You are on page 1of 8
JOURNAT. OF GEOPHYSICAL RESEARCH, VOL. 102, NO. B9, PAGES 20:599-20-502, SEPTEMBER 10, 1997 ‘Lhe physical basis of glacicr volumc-arca scaling David B. Bahr Institute of Arctic and Alpine Research, University of Colorado, Boulder Mark F. Meier Tnstitnte of Avetic and Alpine Research and Department of Gealagieal Sciences, University of Colorado, Boulder Seott D. Peckham ‘Water Resources Division, U. $. Geological Survey, Boulder, Colorado Abstract. Ice volumes are known for only a few of the roughly 160,000 glaciers worldwide but are important components of many climate and sea level studies which require water fux estimates. A scaling analysis of the mass and momentus conservation equations shuws that glacier volumes cau be related by a power law lo more easily observed glacier surface areas. ‘The relatiouship requires four closure choices for the scaling behavior of glacier widths, slopes, side drag and ‘mass balance. Reasonable closures predict a volume-area scaling exponent which is consistent with observations, giving a physical and practical basis for estimating ice volumes. Glacier volume is insensitive to perturbations in the mass balance scaling, but changes in average accumulation arca ratios reflect significant changes in the scaling of both mass balance and icc volume. Introduction Studies of the relation of melting glaciers to sea level rise and other aspects of glacier climate interactions re- guire regional and global catimates of ice volumes and surface arcas. This necessitates cither the impossible tack of measuring the volume ofall glaciers in the world, or using a theoretical technique which can scale up & few local observations to general statements about the global volume of ice. The following analysis solves a key part of many sea level and climate problems by giv- ing a theoretical scaling basis for linking unknown ice valumes ta easily observed ice surface quantities Historically, glaciers have heen studied one at a time, with a fixed set of continum physical rnles and sep- arately measnred houndary conditions appropriate to each glacier (Paterson, 1994). Glacier volumes, for example, are usually acquired by detailed radio echo sounding. Statistical scaling theories like the one pre- sented in this paper, on the other hand, address the properties of many glaciers simultaneously. With this approach, glacier volumes are determined by statisti- cally valid relationships to other kuown quantities, such as surface areas. In parliculas, in Uis paper the Useorel= Copyright 1997 by the American Geophysical Union, Paper number 971301696, (0148-0227/97 /97JB-01690809.00 ically derived relationship between volume and surface area is confirmed by establishing empirically observed trends from over 100 different glaciers For most glaciers (excluding special cases such as surging) the accumulation and ablation of ice controla terminus advance and retreat, iee volume, and ourface geometry. In particular, these quantities all vary with perturbations in the ice mass balance rate b (rate of ice input minus output, due to accumulation and ablation at the surface), and therefore global and regional trends in glacier geometry can be used as indicators of chang- ing climate. In general, on any one glacier, there is an accumulation area at relatively high elevations (b > 0 ona yearly average), and an ablation area at relatively low elevations (<0 on a yearly average), but bal- ance profiles as. a function of elevation or horizontal distance are known im detail for ony a few hndred of the ronghly 160,000 glaciers worldwide (Meter and Bahr, 1996; Haebert et al., 1994). The limited existing observations do suggest that the balance profiles versus elevation are nonlinear and perhaps quadratic (Dyury- erou, 1995; Dyurgerov et al., 1995; Haeberti et al 1994; ‘Meier and Post, 1962). It is not known if quadratic pro- files are characteristic of all or just a few glaciers, but there is some evidence that balance profiles are simi- lar for many glaciers. By the end of the summer melt season, winler snow accumulated ou the surface of a wlacier has melued back Wo au easily observed elevation voulour where b = 0 (called the equilibrium line), and ait photo obser vations of this contour show that the ra- 20,385 386 tio of the acenmnlation area to the area of the entire glacier (accnmnlation area ratio (AAR) falls within a narrow range of values (ronghly 0.5 to 018) for nonealv- ing glaciers in a near steady state condition [Meier and Post, 1962) In the following analysis we demonstrate that the AAR is related to the mass balance profile and that Doth are intimately tied to a volume (V) and surface area (S) scaling relationship. If the mass balance pro- file cranges (periiaps due to a change in climate), then Uke AAR and the exponent 7 in Vo S* also change in a manner predicted by a rescaling of Ue anass aud ‘momentum conservation equations. The exponent °y is insensitive to small changes in the mass balance, but larger changes (easily measured with AAR) can have 1 dramatic effect on ice volumes. ‘The paper is organized as follows. ‘The next sec tion outlines common assumptions and their shorteom- ings for volume-area relationchips. The following cee tion uses a scaling analysis to relate glacier volumes to surface areas. The scaling analysis requires four clo- sures deseribing the behavior of glacier width, slope, side drag, and mass balance. ‘These closures are ex- plored in a following section and summarized in a later section. ‘The sensitivity of volume-area sealing to mass hralance pertmrhations are then discussed, and the con- clusions reiterate the primary volume-area resnlts. Kor reference, the notation section defines variables nsed in the text, Volume Size Scaling ‘The volume ofa glacics is proportional to the product of ts length 2, width w, and an average of its thickness A. In other words, [V] 0 [z][w][A] where brackets in- dicate characteristic values. Similarly, the ourface arca is [5] « [al[w]. With some simple geometrical objects (e, spheres), the length, wideh, and height are ex pected to scale in the same manner, so that [V] « [2° ‘and (S] a [z]2, and [V] oc [S)°/2. For glaciers, howover, data show that (V] 0 [8]? where 7 = 1.36 (see Figure 1) [Zhurovlyen, 1985; Macheret et al, 1988; Chen and Omura, 199], and tor ice sheets -y = 1.28 (Paterson, 1972]. ‘The impheation is that the length, width, and thickness of glacters do not. scale wentacally. Different. sealing hehaviors for length, width, and thickness are usually attributed to a combination of ice dynamics and geometrical boundary conditions [e.., Paterson, 1972]. For ice sheets, with no surrounding valley walls and essentially flat beds, it has long been argued that [2] = fw, and therefore (V] cc [c]?(1] and {S] x (x]?. From arguments which assume perfectly plastic ice and/or uniform accumulation and ablation, it follows that fx 2°", (ti) (2), and then *y = 1.25 fees Paterson, 1972; Nye, 1951]. Similarly, for glaciers confined by mountain valleys it has been suggested that the width is roughly constant, and therefore [Vox fell] while [5] « [2] [e., Paterson, 1994]. Dor perfect plas- BAHK BT AL: GLACIEK VOLUME-AREA SCALING 0.01 1 100 surtace area, km? 10000 Figure 1. V versus § for 144 glaciers, uot iuclud- ing ice caps, plotted on a log log scale. Data were col lected from Emrope, North America, central Asia, and the Arctic (a list of 41 references is available from the authors upon request). Only those sources with recent reliable radio echo soundings were used for the volume estimates. As in a previous analysis with fewer glaciers, [Chen and Ohmura, 1990], the steaigt live fit has regreesion coefficient of 1.36. ‘The squared correlation coefficient is H# = 0100. ‘Fhe data are discussed in more detail by Meier and Bakr {1996] and in a paper in prepa- ration by the authors. Exponents in the range from roughly 1.3 to 1.4 have been suggested by sovoral Rus- sian authors as well (eg., for 103 glaciers between the Altai and Tien Shan mountains, Mucheret et ul. (1988) find an exponent of 1.979 0.066 with an I? ~ 0.95). ticity and flat beds this suggests 7 = 3/2 which is a poor match to the data (Figure 1). However, these previous derivations make assump- tions which are useful only as low order approximations. Glacier beds are rarely flat, ice is not a perfectly plastic solid [Glen, 1958], accumulation and ablation are not uniform, and while intuitively appealing, valley glaciers are not one-dimensional surfaces with widths scaling as a constant. As a simple counterexample, consider a glacier which is completely contained within a moun- iain valley with a characteristic parabolic crass section As the thickness ofa glarter mereases, the width also in- creases as hoc w®, Therefore {w] oc [A]!