Glacier Surge Mechanism: 1982-1983 Surge of Variegated Glacier, Alaska Author(S): Barclay Kamb, C

Home , Surge (glacier)

Glacier Surge Mechanism: 1982-1983 Surge of Variegated Glacier, Alaska Author(s): Barclay Kamb, C. F. Raymond, W. D. Harrison, Hermann Engelhardt, K. A. Echelmeyer, N. Humphrey, M. M. Brugman, T. Pfeffer Source: Science, New Series, Vol. 227, No. 4686 (Feb. 1, 1985), pp. 469-479 Published by: American Association for the Advancement of Science Stable URL: http://www.jstor.org/stable/1694144 Accessed: 01/11/2009 22:59

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=aaas.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

American Association for the Advancement of Science is collaborating with JSTOR to digitize, preserve and extend access to Science.

http://www.jstor.org 1 February 1985, Volume 227, Number 4686 SC I E NCEE

the late 1920's or early 1930's, about 1947, and 1964-1965(4). The recurrence period is thus about 17 to 20 years, and a surge in about 1981 to 1985 was expected. Glacier Surge Mechanism: 1982-1983 In 1973a continuingprogram was initiated to monitor the glacier in its "nor- of Glacier, Alaska mal" state and to follow the buildup to Surge Variegated the anticipated surge (8). The ice mass was found to be already at the melting Barclay Kamb, C. F. Raymond, W. D. Harrison point throughout,ruling out the idea that Hermann Engelhardt, K. A. Echelmeyer, N. Humphrey surging occurs when cold basal ice warms up to the meltingpoint (5). Figure M. M. Brugman, T. Pfeffer 1 shows a map of the glacier and indicates our terminologyfor its component parts and for the longitudinalcoordinate (expressed in "Km") by which we desig- Glacier surging-popularly called recognized as one of the outstanding nate the positions of points along the "galloping"-is one of the most dramat- unsolved problemsof glacier mechanics. length of the glacier. Measurementsre- ic phenomenaof glacier motion. A surg- It is also of wider interest, because of the vealed that from year to year the ice was ing-typeglacier, after flowingalong in an possibilities that glacier surges may im- thickeningin the middle and upper part apparently normal manner for years, pinge upon works of man and that surg- of the glacier (upstreamfrom Km 12)and speeds up for a relatively short time to ing of the Antarctic ice sheet may be a thinning in the lower part (downstream flow rates as much as a hundredtimes factor in the initiationof ice ages and in from Km 12). Above Km 15 there was a the normal rate and then drops back to cyclic variationsof sea level (7). Glacier progressiveincrease in ice flow velocity: an apparently normal flow state. The surging also bears a relation to over- in the middleglacier (-Km 11)it approximately doubled between 1973and 1981, and in the upper glacier (-Km 5) it Summary.The hundredfoldspeedup in glacier motion in a surge of the kindthat approximatelyquadrupled. Summertime took place in VariegatedGlacier in 1982-1983 is caused by the buildupof highwater flow velocities in 1981 were 0.4 to 1.0 pressure in the basal passageway system, which is made possible by a fundamental m/day in the upper glacier (Km 3 to 10) and pervasive change in the geometry and water-transportcharacteristics of this and 0.1 to 0.2 m/day in the lower glacier system. The behaviorof the glacier in surge has many remarkablefeatures, which (Km 12 to 15). The terminallobe, below can provideclues to a detailedtheory of the surgingprocess. The surge mechanismis Km 16, was essentially stagnant. akin to a proposed mechanism of overthrustfaulting. Until 1978the flow increase was more marked in summer than in winter. This indicated that the buildup was having a process repeats itself with a period (typi- thrust faulting and gravity tectonics in marked effect on basal sliding, which is cally in the range 10 to 100 years) that is the earth and to landsliding. thought to be the cause of the summer- thought to be roughly constant for any We have made detailed observations time speedup of "normal" glaciers (9). given glacier of surgingtype (1, 2). This of the surge of VariegatedGlacier that Startingin 1978, wintertimeflow veloci- phenomenon has intriguedand puzzled took place in 1982-1983. Our observa- ties began to show an anomalous in- glaciologists since it was broughtto sci- tions reveal many new facts about the crease, smallbut definite,which suggest- entific attention by the 1906 surge of surge process and provide the basis for ed that increased basal sliding was now VariegatedGlacier, Alaska (3). Evidence some informed reasoning about the occurringin winteras well as in summer. has accumulatedthat a good many gla- physical mechanismof surging.We pre- In 1978 we became aware that short ciers are of surging type, even though sent here a summary of the principal few have been ob- results at an in their evalua- B. Kambis a professorin the Division of Geologi- relatively actually early stage cal and PlanetarySciences, CaliforniaInstitute of served in the act of surging(4). A num- tion. Technology, Pasadena91125. C. F. Raymondis a ber of theoretical professorin the GeophysicsProgram, University of explanationsof surging Washington,Seattle 98195. W. D. Harrison is a have been proposed (1, 5), but progress professorin the GeophysicalInstitute, University of Alaska, Fairbanks99701. H. Engelhardtis senior toward a firm understandingof the phe- Backgroundand PresurgeBuildup researchfellow in the Institutfir Geophysik,West- nomenon has been slow because of a filische Wilhelms-Universitat,Minster, FederalRe- publicof Germany.K. A. Echelmeyeris a research lack of sufficientfield observationspro- Surges of Variegated Glacier are fellow and M. M. Brugmanis a graduatestudent at viding clear indications of the causative known believed on the basis of limit- CaliforniaInstitute, of Technology. N. Humphrey (or and T. Pfefferare graduatestudents at the Universi- factors (6). The surge phenomenon is ed evidence) to have occurred in 1906, ty of Washington. I FEBRUARY1985 469 6001'

17 + 60o0' . Terminal - b 9 . --- loIeo / A?

70 ? +Kj ^ 4 0 ^ 1 2 3 4kmir^_'n eB 4rK > 59 59'- 3. _ g

I I ! i I I . 139020' 139?15' 139?10' 139?5'

Fig. 1. Map of Variegated Glacier, Alaska, showing the longitudinal coordinate scale (0 to 20 Km) by which locations on the glacier are referenced in the text. Light arrows indicate the general direction of glacier flow. Heavy arrows show outflow streams near the terminus. The dashed line shows the boundaries of the part of the glacier that surged in 1983. Crosses indicate approximate positions of reference markers for which ice- flow velocity data are given in the text. Open circles are approximate positions of boreholes drilled to the glacier bed. The point labeled CA is the location of a crevasse whose water content was observed during February through May 1983. The location of a seismometer on bedrock is shown with an x north of Km 6.5; LL designates the location of the lake shown in the cover photo and discussed in the text; LTS signifies the lower terminus stream; CS and TS are other outflow streams.

