'fi.

'O
I

~:,
9

2.8 3.2 Wavelength (Jlm)

3.6

4.0

FlI' 2 The 2.0-4.0 J.LIIl spectrum o IRS7. The solid curve is the estimated position of the continuum (see text). The dashed curve is the July 1980 spectrum where it deviates from 'lhe May 1981 spectrum.

bands stretching from 2.9 to 3.6 J.Lm. wo dominant bands lie at T 3.03 and 3.40 J.Lm,and have optical depths of 0.41 and 0.32 respectively. The increased optical depth of the 3.4 J.Lm feature relative to our previous resultl reflects the improved definition of the continuum. This optical depth would be further increased if the redness between 3.6 and 3.95 11mwere due to absorption rather than to the addition of cool dust. Figure 2 shows ali of the absorption features we believe to be real, and a few (marked?) which are less certaiu. The abtorption clearly breaks up into many discrete features and two of these (at 3.192 and 3.295 J.Lm)are quite narrow. Whilst the broad features centred at 3.03 J.Lm and 3.40 J.Lm presumably arise in solid grains, the narrow features may have a gaseous origino We have found no convincing identification for either, although many hydrocarbons have stretching absorptions in this spectral range. However, laboratory data are not available for even quite simple, incompletely bonded molecules (for example CH3) such as might be found in the interstellar environment. Quite narrow absorption bands are found in the absorption spectra of some solids, such as methane4 and methyI alcohol'. Suitable data on the absorption properties of solid materiaIs are not available, and laboratory work is needed to match the present spectra. We see no evidence for water ice in the available data. The absorption band at 3.03 J.Lm displaced from that of is water (3.06 J.Lm)and is narrower. If water ice is present, it contributes little to the absorption. Similarly, solid ammonia (2.91 J.Lm)s not present. Molecules involving carbon and hydroi gen, on the other hand, can produce absorption near both 3.0 and 3.4 J.Lm, nd are thus particularly attractive identifications. a Organic molecules containing OH and NH bonds also cause . absorption near 3.0 j.l.m(refs 6, 7). The wavelengths of CH, NH and OH vibrations depend criti~ally on the composition of the solid and only a few cases' of astronomical interest have been discussed in the literature. The H bonded solid complex of H20, CH30H and NH3 (rel. 5) has a feature due to NH at 2.97 J.Lm but there is no correspondence with the rest of the spectrum. . It has beeo suggested recentlyS, that surface functional groups attached to reactive sites on small carbon graina may be responsible for the IR features seen in IRS7. Although'there are some interesting wavelength coincidences (aromatic -CH (3.3 J.Lm), -CH3 (3.4, 3.5 J.Lm),-CHO (3.5, 3.65 J.Lmno information has been given on band strengths and shapes to enable a detailed comparison. We have Ipoked at published spectra of specific grain models involving organic polymers which could be possibilities for the organic component of interstellar frains. The polymer~like material in carbonaceous chondrites show absorption due to CH near 3.3 J.Lm and weak absorption at 3.0 J.Lm.Clearly this material.will not produce a fit to the astronomical data even over . a restr.icted wavelength regiolt. However, hydrated silicates o.

