Hostname: page-component-6898cfcb55-zsp5v Total loading time: 0 Render date: 2026-11-11T07:50:43.181Z Has data issue: false hasContentIssue false
  • English
  • Français

Seasonal climatic forcing of alpine glaciers revealed with orbital synthetic aperture radar

Published online by Cambridge University Press:  20 January 2017

Laurence C. Smith,
Richard R. Forster,
Bryan L. Isacks and
Dorothy K. Hall
Laurence C. Smith
Affiliation:
Department of Geological Science, Cornell University, Ithaca, New York 14853, U.S.A.
Richard R. Forster
Affiliation:
Department of Geological Science, Cornell University, Ithaca, New York 14853, U.S.A.
Bryan L. Isacks
Affiliation:
Department of Geological Science, Cornell University, Ithaca, New York 14853, U.S.A.
Dorothy K. Hall
Affiliation:
NASA/Goddard Space Flight Center, Hydrological Sciences Branch, Laboratory for Hydrospheric Sciences, Code 974, Greenbelt, Maryland 20771, U.S.A.
Rights & Permissions [Opens in a new window]

Abstract

The evolution of four dynamic radar glacier zones at the surface of an alpine icefield in British Columbia is monitored using a time series of 35 First European Remote Sensing Satellite (ERS-1) synthetic aperture radar (SAR) images acquired from 1992 to 1994. These zones result from changing wetness and textural properties, and appear to represent: (1) cold snow with no liquid water present; (2) an initial melt front with an upper boundary near the elevation of the 0° isotherm; (3) metamorphosed, rapidly melting first-year snow with a rough or pitted surface; and (4) bare ice. This interpretation is aided by temperature and runoff data, air photographs and field measurements of snowpack properties acquired with two ERS-1 SAR scenes, ice-surface elevations derived from 1:50 000 topographic maps and simulations of radar backscatter from a geometric optics model of surface scattering. Meltwater production is affected by the development of zones (2), (3) and (4), which form, migrate up-elevation and disappear each year between April and September.

Information

Type
Research Article
Information
Journal of Glaciology , Volume 43 , Issue 145 , 1997, pp. 480 - 488
Copyright
Copyright © International Glaciological Society 1997

1. Introduction

Most of the world’s alpine glaciers have been retreating during the last 100 years (Reference Meier,Meier, 1984; Reference Haeberli,, Müller,, Aleand,, Büsch. and Ocrlemans,Haeberli and others, 1989), contributing an estimated one-third to one-half of a 10–15 cm rise in sea level observed since 1900 (Reference Meier,Meier, 1984). Shrinking ice volumes have increased the risk and incidence of landslides and glacier-outburst floods in mountainous areas (Reference O’Connor, and Costa,O’Connor and Costa, 1993) and will have an uncertain effect upon water supply from glacier-fed streams. It is clear that more observations of glacier mass balance, melt patterns and snowline position are needed for climate monitoring natural hazard assessment and water-supply purposes. New techniques for monitoring glaciers with orbital synthetic aperture radar (SAR) will complement existing field programs and also permit temporal study of remote sites where glacier conditions are currently known. SARs penetrate cloud cover and darkness to permit frequent observation of dynamic, climate-sensitive properties at the glacier surface and near-surface.

Knowledge of the timing and volume of meltwater run-off is required for effective management of regulated basins wherever humans and glaciers coexist. For this reason, numerous watershed runoff models have been developed that use meteorological data to predict meltwater production, storage and release from glacial sources (Reference Young,Young. 1985).The amount of melting at the glacier or snowpack surface is a key calculation in all such models. This process is rarely measured directly; melt rates are instead Calculated using an energy-balance approach with ground meteorological observations (see Reference Young,Young, 1985). This approach works well where meteorological data are available near the Study glacier but worsens rapidly when data are extrapolated from distant stations. Regardless of station location, the spatial distribution of melt is assumed to vary as a function of elevation or is modeled using a digital elevation model to calculate the energy balance at each gridpoini on the glacier surface. This limitation has hampered development of physically based distributed runoff models designed to predict glacier runoff.

