Aggregation of snow crystals that were imaged with low temperature scanning electron microscopy. The dendritic crystal (pseudocolored in blue) is one of four basic forms that can found in precipitatingsnow. The width of the image is 2.7 mm.

1

Low Temperature SEM of Precipitated and Metamorphosed Snow Crystals

Collected and Transported from Remote Sites

William P. Wergin1*, Albert Rango2, Eric F. Erbe1 and Charles A. Murphy1

Nematology Laboratory1 and Hydrology Laboratory2, Agricultural Research Service,U.S. Department of Agriculture, Beltsville, MD 20705 USA

ABSTRACT: Procedures were developed to sample, store, ship and process precipitated and metamorphosedsnow crystals, collectively known as "snowflakes", from remote sites to a laboratory where they could beobserved and photographed using low temperature scanning electron microscopy (LTSEM). Snow samples werecollected during 1994-96 from West Virginia, Colorado and Alaska and shipped to Beltsville, Maryland forobservation. The samples consisted of freshly precipitated snowflakes as well as snow that was collected frompits that were excavated in winter snowfields measuring up to 1.5m in depth. The snow crystals were mountedonto copper plates, plunged into LN2 and then transferred to a storage dewar that was shipped to the laboratory.Observations, which could be easily recorded in stereo (three-dimension), revealed detailed surface features onthe precipitated crystals consisting of rime, graupel and skeletal features. Samples from snowpacks preservedthe metamorphosed crystals, which had unique structural features and bonding patterns resulting fromtemperature and vapor pressure gradients. In late spring, the surface of a snowpack in an alpine region exhibiteda reddish hue. Undisturbed surfaces from these snowpacks could be sampled to observe the snow crystals as wellas the organisms that were responsible for the coloration. Etching the surface of samples from these sites,exposed the presence of numerous cells believed to be algae. The results of this study indicate that LTSEM canbe used to provide detailed information about the surface features of precipitated and metamorphosed snowcrystals that can be sampled at remote locations. The technique can also be used to increase our understandingabout the ecology of snow. The results have application to research activities that attempt to forecast the quantityof water in the winter snowpack and the amount that will ultimately reach reservoirs and be available foragriculture and hydroelectric power.

KEY WORDS: Low temperature SEM, snowflake, snow crystal, freeze-substitution, algae.(recieved March 28, 1996: accepted June 12, 1996)

INTRODUCTION

About a hundred years ago, the combination of lightmicroscopy with photography allowed investigators tobegin to magnify and to record the variations thatexisted in the structure of snow crystals(Nordenskiold, 1893; Hellman, 1893; Dobrowolski,1903). In spite of the numerous problems associatedwith working with frozen microscopic particles atsubzero temperatures, an amateur microscopist namedWilson A. Bentley spent nearly 30 years amassingover 6,000 photomicrographs of snow crystals(Blanchard, 1970). However, because of his apparentinterest in symmetrical forms, Bentley concentratedon the flat or two dimensional crystals, namely platesand dendrites, which could be properly focused and

photographed with transmitted light in the relativelynarrow depth of field that existed in his lightmicroscope. Although over 2,000 of thesephotomicrographs were eventually published (Bentleyand Humphreys, 1931), he did not include theatmospheric conditions under which the snow wascollected or the magnifications at which the crystalswere photographed. Furthermore, because thepublication concentrated on symmetrical crystalsrather than on the full range of shapes that Bentleysurely must have encountered, it is somewhatcompromised in its scientific value.These deficiencies were remedied by Nakaya, who

*Corresponding author: Electron Microscopy Unit Bldg. 177B, USDA-ARS Beltsville, MD 20705Telephone: 301-504-9027; FAX 301-504-8923 email: WWERGIN@GGPL.ARSUSDA.GOV

began a light microscopic investigation in 1932 thatlasted nearly 20 years (Nakaya, 1954). In his studies,oblique illumination was used to reveal the shapes andsurface features of all types of snow crystals, includingmany three-dimensional forms such as needles,columns, and graupel. Nakaya not only recordedphotomicrographs of the structural variations of crystalsthat were collected from natural snows, during whichhe carefully recorded environmental conditions, butunder controlled laboratory situations he was also ableto simulate various atmospheric conditions to produceartificial snow crystals that resembled those fromnature. Unfortunately the photomicrographs of thethree-dimensional types of snow crystals were greatly

compromised by the limited depth of field in the lightmicroscope, the structural details of rime and graupelcould not be resolved and other features such as theexistence of double sheets and various skeletal patternswere represented with line drawings because they couldnot be convincingly recorded and illustrated in thephotomicrographs.

To avoid the adverse working conditions thatwere necessary for light microscopic studies and toincrease the resolution of snow crystals, transmissionelectron microscopists attempted to make replicas ofsnow crystals that could be examined in the TEM. Useof replicas further resolved the details about the flatsurfaces of two-dimensional crystals (Stoyanova et al.,

Figures 1 - 3. Snow samples collected at Beltsville, Maryland, elevation 60m above sea level (ASL).