meteoritic origin show a broad absorption feature centred near 3 J.LmlO This may be relevant in view of the possible identifications of the 9.7 J.Lm and 18 J.Lm features in the galactic centre with silicates. We consider it significant that the ycllow component of UV tholin29 looks qualitatively similar to that of IRS7 in the 3.4-J.Lm region: In particular the shoulder at 3.36 J.Lm, he central component at 3.4 J.Lm t and the component at 3.48 J.Lm present in the laboratory data with wavelengths are agreeing to within 0.02 J.Lm. his could indicate the existence T of complex organic molecules in grains21I.:2 However, this material does not produce significant. al1sorption near 3 J.Lm as required by the IRS7 data. It is interesting that the satellite features at 3.36, 3.48 J.Lmare also seen in the spectrum of polyformaldehyde suggesting that H2CO may be present as a structural unie3 . We cannot rule out the possibility that a mix of simple organics might match the observed spectrum. This requires further investigation using laboratory data. However, as our data are of relevance mainly to the properties of dust in the diffuse interstellar medi um it will be more appropriate to look at refractory materiaIs related to the UV component of tholins such as the non-volatile residue produced in the laboratory experiments dsigned to simulate conditions in the interstellar medium described by Greenberg12 and with predictions from the Hoyle-Wickramasinghe modelll The latter comparison has shown a remarkable similarity between the spectrum of IRS7 and that of dried bacteria (Escherichia coli) which will be reported elsewhere'4. Finally emission features are found in this waveband in some H 11 regions and carbon-rich planetary nebulae (for example, NGC7027) 1~. These normaiiy comprise--a strong, narrow peak at 3.29 J.Lm and a weaker, broader feature peaking at 3.40 J.Lm. n I this respect they mimic features in the relevant portion of the absorption spectrum of IRS7. However, the 3.03 and 3.192 J.Lm featur~s are not seen in emission in any objects. An interpretation of the absorption spectrum of IRS7 which simul taneously predicted the more restricted range of the emission bands would be very satisfying.
Recei-fed~' 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13 14. 15. June; accepled 2 October 1981. Wickramsinahe. D. T. & Allen. D. A. Nalu,. 28'7. 518 (1980). Sa8an. C. & Khare, B. N, Nalu,. 277, 102 (1979). Jones. T. & Hyland. A. R. Mon. Nol. R. astr. Soe. 192, 359 (1980). A1lamondola, L. J. & Norman. C. A. Astr. Astrophys. 63, L23 (1978). Hagen W. Allamondola, L. J. & Greenberg, J. M. Astr. AJlTOphys, 86, L3 (1980). Herzbera, G. rnfra,.d and RanuJn S""ctrtJ of Po/yalOmic Mo~c,,~s (vau Nostrand, Amslerdam, 1954). Wexler. A. S. App/. SlH'Ctrosc.Rev. I, 29 (1967). Duley, W. W. & Williams, D. A. Mon. Nol. R. astr. Soe. 1M, 269 (1980. Knaclce. R. F. Nalu,. 169, 132 (1977). Knacke. R. E. & Kralschner, C. Astr. Astrophys. 92, 281 (1981). Hoyle, F. H. & Wickramasinghe, N C. Astrophys. Space SeL 66, 77 (1979). G_nbera, J. M. in 4-Io/uu~s in lhe Ga/actic EnviTOn"",nl (eds Gordon, M. & Snyder, L) (Wiley. New Yorl<. 1973). Wickramasinghe. N. C. Nalu,. 251.462 (1974). Hoyle, F. H., Wickramasinahe, N. c., AI-Mulli. S. Olavescn, A. H. 4: Wickramuinpe, D. T. Astrophys. Spac. Sei. (in lhe press). Willner. A. P., Puetter, R. C., Russell, R. W. & Soifter, B. T. Artrophys. Space Se/. 65,95 (1979).

Fractal dimensions of landscapes and other environmental data

It. A. Burrough
Department o Regional Soil Science, Agricultural University, Wageningen, The Netherlands .

Mandelbrot1 has introduced the term 'fradal' specifically for temporal or spatial phenomena that are continuous but not difterentiable, and that exhibit partial correlations over many scales. The term fraetal strictly defined refers to a series in which the Hausdorf-Besicovitch dimension exceeds the topological dimension. A continuous series, such as a polynomial, is difterentiable because it can be split up into an infinite number of absolutely smooth straight lines. A non-dilferentiable continuous series cannot be so resolved. Every attempt to split it up into smaller parts results in the resolution of still more structure or

Nature Vol. 294

19 November 1981
Table 1 Estimated D values for various environmentalseries Lag IS.2m IS.2m 20m 1m 30cm -lkm -lkm '-lkm -lkm -lkm Various Sm Sm Sm Sm Sm IS jlJJ1 IS 110m Sem SCIV 100m 100m Soom SOOm tOem lkm 3m

141

"

Location Wales England England England Deserts in Afriea. and Ameriea

Property Soil-sodium content -.tone content (both over four direetions). Soil-thiekness of eover loam Soil-electrieal resistivity (4 direetions) Surfaee of airport runway Soil- mean cpne index -.ilt+elay in 0-15 em layer -mean diameter of surface stones ..:...coarsesand fraction in 0-15 em layer Vegetation cover Gold Soil-phosphorus leveI -pH . -potassium levei -bulk density -0.1 bar water Iron ore in rocks -chlorite -quartz -quartz -iron -"ron (E-W) -iron(N-S) -iron (E-W) -iron (N-S) Sea anemones Rainfall Imn ore Groundwater levels Piezometer 1 2 -3 4 Oilgrades Copper grades Topographie heights Soil-sand conte"t -pH Crop yields -"Water table depth