15 A 10-

0 I I I I I I , ' I I ! I 15 30 15 30 15 28 15 30 15 30 15 30 15 30 15 30 15 30 15 30

Dec. Jan. Feb. Mar. Apr. May June July Aug. Sept. 1982

^ 15 E B 10- 0

CD ......

15 30 15 30 15 30 15 30 15 28 15 30 15 30 15 30 15 30 15 30

Oct. Nov. Dec. Jan. Feb. Mar. Apr. May June July 1983

50 - C 40-

30-

20-

10- I 15 30 15 30 15 30 15 30 15 28 15 30 15 30 15 30 15 30 15 30

Oct. Nov. Dec. Jan. Feb. Mar. Apr. May June July 1983

470 SCIENCE,VOL. 227 spurts of accelerated flow velocity were vations during the surge.) Flow-velocity trend. By mid-August the velocity had occurring in Variegated Glacier during measurements beginning on 26 March declined to about 1 m/day. It remained the summertime. These events, which showed that the upper glacier was mov- low through August and September, and we call "minisurges," were documented ing at surging speed, about 2 m/day. The the surging seemed to have stopped, in 1979 to 1981 (10). In an event of this course of the surge motion is shown in ending phase 1 of the surge. kind, the flow velocity increases rapidly Fig. 2. From late May to late June 1982 The surge motions in phase 1 took from -0.4 m/day to 1 to 3 m/day, in an the velocity in the upper glacier (Fig. 2A) place only in the upper glacier. The flow hour or two and then declines to -0.4 rose gradually from 2.6 to 9.2 m/day. On velocity was high and approximately m/day in 10 to 20 hours. Four or five the morning of 26 June the velocity constant from Km 4 to Km 7 and de- minisurges occurred in the early part of peaked at 10.4 m/day and then abruptly creased both up- and downstream from each summer of 1979 through 1981, in dropped to less than half its value in a this reach. At Km 9.5 the velocity was June and early July. The flow-velocity few hours. The motion thereafter was only about half that at Km 7, and the peak of each minisurge propagated as a pulsatory, with five major pulses of lower glacier, below about Km 12, was wave down the glacier over the reach movement (peak velocities of 3 to 7 practically unaffected. The surge motion from near Km 3 to about Km 9, at a m/day) superimposed upon a decreasing extended headward to within about 2 km propagation speed of about 400 m/hour. The velocity increase in a minisurge is accompanied by a simultaneous abrupt rise in the level of water standing in boreholes that reach the glacier bed. In such boreholes, once a hydraulic connection with the basal water system of the glacier has been established, the water level reveals the pressure in the basal water system (11). (By basal water system we mean the system of passageways by which water entering the glacier from the surface or sides and finding its way to the bottom is conducted down along the glacier bed and delivered as outflow at the terminus.) In minisurges the peak basal water pressure occurs synchro- nously with the peak flow velocity and is high enough to "float" the glacier (water level closer than 40 m to the surface in ice 400 m thick). It seems clear that the pulses of high flow velocity are caused by the peaks of high water pressure.

The Surge

The surge began in January 1982. The onset was detected by a marked increase in seismic icequake activity picked up by a seismometer located on bedrock 0.5 km north of the upper glacier (location shown in Fig. 1). (A strong correlation between seismic activity and glacier motion had been established in the minisurges and was later reinforced by obser-