Due to its sensitivity to the presence of liquid water, SAR has strong potential for directly mapping patterns of snow-melt (Reference Donald,, Seglenieks,, Soulis,, Kouwen, and Mullins,Donald and others, 1993; Reference Fahnestock,, Bindschadler,, Kwok, and Jezed,Fahnestock and others, 1993; Reference Hallikainen,, Jääskeläinen,, Kurvonen,, Koskinen,, Herland, Preälä. and Kaldeich,Hallikainen and others, 1993; Reference Rott,, Nagler, and Kaldeich,Rott and Nagler, 1993; Reference Jiancheng and Dozier,Shi and Dozier. 1993; Reference Maxfield, and Stein,Maxfield, 1994). Estimates of snow wetness may be retrieved using Polarimetric C-band SAR data with a backscattering model that considers both surface and volume scattering (Reference Jiancheng, Dozier, and Williamson,Shi and Dewier, 1992, Reference Jiancheng and Dozier,1995; Reference Jiancheng, Dozier, and Rott,Shi and others, 1993). However, relief-induced distortions in SAR data can Severely limit snow-cover mapping in areas of rugged terrain (Reference Haefner,, Holecz,, Meier,, Ntesch, and Piesbergen,Haefner and others, 1994; Reference Rott,Rott, 1994). Over glaciers, areas of wet snow may be observed in SAR imagery from their low backscatter returns Roll and Mätzler, 1987; Reference Bindschadler, and Vornberger,Bindschadler and Vornberger, 1992). Other surface conditions may also be observed. Reference Fahnestock,, Bindschadler,, Kwok, and Jezed,Fahnestock and others (1993) mapped four distinct zones on the Greenland ice sheet using ERS-1 (First European Remote Sensing Satellite) SAR imagery. They were interpreted as corresponding to the dry-snow, percolation, wet-snow and bare-ice glacier facies defined from field measurements by Reference Benson,Benson (1962). High backscattering from icy inclusions buried in the percolation zone during cold conditions has since been successfully modeled as randomly oriented dielectric cylinders embedded in a dry-snow medium (Reference Rignot,Rignot. 1995). Reference Forster,, Isacks and Das,Forster and others (1996) used Polarimetrie SIR-C/S-SAR data to identify climatically driven intra-annual changes in glacier-surface and near-surface conditions in Patagonia. Several multi-temporal ERS-1 SAR studies of glaciers have reported dramatic temporal and spatial variations in back-scattering properties related to glacier-surface conditions (Reference Rott, and Nagler,Rott and Nagler, 1994; Reference Hall,, Williams, and Sigursson,Hall and others. 1995; Reference Rees,, Dowdeswell, and Diament,Rees and others, 1995).

This study presents temporal analyses of 35 ERS-1 SAR images acquired from 1992 to 1991 over an icefield and two outlet glaciers in British Columbia, Canada. Four dynamic zones controlled by conditions of melting and texture are seen in the SAR data. Next, these zones are correlated with daily ground measurements of temperature and runoff in 1992 and 1993, air photographs and field measurements taken during two 1991 ERS-1 SAR acquisitions and ice-surface elevations derived from 1: 50 ООО topographic maps. Finally, simulated radar backscatter values from a simple geometric optics model of radar backscattering are compared with ERS-1 backscatter values from the study glaciers and a nearby snowpack. Model inputs include field measurements of snow properties made within hours of an 8 June 1991 ERS-1 SAR acquisition.

2. The Stikine Icefields, British Columbia

The Coast Range of Alaska and British Columbia is one of the most extensively glacerized regions on Earth, with 88 400 km2 of glacier ice (Reference Meier,Meier, 1984). Glacier mass balance and annual equilibrium-line altitudes in this area may be used as climatic indicators (Reference Pelto,Pelto, 1987). Because the weather conditions during the accumulation season are controlled by a shifting boundary between the Aleutian Low and the Canadian Polar High (Reference Yarnal,Yarunl, 1984), the annual mass balance of glaciers in this area is particularly sensitive to winter atmospheric conditions. Reference Pelto,Pelto (1989) found a strong correlation between the occurrence of winter cyclonic passages and positive mass balances estimated for 90 glaciers in the region.

The Stikine icefields are found in the southern part of the Coast Range. Part of the largest icefield, including two outlet glaciers (56°55′N, 131°00′W), was selected for repeated EAS-1 imaging by the ERS-1 satellite (site 1, Fig. 1). The maximum elevation of this icefield is approximately 1850 m; the outlet-glacier termini are both near 580 m. Daily mean air temperatures at 640 m were observed 30 km away at Bob Quinn Lake. Daily mean meltwater discharge was measured by a Water Survey of Canada gauge located on Forrest Kerr Greek, 11 km downstream of its emergence from the two outlet glaciers. Locations of the discharge and temperature stations are shown in Fig. 1.

