, . awx,.Jts .~~:I~r:~t~ ..; .~.... ~~.~i.onofthe _r~~ h . .

I
;

,NATI2!!~~!!:.:.2!~~I!:...~~~~.

--.-.excess manganese

LEnERSTONATURE--associated to the median

""."", ;"."j,,'i,

vall,~.\ ;md does not spill over into the adjacent '~!dlOugn Il may leak out through fracture

. By~aki~g a con~ervative approach and grouping those plumes tn whlr'J ~~~ "'!?""'!? ':'cc~:- ~~ ~h;:;::;~::-;;;~;:;;:~;, ~.;;: ::;;;~.~~~ . be identified, that is stations 2, 4 and 5; stations 6 and 7; station 8; station 9; and stations 10 and 11. This analysis predicts a frequency of one source per 340 km of ridge; of course the arhlal frequency may be much highel The position ofventing in the rift appears to vary with latitude. Results from stations 8 and 9 are consistent with venting from the east wall, whereas results from slation 10 indicate a west-wall source. Furthermore, these results are consistent with discharge from the valley walls, but they do not eliminate lhe possibility that venting occurs from the valley f1oor. Finally, hydrothermal manganese along this section appears to be confined
Rcccived 21 November 1984: acc;epted 21 Fehruary 1985.

zones. J'in:: anomaues n:'.>(I.-l<:O nere are smatler than those found near venting in th(' P dfic47 However, this difference is an artefact of the regi:: n.'l' .,ampling strategy of this survey and is not a true indication 01' the relative impact of venting from the MAR on geochemical cycles an,: the heat budget. Such Information can be acquired on!y by site-specific work. We thank Caplan R. L. Swanson, officers and crew of the NOAA ship Researcher for their suppori and T. A. Nelsen of NOAA and J. Trefry 01' Florida Institute of Technology for their cooperation. This work was supported by the NOAA Vents programme and the NERC; Cambridge University Earth Scences Series, contribution' 578.
13, Burton, J. D., Maher, W. A. & Stalham. P. J. in nau Metais in Seawater (eds Wong. C. S. Doyle, E. Druland, 11: W., kurtun, J. D. & Goldberg, E. D.) 415426 (Plenum, New York, 1983). 14. Froelich. P. N. tt 01. Geofhim. 'I.\mochim. Arta 43, 1075-1090 (1979). 15. Discholl, J. L. & [)icks~", I,. W. hJrlh plan.,. Sei. Lell. 2~, .185.197 (197ll.
16. Edmond,
II'I~),

I. Rona, P. A. J. Volcano/. geolherm. Ret. 3, 219225 11978). 2. Rona, P. A. ti 01 Geo/. Soe. Am. AbsIr. Progm 14 (7),602 (1982). 3. Rona. P. A. Earrh Sc. R.v.20, 1104 (1984). 4. Klinkhammer. a., Dender, M. & Wei , R. F. Na/ur< 269, 319320 5. Klinkhammer, G. P. Chem. Geo/. 29,211226 (1980).

(1977).

6. Luplun, J. E. el g/. E"grrh planei Sd. Leu. 50,115-127 (1980). ,. Jlllltf'l:. C. J., Jnhm.on. fi. P &. Dclaney. J. H. Gmpltv\. Rn, I I. M71. k7t. {JlJIiH. " h KhulluHlIluer. (;" l:IJcrludd. 11. &. I'ud"'\lll, A NlJIIJ'i' lOl\, 1'K"i Itt~ (Ilnn,. 9. Klinkhammcr, G. Anuly', (,hem, 52, 117-120 (ltJKO'. 10. KJinkhammer,G. P. & Dender, M. L Earrh planeI. S. LeI/.46, 361384 (1980). 11. Dender, M., Klinkhammer. G. & Spencer, P. lJ"pS.a Re..24, 799812 (1977). 12. Lamber!, C. E., !Iishop, J. K. O. Discaye. P. E. & Chesselel. R. Earlh plan.,. Sc. LeI/.70, 2)7248 119841.

J. M.,

VOf1

DlIiTlTTl.

1\, 1...

Md>ull,

R. E. &:

Mell!iurc!\,

(',

I. Ncuurr

29',

187-191

17. S .uu, M. R . ~;"Iln, R. a., !~(l1l,J, P /\., Hutler, I.. W." Nalwiflllr., A. J. Of'oph.vJ. ReJ. Lru, 1,355358 (1974). 18. Shearme. S., Cranan. D S. &. Ron., P. A. Mar. Geol. ~I. 269291 (198)). 19. Rona, P. A. Geoph.vs. Re . Lell. ~, 9939Qh 1\978). 20. Jenkins, W. J., Rona, P. A. & Edmond, J. M. Earrh planei. Sc. Lell. 49, 3944 (1980). 21. Rona. P. A. J. geol Soc, i "nd. 1.17,385402 (1980).