Fig. 2. Glacierflow velocity versus time in the 1982-1983surge of VariegatedGlacier: (A and B) the velocity in the upperglacier in 1982and 1982-1983;(C) the velocity in the lower glacier in 1982-1983. Velocity values are daily averages where available, otherwise longer term averages. The averages are shown as horizontallines over the time intervals over whichthey were measured.Dotted lines show conjectured or interpolated velocities over intervalswhere short-termmeasurements are Fig. 3. Views of VariegatedGlacier near Km 13 takenfrom the same pointbefore the surge,July lacking. The data in (C) were measured on 1982 (A), and during the surge, 4 July 1983 (B). The view is upglacier. The scale, which is markersnear Km 15; (A) and (B) are a compi- identical in (A) and (B), can be judged from the fact that the width of the glacier is lation of data measuredon markersover the approximately1 km. In (A) the positionof the ice surfacein surge, as seen in (B), is markedwith intervalKm 2.8 to 8.3. dashed lines. 1 FEBRUARY1985 471 of the head of the glacier and of its main form an impressive surge front sweeping through the lower glacier to Km 15.6, at tributary, as shown by the dashed lines ponderously down the glacier. The prop- an average propagation speed of 23 in Fig. 1. agating surge front is depicted in Fig. 4, m/day. The forward propagation over Phase 2 of the surge began in October where it is seen as a downstream-facing the period 17 May to 13 June is docu- 1982 with a gradual increase in seismicity topographic ramp about 1 to 2 km long, mented in Fig. 4. During this time the and flow velocity in the upper glacier behind which the ice has thickened by sharply defined leading edge of the front (Fig. 2B). This continued until early Jan- about 100 m from its presurge thickness advanced at a nearly constant speed of uary 1983, when the velocity reached a (which can be judged from the dashed 80 m/day. This speed was maintained, stable level of 5 to 7 m/day, about three line in Fig. 4B). Within the front there is with only a slight decrease, up to the times the wintertime level in phase 1. An a high longitudinal gradient of the flow surge termination on 4 to 5 July. abrupt major drop on 3 February was velocity (Fig. 4A), corresponding to ex- The propagation speed of the surge followed by a gradual recovery to the tremely large compressive strain rates of front can be explained on the basis of stable level. In April the velocity began a up to 0.2 per day. (Compare these with conservation of ice volume (continuity). further gradual but substantial increase, typical rates of c0. 1 per year in nonsurg- If v2 is the flow velocity at the trailing reminiscent of the phase 1 increase in ing glaciers.) Ahead of the surge front edge of the front, h2 is the ice thickness May to June 1982, and reaching a broad the flow velocity is about nil, while be- there, and hl the thickness at the leading maximum of about 15 m/day in mid- hind it the ice is moving at full surge edge (where v 0), then continuity re- June, upon which large, complex fluctu- speed and the longitudinal velocity gradi- quires the front to advance at a speed V ations were superimposed, beginning ent is small and in fact becomes locally given by V = v2h2/(h2- hl). For the val- about 1 June. In the afternoon and eve- extensile as a velocity maximum devel- ues of hl, h2, and v2 seen in Fig. 4, this ning of 4 July the surge entered into ops around the topographic crest of the equation gives V values in the range 50 to an abrupt termination. The velocity surge-front wave from 23 May onward. 100 m/day, in rough agreement with the dropped in a few hours to about a quarter The leading edge of the surge front had observed speed of advance. This agree- of its previous value. There was a slight reached Km 13.5 by 15 February 1982, ment supports the conclusion that the recovery on 6 July, but by 8 July the and by 15 May it had propagated down advance of the surge front is a phenome- velocity had decreased to 1.2 m/day or non of kinematic-wave type, although less throughout the glacier. A slow de- different from the kinematic waves pre- crease continued thereafter. By 26 July viously proposed for glaciers (12). The the velocity in the upper glacier was no details of the surge mechanism govern more than about 0.2 m/day, lower than the propagation speed V via their effect velocities. on and presurge 'a h2 v2. In phase 2 the surge propagated pro- E gressively down into the lower glacier. Flow velocities there, which had been Detailed Flow Features of the Surge low or in the U very (0.2 m/day less) pre- 0 surge condition, rose dramatically as the 0L) Large pulsations or oscillations in the surge front arrived and reached 40 to 60 a surge velocity occurred during July 1982 m/day during June (Fig. 2C). The highest and June 1983; the latter are shown in measured was 65 on a) detail in 5. Three distinct of velocity m/day, 2 Fig. types June 9, for about 2 hours, at Km 14.5. fluctuations are recognizable. (i) Quasi- The surge termination affected the lower regular oscillations with a period of glacier similarly to the upper. By 27 July about 2 days are very pronounced in the the velocity was down to about 0.4 upper glacier (Fig. 5, A and B). From 19 m/day and by September to about 0.2 June to 6 July they have a very regular m/day. period of about 40 hours; prior to 19 June The surge produced great changes in the period is longer and more irregular. E the glacier (Fig. 3). The surface became (ii) Oscillations of shorter period, gener- N intensely crevassed throughout the surg- ally somewhat shorter than 1 day, are 0 ing part (delimited in Fig. 1). In the upper pronounced in the lower glacier (Fig. co 400 glacier, above Km 8, the ice was thinned 5C). (iii) Approximately every 4 or 5 from its presurge condition by as much w days there occur major slowdown as 50 m. Below Km 8, thickening oc- events, in which the velocity, after first curred, up to a maximum of 100 m at Km rising to a peak, drops rapidly, in a few 16. hours, to a level distinctly lower than was reached in the previous several days. The surge termination on 4 to 5 The Surge Front Longitudinal coordinate ((Km) July 1983 was an event of this type, as Fig. 4. Forward propagation of the surge was the major slowdown on 26 June 1982 The enhanced downstream transport front, shown by longitudinalprofiles of ice and the slowdowns that ended the subse- of ice in 1 of the a flow velocity (A) and ice surfaceelevation (B) movement in 1. The phase surge produced as a function of the coordinatei quent pulses phase longitudinal 1983 is general thickening of about 30 m in the in Km (see curvilinearcoordinate scale in Fig. major slowdown on 3 February lower half of the reach affected by this 1). The profiles are shown at five successive also similar except that it was not pre- phase, from about Km 7 to 12.5. In phase times in May and June 1983, at intervalsof 6 ceded by a peak in motion. Major slow- 2 this to as a to 8 days. The position of the ice surface in downs in June 1983 are marked with "bulge" began propagate 1982 is shown in with a dashed wave as the thicken- August (B) arrows in 5. Be- downglacier, and, line, and the approximate position of the downward-pointing Fig. ing increased further, the forward side of glacier bed as determinedby reflection seis- tween slowdown events, there is a ten- the wave steepened and sharpened to mology (8) is also shown. dency toward progressive increase in velocity, superimposed on the oscilla- micity that accompanied these events; mass. Thus about 95 percent of the motions. the propagationvelocity for the 21 Feb- tion is due to basal sliding and about 5 There is a very clear correlation of ruary event was 500 m/hour. percent to internal deformation,the lat- individual velocity peaks, troughs, and On 23 July 1983, well after the surge ter being at least roughly(within a factor slowdowns as observed at different termination, a minisurge occurred in of 2) the amount of internalmotion that points along the length of the upper which the flow velocity quickly quadru- would be expected to take place in the glacier (Fig. 5, A and B). The correlation pled and then dropped back to -0.4 absence of the surging.This result dem- is somewhatconfused as we move to the m/day in -10 hours. It was detected onstrates that the surge motion is caused lower glacier (Fig. 5C) because of the strongly at Km 9.5 and, 7 hours later, at by extreme enhancement of basal slid- increased numberof peaks and troughs Km 14.5. The time delay correspondsto ing-a conclusion often previously sur- there, but in many cases a correlation an average downglacier propagation mised but never before demonstratedby can nevertheless be identified;in partic- speed of 700 m/hour.This is at the upper actual observation. ular, the major slowdown events clearly end of the range of minisurgepropaga- Indirectconfirmatory evidence for the affect the entire surging length of the tion speeds (250 to 700 m/hour)observed large enhancementof basal slidingin the glacier. Close inspectionof Fig. 5 reveals in 1979 through 1981 (10). surge is given by comparison of trans- definite shifts in the time of individual verse profiles of flow velocity as ob- peaks, troughs, and slowdowns at differ- served under nonsurging and surging ent points. In general, these features Basal Slidingin the Surge conditions (Fig. 7). The change from a arrivelater at points fartherdownglacier, quasi-parabolictransverse profile of ve- indicating downglacier propagation of In orderto determinehow muchof the locity across the width of the glacier the velocity fluctuations. From 17 June surge motion is due to internaldeforma- under nonsurgingconditions (Fig. 7, A onwardthe propagationspeed is approx- tion withinthe ice mass and how muchto and D) to a flat profile, representing imately 600 m/hour.Prior to 17 June the basal sliding, on 8 June 1983 a borehole "plug flow" (14), under surge (Fig. 7, B propagation speed is generally faster, was drilled 385 m to the bed (13) at Km and C) is interpretablein terms of a great and in many cases the fluctuationsap- 9.5. The spatialconfiguration of the hole increase in basal and marginalsliding in pear to be almost synchronousthrough- was measuredon 10June and againon 12 the surge. out the surgingpart of the glacier. The June by means of borehole inclinometry Observationsof closely spaced mark- major slowdown on 3 February 1983, (Fig. 6). Over the 2-day interval the ice ers near the south marginof the profilein and also minor slowdown events on 21 moved forward 27.3 m at the surface, Fig. 7C reveal that there is a large veloci- Februaryand 22 and 31 March, showed while the bottom of the borehole lagged ty discontinuitywithin the ice mass near downglacier propagation effects as de- behind by slightly less than 1 m as a the edge. This discontinuity, markedby tected in records of strain, tilt, and seis- result of internal deformationin the ice shear arrows in Fig. 7C, occurs across a