Fig. 1.

Stikine icefields, British Columbia, Canada. Locations of study sites and ground temperature and discharge stations are also shown.

The drifting orbit of ERS-1 during its 1994 geodetic mission allowed ascending and descending SAR acquisitions within a 37 hour period on 8 and 9 June. Helicopter transport to a nearby non-glacerized snowpack (site 2, Fig. 1) permitted measurements of snowpack wetness, surface roughness, grain-size, depth and temperature to be made within hours of the 8 June SAR acquisition. Discharge measurements of meltwater runoff in Bronson Creek, the primary drainage from this snow pack, were collected on 31 May and 1,3,5,6 and 10 June. Hand-held oblique air photographs over the Study icefield and outlet glaciers were acquired within minutes of the 9 June descending orbit.

3. The Ers Satellites and Data Processing

ERS-1 was launched on 17 July 1991 by the European Space Agency and placed in adjustable near-circular polar orbit. It has experienced 3, 35 and 168 day repeat cycles during its various mission phases. Converging orbits at high latitudes permit overlapping scenes to be obtained more often. ERS-1 uses a C-band (5.3 GHz) SAR with vertical transmit and receive (VV) polarizations and a look angle of 23°,producing an 80–103 km swath with a nominal spatial resolution of 25 m and a processed pixel spacing of 12.5 m

From April 1992 to June 1994, 35 descending ERS-1 SAR acquisitions of the study site were downloaded by NASA’s Alaska SAR Facility (ASF) in Fairbanks, Alaska. Four scenes acquired during the 1994 Geodetic Phase were also used for comparison With field observations. Each scene was radiometrically calibrated using ASF-provided software to permit comparison of normalized radar backscatter (σ°) at different limes and also within a single scene. The calibration process removes variations in σ° caused by sensor-antenna pattern, range-to-target and look angle, using a satellite-derived noise vs range function and three calibration coefficients (Reference Bicknell,Bicknell, 1992). Corrections are fitted to the Goddard Earth Model (GEM06) geoid. Image speckle was next reduced for each image, using a 5 × 5 median filter. Reference Rees,, Dowdeswell, and Diament,Rees and others (1995) showed that this process reduces the image local standard deviation in backscatter by nearly 80%, while increasing the apparent mean backscatter by only 0.7 dB. In comparison, backscatter fluctuations due to changing glacier-surface conditions vary as much as 19 dB over the study site. A transect of filtered backscatter values, between the terminus of an outlet glacier and the icefield interior was extracted for each ERS-1 SAR image; these results are presented in the next section.

4. Results

4.1. Time trends in ERS-1 SAR backscatter returns over glaciers

The ERS-1 time series reveals four distinct zones which may be identified by their backscatter characteristics and elevational positions with respect to each other. Following the convention of Reference Forster,, Isacks and Das,Forster and others (1996), we refer to them as radar glacier zones. Radar glacier zones are dynamic on the time-scale of days to weeks and should not be confused with glacier facies which are based upon properties integrated over time periods of years (Reference Benson,Benson, 1962). The four radar glacier zones observed at the study site appear in represent conditions of (1) cold, dry winter snow (hereafter called “dry”); (2) snow that is just beginning to melt (hereafter called “M” for “melt front”); (3) a second phase of wet snow that is probably metamorphosed, roughened and rapidly melting (hereafter called “P2”, see below); and (4) bare ice (hereafter called “bare”). Some general characteristics of these four radar glacier zones are summarized in Table 1. The ranges σ° in values are derived from characteristic backscatter transects through each zone.

Table 1.