/? and {w] 18 not constant. Also, data compiled for some large glaciers ‘hich are not contained within a single valley (e.g., the Barnard, Columbia, Harvard, Knik, Matanuska, Rus- sell, and Tazlina, all in Alaska) show that (1] oc (S]°* ‘where {z] in this case is the average along-channel length [Bakr and Peckham, 1996]. With [5] = [z][u], the observed relationship implies that for these very large ranching glaciers (w) o [x], which is the same as the wid scaling conumouly assumed for ice sheets. To in- ude this and all utler possibilities, in dhe following analyses we will assume that luo ta" o where q is some constant. BAHR ET AL: GLACIER VOLUMB-AREA SCALING With the width ecaling in (1), Ih (2a) [5] o fa}. (2b) Tu the next section a scaling analysis of Wie nas aud ‘momentuin equations will give a relationship between [h] and [2] which is used to relate [V] to [5] and then to predict 7. [x Sealing Analysis From observations, [V] oc [5]! with 7 © 1.36. There- fore (e[](A] (2]"fu}* and [A] oe f2]*—Mwp2, te follows that [A] o (2)? where @ = (q+ 1)(y—1). Our hypothesized relationship for width scaling (equation (1) therefore implies that {f] must be a power law In [c}. (From dimensional analysis this is not unexpected: ‘the Buckingham Pi theorem often gives power law re lationships between characteristic: auantities [Sclmnidt ‘and Housen, 1995].) The coustant é is derived explicitly from a scaling aualysis of due pavtial differential equatious desexibing glacier dynamics. This scaling can be accomplished us- ing many different techniques including the “shallow ice” nondimensionalizations of Hutter [1983], Morland [1984], Fowler [1992] and others, or a stretching sym- metry analysis as detailed in the literature on similar ity methods [e.g., Logan, 1987]. Both approaches give the same results, relating characteristic values for each slacier parameter (thickness, velocity, ete.) to the char- acteristic values of other glacier quantities. There is a large body of observational and analytical evidence sug- gesting that shear deformation is the principal mode of glacier flow with longitudinal compression and exten- sian playing a secondary role, particularly for small val ley glaciers and especially near the bed where the mast significant deformation 1s occurring [Paterson, 1994) By eliminating the typically small nonshear terms in| the flow equations, both techniques show that charac- teristic quantities (indicated by square brackets) scale as follows. Choose an zaxis parallel and a zaxis perpendicular (upwards) to the mean surface profile of a glacier. Let er be the angle the zaxis makes with a horizontal which i perpendicular to gravity. Assume Glen's [1958] non ear flow law. Then from mass conservation (continuity) (ae (Ba) ‘and from momentum conservation (ice flow) 2) o [Allo}"Eae] "AN" FT" (3b) where us is the down-glacier or x component of the ve- locity, gx = gsina is the x component of gravity (9), and p is the ice density. F is a shape factor which ac- counts for drag on the valley sidewalls and glacier bed onast Lye, 1965]. Variable nis a How parameter, with n -» co for perfectly plastic behavior (Paterson, 1994], n = 3 typical for ice, and n = 1 giving a Newtonian constitue tive relationship with viscosity (24) ?. A is a tempera. ture dependent parameter, althongh we will ignore this dependency in the simple scaling. analysis. considered hore; sothermal conditions are reasonable for many of the world’s mountain glaciers (considered in Figure 1), although isothermal conditions ate inaccurate for most {ce caps and ice sheets [Paterson, 1994]. Note that as in other scaling analyses [eg., Paterson, 1972}, equations (Ga) and (3b) assume constants of proportionality which do not change significantly from glacier to glacier. This common assumption is partially justified with data in Bair (199Ta}, Chen au Ofowure (1990), aul iu wore de- (ail by Peckham, in progtess, (Also, the “constant” of proportiouality may not strictly be constant, but may have a distibutiow abut a uvusinal value, a discussed later for mass balance.) ‘The scaling relationships in (3) can be manipulated to express [h} as a function of [2] (neceasary to express [V] 29 a function of [S]).. Combining equations (a) and (Gb) to eliminate [uz] (le) o (Al***LAllp" (gal FY" The constants A, p, and g are