r . , M --" A/i v^ ^/ ^ - IL

0 5-

20- i - I II I A

'g. ...v I j . .i I .....) I I .n J ? I t ....l . ..I ....s t M t *t .i. | t I . i , . 10 15 20 25 30 5 1983 June July Fig. 5. Glacierflow velocity versus time over the period7 Juneto 7 July 1983at (A) a pointin the upperglacier near Km 5.5, (B) a point in the middle glaciernear Km 9.5, and (C) a point in the lower glacierin the reach Km 13.5to 15. Each velocity value is plottedas a horizontalline over the time intervalover which it was measured.(The actual velocity as a functionof time is doubtlessa smooth curve throughthese data lines.) The graphshave a common time abscissa; the ordinatescales are offset as indicated.The downward-pointingarrows mark major slowdowns in flow velocity, which occurredthroughout the surgingpart of the glacier;dashed arrows are shown for the slowdown on 16 to 17 June because of the complexityof this event. Discontinuitiesin the data occur where there is a switch between differentmarkers whose motion is being followed. deep, sharp cleft in the glacier surface, tively modest and the lowest levels thickness of about 400 m; a borehole parallel to the margin and about 50 m in reached were at a depth of 80 to 90 m, water level at a depth of 40 m corre- from it (Fig. 8). The cleft is the surface whereas afterward the fluctuations be- sponds to a basal water pressure equal to expression of a discrete shear surface or came much larger and the lowest levels the ice overburden. Relative to this, the wrench fault within the ice and is analo- reached dropped to 120 to 195 m. In basal water pressures during surge are, gous to such shears sometimes seen in phase 1 of the surge, continuously high at a minimum, within 4 to 5 bars of the sea-ice packs. A cleft of this kind, usual- water levels (35 to 80 m) were recorded ice overburden pressure, and there are ly accompanied by heavy blackening of during June and July 1982 in boreholes at transient fluctuations often up to within the ice surface with rock debris, is seen Km 6.3, whereas in prior years the bore- 1.5 bars of overburden and infrequently at many places near the margins of the hole water levels had ranged over 50 to to as much as 0.5 bar above overburden glacier in surge. No such feature was 210 m except for the high peaks accom- (as on 27 June and 4 July 1983, see Fig. present before the surge. The fault prob- panying minisurges. As surge phase 1 9). After surge and also prior to surge (as ably extends down vertically to an inter- was ending, during July 1982, water lev- observed in 1978 through 1980), the pres- section with the bed, where the internal els in boreholes at Km 9.5 dropped pro- sures are as a whole lower, particularly velocity discontinuity across the shear gressively from 40 m to more than 150 m. the minimum pressures, which are 8 bars surface is transferred to the basal veloci- Superimposed on this declining trend to as much as 16 bars below overburden, ty discontinuity that is the expression of were several high peaks, which coincid- but there are also transient fluctuations rapid basal sliding. ed with the surge pulses that occurred to high levels. The highest of these, in during this period. Similarly, there is a which the basal water reaches or ex- strong correlation between water-level ceeds the overburden pressure, coincide Role of Basal Water peaks and flow-velocity peaks in phase 2 with the occurrence of minisurges, as in 9 and the that occurred on 23 Pressure in the Surge (Figs. 5B). minisurge July It thus appears that the glacier under 1983 after surge termination (see Fig. 9). The record of water level as a function surge is characterized by consistently Further evidence for the role of water of time in boreholes drilled to the glacier high basal water pressures, with peaks in pressure in the surge was the formation bed near Km 9.5 (Fig. 9) shows that the pressure that coincide with pulses in of a lake along the southwest margin of pattern of fluctuating levels underwent a surge movement. The pressure levels are the glacier at Km 15.2 in early June 1983 distinct change at the time of surge ter- best stated relative to the ice overburden (see cover photo and Fig. 1). The lake mination on 4 to 5 July 1983. Before the pressure, which at the borehole site was contained highly turbid water, typical of termination, the fluctuations were rela- about 36 bars, corresponding to an ice basal water charged with comminuted rock debris or "glacial flour," and it filled from water that welled up in a large - -o0 --- I? 27.3 m I.>- 0.100 A crevasse 100 m in from the margin and then flowed outward to the Thus Flow displacement 0.075 - margin. at surface the main source of the water was the 0.050 - E -i-iuu nn basal water system, not marginal a) The 0.025 - streams. highest lake level was co about 75 m below the July 1974 laterally averaged 10 June 1983 \ a) ice surface in the cross profile of the -200 3.0 - B glacier through the lake. The lake 12 June -: : drained away as the surge terminated. It 2.0 for a few hours on the morn- 0(i) / X reappeared 5. -300 t a ing of 24 July, as the minisurge was in 1.0 E progress. - / Flow displacement 0 > There is some evidence that the basal at bed 21 July 1982 1 . 0 A ' m//777777//' water was in midwinter -40077777///(/////////,26-6 ' ' 777 0 pressure high A I ..,, - 1- U 1 2U 3U 4U 10 - 1983. We observed "steam" (condens- Horizontal distance in flow direction (m) ing water vapor) rising from crevasses Fig. 6 (left). Ice motion at depth in the glacier, 10 - near Km 7 on 9 January, suggesting the measured from the of a borehole displacement - l presence of water in the crevasses at drilled to the bed at Km 9.5. - (82-E) glacier 5 shallow depth. On 17 February we suc- The location and shape of the borehole on 10 June and 12 June 1983 are shown, the individ- 9 June 1983 ceeded in lowering a probe to a depth of ual dots representing the individual borehole- 0 42 m in a crevasse at point CA in Fig. 1 inclinometry measurements. The abscissa is 0.25 - D (Km 7.5) and brought up samples of dirty horizontal distance in the flow direction o.20 - water from than 40 m. with an zero. The bore- depths greater (N64?W), arbitrary Since there is no surface source for wa- hole is plotted in projection onto a vertical 0.15 - plane in this direction. Arrows represent the o.10 - ter in midwinter and since surface-de- horizontal component of the forward motion rived water is generally clean, it seems of the over the The 0 05 glacier 2-day period. / 9 probable that the water sampled came July 1983 displacement at the surface was measured by from the bed and is evidence for the surveying to fixed bedrock points. Fig. 7 s o 400 800 N occurrence of high basal water pressure (right). Transverse profiles of ice flow veloci- Transverse coordinate (m) ty at Km 8 measured at four different times, at or before the time of sampling. A two (B and C) during surge and two (A and D) under nonsurging conditions. The location of the recording pressure transducer left in the line is indicated in I the transverse line of crosses at Km 8. The abscissa is profile Fig. by crevasse at point CA indicated that the distance measured in the direction N50"E from the southwest margin of the glacier. The arrows water level in the crevasse rose to a represent marginal shear zones. In (C) these occur as wrench faults in the glacier about 50 m in from the margins. The profile in (A) is from (8). depth of about 38 m in March and April. 