Description of the four radar glacier zones seen in the ERS-1 SAR time series. “Label” is the zone name, as annotated in Fig. 2

The temporal evolution of the four radar glacier zones is shown in Fig. 2, using images acquired on 20 May. 13 July and 29 July 1992. The icefield interior (found near the bottom of each image) slopes downward toward a pair of outlet glaciers. During the winter (not shown in Fig. 2), the entire ice surface is characterized by relatively high σ° values (−3 to −7dB). On 20 May, the icefield interior still exhibited the bright radar returns of this “dry” zone (see Fig. 2). However, at lower elevations the initiation of surface melting labeled “M” in Fig. 2) is indicated by low radar returns (−11 to −22 dB). By 13 July, this melting snow (which appears dark on the images) has risen to the elevation of the icefield interior. The bright “dry” zone has disappeared from even the highest elevations and melting is now occurring everywhere on the icefield. However, at lower elevations, a new radar glacier zone has formed which is characterized by higher values of σ° (−6 to −8 dB), not unlike the returns from the “dry” zone. Because of the low elevation, conditions are known to be melting for this zone of high σ°. As will be explained later, the high returns are likely due to multiple-path surface scattering. Roughness elements are inferred to be centimeter-scale, on the order of the 5.66 cm wavelength of the ERS-1 SAR instrument. This part of the icefield is also below the annual firn limit (i.e. the snow is no older than the previous melt season). This high-backscatter zone is therefore interpreted as a rapidly melting first-year snow-pack with a metamorphosed, roughened surface that is quite wet; we here suggest the term “phase 2 melt” or “P2” for this third type of radar glacier zone. At the time of writing, such bright radar returns from a wet, first-year snow cover have not been reported over alpine glaciers. Both observations and modeling indicate that increasing the surface roughness and/or wetness of a wet snowpack will increase σ° (Reference Jiancheng, Dozier, and Williamson,Shi and Dozier, 1992; Reference Jezek,, Drinkwater,, Crawford,, Bindschadler, and Kwok,Jezek and others, 1993; Reference Rott,, Nagler, and Kaldeich,Rott and Davis 1993; Reference Dowdeswell,, Rees, and Diament,Dowdeswell and others, 1994) but not to levels high enough to explain the −6 to −8dB returns observed over the P2 zone. High σ° has been observed over melting firn (Reference Rott, and Nagler,Rott and Nagler, 1994) but at levels 4– 6dB below those characteristic of P2. Reference Rees,, Dowdeswell, and Diament,Rees and others (1995) noted high ERS-1 SAR returns near the margins of the Austfonna ice cap. Svalbard but these were interpreted as bare, heavily dissected ice. Bright ERS-1 returns can also be seen on Bruarjökull. Iceland (Reference Hall,, Williams, and Sigursson,Hall and others, 1995) but have not been interpreted.

Fig. 2.

Development and upslape migration of four radar glacier zones revealed by the ERS-1 SAR. These zones are interpreted as: (1) cold winter snow with no liquid water present (dry); (2) initial snowmelt (M), with an upper limit around the 0° C isotherm; (3) rapidly melting, metamorphosed and roughened snow cover (P2); and (4) bar ice. SAR illumination is from the top of each image.

A fourth zone (labeled “bare” in Fig. 2) with backscatter values (−10 to −12 dB) between those of the M and P2 radar glacier zones can be seen immediately down-glacier from the P2 zone on 13 and 29 July ERS-1 SAR images. Its upper boundary moves steadily upglacier throughout the much season and it is not replaced by any new type of radar signature. This zone is interpreted as bare glacier ice. The interface between the bare and P2 zones is therefore believed to represent the position of the summer snowline, which steadily rises in elevation throughout the summer. By 29 July, almost all of zone M in the icefield interior has metamorphosed into P2 and at lower elevation P2 has been replaced in turn by bare ice.

An oblique hand-held air photograph of the study Site and an ERS-1 SAR image acquired moments later are shown in Fig. 3a and b. Cloud shadows create a mottled appearance over pan of the air photograph. Although the glacier surface was observed to be completely, snow-covered and featureless except near the terminus, the ERS-1 SAR image reveals a distinct interface between the dark M zone and a developing P2 zone. This confirms that (1) SAR can identify variations in glacier-surface conditions that cannot be seen in air photographs, and (2) the relatively bright P2 radar returns described earlier are not produced by bare glacier ice. Furthermore, since the bright P2 region in Fig. 3b is well below the annual firn limit, it is also concluded that (3) these P2 returns are produced from a first-year melting snowpack, not firn. The beginnings of a bareice zone on one outlet glacier can be seen in both the air photograph and the SAR image.

Fig. 3.