Figure 1. Stereo pair illustrating a snowflake composed of numerous snow crystals that consist mostly ofirregular hexagonal plates. Sampled at 00C air temperature.

Low Temperature SEM of Snow Crystals

1987; Takahashi and Fukuta, 1987), however, theability to image intact, three-dimensional snow crystalsremained elusive. Recently, Wergin and coworkers(Wergin and Erbe, 1994a; 1994b; 1994c; Wergin et al.,1995a; 1995b) used an SEM equipped with a cold stageto image various forms of newly precipitated snowcrystals that were collected outdoors at their laboratorysite. Their collection and processing proceduresprovided the first clearly focused images of intact snowcrystals. In addition, this technique was used to recordstereo images of single snow crystals as well as theiraggregates, which are more commonly known assnowflakes (Wergin et al., 1995b).

During this same time period, Wolff and Reid(1994) made an attempt to collect, ship and imagenewly precipitated snow crystals from remote sites.Snow samples were collected in Greenland, stored inliquid nitrogen and transported to England forobservation. However, because of difficulties inpermanently mounting the samples, most of thespecimens were either lost or broken during transportand the authors were only successful in recordingfragments of snow crystals that had been collected atthe remote site. Because snow scientists, includingglaciologists, geophysicists and ice physicists, arefrequently interested in the metamorphism of snowcrystals that occurs in remote winter snowpacks andbecause of our previous success in collecting andimaging newly precipitated snow crystals at ourlaboratory site, we were encouraged to developprocedures to collect, store and ship newly precipitatedas well as metamorphosed snow crystals from distantsites. The results of these efforts are presented in thecurrent study.

2. MATERIALS AND METHODSSnow was collected during 1994-96 from sites

near the following locations: Beltsville, Maryland;Davis, West Virginia; Loveland Pass and Jones Pass,Colorado; and Fairbanks, Alaska. The samples, whichwere obtained when the air temperatures ranged from+180C to -110C, consisted of freshly fallen snowflakes,as well as, snow crystals that were collected from thewalls of snowpits

Figure 2. Snowflakes consisting of crystalline needles encumbered withaggregations of short prismatic crystals. Sampled at -50C air temperature.

Figure 3. Dendritic form of snow crystal lacking sharp edges. Because ofthe air temperature at the time of collection, this crystal may haveundergone some sublimation or melting. Sampled at 00C air temperature.

William P. Wergin et al.

excavated in winter snowpacks measuring up to 1.5min depth. The snowpits were established in pristineareas, which were located from about 50m to nearly10km from plowed roadways. Sleds, snow shoes, crosscountry skis, snow mobiles and a snow cat were used totransport the sampling supplies, including liquidnitrogen, to these remote sites.

The collection procedure consisted of placinga thin layer of liquid Tissue-Tekr, a commonly usedcryo-adhesive for biological samples, on a flat copperplate (15mm x 27mm). The Tissue-Tek and the plateswere precooled to ambient outdoor temperatures beforeuse. Newly fallen snowflakes were either allowed tosettle on the surface of the Tissue-Tek or were lightly

dislodged and allowed to fall onto its surface. Next, theplate was either rapidly plunged into a styrofoam vesselcontaining LN2 or placed on a brass block that had beenprecooled with liquid nitrogen (LN2) to -1960C. Thisprocess, which solidified the Tissue-Tek, resulted infirmly attaching the sample to the plate. When sampleswere obtained from snowpits, a precooled scalpel wasused to gently dislodge crystals from the pit wall ontoplates, which contained Tissue-Tek, that were rapidlyplunged in LN2. The frozen plates were inserteddiagonally into 20 cm segments of square brasschannelling and lowered into a dry shipping dewar thathad been previously cooled with LN2. The dewarcontaining the samples was conveyed from the

Figures 4 - 8. Samples collected on Bearden Mountain near Davis, West Virginia, elevation 1150m ASL.

Figure 4. Stereo pair illustrating a snowflake consisting of dendritic snow crystals. Sampled at -110C air temperature.

Low Temperature SEM of Snow Crystals

snowpit sites and then either transported by van (fromWest Virginia) or shipped by air (from Colorado andAlaska) to the laboratory in Beltsville, Maryland. Uponreaching the laboratory, the samples were transferredunder LN2 to a LN2 storage dewar where they remainedfor as long as nine months before being furtherprepared for observation with LTSEM.