474 SCIENCE,VOL. 227 The foregoing observations, particu- ing to the flood peaks in Fig. 10B. There through July of 1982 and 1983 (Figs. 2 larly when considered collectively, dem- is also a diurnal cycle of water input, and 5). A particularly suggestive in- onstrate a well-defined influence of basal which is not noticeably reflected in the stance of this is the major slowdown on water pressure on the motion of the outflow discharge. 26 June 1982, which occurred after the glacier and, taken together with theoreti- There does, however, appear to be a first 3 days of warm weather and rapid cal considerations (see below), make a relation between weather-related water melting of the season. On the other hand, strong case for the conclusion that high input and surge motion, although some- the slowdown on 3 February 1983, and basal water pressure is the direct cause what indirect and complex. A gradual the surge buildup in January 1982 and in of the high flow velocities in the surge. increase in low-level water input as win- November and December 1983, are phe- tertime conditions give way to spring nomena outside the above relationship, seems to be reflected in the progressive because they are correlated not with Discharge from the Basal Water System increase in surge motion in May through water input but rather with the lack of it. June of 1982 and 1983 (Fig. 2). Larger The surge produced substantial During the surge, floods in the outflow water input, however, results in surge changes in the outflow-stream pattern of streams at the terminus correlated close- pulses followed by slowdowns and ulti- the glacier. Prior to the surge, most of ly with rapid slowdowns of flow velocity mate termination, as occurred in June the outflow was in stream TS (Fig. 1). As in the surging part of the glacier. This was first observed in the major slowdown on 26 June 1982, which was accompanied by a particularly large flood. The correlation was observed repeatedly in the subsequent surge pulses in phase 1 and throughout phase 2. Figure 10 is a comparison between the flow velocity at Km 16.5 and the discharge of the lower terminus stream (LTS) over the period 17 June to 6 July 1983. Flood peaks occurred on 17, 21, 25, and 30 June and 5 July, in coincidence with major velocity slowdowns. Although the minor slowdown on 28 June (Fig. 10) was not asso- ciated with a flood in the LTS stream, there was a flood in the CS stream (Fig. 1) at this time, shown by the dashed line in Fig. IOB. There is a good correlation also between smaller floods and velocity slowdowns earlier in the season, from February to mid-June. In particular, the slowdown on 3 February 1983 was accompanied by a marked flood in the otherwise small and snow-covered TS stream, observed by an automatic cam- era. The termination of the surge was accompanied by a particularly spectacular flood. Not only was the flow of the principal outflow streams greatly increased, but, in addition, highly turbid water gushed and spurted from numerous cracks, crevasses, and holes along the leading edge of the surge front, particularly along its south side (dashed line south of Km 17 in Fig. 1). In an ice cliff above the TS outlet (Fig. 1), water spurted out over a vertical range of 20 to 30 m above the stream level. The flood peaks that correlate with the surge slowdowns and termination in 1982 and 1983 are a manifestation of rapid release of stored water from the glacier; only indirectly do they reflect water input to the glacier from rain or the melting of ice and snow. The water input is 8. Aerial view of wrench fault on the southwest side of the at Km variable in accordance with in Fig. marginal glacier 8.5, changes looking southeast. The fault is the vertical black line in the lower center. It is a deep cleft the weather, but the weather-related in- blackenedby rock debris. Ice to the left is movingtoward the observerat -10 m/daywhile ice put did not have high peaks correspond- to the right is practicallystationary. 1 FEBRUARY 1985 475 the surge front moved past this outlet, its by means of the standard water tracer dye tracer injection and at increased sam- discharge was greatly reduced and the rhodamine WT, detected fluorometrical- pling intervals thereafter (Fig. 11A). Trac- outflow appeared first in the newly ly at the fraction of a parts-per-billion er first began to appear in detectable formed outlet, CS (Fig. 1) and later, as level. This tracer is in common use for amounts in the outflow about 24 hours CS became shut down, in LTS. hydrologic studies of domestic water after injection. The concentration During the surge, the turbidity of the systems and has been used for similar climbed to about 1.7 ppb by about 50 outflow-stream water was much higher measurements in other glaciers (15). hours after injection and remained in the than before or after the surge, reflecting A borehole drilled 385 m to the bed at range 0.7 to 2.2 ppb until about 130 a much higher content of glacial flour. Km 9.5 on 8 June became strongly cou- hours, after which it slowly decreased, pled to the basal water system as evi- reaching the detection limit (-0.05 ppb) denced by strong pull-down of instru- about 15 days after injection. The mean Subglacial Water Flow ments into the hole, indicating a rapid transit time for the dye was about 90 rush of water down the hole to the bed. hours. The average flow speed of water The flow of water in the basal water While this condition obtained, at 11:20 in the basal system was therefore about system was studied by water-tracer ex- on 9 June, the tracer dye was introduced 0.02 m/sec over the 8-km interval from periments initiated on 9 June 1983, dur- into the borehole. The return of tracer at Km 9.5 to stream CS. The broad disper- ing surge, and on 14 July 1983, after the outflow stream CS, which was the sion of transit time (Fig. 11A) was ac- surge termination. The water transit time main outflow stream at this time, was companied by a lateral dispersion of between a borehole at Km 9.5 and the followed by collection of water samples tracer in the basal water system, shown terminus outflow streams was observed every hour for the first 24 hours after by the fact that the tracer appeared in all three outlet streams (CS, TS, and LTS). The foregoing behavior contrasts greatly with that observed in the water tracer experiment after surge termination. Tracer was introduced at 10:00 on 16 July via a borehole of depth 355 m located at Km 9.5. The observed tracer return in outflow stream LTS (now the Fig. 9. Water level in main stream) is given in Fig. lB. The a) Il- i\boreholes near Km tracer appeared at LTS as a sharp peak 0 - (a / 10 9.5, measured with (of peak concentration 40 ppb) 4 hours L- - (a downhole pressure after The flow I - ttransducers. n Record injection. average speed 0 over the 10-km interval from Km 9.5 to ll! I l I over 9 to 26 June is .0rr LTS was therefore 0.7 m/sec. This re- I from borehole 83-E, .0 I 26 June to 18 July sembles flow measured by a simi- a)0. speeds (1 from borehole 83-F, lar technique in normal alpine-valley gla- and 18 to 29 July from ciers No detectable tracer borehole 83-G. (15). appeared in streams TS or CS. These experiments show that a dra- matic change took place in the basal water system of Variegated Glacier from the surging to the nonsurging state. In June 1983 July the nonsurging state, the water flow speed was approximately the same as in normal, nonsurging-type valley glaciers, in the flow in the 50 whereas, surging state, basal water system was greatly retarded and lateral dispersion of flow was enhanced. c 40 -o

