSASjBqM 'uiSBq b jnoqSnojqj pjjba si diqsuoijEpj adXj-jfOBH b jBqj Smuinssv *<» *- (o)3] usqM 3_£ ~ (a, '£>_. 'siJBqj 'jiunj azis sjiuqui sqj ui (o '_^)j. joj JOiABqsq mbi J3Mod bS"3*!3 (03) 3im^ '(tl) ui ibjSsjui sqj jo Xjiuqui jb sousSjsauoosqj ssjnsua '(51) 'uoijipuoo jsju aqx -[9651 <•/_•/_> __irju__jv !t66Il-lvj3uppjdjA^-2-3] JoiABqaq Smpos SuizXjbub uaqM uoijdumssBuounuooB si)(8i) uqiiBnbg -jubjsuoo ajqBjms b si 0 ajaqM

. O( 0 3 ) \ \ o = ( x ) 6 r x <

:saijjadoj\l SuiMoqoj aqj saysijBSXub jo azis ajiuq aqj joj sjunoooB jBqj uoijounj

3uqBos b ^ (r)_J puB qjSuai oijsuajOBJEqo b si (o)"/ ajaqM

[(»yi /X]6(5X/ l ) = ( t> '£\:adXj ?uiMoqoj aqj jo uuoj 3ui|bos b Xaqo oj patunssB aq

11!^ (» '^)V uoijnqujsip aqx -rp(o «_:)j. »/ = (v (gj)rj 'ajd

-uiBxa joj 'aouapaaoxa jo saijqiqBqojd Suipuodsajjoo aqj (^)jjpuB (v '^)jj qjiM ajouap jaqjjnj rreqs a^ 'SBajB Suijnqujuooaqj jo (jpd) uoijounj Xjisuap XjijiqBqojd aqj si (o)d ajoqM

(LI) (D)d(D <X)lLVp =(%)U.PXq uoaiS uaqj si uoijnqujsip qjSuaj aqx -v BajB Suijnquj

-uoo uaAiS b qjiM sjuiod ujojj % 'sqjSusj uibsjjs sqj jo uoij-nqujsip XjqiqBqojd rEUoijipuoo sqj (v '^)jz. Xq mou sjousq*7 = 5T SBM 9uo ' y = v ajsqM 'uissq sqj jo sjnsojo sqj jb 'puBX sb qj§usi uibsjjsuibui sqj sjousp sm lv bsjb jo suiSBqqnsjoj jBqj sziSBqduis sm 'jsdBd snjj jnoqSnojqj psjdopB uoijbjousqj oj sjbisj ox -soioqo uiopuBJ b ssjjbui 'Xjbsssosu ji 'puBqjBd jssSuoj 3M1 sssooqo suo uoijounf b jb sbsjb SuijnqujuooIBnbs jo ssbo sqj u_ psqoBSj si sojnos b jijun bsjb SuijnqujuoojsSjbi b SuiABq sjis sqj SuTMOnoj uoijounf Xub jo uiBSjjsdn qjBduibsjjsuibui sqj saugap auo 'XqBoiuqoax "UiSBq aqj jo XiBpunoqaqj oj juiod aqj uiojj 5uoMjau aqj qSnojqj pajnsBaui aouBjsipuibui aqj sb pauqap si 'gj Xbs 'juiod Xub jb qjSuaj uibbjjs aqx

's»AI3a uiojj BjBp qSnojqj pajBpqBA ajB suoijBnbaSutjinsaj aqj puB diqsuoijBjaj b qons jo sjuauioui juajajjip

^ 0 \ a \•Xija-uioaS sdoisqiq oj [BiAnij ujojj uoijisubjj aqj oj SuiVo uajjojq

Si Suipos aqj sapos jreuis XiaA aqj jy -sspBosp 3 ^-sb^i jb jojSuqBos sjduiis psuqsp qsM e SMoqs joid sqx -pJBMUMOp psjjiqs'Xl3AlJ03dsSJ 'C pUE 'V <£ % m U JOJ „_• x »<. _„3j)/"<_^)oijbj sqj Moqs ssAino jnoj SuiMoqoj sqj tjusiuoui jsjy 3qjSMoqs SAino doj oqx JSATy sjjopuBXnQ joj bsjb qjiM sqjSusjuibsjjs jo sjusuiouj sqj jo Suqsos sqj jo sisXjbuv *e 3-nSij