largely independent of 2 (they do not vary with the size of the glacier), so wt? ge lla] yo OE @ To completely express {h] a6 a function of fa], closures for [sinal, [F], and [6] must be hypothesized. For the moment we propose that [sina] «x [r}-" (5a) Fleer 6) box [x]”™ (5e) for some constants r, J, and m. ‘The power law forms are consistent with expectations fram the Buckingham Pi theorem [Sehmidt and Housen, 1995] and wil be jus- tified in the next. section. Substituting (5) into (4), Un fay RS © ‘The volume-area relationship ean now be constructed from (1), (2), and (6) (V] & felteltrl [V] x faite eS IV] o< [gy aan 7) ‘Therefore 7 Is fixed for any choice of the scaling expo- nents g, r, f, m, and nm. Appropriate values are sug~ gested in the following section. 20,358 Choices for Closure ‘Ty complete dhe relationship between volume and surface area, four closure choices suust be nade, one for ‘each of the scaling exponents related to glacier width (a); slope (r), side drag (f), and mass balance (m) (the ‘exponent n is fixed by the choice of a constitutive law). These closures are each addressed separately. Width Closure As discussed above, q = 0 (width scales as a constant) is inappropriate, and g = 1 (width scales with length) is expected for large ive caps and ive sheets. A study of, surface area, length, aud widil data from au inventory of over 24,000 Eurasian glaciers and 5400 Alps glaciers suggests that [uu] « [2]" [Dahr, 1997b]. Tn other words, q = 06, agrees with available valley glacier data, so this is selected as the most likely value, Slope Closure If surface slopes are small (as with most very large glaciers), then as detailed in shallow-ice approxima- fl fal ) (essentially “tise over rm”) [o.g., Hutter, 1983]. In this case, q is not independent of the other scaling constants, gad from (60), (6) and (8), [e]-" ye, [sina] ox (l-m+n-nf) a@+ I This will be referred to as “shallow slope sealing.” On the other hand, for the large surface slopes typ- ical of steep cirque glaciers, [sina] cannot be rescaled. In these cases the surface may follow the unpredictable bed topography, so that r = 0 seems most appropriate (Ukis unscaled version will be referred to as “steep slope scaling"). However, longer glaciers typically have shal- Tower slopes, su au inverse sealing with length, such as r= 1, may also be appropriate. Without some well defined bed profile, choosing au exact value for slope closure is difficult. @) Side Drag Closure ‘The shape factor can be approximated as P= Ad/(hP) where Az is the cross-sectional area, and P is the perimeter length excluding the surface (i.., perimeter of contact between ier and hed) [eg Budd, 1969). In tuitively, the cross-sectional area divided hy the basal perimeter is just a measure of the glacier width. "This Js shown rigorously by assuming a roughly polynomial shaped cross section, fh o w? for some constant f (with 5 = 2 for parabolic channels). ‘Then the perimeter is given by the arc length formula BAHR ET AL.: GLACIBR VOLUME-AREA SCALING which scales as [fh] (because P is essentially the integral of a derivative of A). Similarly, the cross-sectional area (20) Ae a Cb af mean wr lo which scales as (w}?"'. Therefore the shape factor sealles as [F] 0 [Ae]/((A][PI) 0 fw)-*, or [F] [u]/(h. (aa) In other words, the scaling exponent for the side drag is not independent of the other exponents which already describe the scaling of (w] and {A}. From (6) and (12), [a|-t — [a]? EHF". Solving for f gives J=1/2(m+nr+1-a(n+2)) (aa) ‘The approximation F = Ae/(hP) ia valid when the half-width divided by the thickness is not too large, 4 good approximation for many valley glaciers (Budd, 1060]. When the halfwidth is large relative to the thick ness, there is little side drag, and the shape factor tends to H = 1 [Nye, 1965]. ‘Therefore for ice caps the side drag sealing exponent is expected to be f = 0. Althongh onr goal ss to predict the volume-area seal- ing exponent 7, i observed value of 7 16 for valley glaciers (Figure 1) can be used to suggest: a reasonable value for f. Recall (h] 0 (VI/[S] 2 fz}("*P0-, so (w}/{A] ce [al Y, For g % 0.6, as suggested above, fu]/[A] 0c (21°, In other words, the data show hat fs 0 is expected. For g = 0.6 and r = 0 as sug- ested above, and for m = 2 as suggested below, (13) abso predicts that f= 0 (when n= 3). As a reasonable closure value, Uerefore