(a) Hand-held oblique air photograph of the study ice-field and outlet glaciers acquired within minutes of the ERS-1 SAR acquisition shown n (b). Mottled appearance at left is from cloud shadows. Arrow ponts to nunnatak for reference in (b), (b) ERS-1 SAR image of the area photographed in (a), revealing an nterface between a low-return initial melt (M) and a developing P2 radar glacier zone. This interace cannot be seen in the air photograph (note reference arrow at nunatak). Transect used to derive backscatter profiles in Fig. 4 is also shown. SAR illumination is from the bottom of the image.

Our interpretation of the four radar glacier zones illustrated in Fig. 2 is based upon the complete ERS-1 SAR time series over the study site. These scenes show the up-elevation migration of the four zones to be continuous and consistent throughout the 1992 and 1993 melt seasons (April September , Radar backscatter transects from the terminus of one outlet glacier to the icefield interior are presented for all 1992 and 1993 ERS-1 SAR acquisitions in Fig. 4. The transect is 21 km in length and is located on Fig. 3b. Fig. 4 shows that the dark initial melt front (M) was present by mid-April in both 1992 and 1993. The M zone climbs rapidly in elevation and metamorphoses into the bright P2 zone,which is later encroached upon by bare glacier ice. The development and up-slope migration of these zones can be easily observed until late-September, at which point the accumulation of new snow and/or refreezing causes radar returns over the entire icefield to become erratic.

Fig. 4.

Time series of ERS-1 SAR σ° transects between an outlet-glacier terminus (0 km) and the icefield interior (21 km). Transect location is shown in Fig. 3b. Inset contains the transect topographic profile and a sample σ° transect showing the bare. P2 and M radar glacier zones.

The topographic profile (inset.Fig. 4) of the radar back-scatter transects was derived from the 1: 50000 scale 104 B/ 11 and 104 B/14 Canadian topographic maps. The ice surface descends smoothly from 1830 to 580 m with a mean slope of 2.9°, except for a short reach at about 19 km where a slope of 16° is found. Assuming equal slope in both the range and azimuth direction, the absolute error in σ° resulting from the pixel-area difference between the actual ice surface and a spherical earth model (Reference Van Zyl,, Chapman,, Dubois and Shi.van Zyl and others, 1993) is calculated at less than ± 1 dB over nearly all of the ice surface. This absolute error is much less than the 19 dB range σ° variation observed between the four radar glacier zones (−3 to −22 dB). The effect of a sleep icefall on the illuminated pizel-size calculation is found at 19km, where a consistent spike of high σ° can be seen in most of the backscatter transects (Fig. 4).

4.2. Correlation of radar glacier zones with temperatures and runoff

Daily mean temperatures recorded at Bob Quinn Lake in 1992 and 1993 were used to calculate daily elevations of the 0°C isotherm, assuming a temperature lapse rate of 6.5° C km−1. Upper and lower boundary elevations for each radar glacier zone were estimated from 1:50 000 topographic maps; these elevations are plotted with the isotherm elevations in Fig. 5 for the 1992 and 1993 melt seasons. Estimated vertical elevation errors resulting from this procedure are also shown (except where boundary elevations exceed the maximum icefield elevation). Errors in point elevations determined from topographic maps are nearly insignificant, particularly since boundaries between radar glacier zones are not always sharp and may occur over a range of elevations as large as 200 m.

Fig. 5.

Time series for the 1992 and 1993 melt seasons of: (1) boundary elevations between the dry, M, P2 and bare radar glacier zones (each point is derived from a single ERS-1 SAR image); (2) daily elevations of the 0° isotherm calculate from ground observations of air temperature; and (3) daily mean meltwater discharge measured in Forrest Kerr Creek.

Winter images over the icefield are characterized by uniform and high values of σ° (−3 to −7 dB), in agreement with the results of Reference Fahnestock,, Bindschadler,, Kwok, and Jezed,Fahnestock and others (1993). Reference Rott, and Nagler,Rott and Nagler (1994) and Reference Rees,, Dowdeswell, and Diament,Rees and others (1995) High winter backscatter intensities likely result from low dielectric losses in dry snow and volume scattering from sub-surface ice lenses and pipes (Reference Rott,Rott, 1994). The M zone appears by mid-April in both years. Its upper boundary is approximately coincident with the elevation of the 0° C isotherm until the isotherm exceeds the maximum elevation of the icefield. At this point, the upper limit of the M zone equals the maximum icefield elevation; its progress could undoubtedly be monitored further in time if the topography were higher. Metamorphosis of the M zone into P2 begins around 1 June. Bare glacier ice is first seen in early-July (1992) and late-June (1993). All four radar glacier zones rise steadily in elevation throughout the melt season until their obliteration in September.