Basal Cavitation 30

Measurements of the position of a 60 marker stake in the glacier at Km 8 show that the glacier surface there dropped a 40 a) about 10 cm at the time of surge termination 12). in the E 20 (Fig. Similarly, major a slowdown on 26 June 1982 the glacier surface at point T at Km 6.5 (Fig. 1) dropped 70 cm. These vertical motions 20 25 30 | 1 may be the surface indication of the June 1983 July closing up of cavities between the base Fig. 10. Time variations of the discharge Q of the terminusoutflow stream LTS (B) plotted of the ice and the bed. Such cavities form the variationin flow v at Km 16.5 over the 17 June to 6 alongside glacier velocity (A), period when a slides over an July 1983.The dashed curve in (B) shows the dischargeof outflow streamCS, multipliedby 2, glacier irregular over the period27 June through2 July; it follows the curve for LTS over 27 June to 10:00on 28 bed with sufficient speed and at a suffi- June. ciently high basal water pressure, by a process known as ice-bedrock separa- plicable to the entire surging part of the semicircular tunnel, a few meters in di- tion or basal cavitation (16-18). The cav- glacier, the volume of water released ameter, carved by frictional melting into ities close by collapse of the ice roof would be about I x 106 m3, comparable the ice directly overlying the bed. The when the water pressure falls or the to the amount that might be released water pressure is well below (-10 to 20 sliding speed decreases, or both. In the from basal cavities. bars below) the overburden pressure of surge, the opening of these cavities is not The closure of basal cavities would ice, and the tunnel is kept open by the detectable by the method of Fig. 12, increase the area of contact between the frictional heat generated by viscous dis- because it takes place too slowly, except glacier and its bed, which would neces- sipation in the flow of water through it, possibly for the uplift of about 5 cm sarily increase the resistance to basal which enlarges the tunnel by melting at a between data points 1 and 2 in Fig. 12, sliding. This accords with the observed rate that compensates for the closure of which occurred in connection with the correlation between floods and slow- the tunnel by quasi-viscous creep of the speedup just before the onset of surge downs. ice under the ice overburden pressure termination. But in the minisurges ob- less the water pressure. We will call a served in detail in 1980 (10), each quick basal water system of this normal type a The Basal Water in speedup was accompanied by a rapid System "basal ice tunnel system." For Variegat- uplift of 5 to 10 cm, followed by a slower Surge and Nonsurge ed Glacier in the postsurge condition in drop back to the original level, which mid-July 1983, the diameter of the main constitutes an observation of both the In the nonsurging state, the transport tunnel was probably about 4.5 m, to opening and the subsequent closing of of water in the basal water system is account for the observed outflow dis- basal cavities. Iken et al. (19) have simi- thought to be describable by the model charge of about 20 m3/sec. larly interpreted surface uplifts and discussed by Rothlisberger (20). There is Water transport through Variegated drops connected with flow-velocity fluc- only one main conduit, or at most only a Glacier in surge cannot be in a normal tuations in Unteraar Glacier, Switzer- few, carrying most of the subglacial dis- tunnel system, because the observed av- land. charge. It can be approximated as a erage water-flow speed (0.02 m/sec) is An interpretation of the drop in Fig. 12 in terms of basal cavitation is subject to the qualification that a drop of compara- Hours aafter injection ble amount would be expected on ac- 0 100 200 300 400 count of the strain wave that accompa- A t nied the downglacier-propagating wave of velocity slowdown. In the minisurges 2- of 1980, we observed examples in which the strain that accompanied the uplift and subsequent drop was of such a char- acter that the uplift and drop must have been due to changes in basal cavitation, but in the surge termination in 1983 the 0 drop could have been due to strain, at least in part. If water-filled basal cavities open, this provides a means for storage of water at 10 15 the base of the glacier, and, if such : 1 983 June cavities close at the surge termination o Hours and the major slowdowns, water would c 0 20 0 be expelled, contributing to the floods Q"40 or B . ' that accompany these events. If the areal t B T average of the cavity height is 0.1 m, as the drop in Fig. 12 may indicate, the volume of water stored is about 1.7 x 106 m3 over the surging part of the glacier, about 17 km long and 1 km wide. Release of this water over a period of 20 hours, as in the outflow-stream flood at surge termination, would add an average discharge of about 24 m3/sec, which is comparable to the observed flood (peak discharge 50 m3/sec, Fig. 10). However, this contribution to the flood would be difficult to disentangle from that due to the release of intraglacially stored water. If the intraglacial void space that is con- 16 17 18 nected to the basal water system in such 1983 July a way that the contained water can drain Fig. 11. Results of water tracer experiments under surging conditions (A) and after surge termination out amounts to a bulk porosity of 0.1 (B). The tracerdye (rhodamineWT) was introducedinto the glacierin a boreholeat Km 9.5 at the times indicated the arrows. Tracercontent of terminusoutflow streamCS and if the -50-m in water by (A) percent drop is plotted over the period9 June to 1 July, showing the results of the first injection. (B) Tracer level at surge termination (indicated by contentof streamLTS is plottedover the period 15 to 19 July, showingthe results of the second the borehole water levels, Fig. 9) is ap- injection. Note the differentabscissa and ordinate scales for (A) and (B). much too slow in relation to the outlet- there is more than one tunnel, the larger stream discharge at the time of observa- N ones will grow at the expense of the a) 2.0 a water-flow co so that the evolves into tion (7 m3/sec). To permit _ smaller, system 0.02 m/sec down the actual 'a-o one with a tunnel or at most speed of only 0 only single 0 ?... 3 \ 10-cm drop gradient (0.1), a tunnel would have to be 0 very few (20, 25). In contrast, because of *a 1.0 only --1 mm in diameter and would carry 0 11 J the direct relation between water flux U) a discharge of only -10-8 m3/sec. Thus (D and pressure, passageways in a linked- there must be a are stable many water-conducting 0) cavity system against pertur- 0 m passageways, distributed across the -0 n bations in which some passageways are 0 20 40 60 width of the glacier bed. A limiting case c enlarged at the expense of the others; as Longitudinal coordinate x (m) of this would be a continuous gap, about a result, the linked-cavity system can of a marker 1 mm thick, between the ice and its bed, Fig. 12. Vertical displacement exist stably as a system of multiple inter- in ice at the center line near which is similar to the model emplaced glacier distributed over proposed Km 8, plotted against horizontal displace- connecting passageways for surging by Weertman (21). However, ment. The reduced vertical coordinate z' is the width of the glacier bed. even in this limiting case the water flux the actual elevation z reduced by an amount carried by the system would be much too correspondingto motion down a sloping line of constant 0.125; thus z' = z - small, only -0.02 m3/sec. From this we slope The Surge Mechanism 0.125x - zo, where Zo is a fixed arbi- are forced to conclude that the system traryconstant; x