we will use f= 0. Mass Balance Closure ‘The mass balance i isa funetion of position 2 hecause it is a climatically imposed boundary condition which can vary from point to point. As described elsewhere [e.g.. Jéhannesson et al., 1989], the characteristic value of the mass balance {}) is given as the magnitude of the mass balance at the terminus. However, if a glacier grows longer, then the terminus will move to a lower ele- ‘ation and (i will be correspondingly larger. Therefore (O] depends on the characteristic length. Rearranging (4) shows that the length dependence must be power Jaw in form, as loug, as the slope aud side drag also scale as power laws; recall [A] x [V]/{5] w [2], thir isin o)*LFI ta] JrHNab Dwr —t @ fh « (aa) BAHK ET AL: GLACIER VOLUME-AREA SCALING. Again, a power law form is a reasonably expected con- sequence of the Buckingham Pi theorem. Field data suggest that typical glacier mass balances have the approximate form be tut” +00 (15) with m re 2 [Dyurgerow, 1995; Dyurgerov et al., 1995; Hacherli ct al., 1994; Meier and Post, 1962] and posi tive constants cm and >. For each individual glacier, this form can be thought of as the dominant terms in 1a power series expansion of the actual mass balance profile. In other words, the real mass balance profile on any particular glacier will have some complicated shape represented by the expansion (a6) ‘The dominant terms which deseribe the overall shape are given in (15). Note that em and cp are constants for ‘any one glacier but can have different values for different, glaciers Assume for the moment that two different. glaciers have similarly shaped bit different mass halance profiles given by b= —rqr™ + ro (for the first glacier) and b= ~émz™-+ép (For the second glacier). Let the glaciers have lengths L and L (see Figure 2). If both glaciers are in balance then (17) and by integration (18) ablation =< b ~~ oq x" + 60 Diagram of the balance rate profile and ae- gson and ablation areas equal for steady state 20,359 ‘Therefore, at the terminus, b(L) = —CmL™ + a = —mco. Using (18) and an analogous equation for the second glacier, ¢m and é,, can be eliminated, and é a) 1 (m+1)(e/L)™ 0) 57 \meo)\1- (m+ aye@yiym If glacier lengths are sclected as the characteristic lengths, then the second factor is O(1), which sug gests that the appropriate choice for tho characteris tic maso balances should be the magnitudes at the ter minus [6] = -m{co] and [2] = ~m{é]- This choice agrees with previous analyses which also select the bal- ance at the terminus as the characteristic mass balance (Johannesson et ul, 1989). Equation (18), however, shows an alternative way to ‘write the characteristic balance at the terminus. mri [8] = -mfea] — -femla” (20) (where glacier length is selected as the characteristic length [x]). Both parts of (20) express the character- istic mass balance as the magnitude of the balance at the terminus, The right-hand side of equation (20) just makes the hidden dependence om characteristic leugth explicit, and this is the form we will use to help express the characteristic volume as a function of the charac- teristic surface area in (7). ‘As an important side note, recall that b is typically expressed as an increasing function of altitude =. With 1 change of coordinates that follow the aurface of the slacier 2(z), b can be expressed just as easily as a func tion of length along a glacier, as done above. Note that when written as a function of altitude, no one would suggest that glaciers with a larger altitudinal range have larger mass balances. This is because the activity index ‘b/d has a random distribution of possible values. In ‘other words, the slope of 6(z) can change from site to site depending on climate, so two glaciers with the same altitudinal range may span very different ranges of mass balance. Similarly, the slope of h(x) (roughly cy.) has a random distribution of possible valnes. We refer to this slope Gb/0tr as the “balance index.” Two glaciers with the same length but different balance indices may span entirely different ranges of mass balance. Likewise, de- pending on the balance indices, a longer glacier does not always have a larger characteristic mass balance than a different shorter glacier. The characteristic mass bal- ance must still scale with length (equation (20)), but Une comstants of proportionality [cm] can change (ran- domly with climate) for different xlaciers inaly, to suggest an appropriate closure exponent 'm for the mass balance, we consider accumulation atea ratios which can be expressed as a function of m. Sup- pose a glacier runs from x = Oat its head tox = L atits terminus, and suppose the equilibrium line is at (igure 2). Assume a balance rate profile given by (15). 