Fractal dimension or vegetation and the distribution or arthropod body lengths

D. R. Morse*, J. H. Lawton*, M. M. Dodsont & M. H. Willialhson*
Departments of *Biology and tMathematics,
Y..,t'k Y0i 3i:Ji:J) Li,,"

University of York,

transects across a C;-i ai r~:.f~. An alternative widely used method (see ref. 1, p. 130), which has some practical advantages, is as follows. Enclose the ~bject (or a subsection of it) in a square of si de S and divide the square inlo (SI A)2 squares of side . Let N(A) be the number 01' sCl;;an;, '.he edge of the object enters and plot log N(A) against lo!,! 1,1 A. Suppose that for small values of A, the graph is almost linear with slope D. Then D can be inter;.. ,~~~.j a.s th~~ f,.;tc:utl dlrnens10n, . Slnce

I
~

I
1

Following Mandelbrotl, recent studies~ demonstrate that some nalural surfaccs are fractal. Hcre wc show that transccts across vel?ctation are fractal, and considcr 00(' J}ossible conscqllcncc of this ohscrvation for arthropods (mainly .nsects) Iiving on plant surfaccs. An important future of a fractal curve or surface is that its lenglh or arca, rcspectively, becomes disproportionately large as the unit of measurement l!'l decreascd I. This suggests that if vegetation has a fractal structure, thcre is more usable space for smaller animais Iiving on vegetation than for fllrger animais. Hence, there should be more individuais with a smalI body length Ihan a large body length. We show that this is the case, and that relative numbers of small and large individual arthropods collected from vegetation are broadly conslstent with theoretical predictions originating from the fractal nature of vegetation7 and individual rates of resource utilizatlon. . Mandelbrotl,4 asks, how long is the coast 01' Britain? Measuring its lenglh by stepping round with successively shorter divider slep-lenglhs allows finer and finer resclulion af lhe coas'tline; if the log of divider step-length is plotted against the log of total length measured, a straight line ofnegative slope is obtainedl,4,8. Mandelbroll4 inlerprets this as evidence lhat coastlines are fraclal and lhat lhe slope of the graph is an estimale of 1- D, where [) is lhe fraclal dimension 01' lhe coastline, The relationship between D, step-Iength r and lhe perceived length of the line I( r) measured with step r is given by
l(r)OCr(I-D)
l

(I)

I
j

I
i

By definition , for a line (that is, a transect across a surface), D must lie in the range 1 ~ lJ ~ 2 and for a surface, 2 ~ D ~ 3. The fractal dimension of vegetation might also be measured hy stepping along a transect over the slIrface wilh a pair 01' dvidas sei to a series 01' step-lenglhs and lIsing cqualion (I). Th:s ':..~~ ~:I'"It~(.thvJ u::H.,J lU liit.::t;,lI(C; ~i1c "l dt,;ta; dlm~n"lnl) ""

The method is explained in gr.: ter detail in Fig. I legend, together with an example. Photographs were tilken of a variety of plants during early spring. The plants were eilher leat1ess, JUSl coming into leaf or evergreen. By careful choice or plant-parts and camera focal length, the resulting photographs were approximately two dimensional, greatly simplifying their interpretation under the grid (Fig.la). Two ditlerent arbitrarily sized . grids (128 mm or 256 mm along one ed;;e) were partitioned into 2n squares along each edge. Depending on the grid size, n varied from 2 to 6 or 7, down to. a square size of 2 mm. The grid was randomly repositioned several tim,~s to give semi-independent estimates 01' the fractal dimer.,ion 01' the same piece 01' vegetation. The slope of the resulting !il1c (for cxample, rig. 1h) was estimated by legres~ion unulysi". There are several points to note in the interpretation of such graphs. One is that when the number 01' divisions is small, it is prohable that ali lht> sqllar.:s will he entered, hence the slop. 01' the graph will be n == 2. This is an artefacl of the melhod; consequently, in the prescnt study ali points that fell on fhe Iille y = 2x (on a log/log pIo:) were omitted when estimating the slope 01' the line. At tht: other extreme of resolution, when a very large number 01' sqllares is used, the slope of the graph will in general fali to I. This could be due 10 a lack of resolution in the photograph or pccallse irregularities in the outline of the plant no longer OCCI.'r <il '!Iat scale, Taking a photograph at a much larger scale (higher ma~nification) resolves this prohlem. Thc graph may he convcr.'1 hClween the Clllrcml:s 01' [> == 2 1I11d [) - I. or it could h,\'c lwo straight subre/l.ions characterized ")' di!Jere!ll ~;Iopes (hg. 1 iJ), Similar changt:s in fractal