Preparation for LTSEM examinationTo prepare the samples for LTSEM

observation, the brass channelling was extracted fromthe storage dewar and placed in styrofoam workchamber that was filled with LN2. A plate wasremoved from the channelling and attached to amodified Oxford specimen holder. The holder, whichcontained the plate, was transferred to the slushchamber of an Oxford CT 1500 HF Cryotrans systemthat had been filled with LN2. Next, the holder wasattached to the transfer rod of the Oxford cryosystem,moved under vacuum into the prechamber for etchingand/or sputter coating with Pt or Au/Pd and theninserted into a Hitachi S-4100 field emission SEM thatwas equipped with a cold stage maintained at -1850 C.Accelerating voltages of 500 V to 10 kV were used toobserve and record images onto Polaroid Type 55 P/N

film. To obtain stereo pairs, a stage tilt of 6o wasintroduced between the first and second images.

Preparation for TEM observationAfter snow samples containing the alga cells

were observed and photographed in the LTSEM, thespecimen holders were withdrawn into the prechamber,removed and then transferred under vacuum to LN2 tobegin a freeze-substitution procedure. The flat plates,on which the samples were mounted, were thentransferred to 45 ml cryogenic vials that were filledwith a solution of 2% (w/v) osmium tetroxide inacetone precooled to -1960C. The vials were placed inwells that had been drilled in an aluminum block. Thealuminum block was put into a precooled brasschamber and then placed into an insulated encasement,which had been precooled with LN2. The insulatedencasement was filled with dry ice to maintain atemperature of -800C. The temperature of the sampleswas monitored with a thermocouple that was placed ina vial in the center well of the aluminum block.Substitution with the osmium solution was allowed toproceed for 3 days. Subsequently, the solution wasslowly warmed (2h at -600C, 2h at -180C, 2h at 40C and2h at room temperature) and the substitution mediumwas replaced with fresh

Figure 5. A hexagonal dendrite having a central hexagonal plate. The arms of the dendrite are branched and contain pronounced ridges andsmall depressions. Sampled at -110C air temperature.

William P. Wergin et al.

William P. Wergin et al.

acetone. Next, the samples were detached from theplates, gradually infiltrated with Spurr's low viscosityresin and cured at 600C. The embedded samples werethin sectioned on an American Optical Ultracutultramicrotome and mounted on copper grids. Toimprove contrast, sections were post stained in 2%aqueous uranyl acetate and Reynold's lead citrate.TEM observations were performed on a Hitachi H500-H operating at 75kV.

3. RESULTS

Snowflakes - Aggregations of Snow CrystalsA snowflake consists of one or more snow

crystals that either simultaneously grow together ormake contact and become associated as they movethrough the atmosphere. The snow crystal aggregatesthat form snowflakes occur as variations of plates,needles, columns or dendrites. The specific shape thatprevails at a given time is largely dependent upon thecloud temperature at which the crystal forms.Snowflakes that largely consisted of aggregated plates(Fig. 1) and needles (Fig.2) were collected outside thelaboratory at Beltsville, Maryland when the airtemperatures at ground level were 00C and -50C,respectively. These samples were attached to plates,plunged in LN2, taken into the laboratory, transferred to

the cryosystem for coating and inserted into the SEMfor low temperature observations; these samples werenot stored or transported.

The snowflakes, which were sampled whenthe air temperature was 00C, measured 1 to 5mm acrossand were composed of randomly aggregated variationsof flat hexagonal plates (Fig. 1). The individual platesvaried from 0.4 to 0.8mm across. Snowflakes that werecollected at -50C were generally smaller and consistedof bundles of needles that were encumbered with shortprismatic crystals (Fig. 2). The needles were 1.0 to2.0mm in length whereas the diameter of the bundleswas about 0.2 to 0.3mm; the short prismatic crystalsthat were present on the surfaces of the bundlesmeasured about 0.1 mm.

Dendritic forms of snow crystals were onlyoccasionally found in the Beltsville samples (Fig.3).This type of crystal was easily distinguished by its sixarms that were frequently branched. The structuralsimilarity that often existed between the six arms of thedendritic forms provided hexagonal symmetry thatcharacterized this type of crystal. The dendritescollected at Beltsville generally lacked the sharp edgesthat characterized

Figure 6. Stereo pair illustrating a snowflake that has become encumberedwith supercooled water droplets or rime. Sampled at -20C air temperature.

Low Temperature SEM of Snow Crystals

many of the other types of crystals. Furthermore, thedendritic arms generally exhibited less branching andsimilarity to one another compared to those fromdendrites that were obtained from colder sampling sites.

Samples of snowflakes that were obtainedfrom West Virginia, were plunged in LN2 at thecollection site, transferred to either a dry LN2 shipper ora LN2 storage dewar and transported by van to thelaboratory in Beltsville, MD for storage andobservation. At -110C, a snowflake sample obtainedfrom West Virginia was composed of overlapped andinterlocked dendritic snow crystals (Fig. 4). Althoughthese crystals appeared more fragile than the hexagonalplates or the needles, neither the collection procedurenor transport in the dewars seemed to result in anysignificant breakage or damage. Although snowcrystals having broken arms were occasionallyobserved, this type of damage could also result fromcollisions with other crystals as they descended throughthe atmosphere.