is the horizontaldistance in consists not of conduits or gaps of uni- the direction of glacier flow (N50?W), with The immediate cause of the high gla- form width or thickness but rather of arbitraryzero. The drop of 0.1 m that occurs cier flow velocities in the surge is high 4 and shown cavities linked by nar- between measurementpoints 6, basal water pressure, which causes a relatively large the offset between the lines fitted orifices that conduct the by sloping increase in basal The ob- row, gaplike to the points before and after, may be due to great sliding. water flow from one cavity to the next. the closure of basal cavities when the surge servational basis for this conclusion was The overall flow rate (water discharge) is ended. The measurementtimes of the num- given above. According to theoretical controlled by high-speed flow through bered points are as follows: 1, 4 July 1983, discussions of the basal sliding process 10:34; 5 the narrow orifices, whereas most of the 2, 15:11;3, 17:24;4, 20:09;5, 22:04;6, (18), increased basal water pressure July 1983, 12:11;and 7, 13:49. travel time for water transport through causes increased sliding by reducing fric- the system is taken in slow-speed flow tion and by increasing basal cavitation, through the large cavities. and also by exerting a direct effect of the We believe that this system of linked comparison with a tunnel system, the integral of pressure over the irregular cavities forms by the process of basal enlargement of the passageways by fric- undulating base of the ice mass. cavitation (16). The cavities, which form tional melting is greatly reduced, be- Theoretically, if the basal water pres- on the lee sides of bedrock protuber- cause the total ice surface area of the sure were to reach and remain at the ice ances, probably have horizontal dimen- cavity roofs is much greater than the roof overburden pressure, the glacier would sions of a few meters and heights of a area of a tunnel carrying the same dis- be floated off its bed, which would cause meter or less. The connections between charge. The total frictional heat, which the sliding velocity to increase practical- them, which are in places where cavita- for the same discharge and same hydrau- ly without limit. A more limited effect of tion occurs only marginally, must be lic gradient will be the same in both this kind must be produced when the only a few centimeters high to throttle systems, is thus spread out in the linked- basal water pressure approaches but the flow. cavity system over a much larger ice does not reach the ice overburden pres- A quantitative theoretical model of surface area and the rate of enlargement sure. How closely the water pressure such a system has been constructed (22), of the cavities by melting is proportion- must approach the overburden to give which is able to account for the observed ately reduced. To keep the conduits sliding velocities of the magnitude ob- discharge of 7 m3/sec and transit time of open enough to provide the hydraulic served in the surge (--5 m/day in the 90 hours. These figures require that the conductivity needed to carry a given upper glacier, -50 m/day in the lower) is cavities have an average area of 200 m2 discharge therefore requires a much from the theoretical point of view (18) a in lateral cross section, which means that higher basal water pressure in the linked- complicated question. Empirically, on their average height is 200 m2 - 1000 m cavity system than in the tunnel system. the basis of the evidence presented (the glacier width) = 0.2 m. The opening In the linked-cavity system, an in- above, we conclude that basal water or closing of such a cavity system would crease in basal water pressure enlarges pressures consistently within 4 to 5 bars store or release water as discussed earli- the cavities and therefore increases the of the overburden, and with transient er and would cause the glacier surface to basal water flux, under otherwise fixed fluctuations close to the overburden, are rise or fall by 0.2 m, which is of the order conditions (24). This behavior is oppo- sufficient to cause these high sliding ve- of magnitude of the drop actually ob- site to that of a normal tunnel system, in locities. In minisurges, the peak basal served at surge termination. The wide- which an increase in basal water flux water pressures are higher, normally at spread distribution of the passageways results in a decrease in water pressure in or above overburden, while the peak across the glacier bed accords with the the tunnel or tunnels, once transient ef- sliding velocities (1 to 3 m/day) are lower greatly increased turbidity of the outflow fects have died out and a steady state is than in surge; this is probably because water during surge: by comparison with established; this is so because the in- the high pressure does not gain access to a normal tunnel system, it allows a much creased melting caused by increased wa- more than a fraction of the area of the more extensive access of the basal water ter flow must be balanced by an in- bed in the relatively short duration of the to the source of comminuted rock debris creased viscous closure rate for the tun- high-pressure peak (-10 hours). that is generally present between the nel, which requires a drop in the water The role of high water pressure in base of the ice and the underlying bed- pressure (20). One consequence of the causing the rapid basal sliding motion in rock (23). inverse relation between water flux and the surge of a glacier is analogous to its In the linked-basal-cavity system, by pressure in a tunnel system is that, if proposed role in the mechanics of over- 478 SCIENCE, VOL. 227 Nature (London)283, 629 (1980);0. Q. Bowen, thrustfaulting and landslides (26). Land- tween flux and pressure in a normal ibid., p. 619. slides sometimes show of tunnel water in 8. R. Bindschadler,W. D. Harrison, C. F. Ray- periods greatly system, high pressures mond, R. Crosson,J. Glaciol. 18, 181(1977); R. enhanced movement that are called this system can be maintained only in Bindschadler,W. D. Harrison,C. F. Raymond, when the water flux is low. C. Gantet, ibid. 16, 251 (1976). surges (27). winter, very 9. S. M. Hodge, ibid. 13, 349 (1974). The key element of the surge mecha- It thereforemakes sense in a generalway 10. B. Kamb and H. Engelhardt,unpublished paper;C. F. Raymondand J. Malone,unpublished nism (and of the overthrustand landslide that the time when the pressure might paper; N. Humphrey,C. F. Raymond, W. D. mechanisms also) is what enables the rise high enough to cause the transition Harrison,unpublished paper. The name "minisurge" shouldnot be taken to imply that motion high basal water pressures to be built up to the surgingstate is in winter, which is events of this kind necessarily occur only on and maintained. The difference in the when the started. surge-typeglaciers or that their mechanismis surge actually necessarilythe same as the surge mechanism. basal water system between nonsurging Althoughthe foregoingideas about the 11. S. M. Hodge, J. Glaciol. 16, 205 (1976);ibid. 23, and is 307 (1979);H. F. Engelhardt,W. D. Harrison, surging conditions crucial. The fundamental differences between the B. Kamb,ibid. 20, 493 (1978);H. Rothlisberger, pressure in the system is the result of a basal water systems in the nonsurging A. Iken, U. Spring,ibid. 23, 429 (1979). 