_ui>_] vaiV0001

lOOO'O

1000

10*0AH

AH

5

AAVT S.MDVH NO "TV 1H NOOIH

O fef\ *y1oz.ee

Page 5: OriljH-ack's law - The University of Vermontpdodds/files/papers/others/1996/rigon1996a.pdf · OriljH-ack's law Riccardo Rigon,1-2 Ignacio Rodriguez-Iturbe,1 Amos Maritan,3 ... surements

'-TO < H 3J3qM 'z a.qsx ui sjuauiajnsBaui aqj qjiM puB[_66T] «oyuHzodvS P™ t»oyN Xq payodaj sjuauiajnsBaui aqj g ajnSy jo uo.jdBO aae uaAm _-yoai Sn t joi EajP mTlV°}qjiM juatuaajSB ui 'og'O P™ PL'O uaaMjaq '// 'juauodxa jsjn__ uiBaiislo S,oL-a,n TElA^l??* ?** *****paugap-ipM b psq [^661] :jv id zdnbsDA-zsvrfj Xq paipnjs saus-punoq msBq aqx -jpsji «£• Pub 'T7 oj puoiyodojd si qoiqMArepunoq aqj oj jajauisip aqj uiojj XqBUoSoqjjo pajnsBamaouBjsnp umuipcBui aqj uaaMjaq (6) oj jBnuns uoijBpj Suiposb Xq pazuajoBJBqo si srpij pjobjj omj aqj jo XjnigjB-jps aqx•uiSuo uourmoo b uiojj Suijjbjs spjobjj auujB-jps omj jo aunjuoisyjoo jsjg aqj jo XSopuB aqj sb papjBSaj aq ubo puB "^ qjiMsappuioo juiod SuiSjaui aqj Pub jajjno aqj Suiuiof auq jqSiBJjsar_L 'J! uiojj juiod jsaqjjBj aqj ui SuiSjaui pu*B japno aqjuiojj SurjjBjs spjobjj auujB-jps omj jo pasoduioo si uissq aqjjo Xrepunoq aqx -sjoafqo auujB-jps ajB qjoq jBqj papnpuooXaqx -suiSBq jbau jo sauBpunoq aqj pus puuBqo uibui aqjjo saiuadoid Suubos aqj paipnjs [p^\] 70 id zmbsvA-zsDff/

"SWHa moiJ sjuauiajnsBauiqjiM juaraaaiSB pooS XpAijBpi rn si siqj *_ a.qsx ui UMoqs sy

(oe) 1 = V + E.

saqduii(63) snqx *(l + H)/7<P = H Jeqj osp aAiasqo 3m (p\) uiojj

(63 ) ( I + / / ) / (7 t f> - I + H) = _ /

sb passaidxaaq ubo g/ juauodxa aqj jBqj paMoqs [9551] 70 ip ifoftwjo'Z m*± ui payodaj a_B ..iujj Q03 UBqj jaSjBj bjb ssajE asoqMI aiq«_L ui suisEq aqj joj cj j0 sanpA sqx -sjios pus 'uoijbj-sSsa 'ajBuiqo 'suopipuoo oiSoioaS jusjsjjm Xisa jo suissq jojgp'O pus etrO ussMjsq aSuBJ aqj uiXpuanbajj jsoui si _* ajaqM

uiBajjs jo sjuauioiuaqj jo Smpos sqj jo sjsXpuv > aingy

ro0001 01

100

10000

1000

AH

AH

. . - S J I B J S pjoj £ sinSi_r jo uoijdBo sss -jsai^ suqof joj bsjb qjiM sqjSuaiUJE3JJS jo SJU3UIOUI sqj jo Sujpos sqj jo sisXpuy *'S -JnSitf

(83) _•_» x [_■< y]j

[_uiM] vajy0001

10000

1000

10*0

AH

r o vAH

mbj jsMod sqj SMoqoj jjJOMjsu sSbuibjp b ui juiodussoqo XjiuopuBJ b oj bsjb Suijnqujuoo sqj joj uoijnqujsipXjqiqBqojd sqj jBqj punoj [q366l] _» 19 dqmn-zanSupo^