L__

--------

1001 Jb'155

loi j(1cw

magnlflcallon)

lJ-:--ii

100 1,300

No. of squares 00 one

side of grid

flg. 1 a, Photographs of plants at various magnifications were placed under a grid. The number of squares entered by the outline of the plant were counted, starting with a course grid of two large squares on one side, then 2" squares, with n varying from 2 to 6 or 7, depending on the grid size. For ease of representation, the plant's leaves in this figure are drawn fiat; in reality they are orientated at ali anglts with respeet to the grid. Also for c1arity, the progressivelyfiner divisions are only ilIustrated in one comer of the figure. The logarithm of the number of squares entered by the outline of the plant was then plotted against the logarithm of the number of squares along one side of the grid, as in b. The slope of the line equals the fractal dimension, D. b. Data gathered in this way for Virginia creeper, photographed without leaves in early spring. The twigswere photographed at one scale, then parts of the samc twigs were rephotographed at a higher magnification, permitting O to be estimated at two levclsof resolution,

dimension at different scales have been found in other natural objects 1,3.11. Table 1 presents data on the fractal dimension of a selection of pIants. Estimates of D range from 1.28, for a close-up photograph of Virginia creeper, to 1.79 for cotoneaslet. Those plants which one might, a priori, consider to have a more complex growth form have a higher fractal dimension. In those plants which were photographed at two scales, estimates of the fraetal dimension are lower at the higher magnific~tion, as would be expeeted from the above argument. The mean fractal dimension of the samp:es iu Table 1 is 1.44. For ease of calculation, assume that, typical1y, D = 1.5. It fol1ows from equation (1) that for an order-of-magnitude decrease in

uler ienglh, (he expected distance between two points on a linear transect (of dimension D = 1.5) across a fraetal surfaee increases hy a factor of jlo = 3.16. The difficulties of measuring the fractal dimension of a surface are considerable and no data are yet available. However, squaring the increase in linear distance (that is, adding the fractal dimensions of the orthogonal transects; ref. I, p. 365) gives a heuristic upper bound estimate for the expected increase in surface area. Therefore, the maximum expected increase in surface area is 3.162 = 10.0 for an order-of-magnitude decrease in ruler length. Bearing in mind the disconnected character of the surface of vegetation, an estimate for the lower bound ofthe fractal dimension is obtained by adding I to the linear fructal dimension (for example, 1.5+ I; see rer. 1, p. 365). This holds exactly under some circumstances, and in particular when the surface is flat in a direetion transverse to the cross-section( which is dearly not the case for vegetation). This lower bound predicts a 3.l6-fold increase in surface area for an order-of-magnitude decrease in ruler length, when D= 1.5. Now consider the implications of the fractal nature of plant surfaces for the animais living on them. As a first approximation, substitute animai body length (L) for step-Iength (A) in equation (1). Ifthe way in which animais perceive and use their environment is proportional to their body length12, then for a homogeneous fractal surface having transects with D = 1.5, the area perceived by animais 3 mm long may be up to an order of magnitude greater than the area perceived by animais 30 mm long, for the same reference area. This increase in available space for animais of smaller body length may be combined with a consideration of the way Jn which metabolic rate scales with body length 13,14 to make predictions about the distribution of body lengths of animais living on vegetation. The metabolic rate of individual animais 13,1~ scales approxiw"'di dS lle 0.75 power of body weight, W,that is, as (L';tI~. Next, suppose that population densities are approximately prop;):1ional tothe rcdprucal of individual rates of resource utilization (that is, to metabolic rate-I; see, for example, ref. 13). Then, if use of resources per individual is proportional to WO.75, it fol1ows that population density, N, scales as (e)-0.75. Hence, ali other things being equal (especial1y the rate of appearance of new resources), a 10-fold decrease in body length results in a (103).75 = 178-fold increase in the density of individuais. This increase in density may be combined with the expeeted increase in the available surface area, outlined above, for.a given decrease in body length, to prediet reiative numbers of individual animais of different body lengths Iiving on the surface of vegetation. For an order-of-magnitude decrease in body length, such calcula-