Daily mean discharges of meltwater runoff from Forrest Kerr Creek are also shown in Fig. 5. Flow routing was not applied to these data. Despite detection of internal melt-Water storage in these glaciers (Reference Tangborn,Tangborn, 1984), runoff is strongly correlated with air temperature. The appearance of the M zone precedes significant increases in stream flow by about 2 weeks.These higher flows are associated with for presence of a P2 zone but the temporal sampling of ERS-1 is not line enough to determine whether P2 conditions must develop before large meltwater volumes are produced, However, it can be concluded from Fig. 5 that increased meltwater production precedes the appearance of bare glacier ice by at least 1 month.

4.3. ERS-1 SAR returns and field measurements from a melting snowpack

Helicopter transport to a non-glacier snowpack locked approximately 20 km from the primary study area (Fig. 1) permitted characterization of physical snow conditions within hours of an ERS-1 SAR acquisition on 8 June 1994. Nine snow pits were dug along a 500 m transect between the snowpacks maximum elevation and its lowest edge, field measurements of snow depth, wetness, temperature, surface roughness and grain-size collected from these pits air given inTable 2. The surface roughness is characterized by the Standard deviation of the surface roughness is characterized by the standard deviation of the surface height (rms height σh) and the correlation length (lc ) (Reference Ulaby,, Moore, and Fung,Ulaby and others, 1982). The σh, values at all locations are large enough to characterize the surface as “not smooth” based upon both the Rayleigh and the Fraunhofer criteria (Reference Ulaby,, Moore, and Fung,Ulaby and others, 1982). The horizontal resolution of surface profiles was insufficient to separate lc for several of the pits. The volumetric snow wetness (Wv) was measured with a Denoth dielectric meter (Reference Denoth,Denoth, 1909) near the surface and at depth. Mean surface snow-grain diameters were estimated from photographs of representative samples. High wetness values, observations office water, and 0° C temperatures throughout the pits indicate that the snow pack was wet and melting at all depths (Table 2).

Table 2.

Results of snow-pit measurements collected from a non-glacier snowpack within hours of an 8 June 1994 ERS-1-SAR acquisition. Surface roughness near the snow pit is characterized by the standard deviation of the mean surface height(σ°) and the correlation length (lc). Snow depths are based on five probe measurements made in the general areas of each snow pit. Volumetric snow wetness (Wv) is calculated from a capacitance wetness meter reading and snow density at upper (25 cm) and lower (26–34 cm) snow depths. Snow wetness calculations for pits 6–9 use the mean density of pits 1–5. Error bounds on mean snow depth and grain-size are one standard deviation. Snow temperatures were measured between 26 and 34 cm in depth

Low backscatter returns (≈−20dB) were recorded everywhere over the test site despite the range in snowpack properties given in Table 2. Ascending ERS-1 SAR images acquired over this snowpack on 31 May. 8 and 25 June are given in Fig. 6. Wet-snow extent at the test site is greatest on 19 May, reduced on 8 June, and nearly gone by 25 June. Meltwater discharge in Bronson Creek, the primary drainage from the test snowpack, was measured on 31 May and 1, 3, 5,6 and 10 June using Price AA current meters. Measurements were carried out at approximately the same time each morning. Flows rose Steadily during this time period (4.5, 4.8, 5.3. 6.0, 6.4 and 6.8 m3s−1). Discharge data were not collected during the 19 May and 25 June ERS-1 SAR acquisitions. However, the data do Suggest that meltwater production increased while the extent of dark, melting snow at the test site and surrounding mountain slope decreased (Fig. 6).

Fig. 6.

Multi-temporal ERS-1 SAR images of the study snowpack (see arrow) and surrounding area. The extent of low radar returns is greatest on 19 May 1991, intermediate on 8 June and diminished by 25 June. Field measurements of snowpack properties given in Table 2 were collected within hours of the 8 June image. Location of meltwater-discharge measurements made in Bronson Creek is also shown. Radar illumination is from the top of each image.

5. Simulation of Backscatter Returns Using a Geometric Optics Model

This section presents simulated values of σ° generated by a model of radar backscattering from a wet snowpack. Field measurements described in section 4.3 are used as inputs. Simulated values of σ° agree with observed values from the M zone but are lower than those from the P2 zone.