20360 AHR RT AL Note that the balance profil is concave downwards, 28 observed for real glaciers [Dyusgerey, 1996; Dyuryeron thal, 15; Macher a, VMN, Mer and Post, 182] By aasuming appremimatsy steady state conditions, the total mass balance over the length of the glacier 0, and gs derived in (18), ca/ém = L™/(m-+1). Note that 6(req) = 0, so from (15), co) and then by eliminating ep and em, L t4-—. 22) 1 in ym ”) ‘Now recall that the accumulation area ratio is the accu ‘mulation area divided by the total glacier area. There- fore for a given characteristic or constant width, f AAR = “UIE 1 2 aan = (73) Accumulation area ratios are observed to fall approx- imately within the range 0.5 to 0.8 with typical values near two thirds for small mountain glaciers in a steady state balance [Meier and Post, 1962]. From (23), this 11. However, glacier inventory data for over 24,000 Eurasian glaciers gives an aver~ age AAR of 0.578, and inventory data for over 5400 glaciers in the Alps gives an average AAR of 0.580 [Bahr, 1997b]. ‘These ARs both correspond to ex- ponents of m = 20. This agrees with our intuition about the quadratic shape of balance profiles [Dyurg- rou, 1995; Dyurgerow et al., 1995; Hacberli et al., 1994 Meier and Post, 1962], 20 m = 2 is selected as a reason. able closure value. (23) spans mm = 1 lo 1 Closure Summary Although many closures are possible, for valley gla- ciers q ~ 0.6 (width exponent) and m = 2 (mass balance exponent) are both suggested by data. The value f = 0 {for side drag may also be appropriate. There is no sup- porting data for the slope closure, but of the reasonable values discussed, only the steep slope scaling r= 0 is consistent with the other closure ehuices for gsm aud f Fur ice caps aul ice sheets, theory requires g — 1 and J =0. Shallow slape scaling is reasonable, so from (9), r= (4= m) /B. If we assume that ice caps and ice sheets follow a roughly “square root” shaped surface profile as suggested by data [e.g., Paterson, 1972], then rm 1/2, In that case m = 0, a2 frequently assumed clocwhere [¢-g.5 Paterson, 1972]. GLACIER. VOLUMIE-ARBA SCALING With these closure valnes, volnme-area sealing is pre- dicted fram (7) with y = 1+ ((1+m-+n(f+1))/ (2+ 1)(n+2)). For valley glaciers the closnre choices predict 7 = 1.375, and for ice caps closure choices predict 7 = 1.25. ‘These are both in excellent. agree- ‘ment with all of the available data (Figure 1) (Paterson, 1972; Chizhow and Kotlyakov, 1982; Zhurovlyeu, 1985; ‘Macheret et al., 1988; Chen and Ohmura, 1990; Meier and Bahr, 1996]. Other closure chotces for g, m, f, and the sane values for 7. However, for val ley glaciers, if we asume hal q = 0.6 aud m = 2 are roughly fixed by data, then f aud r must have opposite signs and similar magnitudes to give a scaling exponent anywhere neat the "7 ~ 1.36 which is supported by the jearea data. Doth f and r are expected to be ve (sealing inversely with length as shown in (5)), 20 f and r must both be near 0, as was assumed above. Aehough the accumulation arca ratio date docs not support the value of m = 0 for valley glaciers, this ia ‘ common assumption. In that case, for q = 0.6, the volume-area scaling exponent of 1.36 can only be repro- duced if f+r ~ 0.6. As discussed above, f must be near Owhen 7 = 1.36 and when q=0.6. Therefore r ~ 0.6, which is a reasonable exponent for slope scaling. Again, however, m =U) seems unreasonable; valley glaciers al- ways have an accumulation area and an ablation area, which requires some kind of change in the mass balance with length (Le., nonzero m; see (15) Although we have assumed throughout this paper that n = 3 in accordance with Glen's flow law, the scaling predictions can change with different m. If per- fect plasticity is a reasonable assumption for ice sheets [iNye, 1951], then v approaches infinity, and the volume- area scaling exponent immediately reduces Lo y = 1.25, as observed. Shallow slope scaling is still assumed, but ‘because m is so small relative to the infinite magnitude