Species Barberry, Berberis vulgaris L.(evergreen) Virginiaereeper, Parlhenocissus Iricuspidala (Sieb. and Zucc.) Planch. (twigs and buds) Weepingelm, Ulmus glabra forma pendula (Loud.) Rehd. (twigs and buds) Cotoneaster, Coloneasler horizonJalis De~aisne (twigs and leaves) Ivy, Hedera helix L. (evergreen) Yew, Taxus baccala L. (evergreen) Silverbireh, Belula pendula Roth (lwigs and (eaves) Downy birch, Relula pubescens Ehrh. (twigs and feaves) Ash, Fraxinus excelsior L. (twigs anJ !eaves) Sycamore, Acer pseudoplatanus L. (twigs and leaves)

Magnification High Low High Low Low High Low Low High Low Medium Medium Medium Medium

No.of estimates 3 3 6 3
6

Mean D 1.46 1.43 1.28 1.S5 1.41 1.35 1.79 1.39 1.47 1.68 1.40 1.40 1.42 1.31

s.d. 0.018 0.042 0.078 0.009 0.111 0.019 0.093 0.050 0.042 0.099 0.040 0.035 0.083 0.023

3 6 18 3 3 3 3 3 3

Photographs or branehes and twigs of a selection of woody plants, with or without leaves, were taken from the University of York eampus or SkipwithCommon, North Yorkshire.High-magnificationestimateswere derivl'<1rnrn "'n~ t:t' p~~~~r::t:h: f :;ftwigs":O em loog; IvW-11l3l!'lif,cctliun estimat,;;,were from pnotographs oftwigs> 50em long. Single medium estimates were made from speeimens >25 cm long. The state of each plant when photographed and the plant part photographed is indicated in p::!renthesesarte; lhe speeies name. (Note lhat ~"mr!e tran!ects !~e::t;;bl) tended to foU:;','; :":";6" ;;a;: lher princip axes' in the natural structure of the vegetation. This could introduce some bias into lhe estimates of i their fraetal dimension.)

LETTERSTOATURE-- N
Fia. 2 Data on the number of individual arthropods (mainly insects) of different body lengths collected from vegetation. a, Understorey foliage in primary forest, Finca Tabogo, Costa Rical5; b,Osa secondary vegetation' (.) and Kansas secondary vegetation (O); c, Tabogo primary riparian vegetation (.) and Icacos vegetation (0)'6; d, understorey foliage in cacao plantations in Dominica (.) and at Finca La Lola, Costa Rica (O)"; e, Birch ( Betula fluhptN'n<.' !r . ~! Sld~w!t!l ~ Common, North Yorkshire. In a-d, arthropods were collected bv sweep nel, in e by pyrethrum knock-down of the tree canopy. The lower bound prediction that, for an order-ofmagnilude decrease in body lenglh, lhere should be a 560-fold increase in lhe number of individuaIs, is indicated by the lower dashed line on each graph; the upper bound prediclion-a 1,780-fold increase for an order-of-magnitude decrease in body length-is shown by the upper dashed line. Regression lines were fitted through the log-transrormed data, the anti-Iogarithm ofthe slopes of the Iines were taken and are as follows: a, 390; b, 1,040 (.) and 700 (O); c, 360 (.) and 980 (O); d, 21,800 (.) and 530 (O); e, 1.440. The decline in the numbers or individuais of < I mm body length may be due to a variety of factors", not least of which is the difficulty in sampling very small arthropods. Consequently, when filtng regression lines to the dala, in a-d lhe I mm body lenglh dala point and in e the firsl three dala points were omitted.

-_._----------.
\

733

\ \ \ \ \ \ \ \ \ \ \
I,
\ \

a
o

b
\

\ \ \

c \

.
.,
o

\ \ \

\ \ \ \ \
I

i \
\ \

\
o

\ \ \
\ \

\ \ \ \

\ \
\ \

\
\ \
\
I,
\ o.

.
I,

.... \ \
u

\ \
\

.0

~ ~ ~

\ \

,.
\

-.

. ...

\
\

~
\ \

i .".

~
\ \ \

\ \
,o.

" ..". I.:.:

10

..
o

\
\

I,

\
\

..

\
1U~ \ \ \ \

' .\;

10,000

\ \ \ \
I, I,

1.000
00

\ \ \

100

\ \ \ \

.. .
00

\
\ \ \ \ \ \
I,

I, I, I, I, I,

I,

o.

o o oo

\ I,
\ \
\ \
I,

0'1,

10

\ \

.
\
\

\
I, I,

. ....