12. J. F. Nye, Proc. R. Soc. LondonSer. A 256, 559 balance between buildup of pressure and surging states provide a working (1960). from water input and drawdownof pres- hypothesis for understandingthe surge 13. The hot-waterdrilling method was used (P. L. Taylor, U.S. Army Cold Regions Res. Eng. sure by outflow. Since increased pres- mechanism, numerous features of the Lab. Spec. Rep., in press). An indicationthat sure tends to the remain the bed was reachedwas a suddenonset of rapid increase cavitation and surge to be understood.What are waterflow down the hole when the drillreached hydraulicconductivity of the basal water the switchover processes in the basal 385 m and stopped advancing. it seems at first water that start the in win- 14. J. F. Nye, Proc. R. Soc. LondonSer. A 207, 560 system, sight paradoxical system surge (1951).This type of flow is also sometimescalled that the higher basal water pressure in ter and terminateit in summer?What is "block flow" or "Block-Schollenflow" [R. P. Sharp, Geol. Soc. Am. Bull. 65, 829 (1954)]. the surgingstate can be accompaniedby the water source, intra- or extraglacial, 15. H. Lang, Ch. Liebundgut,E. Festel, Z. Glets- increased retention of water in the gla- that in midwinterprovides the water to cherk. Glazialgeol. 15, 209 (1979). 16. The phenomenonof basal cavitationin glaciers cier, as is necessary to maintain the pressurize the basal water system and is distinctfrom the phenomenonof cavitationin water However, the initiate the How in the hydrodynamics,although the two are similarin higher pressure. surge? important that both involve the opening up of cavities linked-basal-cavitywater system of the surge mechanism is the water produced when the local pressuredrops low enough. For state has characteristicsthat can frictional due to basal the literatureon observationsof basal cavitation surging by melting sliding? see (17) and for theories of the process see (18). maintainthe higherbasal water pressure. How does the surge mechanism control 17. H. Carol, J. Glaciol. 1, 57 (1947); R. Haefeli, ibid. 1, 496 (1951); R. Vivian and G. Bocquet, An essential characteristicis the stability the advance of the surge front? What ibid. 12, 439 (1973);B. Kamband E. R. LaCha- of the linked-cavity system against the causes the oscillations or fluctuationsin pelle, ibid. 5, 159 (1964). 18. L. Lliboutry, ibid. 7, 21 (1968); J. Weertman, preferential growth of larger passage- surge motion on various time scales and ibid. 5, 287 (1964); B. Kamb, Rev. Geophys. as occurs in a normal tunnel what controls their down- Space Phys. 8, 673 (1970);A. Iken, J. Glaciol. ways, sys- propagation 27, 407 (1981). tem. Without this stability, the linked- glacier? What is the relative importance 19. A. Iken, H. R6thlisberger,A. Flotron, W. Hae- would into a of basal water volume versus in berli, J. Glaciol. 29, 28 (1983). cavity system degenerate pressure 20. H. Rothlisberger,ibid. 11, 177 (1972). tunnel system, as indeed it does in the causingthe large basal slidingrates in the 21. J. Weertman,Int. Assoc. Hydrol. Sci. Publ. 58, 31 (1962);Can. J. Earth Sci. 6, 929 (1969). surge termination. There are hints of surge? What are the roles of the numer- 22. B. Kamb, in preparation. instabilityin the major slowdowns prior ous prominent structural features- 23. We have observed this basal debris zone many times at the bottom of deep boreholesin Varie- to surge termination, but in these the wrench faults, thrust faults, folds on gated Glacier, by means of borehole television. itself robust various and intense It is very similar to the "active subsole drift" linked-cavitysystem proves scales, crevassing- layer observed earlier at the bottom of Blue and recovers. Theoretical modeling produced by the surge and doubtless Glacier,Olympic Mountains, Washington [H. F. shows connected Engelhardt,W. D. Harrison,B. Kamb, J. Gla- that the primaryfactor responsi- somehow with the surge ciol. 21, 469 (1978)]. ble for the stability is the effect of rapid mechanism(28)? 24. This is shown by the detailed physical model and theory of the linked-cavity system (22), basal sliding, which provides a powerful We hope that our observations of the whichtakes into accountthe effects offrictional- offset to the enlargementof the cavity- 1982-1983 surge of Variegated Glacier heat meltingand is the counterpart,for this type of system, to the theoryof R6thlisberger(20) for connecting orifices by frictional melting and the resulting inferences about the the tunnel system. and is able to stabilize the orifices at a mechanism will be useful in the 25. R. L. Shreve, J. Glaciol. 11, 205 (1972). surge 26. M. K. Hubbertand W. W. Rubey, Geol. Soc. size only modestly enlarged from what furtherefforts needed to understandthe Am. Bull. 70, 115 (1959). would be in the absence of of the 27. D. K. Keefer and A. M. Johnson, U.S. Geol. they melting many intriguing aspects glacier Surv. Prof. Pap. 1264 (1983), p. 47. (22). surge phenomenon. It will be necessary 28. Structuralfeatures of apparentlysomewhat si-m- ilar types were observed by Dolgoushin and From this perspective, what is re- to determine how general the inferred Osipova in Medvezhiy Glacier and were inter- quired for a glacier to be in the surging surge mechanism is and whether there pretedby them as principalagents in the surge mechanismof that glacier (6). state is a kind of "bootstrapping" in are other types of glacier surges in which 29. This work was made possible by grantsDPP-82- which basal water is need- other mechanisms 09824,DPP-82-00725, EAR-79-19424, and EAR- high pressure play a predominating 79-19530from the NationalScience Foundation. ed to cause rapid sliding, while rapid role. It was carriedout with permissionof the U.S. is needed to the Forest Service (Tongass National Forest) and sliding provide cavity the U.S. National Park Service (Wrangell-St. stabilitythat permitsthe high basal water Referencesand Notes Elias National Park). Essehtial contributionsto this work were made by pilots of Livingston pressure to be maintained. 1. M. F. Meier and A. Post, Can. J. Earth Sci. 6, Helicopters, Iiic., Junealu,and Gulf Air Taxi, We visualize that in the surge front as 807 (1969). Inc., Yakutat. We acknowledge the able and 2. W. S. B. Paterson, The Physics of Glaciers dedicatedefforts Of many co-,workers and assist- it advances the great increase in basal (Pergamoh,Oxford, 1981),p. 275. ants in the fieldwork,including H. Aschmann, slidingof the ice, shoved forward the 3. R. S. Tarr,Geol. Soc. Am. Bull. 18, 262 (1907). M. Balise, M. Hendrickson, R. Jacobel, T. by 4. A. Post, J. Glaciol. 8, 229 (1969). Johanesson, M. Magnesson, E. Mezger, C. surgingmass behind it, destroys the tun- 5. See (2), p. 287. Moore, P. Mullen, M. Richards, D. Sabol, P. nels of the normal basal water 6. The most extensive and detailedobservations of Schweizer, E. Senear, D. Sollie, B. Svendson, system glacier surging up to now have been on the C. Tobin, and P. Titus. A. Iken contributed that exist below the front and at the same MedvezhiyGlacier in Soviet centralAsia: L. D. greatlyto the work. The photo in Fig. 3A is due Dolgoushin and G. B. Osipova, Int. Assoc. to B. Krimmel, that in Fig. 3B and the cover time causes widespreadbasal cavitation, Hydrol.Sci. Publ. 107, 1158(1973); ibid. 104, 292 photo to E. Mezget, and in Fig. 8 to A. Iken. thus generatinga linked-cavity network (1975); PulsiruyoushchiyeLyedniki [Pulsatory Equipmentand help in carryihg out the dye- (surging) Glaciers] (Gidrometeoizdat, Lenin- tracerexperiment were providedby the staff of that takes over as the basal water system grad, 1982). the U.S. Geological Survey, Tacoma, Wash., above the front. 7. A. T. Wilson, Nature (London)201, 147(1964); particularlyB. Krimmel.Contribution No. 4025 Can. J. Earth Sci. 6, 911 (1969); J. T. Hollin, from the Division of Geological and Planetary Because of the inverse relation be- ibid., p. 903; Quat. Res. (N.Y.) 2, 401 (1972); Sciences, CaliforniaInstitute of Technology. 1 FEBRUARY1985 479