Different forms of growth were frequentlycombined into a single snow crystal. In Figure 5, thehexagonal dendrite is the largest and most predominantform of snow crystal. However, closer examinationrevealed the presence of a small, flat hexagonal plate inthe center of the dendrite. This dendrite also exhibitedother distinct structural features that were frequentlypresent on this form of snow crystal. Arms were oftenbranched; in this crystal, a left and right pair ofbranches emanated from the main axis of each armabout two thirds from its base. A slightly raised andcontinuous midrib formed along the center of the armsand extended into the branches. Lateral to the midribwere a series of depressions, consisting of elongatechannels and circular cavities. These depressions werebelieved to correspond to negative crystals that hadbeen previously described in ice (Hobbs, 1974). Thedepressions frequently exhibited bilateral arrangementalong the axis of the arms and branches. Because theoccurrence and arrangement of ribs and depressionswere fairly consistent in each of the six arms and theirbranches, these structural features reinforced theimpression of hexagonal symmetry that wascharacteristic of this type of crystal.

Further Resolution of Surface FeaturesAlthough numerous light micrographs of

snow crystals have been recorded, clear illustrations ofadditional structural features, such as rime deposits andmicrocrystalline growth, are beyond the resolution ofthat technique. During formation and descent, snowcrystals frequently encounter super-cooled waterdroplets that freeze onto their surfaces. These particles

Figure 7. Dendritic crystal with central hexagonal plate and armsterminating with plate-like sectors. Small raised hexagonal crystallineparticles can be found on the surfaces of these terminal plates. Sampled at-80C air temperature.

did not exhibit crystallographic features butalternatively appeared as random accumulations ofsmall spheres measuring 0.02 to 0.06 mm in diameter(Fig. 6). These frozen droplets are referred to as rime.When the droplets continue to accumulate until theoriginal crystal is no longer distinguishable, the mass ofrandomly arranged, frozen microdroplets is referred toas graupel (Brownscombe and Hallet, 1967).

LTSEM also resolved the multiple layeringand microcrystalline growth that occurred in many ofthe snow crystals (Figs. 7 and 8). A hexagonal dendriteobserved with the light microscope is generally thoughtof as a single crystal. However, a distinct, raisedhexagonal plate was also frequently incorporated intothis type of crystal (Figs. 5 and 7). The plate wascentrally located and positioned so that its six apiceswere aligned with the arms of the dendrite. In additionto the central plate, the plate-like extensions at the endsof the arms contained raised particles. Unlike the rime,which tended to be spherical, these particles consistedof microcrystalline hexagonal structures, measuring0.04 to 0.06 mm. Even closer examination

William P. Wergin et al.

Figure 8. Apex of the hexagonal plate, delineated in Figure 7, illustratinga developing microcrystalline extension.

of the edges a crystal, revealed microcrystallineformations that probably contributed to further growthand layering along the arms of the dendrite (Fig. 8).

MetamorphismIn Colorado and Alaska, snow samples were takenfrom the accumulated winter snowpack to illustrate thetypes of transformations or metamorphism thatoccurred after the snow crystals reached the ground andwere subjected to microenvironmental variations intemperature and vapor pressure. Snow crystals thatwere sampled several cm below the surface ofsnowpacks, which were not subjected to significanttemperature differences between the ground and the air,underwent changes that are referred to as lowtemperature metamorphism (Figs. 9 and 10). Theinitial changes were characterized by sublimation of thefine delicate structures on the edges and surfaces of thesnow crystals (Fig. 9). The sharp surface angles thatcharacterized the crystalline features of a snow crystalwere lost. As this process proceeded, the surface of thesnow crystals became smooth and sinuous. In addition,adjacent crystals bonded or sintered to one another. Atlower depths where this process had proceeded forlonger times, the original forms of the crystals were nolonger discernable; they appeared rounded and wellbonded to one another, resulting in compaction of the

snowpack and somewhat reducing the air space thathad been present between the crystals (Fig. 10).

A second type of metamorphism occurred inthe snowpack when the temperature of the ground wassignificantly greater than the air temperature at thesurface of the snowpack, i. e., when a temperaturegradient of at least 100C/m had been present. In thissituation, high temperature gradient metamorphismresults from a heat flux moving upward through thesnow pack (McClung and Schaerer, 1993).Sublimation

Figures 9 - 16. Samples from either Colorado or Alaska, shipped by airto the laboratory and stored in LN2 prior to imaging.

Figure 9. Snow crystals subjected to low temperature gradient. The finedelicate features on the surface of snow crystals are no longer apparent.The crystals exhibit smooth sinuous contours and have become bondedwith one another. Sample was collected from a snowpit 5cm below thesurface near Jones Pass, Colorado, elevation 3150m ASL, air temperature-60C.