of n, the value of m is irrelevant in this case. Discussion or most glariers, mass balance isthe houndary con dition which mist ultimately drive changes in glacier vohimes. Homver, fom experience and modeling ex. periments, we know that the volume is relatively in- sensitive to variations in the shape of the mass bal- ance profile {e.g., Paterson, 1972]. Examining (7) shows why, The mass balance scaling exponent m changes the volume-area exponent by an amount m/(q- 1)(n +2). For n = 3 and the observed valley glacier value of @ = 06, this means that mass balance contributes only m/8 to the total voluime-area scaling exponent. For ice caps where q = 1, the contribution reduces to m/10. Therefore w 10% ‘chau in on will only alter the volume-atea scaling exponent by roughly +1/100. Small changes in m will have aluost no effec on glacies volumes. While m/8 and m/10 can be small, they ase not in- significant. If m = 2 as suggested by valley glacier BAHR ET AL. GLACIER VOLUME-AREA SCALING data, then the mass balance sealing contributes 0.25 to the total of 7 ~ 1.36, This is a contribution of roughly 20% of the total volume-area sealing exponent. So small ‘changes in m may be insignificant, but the magnitude of ‘m determines a sizable fraction of the volume-area ex- ponent. Hence large changes in m will cause noticeable ‘changes in y and glacier volumes TLarge changes in m ean be tracked by AARS. In gen- eral, the AAR. is not sensitive to small changes in m (equation (23)). However, any measured changes m the long-term average AAR must indicate larger changes in m. A reduction from an average AAR = 0.58 to AAR = 0.50, for example, indicates a change from m = 2 to m= 1. For valley glaciers, this could in- dicate a change from the current value of 7 = 1.36 to 1 = 1.24, Fora typical valley glacier with asurface area of 10km?, this decrease in the AAR would mean a most 25% reduction in ice volume. Observed change: average AAR therefore indicate changes in the volume- area exponent and may indicate significant changes in lacior volume. Conclusions Characteristic glacier volumes depend on the product. of a characteristic length (x), width (w], and thickness [i]. The preceding analysis shows that these dimensions can be rewritten to express the volume as a function of the characteristic surface area [S], and four closure choices for the widih, slope, side drag, aud avass bal- ance. Although many closures are possible, reasonable choices (supported by data) predict that [V] « IST", with 7 = 1.375 for valley glaciers and 7 ~ 1.25 for ice caps and ice sheets. These predicted scaling exponents are supported by volume-area data (Figure 1). Changes in the volume-area scaling can be the result of changes in the specified width, slope, side drag, or ‘mass balance sealing. Of these, the mas balance seal- ing behavior is most likely to be altered by changing climate. Glacier volumes are insensitive to small per turbatione in the mace balance ecaling exponent m, but ‘measurable changes in average AAR are the result of much larger changes in m. ‘These large changes can indicate significant changes in -y and glacier volumes. AARs have the disadvantage of being noisy functions ‘of time and space {Meier and Post, 1962; Dugdale, 1972; Purhish and Andrews, 1984), so multiple venr or regional averages wonld he neressary to establish trends which might indicate significant changes in ice volume. How- ever, even with the tronble of multiple year averages, ‘measuring AARs and [5] from satellite images (or aerial photographs) is significantly less costly than detailed radar measurements of ice volume. ‘The volume-area_ scaling relationship gives both a practical and physi- cally based method for estimating glacier volumes. Notation 2 length along “horizontal” axis, 20361 : height along “vertical” axis, w width, h thickness. v volume. s surface area, e density. Ie eravity. 4 constitutive faw parameter. n constitutive law parameter Us velocity (down-glacier), F shape factor (side drag). Ac cross-sectional area. P perimeter along bed. i mass halance rate, m, 60 balance profile parameters, a axis angle. AAR accumulation area ratio. oq © position of equilibrium line. 6 cross-sectional shape exponent. 