Das lag . O
1.7-1.9 1.1-1.8 1.6 1.4-1.6

D at max.slope

Ref. 7 7 7 7 8 9 9 9 9 9 2 10 10 10 10 10

..

l.st
1.9* . 1.8* 1.8* 1.8* 1.6* 1.9 2.0* 1.5* 1.6*

South Afriea ( Australia

l.U !

I.S*
1.5* 1.6 1.9 1.6

Frane

..

I.S
1.7 . 1.8 1.6 1.9 1.6 1.7 1.4 1.6 1.7 1.8 1.3 . 1.7 1.7

France Chad' Mauritania Ivory Coast

11 11 11 11 11 11 11 11 12 13 2 2

Canada Chile France USA Worldwide India Estimated

1 day lday . 1 day 1 day 60cm 2m 10m 10m 10m 1-I,ooOm 250m'

1.3~ LI 1.1

2
2 2 2' 2 2 14 14 15 16

1.S
1.6-1.8 2.0 1.6-1.8* 1.6'*

from variogram.

* Estimated

from bloek variance.

t Estimated from power speetrum.

Estimated from covarianee.

roughness. For a linear fradal fundion, the Hausdorf-Besicovitch dimension D may vary between 1(completely difteren- t1able) and 1(so rough and irregular that it eftediveiy takes up the whole of a two-dimensional topological space). For surfaces, the corresponding. range for D Iies betw41en 1 (absolutely Imooth) and 3 (infinitely erumpled). Because the degree of roughness of spatial data is important when trying to make lnterpolations from point data sueh as by least-squares fiUing or kriging2, it is worth examining them beforehand to see if the data contain evidence of variatiQII over difterent scales, and how Important these seales might be. Mandelbrot's work1 suggests that the fradal dimensions of eoastlines and other linear natural phenomena are of the order of D = 1.1-1.3, implying that lo"g range efteds dominate. I show here that published data on many environmental varillbles suggest that not only are they fradals, but that they may have a wide range of lradal dimensions, lncluding values that imply that interpolation mapping may not be appropriate in certain cases. . Berry and Lewis3 have shown that the Weierstrass-Mandel. brot fraetal functiol1' (WMF) '.
00

estimate the fraetional dimension Dof a'real series, either from the slope of the log-Iog plot of the power spectrum as t -+ O,or from . d log V(t) . d log t 4-2D . .' (t ..O)

W(t) -

__

1:

00

[(l-ei""')el"'] Y <2-D) .

The variance of increments (or the half thereof, the semivariance) illmueh used in geostatistieal studies where, computed over distanees, it is referred t,oas the variogram2 Computing the variogram is the first step in the interpolation proeedure known as kriging which is used to asstst estimation of mineral reserves, eontouring groundwater surfaees, and so on. Thus many published data are available in this form and it is then a simple malter to calculate their dimensions D relative to the sampling interval, used. For a second-order stationary series, the variance of increments at a given tag is equal to twiee the difference between the varian<:e of the series and the covariance. Thus, D values may also be eomputed from eovariances. In agriculture and soil science, many data have been published in the form of block varianees; that is the varianeewithin blocks of ~qual size plotted against block size. Beeatise Yates4 showeQ that the varianee for a bloek ola given sizek is

_ (1 < D

< 2, y > 1, tfJ = arbitrary phases)

.si .k'(k +1)

k k~l

(k -h +l)Vh

has a r'wer' spectrum P(w) that varies ap-proximately' as W-<S-2D, and a variariee of 'inerements. V(t) = ([W(to) - W(to+ t)]2) that varies as (4-2Dat lhe origin.lf D >: 1.5, V(t) is itself a fractal function. These results'allow us to

where si is the bloek variance f 1:lloeklength k, ando Vh the varianee ineremen,t for lag h, the' slope of log Vh versus .Iogh (h .. O) eari also be estimated Irom these data.