Low Temperature SEM of Snow Crystals

occurred on the lower surfaces of crystals followed byrecrystallization of the water vapor on upper lyingsurfaces (Fig. 11). With time, this process resulted inthe formation and growth of large crystals known asdepth hoar, which was found near the base of thesnowpack (Fig. 12). The crystals of the depth hoarwere frequently hollow or had internal, parallel arraysof facets or steps, which resulted from the refreezing ofsuccessive molecular layers of ascending water vapor(Fig. 12). Unlike precipitated snow, depth hoar crystalsgenerally were not hexagonally symmetrical.Furthermore, the depth hoar was not significantlysintered or bonded to adjacent crystals.

In spring, the snowpack in Colorado wassubjected to a day/night fluctuations in temperaturesthat produced changes referred to as melt-freezemetamorphism

Figure 10. Rounded and well bonded snow crystals that result fromlonger exposures to low temperature gradients. Sample was collectedfrom a snowpit 19cm below the surface near Loveland Pass, Colorado,elevation 3600m ASL.

Figure 11. Large crystal that exhibits rounding and faceting and isbelieved to result from both low (rounding) and high (faceting)temperature gradients in the snowpack. Sample obtained near Jones Pass(Henderson Mine), Colorado, elevation 3150m ASL, air temperature -60C. Collected at 84 cm below the surface.

Figure 12.. Large depth hoar crystal. Sample was collected 90 - 100cmbelow the surface of 1m snowpit at Ester Dome, near Fairbanks, Alaska,elevation 700m ASL, air temperature +40 C.

William P. Wergin et al.

Figures 13 - 16. Samples were collected near Loveland Pass, Colorado,elevation 3600m ASL, air temperature +180C, shipped to the laboratoryand stored in LN2 prior to imaging.

Figure 13. Melting surface layer of a snowpack. A film of water appearsto bond the rounded snow grains into a cluster.

During the day when the temperature was abovefreezing, this process resulted in the formation ofspherical snow grains that were covered with a film ofsurface water and consequently, were only weaklybonded to one another. However, at night when thetemperature was below freezing, the water film frozeand the grains became well bonded to one another. Oursampling procedure, which refroze the surface water,produced images that simulated those found at night(Fig. 13). Lightly etching the surface of this type ofsnow sample revealed different morphological patternsresulting from dissimilarities between the water-ice ofthe crystals and that formed by the film of surface waterwhich was frozen during the sampling procedure.

Red SnowIn Colorado, the surface of the snowpack that

was undergoing melt-freeze metamorphism exhibited areddish coloration. Fracturing surface samples fromthis snowpack, revealed numerous cross sections ofcircular structures measuring about 5 to 25µm indiameter (Fig. 14). These structures, which occurredjust beneath the uppermost film of water, could also be

exposed by etching a few microns of water-ice from thesurface of the sample. The structures appeared toresemble cells, had peripheral layer consisting ofdensely packed granules but showed no significantevidence of the formation of internal ice crystals (Fig.15).

To help further identify these structures, thesamples that were observed with LTSEM wereremoved and processed, using freeze-substitution, forTEM observations. Cross sections of these structuresexhibited a large central chloroplast with numerousstarch granules that tended to be localized around theperiphery. The parietal layer of cytoplasm contained anucleus, mitochondria, dictyosomes, ribosomes andseveral types of vesicles (Fig. 16). Because the presenceof these cells was not expected at the time the snow wascollected, no additional samples were taken in anattempt to isolate, culture or positively identify thisorganism.

4. DISCUSSION

Results form this study indicate that snowsamples can be successfully collected at remotelocations, stored and shipped to a laboratory forobservation with LTSEM. The structural details ofthese snow crystals were similar to those from newlyprecipitated samples that were previously obtained atthe laboratory site and directly observed by thistechnique (Wergin and Erbe, 1974a; 1994b; 1994c;Wergin et al., 1995a; 1995b). These observationssuggest that neither the storage conditions nor theshipping procedures affect the general structure ofprecipitating snow crystals. Because themetamorphosed snow samples observed in this study,were handled in the same manner, we feel that theirstructural features were not altered.

Observing and recording images of snowcrystals with LTSEM have numerous advantages oversimilar studies using light microscopy (Bentley, 1931;Nakaya, 1954) or hand lenses and photomacrography(LaChapelle, 1969; Armstrong, 1992; McClung andSchaerer, 1993). In light microscopic studies, samplesmust generally be observed and photographed at fieldsites in subzero temperatures or in specially designedcold laboratories. The classical photographs of Bentley(1931) were taken in an outdoor shelter; whereasNakaya (1954) constructed a laboratory that wasmaintained at -30 to -450C for this purpose. Even atthese temperatures, humidity, body heat, sampleillumination and variable pressure gradients can affectmelting, recrystallization or sublimation of the sample.Alternatively with LTSEM, cryofixation, storage,

Low Temperature SEM of Snow Crystals

transportation and observation of the samples areaccomplished at LN2 temperatures. Therefore: 1)melting was not a problem; 2) recrystallization does notoccur (Robards and Sletyr, 1985) and; 3) sublimation inthe SEM, which was calculated to be less than 10-13

cm/sec at -1500 C (Robards and Sletyr, 1985), wouldnot be detectible. Although the cost of equipmentrequired for LTSEM is greater than that needed forlight microscopy, the former technique does not requireconstruction of a cold laboratory on site and the basicprocedures are similar to those routinely used forobservations of frozen, hydrated biological specimens(Wergin et al., 1996).