7 volume-surface area exponent, @ hueight-length exponent, t side drag exponent. m Dalauce rate exponent q width exponent. r slupe expouent Acknowledgments. We thank Tad Pfeffer for many helpful discussions, and we thank Per Holmlund, Charles Raymond, Joe Walder, and an anonymous reviewer for ex- ceptionally’ thorangh and helpfil comments which signif- ‘cantly improved this manuscript. Helpful comments were also made by Richard Hindmarch on an earlier draft of this paper. References Bahr, D. B., Global Distributions of Glacier Properties: A Stochastic Sealing Paradigm, Water Resour. Res, 93, 1660-1679, 19972, Bahr, D.B., Width and length scaling of glaciers, J. Glaciol, in prese, 1997b. Bahr, D. B. and S. 1. Perkham, Ohservations and analysis vf self-similar branching topology in glacier nevworks, J. Geophys. Res., 101, 25511 25521, 1996 Budd, W. F., ‘The dynamics of ice masies, Sei. Rep. 108, Aust, Nati, Autaret. Res. Exped., Melbourne, 1969, Chen, J. and A. Ohmura, Estimation of alpine glacier water resources and their change since the 187Us, JAHS Publ. 199, 127-195, 1990, Chizhov, O. P'and V. M. Kotlyakov, Studios of the Antare- tic and present-day concepts of global glaciation, Ann, Glaciol, 9, 80-58, 1982 Dugdale, 2 'K., A statistical analysis of some measures of he state of a glacier’s “health,” J. Glactol., 1(61), 73-79, 1072, Dyurgerov, M. B., Changes of the ‘activity index” of the wuortheru leiaisphere glaciers during global warmaing (ab- stract), Bos Trane. AGU, 76(16), Fall Mect. Suppl., F208, 19, Dyurgetoy, M., M. G. Kunakovieh, V. N. Mikhalenko, C. Lin, The mass balanco behavior in difforent regions, in Tien Shon Glaciers, edited by M, Dyurgerov, Chap. 7, pp. 121-129, Tuuss. Acad. of Sci., Moscow, 1996. 20,362 Fowler, A. C., Modeling ice shoot dynamico, Geophys. Ae ‘rophos. Bid Dyn, 6, 29-65, 1992 Pusbish, DJ. aud 3. Anew The use of hy psometry to indicate long. tern stability and seeponso of valoyglacore to changes i mass transfer, J. Glacial, 90, 199211, 1984 Glen, 3. Ws The ow ln ofc: A dicutson of the acsumap- tions made in glacier theory, their experimental founda tlons and conséquences, JASH Publ 47, 171-183, 1958. Macbesly, W., B. Heeren, aud M, Hoclle (Eis), Glacier ‘Mass Belance Bulletin (1092-1909), Bul. 8, Int Assoc of Hydrol, Se. - UNESCO, Zurich, 1998 Hotter, K, Theoretical Glacisiogy,D. Reidel, Norwell Mass, "aK Johansson, T., C. Raysoud, and B. Waddingion, Tine ‘cal for adjustment of glacire to changes i mae bal ance, J. Glaciol, 35,3503, 1989. Logan, J. D., Applied Mathematics, A Contemporary Ap- proach, chap. 7, John Wiley, New York, 1987. Macheret, Y. Y. B.A. Cherkasov, and LI, Bobrova, Tls hina i ob'em Iednikow, Djungaralogo Alatan po. dan hiy aeroradiorondiewaniya, Mater Clyatenlogirkeskth Tested. Kronika Obmushdenina, 62, 59-7, 1968 Meier, M. F- and D. B. Babr, Counting glaciers: Use of Scaling methods to estimate the number and size distri- Buti ofthe places uf Uke worl iu Glaciers, fe Sheets and Volcanoes: A Tribute to Mork E. Meier, Edited by 8 Golbeck. Spee. Rep. 90-27, 89-94, Cold Regions Research tnd Engineering Unborotory, Hanover, 196. Meier, MF) and A'S Post, Recent variations in mass net Duuigets of laces in western Noreh Arterica, 45H PUL 58, 63-77, 1962 BAHR ET AL: GLACIER VOLUME-AREA SCALING Morland, L. W., ‘Thermomechanical balances of ice sheet flows, Geophys. Astrophys. Fluid Dyn., 29, 237-266, 1984, Nye, J. F., The flow of glacters and ice sheets as a problem in plasticity, Proc. R. Soe. London, Ser. A, 207, 594-372, 19851 J. F., The flow of a glacier in a channel of rectangular, ‘elliptic or parabolic croas ccction, J. Glaciol, 6, 681-690, 1968. Palersou, W. 8, B., Laurentide Icv Sheet: Estimated vol- ‘umes during Late Wisconsin, Rev. Geophys. 10, 885-917, 1912. Paterson, W. S.D., The Physics of Glaciers, Srd ed., Perk ‘amon, Tarrytown, N. Y., 1994, Schmidt, R. and K: Housen, Problem solving with dimen- sional analysis, Ind. Phys.,1(1), 21-24, 1995. Thorovlyev, AB, Korrelyatsionniy metod ozenky sapasov da vlednikakh, Mater. Glyatstologicheshith Issled. Khro- nika Oboushdeniya, 68, 241-249, 1986. DH Fahr, MF. Meier, Institute of Arctic and Alpine Research, Campus Box 450, University of Colorado, Boul- der, CO '80300 0450. (e mail: David. Bahr®colorado.eduy Mark, Metertleolorado edt 8. D. Peckham, Waler Resourcis Division, U. S. Geolog- ical Survey, 3215 Marine St., Boulder, CO $0303 1066. (c ‘mail: pecktam#@usgs.gov) (Recetved August 7, 1990; revised May 13, 199/; ‘accepted June 6, 1997.)

You might also like