•sqjSusj uibsjjs jo uoijnqujsip sqj puB sbsjb Sui-jnqujuoo jo uoijnqujsip sqj qSnojqj suiSBq jsau jo sjnjomjsPjobjj sqj oj psjBpj mou si SAoqs sb pszipjsusS mbi s.xobh

sinseg J3ai^ joo.injDru}§ IBJDBJj aqj pue A_eq st_pBH f

•psjBUiijssjspun si [966I] 70 ;. umuvflXq payodaj //jo sSubj sqj snqj Pub 'X Mopq sjb ssnpA sqj jojsojai *(j>l) ui sb //jo uoijBnpAS sqj ssaiS uuinpo qjjq sqj Pub'SAoqB psquossp sb psjBnpAS // jusuodxs jsjn_-j sqj jo anp\sqj sjjodsj 3 siqBx jo uuinioo qynoj sqx ■XjiuyjB-jps ssqduitI > // sjsqM 1 - tu/i = // sonpojjui sm (qi) oj snoSopuBXbm b ut -juspujsoo ousuinu b si q Pub g-Q < ,_/ 3j3qM

(LZ) mV(I = °^)sb bsjb Supnqujuoo qjiM sspos jusuiouj jsjg sji 'jBjnoiUBd

UI 'X aiqBUBA uiopuBJ aqj jo auo aqj qjiM XSopuB ajajduiODui pazXpuB XqBuuoj aq ubo uoijnqujsip sj_ -uiSBq aqj apisutjuiod uasoqo XiuiopuBJ b si japno asoqM uisBqqns b jo jajauiBipaqj sjuasaidaj jBqj aiqBUBA uiopuBJ aqj sb % auqap mom

•soijsijbjs jsnqoj Xioa qjiMpazXpuB ajaM suiSBq SuiddBysAOuou siqBjms qoiqM ui ssuopuiSuo sqj oj ssop ssnpA sppiX sjusuinSiB Suipos qSnoiqjpsuysp jusuodxs s.jpBj-i jo jusuisssssb SAoqB sqj snqiT oj pnbs si 5 jBqj SuijssSSns '(93) ui ' y 'soijbj jusuioui aqj jo

xz.ee AAVT S.-SDVH NO "TV 13 NOOIH

'

Page 6: OriljH-ack's law - The University of Vermontpdodds/files/papers/others/1996/rigon1996a.pdf · OriljH-ack's law Riccardo Rigon,1-2 Ignacio Rodriguez-Iturbe,1 Amos Maritan,3 ... surements

3372 RIGON ET AL.: ON HACK'S LAW

Table 2. Values Relative to the Rivers in Table 1 WhoseArea Is Larger Than 200 km2River Basin <t>L h H (*_/*) - B y 0//i

Guyandotte 1.06 0.56 0.92 0.89 0.43 0.74 0.77Little Coal 1.06 0.56 0.92 0.89 0.41 0.74 0.73Tug Dry Fork 1.07 0.54 1.00 0.98 0.41 0.77 0.76Johns 1.06 0.60 0.75 0.77 0.40 0.67 0.67Big Coal 1.05 0.56 0.89 0.88 0.42 0.75 0.75Racoon 1.02 0.52 1.00 0.96 0.45 0.89 0.86Pingeon 1.06 0.55 0.92 0.93 0.45 0.75 0.82Moshannon 1.02 0.52 0.96 0.96 0.45 0.86 0.85Brushy 1.04 0.54 0.96 0.93 0.43 0.82 0.80Rock Castle 1.06 0.55 0.92 0.93 0.43 0.79 0.78Sturgeon 1.03 0.52 1.01 0.98 0.46 0.92 0.88Island 1.07 0.54 0.96 0.98 0.43 0.78 0.80Wolf 1.07 0.55 0.92 0.96 0.41 0.75 0.75

P[X > /] « p[Ah > /] = P[A > /'"•] (31)

Here <£_, sinuosity exponent; h, Hack's exponent; H, Hurst's exponent; B, power law exponent for areas; y, power law exponent forlengths. Data were obtained after extracting the drainage networksfrom DEMs using the threshold A, = 50 pixels as the maximumcontributing area necessary to maintain a channel. The exponent h wasestimated by fitting (_?)_. versus the contributing area in a log-log plot.The value of H was obtained from (__,)_ ~ _<I///>+1 where (_?„>_ isthe expected value of diameters with the contributing area, a. Incolumn 5, (14) is used to estimate H from the values of cf>L and hwhose values are reported in columns 2 and 3, respectively. In column8, y is estimated according to (31) being B and h as given in columns6 and 3, respectively.

This argument, (29), and (30) clearly show that the self-affine characteristics (e.g., Hurst exponent) of the basin boundaries are intimately related to Hack's exponent and the distribution law of contributing areas.