Table 1 presents a selection of data so analysed, giving the Itudy location, the type of environmental variable, the lag interval used for sampling and the estimated D value assuming that the real data are but a series of regularly spaced samples of a realization of the Weierstrass-Mandelbrot function over onedimensional space or time. o The data support Mandelbrot'sl assertion that D values of landscape and other data may range over mariy values. It is evident that most of the values reported here exceed 1.5, and many are greater than 1.8.-Note that one of the sm<?othest lurfaces imaginable in a landscape, a new airport runway, has a relatively high D, presumably because variations over long distance are low in amplitude. These data do not conform to a lingie roughness madel as proposed by Sayles and Thomas,' indeed, as Berry and Hannay6 have commented; much wider ranges of roughness \1r randomness are to be expected. These results should not be accepted uncritically, however. First, it is important to realize that the Weierstrass-Mandelbrot function is just one of a class of 'model' fractals. Its peculiarity is that it has a discrete and geometric spectr.um, and this might make some of its properties non-universal among other fractals with the same D (M. V. Berry, personal communication). Second, although some environmental data do appear to display the fractal property of statistical self-similarity at ali scales, there are also inany that show self-similarity over a limited range of scales, or over a few widely separated scales. For example, a variable with a highly regular spatial variability has a variogram that exhibits parabolic behaviour near the origin2 This is a good example where d log V(h)/d log h becomes less Iteep as h ..O, and it indicates that the ultra-short, short and middle range variations are trivial compare<! with the main variations seen at larger scales. In fact, computing D as h ..O will in this case estimate D for the ultra-short range scales of variation. Third, the form of the variogram is often highly dependent on sampling direction and sampling interval. It is well known that when a particular sample spacing tends to match the scale of a spatial pattern, the variance of increments can falI rapidly. Accordingly, when a sampling interval matches a particular scale of a phenomenon in the landscape it perceives an apparently lower D. If this is so, it would seem more appropriate to estimate D values from the parts of the variogram showing maximum slope. Table 1 also contains these data for those phenomena showing a m8Ximum slope at positions other than h ...O. Tbe results presented here suggest that Mandelbrot's D can be used as a useful indicator of the complexity of autocorrelations over many scales for natural phenomena.o However, although many natural phenomena do display certain degrees of statistical self-similarity over many spatial scales, there are others that seem to be structured and have their levels of variability clustered at particular scales. This behaviour does not exclude them from the fractal concept. MandelbrorJ considers that it is quite acceptable to have a series of zones of distinct dimensions connected by transition zones. If this is reasonable, it o means that the examination of D values would be useful for trying to separate scales of variation that might be the result of particular natural processes. Moreover, identifying such scales could be oi enormous practical value because one could then taiJor sampling to a particular scale range oi the phenomenon in question, thereby improving the efficiency of expensive field investigations and the resulting interpolations. The high leveI of the D values for some soil and geological data reported here would seem to question the wisdom of interpolation mappingin certain instances, however, and it would seem worthwhile to use D values as a guide to how further mapping and interpolation should proteed ..
a.ce~v.cct U.Juno .ccepted 30 September 1981.
1. 2. 3. 4. S. M.nclelbrot, B. FrtJC,dls.Fomt, aana a/ld D/_nsloll (Freeman. Sen Franc:lsco. 1977). Joumel, A. O. &. Huijbrelta, Ch. J. Mln/II' Ocostar/stlcs (Acedemic, London. 1978). BeR)', M. V. &. Lewis. Z. V. 1'rrK. R. Soe. A370. 459-484 (1980). Y.tOl, F. 1'111/. T'rrJIIS. R. Soc. AUI. 345-77 (1948). Se)'lea;R.A.I< Thomu, T. R. NG'IUW J71,431-434 (1978) . .~. . .

6. 7. 8. 9. 10. 11. 12.

13. 14. 15. 16.