Information that was obtained from LTSEMobservations was less confusing and easily surpassedthat which was achieved through examination with thelight microscope or a hand lens. The LTSEM providedsurface information that was not confused by anyinternal structure of the crystals. Internal structurecould be observed if the crystals were cryofractured.Alternatively, the illustrations obtained with lightmicroscopy (Bentley, 1931: Nakaya, 1953), which usedtransmitted or oblique illumination, respectively,provided images that were composed of surface as wellas internal structures. As a result, surface patternscould not be readily distinguished from internalfeatures. For example, Bentley described "bubbles" inthe dendritic crystals, which we believe corresponded to"negative crystals" (Hobbs, 1974) that were commonlypresent on the surface of the dendrites that we observed.

The LTSEM also had a depth of field severalorders of magnitude greater than that of the lightmicroscope; furthermore, this instrument allowed us torecord and view true, three-dimensional images of thespecimens. This greater depth of field allowed clear, infocus images of large crystals such as depth hoar andgraupel, which had considerable surface topography.Furthermore, three-dimensional images provide thepotential to obtain quantitative data (stereometry), suchas crystal size and volume. Neither of these featureshave been reported in light microscopic investigations.Finally, many of the images that are illustrated in thepresent study were recorded at magnifications similar tothose which have been photographed with a lightmicroscope; however, the resolution of the LTSEMeven at magnifications of only a few hundred times wasmuch greater than that which has been published forlight micrographs. Consequently, rime droplets and"skeletal" structures, which could only be surmised orillustrated by line drawings in the light microscopicinvestigations can now be clearly illustrated withimages obtained in the LTSEM.

Figure 14. Fractured surface of melting snow. Sample was also etched tosublime the surface water-ice. This procedure reveals the presence ofsmall spherical bodies believed to be a green alga, possiblyChlamydomonas nivalis

Comparisons of Crystal TypesAlthough storage and transport did not appear

to adversely affect the structure of the snow crystals, thetypes of newly precipitated crystals that were observedin the samples from Maryland and West Virginia diddiffer; hexagonal plates and needles were present in theMaryland samples and graupel and dendritic formswere found in the West Virginia snow. Thesedifferences are not believed to be related to samplingprocedures, differences in air temperatures or storageconditions but are probably associated with variationsin the cloud temperatures that occurred during crystalformation. This observation would be consistent withthose of other investigators who concluded that theshapes of newly formed snow crystals arepredominantly influenced by the temperatures at whichthey were formed (Nakaya, 1954; Hobbs, 1974; Mason,1992).

The samples from Colorado and Alaska werechosen to illustrate the metamorphism that occurs inthe snowpack. The observations of depth hoar crystals,which develop under high temperature gradients, wereconsistent with the observations that had beenpreviously

Figure 15. Fractured and etched surface of melting snow illustrating a single cell believed to be a green alga causing the "red snow" phenomenon.

Figure 16. Transmission electron micrograph of an algal cell from one of the samples of red snow that had been observed in the LTSEM. After thesample was removed from the SEM, it was freeze-substituted in acetone-OsO4, embedded and sectioned. The TEM image exhibits fine-structural featurescharacteristic of a green alga. These include the large central chloroplast (C) with numerous starch (S) granules and a peripheral layer of cytoplasmcontaining, a nucleus (N), mitochondria, ribosomes and vesicles.

made by light microscopy (Armstrong, 1992); namely,the depth hoar consists of crystals that attain sizes ofseveral mm, were highly faceted, hollow and show littleor no sintering or bonding to adjacent snow crystals.Alternatively, low temperature gradients lead to snowcrystals that are rounded or sinuous and highlysintered; the snowpack has smaller air spaces thanconditions described above (Armstrong, 1992). Thesecharacteristics were well preserved in the samples thatwere obtained from the snowpack in Colorado (Fig. 9).

Finally, melt-freeze metamorphism, whichnormally occurs in spring, results in spherical snowcrystals that are bonded in clusters; the strength of thebonds is affected by the amount and status of water thatis present (McClung and Schaerer, 1993). Suchspherical snow crystals were evident in the surfaceformations that were collected in spring from Colorado.Although these samples where obtained under meltconditions during the day, plunge freezing them inliquid nitrogen froze the film of surface water that waspresent during the day melt. Etching these sampleshelped to distinguish the two types of water-ice thatwere present.