\ \ \
\
10
I,

\ \ \

100

Body length, L (rT'r,)

tions predict an increase of between 178 x 3.16 = 560-fold and 178 x 10= 1,780fold in the number of individual animais 7. Data on the number of individual animais of different body !e:-:s!!':~ ~;}:l':vt\;J r.o", lhe surface 01' vegetatin are scaCC. Figure 2 shows such data for sweep-net samples of terrestria! arthrpJ" 1<_17 all originai data collected by pyrethrum knockdown of a tree (Refuta pubescens) canopy. Ali the data show a steep increase in the number of individuais for an order-ofmagnitude decrease in body length. in general agreement with OUI" predictions (Fig. 2). The above heuristic calculations may be reversed, allowing prediction of the expected fractal dimension of the surface of vegetation, given the mean ofthe fitted slopes for the distribution of individuaIs' body lengths shown in Fig. 2. This calculation results in a prediction of the fractal dimension of the surface rlf vl'l'.l'lalilll1 111' f) -, 2,711. C1early, approximate agreement betweell data and our predictions does not prove that the fractal nature of vegetation contributes to a steep increase in number of individuaIs as anhropods get smaller. Moreover. dilferent values for IJ (1.281.79; Tahle I) and the metabolic rate exponent (0.62-0,86; refs 13,14) lead to dilferent predicted slopes in Fig. 2, as will the precise form oflhe relationship between the fractal dimension 01' surface anel transeClS across that surface. li remains to oe seen whether detailed differences in the fitted slopes apparent in Fig, 2 can be explained by differences in the fractal dimension of particular plant surfaces, and by the biology of associated arthropods.

We know of no oth'~r aitempts to explain patterns in the body size distributions 0(' ;mhropods Iiving on plant surfaces, although previoll'l S;jiln~. h2~'e examii .. ~1,<;..ic~/rre4lJl:llcy J istributions for animaIs of different sizesl2,18, Theoretical and emoirical Ii"k~ hl"h''':'P''l ?:.~~er~~ ir. !!':: ;;"i.~,lvf "pecies and number of individuab n:' different body lengths constitute an important unsolved l:'w'.\lem. D.R.M. gratefully ack.llowledges the financial support of the NERC. Colleagues at York and ?rofessor B. Mandelbrot provided helpful criticism of the manuscript.

t. Mandelbrot. B. B. 111(" "'''''aNaf G(!(tn("ry lJf Na,urt> (Freeman, New York. 19)13). 2. Avnir, D., Farin, D. &. I'feifu. f', N,ullrr ~O". 7111-2".1 IIQH 1. r\h,"lrlhl'II.1I 11" p" ,lt. I: '-.\ 1';IUIl&l\, :\ ,I N"h"~ .tUN, '.!I I!~ ll~N~I. u .t Matltttlh . lI. H. S"lt"tll'f.1~6. \b r.JH (l"011. , 5. Ihnh)ughs, 1'. A. NUIII'~ 2'1.1. ~,.q!"2 (~gl). 6. Ilradhury, D. O., Rri<hrll, h. E. & Grren, O. O. Mar. Eco/. Str. 14,295-296 7. lawlon, J. H. in Piam Surfw"" (rrl. Junirer, R. P.. & Southwood. T. R. E.I (Edward

"'"g.

(1984). Arnold,

London. in lhe pFess).

K. Rkhilrds(lO. I" I", (il''' .. \)'_". 6, t:t(J-I~" (1'161). 9. ')ltUmSilflncr. 1)., 1.01'11. . & Wt:lbd, E, H.. J. Miatn G 10. Rigllul. J, I'. }, MiI',tJJf. 1.11. lI~.:'''ft 1 m,,).
1

., 121. ~I 61 (l9MI),

11. Orford, J. n. & Whalley, W ll. c\.dim'.'''/''K.V 30, 655-668 (1983). 12. "by. R. M. in [)ivtrsi,.v af 1o,,," I'a,mas (eds Mound. I. A. ,t Walolf, N') 188 204 i1l1ackwrll, O,ford, 197Kl U. r~'cl.). R. H. Til.:E(:u;ug;(:uiIm;'[ic""o"J olBod.v Sizt (Cambridge Universily Press. 1983). 14. Schmidl-Nielsen, K. Sca/;nl(. \Vh~' ;~ .nimal Siu ,fO Imp()"unl? (Cambridsc ' Univcrsil)' Pre , 1984). 15. Janzen, 1>. H. & Schoener, T. W. /'eto'''R.l 49, 96-110 (1968). 16. Jan.en, D. H. Eco/"KY 54, 659-686 11973). 17. Andrews, R. M. Brrviora 454, l-SI. (1979). 18. HUlchinson, G. E. & MacArthur. R. fi. AlO. Na'. 93, 117-125 (\959).

1__

----------