Berr)'. M. V. &. Hann.)'. J. H. Na'u~ 173, '73 (\978). Burleu, T. &. Wehsler, R. J. Sol/ Sei. 31, 315-42 (1980). Jenkins, G. M. &. W.lls. D. O. S(Jtral "'"a/v,is (Holden D.)', London, 1968). Milehell. C. W. Wehsler. R., Beekell. P.II. T. &. Clilford, C. OCO". J. 145.72-85 (1979). Wehsle" R. &. Ouller. O. E. A,..". J. Soi! R .. 14, 1-24 (\976). Serra. J. Mlnc,. o.posl,a 3, 135-154 (1968). Kooijman. S. A. L. M. Spo,/al alld Tcmpo,al Alla/)',I. jll Ecolo,v Vol. 8 (edl Corlliaek. R. M. I< Ord, J. K.) 305-332 (Internalional CooPr.tive Publilhinl House. Mar)'l.nd, 1979). Delflncr. P. I< Delhomme, J. P. Dlsplay alld "'"a/)',/. 01 SpatItJl DaID. (eds O J. C. It MeCull.lh, M. J.) (Wile)', London. 1975). Campbell, J. Sol/ Sei. Soe. A",. J. 41,460-4 (1978). Fairfleld Smith. A. J. "'grlc. Sei. 18, 1-23 (1938). Hillel, D. App/lca'/ons 01 Sol/ Phy,lcs (Acldemic: New Vork, 1980).

'Reactor-released radionuclides' in Susquehanna River sediments

C. R. Olsen"', I. L. Larsen*, N. H. CutshaU"', J. F. Donoghuet, O. P. Bricker:\: & H. J. Simpson
Enviroomental Sciences Division, Oak Ridge National Laboratory, Oak Ridge, Tennessee 37830. USA t Smithsonian Institution, Washington DC 20560. USA :I: Maryland Geologica} Survey, Baltimore. Maryland. USA Lamont-Doherty Geological Observatory and the Department of Geological Sciences of Columbia University, Palisades, New York 10964, USA

Three Mile Island (TMI) and Peach Bottom (PB) reactors have introduced 1.l7Cs, 1.l4Cs, GCo,5lICo and several other anthropogenic radionudides into the lower Susquehanna Rlver. Here we present the release history for tht'se nuclides (Tl1ble 1) anil radionuclide concentration data (Table 2) for sediment samples collected in the river and upper portions of the Chesapeake Bay (Fig.l) within a few months after the 28 March 1979 loss-of-coolant-water problem at TMI. Although we found no evidence for nuclides characteristic of a ruptured fuel element, we did find nuclides characteristic of routine operations. Despite lhe TMI incident, more than 95% of the total 1UCs input to the Susquehanna has been a result of controlled low-Ievel releases from the PB site. 134CSactivity released into the river Is eftectively trapped by sediments with the major zones of reactor-nuclide accumulation behind Conowingo Dam and in the upper portions of Chesapeake Bay. The reported dlstributions document the fate of reactor-released radionudides and their extent of environmental contaminatlon in the Susqueh~nna-Upper Chesapeake Bay System. During the past 14 yr, five nuclear power reactors have operated for varying lengths of time at two sites on the Susquehanna River (Fig. 1). The first reactor, PB 1, was a 40-MW, gsS-cooled rea'ctor and operated from March 1966 to October 1974. Three others, PB units 2 and 3 (boiling-water reactors), and TMI unit 1 (a pressurized-water reactor): began power production in 1974 and each produces -1,000 MW. TMI unit 2,a twinofTMU, was activated on 30 December 1978 butshut down on 28 March 1979 after the loss-oi-coolant incident. As with the other nuclear power stations, minor amounts of 137Cs, 134Cs,60Co, 'ICO and other radionuclides are released with the coolant-water etHuent. The radionuclide release history of these reaetors has been compiled using Nuclear Regulatory Commission documentsl and is summarized in Table 1. The Peach Bottom plant has contributed mos! of the reactor-prodced radionuclides introduced into the Susquehanna River, and from 1975 to 1979, the PB reactors have released >95% of the total 134Csinput to othe river. In addition to reactor releases 137Cs(half life ai - 30 yr) has been introduced intp the Susquehanna-ehesapeake Bay system as global fallout from atmospheric nucler weapons testing. The major inftux of fallout l37Cs occurred between 1962 and 1964. Reil2 used the distribution of fallout 137 s in Chesapeake Bay to C study the hydrography of the estuary. Although otherfission produc~ and neutron activation nuclides, s~ch as 134CS,60Co and '8CO are also produced during weapons tests, their shott half lives (-2. 5 and 0.2 yr, respectively) and low yields cause the