Red SnowDuring spring in alpine regions, the surface of

the snowpack frequently has a colored appearanceconsisting of light hues of green, salmon, orange or red.The coloration is generally associated with the growthof a specific species of algae, which along with manyother organisms can be found near the surface of themelting snowpack. Hoham (1992) reported that inwestern North America the appearance of red snow wascaused by the motile unicellular green algaChlamydomonas nivalis. The red coloration resultedfrom the high accumulation of carotenoids, especiallyastaxanthin, that occurs under certain alpineenvironments. The carotenoids probably function asphotoprotectants for chlorophyll under the highirradiation levels that occur at high altitudes (Czygan,1970). Our summer collection site was similar to thatsurveyed by Hoham. When the cells that weencountered in the red

Low Temperature SEM of Snow Crystals

snow samples were prepared and observed in the TEM,they exhibited fine structural features that were verysimilar to those recently described for anotherChlamydomonas species, namely Chlamydomonasreinhardtii (see Fig. 3a, Pérez-Vicente et al., 1995).Although we did not culture these cells for taxonomicidentification, the previous survey by Hoham and thefine structural features observed in our TEMobservations suggest that the cells found in our sampleprobably represent the green alga Chlamydomonasnivalis, and that the LTSEM technique may allow insitu studies of these and other microorganisms that arecommonly found in the spring snow pack.

Importance of Snow Crystal StructureSnow, which may cover up to 53% of the land

surface in the Northern Hemisphere (Foster and Rango,1982) and up to 44% of the world's land areas at anyone time, supplies at least one-third of the water that isused for irrigation and the growth of crops (Gray andMale, 1981). For this reason, calculating the quantityof water that is present in the winter snowpack is anextremely important forecast activity that attempts topredict the amount of water that ultimately will reachreservoirs and estimates how much water will beavailable for agricultural purposes during the followinggrowing season. In 1980, Castruccio et al. (1980)estimated that improving predictions by 1% wouldresult in a $38 million irrigation and hydropowerbenefit for states in the western U.S.A. Today thisfigure has probably doubled as a result of inflation.Remote sensing approaches, using microwave data,have been successfully tested in certain situations tocalculate areal water equivalent of the snowpack priorto melting (Rango et al., 1989; Goodison et al., 1990).Unfortunately, these estimates can be easily confoundedby the sizes and shapes of the snow crystals or grainsthat constitute the snowpack. Our results indicate thatlow temperature SEM can be used to illustrate the sizesand shapes of snow crystals and to follow theirsubsequent metamorphisms that are influenced bytemperature, relative humidity, vapor pressure andtime. Current microwave radiative transfer modelsassume that the snow crystal is spherical. Ourinformation concerning the shapes of snow crystals cannow be used in developing new radiative transfermodels to determine more accurately microwavescattering in the snowpack and thereby assess the snow-water equivalent that is present.

In conclusion, LTSEM provides a newtechnique for observing various types of newlyprecipitated and metamorphosed snow crystals that canbe collected, stored and shipped from remote sites. Thetechnique has resolution, depth of field and stereology

that cannot be achieved with the light microscope. Inaddition, application of this technique can be used toobserve the process of sublimation of snow crystals(Wergin et al., unpublished), to gain information abouticicles, ice fabric, and frost (Wergin et al., 1996) andoffers the potential for x-ray analysis to acquire dataabout the elemental composition of condensing nucleiand particulate pollutants that may becomeincorporated into snow and ice.

5. ACKNOWLEDGEMENTSThe authors thank Christopher Pooley for

converting the SEM negatives to the digital images thatwere used to illustrate this study. We thank RichardArmstrong and Rod Newcomb for assistance inselecting appropriate snowpits in Colorado and AnneNolin and Beth Boyer for assisting in data collection atLoveland Pass. We were also assisted by Dorothy Hall,Jim Foster and Carl Benson in the snowpits nearFairbanks, Alaska. This investigation was partiallyfunded by the NASA Goddard Space Flight Center.

REFERENCESArmstrong, R.L . (1992) The mountain snowpack: pp. 47-83 in:The Avalanche Book, ed. by B.R. Armstrong and K. Williams.Colorado Geological Survey, Denver.

Bentley, W.A. and Humphreys, W.J. (1931) Snow crystals.McGraw-Hill Book Co. Inc., New York, 227 pp.

Blanchard, D.C. (1970) Wilson Bentley, the snowflake man,Weatherwise, 23, 260-69.

Brownscombe, J.L. and Hallet, J. (1967) Experimental and fieldstudies of precipitation particles formed by the freezing ofsupercooled water, Quarterly J. Roy. Met. Soc., 93, 455-73.