Now consider the random variable defined as the distance,i_, measured through the network from a randomly chosenpoint to the boundary of the basin. The exceedence probabilitydistribution of this distance may be derived from Hack's lawand the power law distribution of contributing areas.

Calling _6 the random stream length to the divide, we have

where /3 is defined in (28) and h is Hack's exponent, assumedhere to hold also for the internal structure of a basin. Thisholds for £ = 1 in (22). The power law of lengths defined in(31) has been verified in numerous basins with excellent fittingthroughout several logarithmic scales. It links the aggregationpattern of areas with the stream length structure of the network and Hack's law.

Table 2 shows a synthesis of the experimental analyses conducted on the DEMs of the basins in Table 1 with area largerthan 200 km2. The exponent y falls between 0.70 and 0.90 forvalues of /3 between 0.41_to 0.46 and h between 0.52 and 0.60with an average of 0.55. Tne average exponents of the probability distributions are 0.79/ for the lengths and 0.43 for totalcontributing areas.

Figures 6 and 7 show.examples of P[X £_ /] for the riverbasins whose DEM features have been described in Table 1. Inparticular, Figure 6 shows the detailed results obtained for theBrushy Creek basin where w_\have used a support area of 50pixels to define the drainage networks and the lengths aremeasured in kilometers. In this case, the exponent of the powerlaw of lengths is 0.8, the value of /3 is 0.45, and the value of hresulting from (31) is 0.56. Figure 7 shows the same results forthe Little Coal River basin.

. o4

500.1000.

5 1 0 . 2 0 . 5 0 . 1 0 0 .

Stream Length [km]

Figure 6. (a) Brushy Creek with support area 50; (b) P[A > a ] versus a (0 ~ 0.45); (c) P[S£ > L] versus L(y-0.82).

Page 7: OriljH-ack's law - The University of Vermontpdodds/files/papers/others/1996/rigon1996a.pdf · OriljH-ack's law Riccardo Rigon,1-2 Ignacio Rodriguez-Iturbe,1 Amos Maritan,3 ... surements

RIGON ET AL.: ON HACK'S LAW 3373

1000.

0 . 5 1 5 1 0 .

Stream Length [km]

50. 100.

Figure 7. (a) Little Coal River with support area 50; (b) P[A > a] versus a (0 0-41); (c) />[__ > L]versus L (y ~ 0.74).

The consistency of the measured exponents suggests that thescaling theory developed in this paper is a comprehensivedescription of the planar organization of basins.

5. ConclusionsThe results of this paper suggest the following conclusions:

1. Hack's law appears to hold, in a statistical sense, forany point inside a basin.

2. The statistical framework developed here predictsthat the ratio of consecutive moments of the distribution ofstream lengths to the divide from points with the same contributing area should follow a power law as a function of areawith the exponent being close to Hack's h. This prediction isconfirmed in natural basins.

3. Hack's exponent, the self-affinity of the river basin,and the fractal sinuosity of its channels are related by H =4>Jh - 1. For elongation to occur, H needs to be smaller than1. This imposes constraints between Hack's exponent and thefractal characters of rivers. The values of h and <\>L obtained innatural river basins yield// < 1 and thus confirm the tendencyto elongate.

4. Hack's relationship and the self-affine character ofbasin boundaries are related. The values obtained for H and</>_ in real basins yield excellent agreement with the commonlyaccepted>alues of h.

5. The exponents /3 and h are found to be dependentvariables through /3 -f h. = 1. Estimations of /3 and h innatural basins tend to confirm this dependency.

6. The stream length from any point to the divide followsa power law distribution whose exponent, y, is the ratio of theexponent of the power law describing the contributing area at

any randomly chosen point, /3, and Hack's exponent, h. Analyses of DEMs yield an excellent fit to the power law for thestream lengths and moreover confirm the relation between y,/3, and h.

Acknowledgments. This work was supported by C.N.R. "GruppoNazionale per la protezione dalle catastrofi idrogeologiche," Linea 3,Progetti MIEP-METEORAIN, by funds provided by the R. P. GregoryChair in Civil Engineering at Texas A&M University, and byM.U.R.S.T. 40% "Processi Idrologici fondamentali." Their support isgratefully acknowledged.