Castruccio, P.A., Loats, H.L., Lloyd, D. and Newman, P.A.B.(1980) Cost/benefit analysis for the operational applications ofsatellite snowcover observations (OASSO): pp. 239-254 in: NASAConf Publication 2116, ed. by R. Rango and R. Peterson. Sci. Tech.Info. Off.

Czygan, F.C. (1970) Blutregen und Blutschnee: Stickstoffmangel-Zellen von Haematoccus pluvialis und Chlamydomonas nivalis,Archiv Mikrobiolgie, 74, 69-76.

Dobrowolski, A. (1903) La niege et le givre Expédition AntarctiqueBelge: pp. 1-78 in: Résultats du voyage du S.Y. Belgica en 1897-1898-1899, ed by J.E. Buschmann, Antwerp.

Foster, J.L. and Rango, A. (1982) Snow cover conditions in thenorthern hemisphere during the winter of 1981, J. Climatology, 20,171-183.

Gray, D.M. and Male D.H. (1981) Handbook of Snow:Principles, processes, management and use. Pergamon Press,Ontario, 776 pp.

Goodison, B., Walker, A.E. and Thirkettle, F. (1990)Determination of snow water equivalent on the Canadian prairiesusing near real-time passive microwave data: pp. 297-316 in Proc.

Workshop on Application of Remote Sensing in Hydrology.National Hydrology Research Institute, Saskatoon, Canada.

Hellman, G. (1893) Schneekrystalle. J. Muckenberger, Berlin, 27pp.

Hobbs, P.V. (1974) Ice physics. Clarendon Press, Oxford, 837 pp.

Hoham, R.W. (1992) Environmental influences on snow algalmicrobes: pp. 78-83 in Proc. 16th Annual Western SnowConference, ed. by B.Shafer. Jackson Hole, Wyoming.

LaChapelle, E.R. (1969) Field guide to snow crystals. Universityof Washington Press, Seattle, 101 pp.

Mason, B.J. (1992) Snow crystals, natural and man made,Contemporary physics, 33, 227-243.

McClung, D. and Schaerer, P. (1993) The avalanche handbook.The Mountaineers, Seattle, 271 pp.

Nakaya, U. (1954) Snow Crystals. Harvard University Press,Cambridge, 510 pp.

Nordenskiold, G. (1893) The inner structure of snow crystals.Nature Land., 48, 592-94.

Pérez-Vicente, R., Burón, M.I., González-Reyes, J.A.,Cárdenas, J. and Pineda, M. (1995) Xanthine accumulation andvacuolization in Chlamydomonas reinhardtii cells, Protoplasma,186, 93-98.

Rango, A., Martinec, J., Chang, A.T.C. and Foster, J.L. (1989)Average areal water equivalent of snow in a mountain basin usingmicrowave and visible satellite data, I.E.E.E. Trans. Geoscienceand Remote Sensing, 27, 740-745.

Robards, A.W. and Sleytr, U.B. (1985) Low temperaturemethods in biological electron microscopy. Elsevier, Oxford, 551pp.

Stoyanova, V., Genadiyev, N. and Nenow, D. (1987) Anapplication of the replica method for SEM study of the ice crystalinstability, J. Phys. (Paris), 48, 375-381.

Takahashi, T. and Fukuta, N. (1987) J. Phys. (Paris), 48, 405-411.

Wergin, W.P. and Erbe, E.F. (1994a) Can you image a snowflakewith an SEM? Certainly! Proc. Royal Microsc. Soc., 29, 138-140.

Wergin, W.P. and Erbe, E.F. (1994b) Snow crystals: Capturingsnowflakes for observation with the low temperature scanningelectron microscope, Scanning, 16, IV88-IV89.

Wergin, W.P. and Erbe, E.F. (1994c) Use of low temperaturescanning electron microscopy to examine snow crystals. Proc. 14thInt. Cong. Electron Microsc., 3B, 993-994.

Wergin, W.P., Rango, A. and Erbe, E.F. (1995a) Observations ofsnow crystals using low-temperature scanning electron microscopy,Scanning, 17, 41-49.

Wergin, W.P., Rango, A. and Erbe, E.F. (1995b) Threedimensional characterization of snow crystals using low temperaturescanning electron microscopy, Scanning, 17, V29-V30.

Wergin, W.P., Rango, A. and Erbe, E.F. (1996) Use of lowtemperature scanning electron microscopy to observe icicles, icefabric, rime and frost, J. Microsc. Soc. Amer., Proc. In Press.

Wergin, W.P., Yaklich, R. W. and Erbe, E.F. (1996) Advantagesand applications of low temperature scanning electron microscopy.In: Focus on Modern Microscopy. World Scientific Publishing Co,Ltd., New Jersey. In Press.

Wolff, E.W. and Reid, A.P. (1994) Capture and scanning electronmicroscopy of individual snow crystals, J. Glaciology, 40, 195-197.