ReferencesDurret, R., H. Kesten, and E. Waymire, On weighted heights of ran^1 - ~ 4 o j n _ j _ ^ , V J J S _ Q n . P h 9 6 : ^ 2 2 3 - ^ J _ _ 2 1 . - _ - ~ 'Feder, J., Fractals, Plen_rn7N_w""Yorleri988.Gray, D. M., Interrelationships of watershed characteristics, /. Geo

phys. Res., 66(4), 1215-1233, 1961.Hack, J. T., Studies of longitudinal profiles in Virginia and Maryland,

U.S. Geol. Surv. Prof. Pap., 294-B, 1, 1957.Ijjasz-Vasquez, E., R. L. Bras, and I. Rodriguez-Iturbe, Hack's relation

and optimal channel networks: The elongation of river basins as aconsequence of energy minimization, Geophys. Res. Lett., 20(15),1583-1586, 1993.

Ijjasz-Vasquez, E., R. L. Bras, and I. Rodriguez-Iturbe, Self-affine "T_r^"scaling of fractal river courses and basin boundaries, Physica A, 209, \288-300, 1994.

Jenson, S. K., and J. O. Domingue, Extracting topographic structuresfrom digital elevation models, HydroL Proc., 5(1), 31-44, 1988.

Langbein, W. B., Topographic characteristics of drainage basins, U.S.Geol Surv. Prof. Pap., 968-C, 1947.

Mandelbrot, B. B., The Fractal Geometry of Nature, W. H. Freeman,New York, 1983.

Maritan, A., A. Rinaldo, R. Rigon, A. Giacometti, and I. Rodriguez-Iturbe, Phys. Rev. E, 53, 1510, 1996.

Page 8: OriljH-ack's law - The University of Vermontpdodds/files/papers/others/1996/rigon1996a.pdf · OriljH-ack's law Riccardo Rigon,1-2 Ignacio Rodriguez-Iturbe,1 Amos Maritan,3 ... surements

3374 RIGON ET AL.: ON HACK'S LAW

4___akjn, P., J. Feder, and T. Jossang, Simple statistical models of rivernetworks, Physica A, 176, 409-429, 1991.

Mesa, O. J., and V. K. Gupta, On the main channel length-arearelationships for channel networks, Water Resour. Res., 23(11),2119-2122, 1987.

Montgomery, D. R., and W. E. Dietrich, Channel initiation and theproblem of landscape scale, Science, 255, 826-830, 1992.

Muller, J. E., Re-evaluation of the relationship of master streams anddrainage basins: Reply, Geol. Soc. Am. Bull.,-84, 3127-3130, 1973.

/NIkoraTy. I., and V. Sapozhnikov, River network fractal geometry andf itscomputer simulation, Water Resour. Res., 29(10), 3569-3575,

JPeckham, S., New results for self-similar tree with application to riverv networks, Water Resour. Res., 31(A), 1023-1029, 1995.

Rinaldo A., I. Rodriguez-Iturbe, R. Rigon, R. L. Bras, E. J. Ijjasz-Vasquez, and A. Marani, Minimum energy expenditure and fractalstructures of drainage networks, Water Resour. Res., 28(9), 2183-2195, 1992.

Rodriguez-Iturbe, I., A. Rinaldo, R. Rigon, R. L. Bras, A. Marani, andE. J. Ijjasz-Vasquez, Energy dissipation, runoff production, and thethree-dimensional structure of river basins, Water Resour. Res.,28(4), 1095-1103, 1992a.

Rodriguez-Iturbe, I., E. Ijjasz-Vasquez, R. L. Bras, and D. G. Tar

boton, Power law distributions of mass and energy in river basins.Water Resour. Res., 28(3), 988-993, 1992b.

Shreve, R. L., Statistical law of stream numbers, /. Geol., 74, 17-371966.

Tarboton D. G., R. L. Bras, and I. Rodriguez-Iturbe, On the extractionof channel networks from digital elevation data, Hydrol. Proc., 5(5)81-100, 1991.

A. Giacometti, Dipartimento di Scienze Ambientali, Universita diVenezia, D. D. 2137 Venezia, Italy.

A. Maritan, International School for Advanced Studies and IstitutoNazionale di Fisica Nucleare, Trieste, Italy. \

R. Rigon and R. Rodriguez-Iturbe, Department-of Civil Engineering, Texas A&M University, CETTI Building, Rm. 201, College Station, TX 77843. (e-mail: [email protected]; [email protected]_)

A. Rinaldo, Istituto di Idraulica "G. Poleni," Universita di Padova,1-35121 Padova, Italy, (e-mail: [email protected])

D. G. Tarboton, Department of Civil and Environmental Engineering, Utah State University, Logan, UT84322-4110.

(Received March 4, 1996; revised July 30, 1996;accepted August 2, 1996.)