In snow packs with high liquid-water content and a relatively rough surface, C-band surface scattering dominates over volume scattering at all incidence angles (Reference Jiancheng, Dozier, and Williamson,Shi and Dozier, 1992). Here, we use a simple surface-scattering model (Reference Jezek,, Drinkwater,, Crawford,, Bindschadler, and Kwok,Jezek and others. 1993) to (1) compare ERS-1 backscatter intensities with values modeled for a wet alpine snowpack where ground measurements were obtained (Table 2); (2) estimate the range of surface conditions for the M zone; and (3) aid interpretation of the P2 zone. Model results are constrained by input values of snowpack surface roughness and dielectric constant. Surface roughness is parameterized as the rms slope of the surface. Assuming a Gaussian distribution of slopes, the rms slope

can be calculated as 1.14σh/lc , (where σh, is the standard deviation of the height and lc , is the correlation length). The dielectric constant is derived from the snow wetness and density through an empirical Debye-like model (Reference Ulaby,, Moore, and Fung,Ulaby and others, 1986).

Simulations of σ° as a function of surface roughness for snowpacks of 6, 10 and 15% wetness are shown in Fig. 7 average wetness measured in the field was 6%).The input snow density of 0.52 Mg m−3 was obtained by averaging held measurements (Table 2). The input local incidence angle of 35° combines the ERS-1 incidence angle (23°) and the ground slope at the test area (12°). Ranges of rms slopes measured at the snow pits and ERS-1 SAR backscatter values recorded over the test site are also shown in Fig. 7. The σ° values have been decreased by 2 dB to collect for the local slope effect (Reference Van Zyl,, Chapman,, Dubois and Shi.van Zyl and others, 1993).

Fig. 7.

Simulated C-VV σ° curves for three wet snowpacks (liquid water contents: 6, 10 and 15%), using a geometric optics model of radar backscattering. Increased backscatter is associated with increased surface roughness and liquid-water content. The ranges of surface roughness measured on the ground on 8 June 1994 (A) and observed σ° recorded hours later (B) are also shown. Model value of σ° simulated using average density (0.52 Mg m−3) and incidence angle (35°) measured at the test site is shown at (C).

Model results for the mean snow conditions measured at the site (wetness = 6%, roughness slope = 17°, density = 0.52 Mg m−3 are within 2 dB of the range of terrain-corrected σ° values measured by ERS-1. Also, the model predicts the backscatter at the minimum value of measured surface roughness to be well below the noise level of this ERS-1 scene. Over the test glaciers and icefield, ERS-1 σ° values (−11 to −22 dB) from the M zone are contained within the simulated range of 6–15% snowpack wetness. These results suggest that the model’s assumptions of single scattering and Gaussian slopes are valid for freshly melting snow.

However, the geometric optics model fails to simulate the higher returns (−6 to −8 dB) from the P2 zone. Forcing Snow-Wetness values above 15% is physically unreasonable and increasing surface slopes only decreases σ° Increasing snow density to 0.7 Mg m−3 increases the 15% wetness return by only 1 dB. Changing the incidence angle to 23° and terrain correction for the mean glacier slope (3°) also increases the returns only slightly. The highest σ° value the model predicts for snow wetness not exceeding 15% is only −13 dB.

This failure may result from the model’s assumption of single-path scattering. To simulate σ° returns from P2-type surfaces, a model that considers multiple scattering, such as the integral equation model (Reference Fung,, Li, and Chen,Fung and others, 1992), may be required. Rough, wet-snow surface conditions that could induce multiple-path scattering (and hence strong back-scatter) include sun cups, penitente snow, ploughshares, perforated crust and suncrust. A hydrological mechanism that may induce multiple-path scattering and also promote a rough surface is the draining of surface meltwater into the snowpack. This process can form vertical pipes of liquid water in the sub-surface, creating areas of sharp dielectric contrast and strong radar returns (Reference Kattelmann,Kattlemann, 1995; Reference Jiancheng and Dozier,Shi and Dozier, 1995). Development of a vertical drainage system would also contribute to the increased melt water discharge observed during P2 conditions (Fig. 5).

6. Discussion

Radar glacier zones are formed by important physical processes that occur at the glacier surface, particularly surface melt and snow diagenesis. While this information cannot be used to estimate glacier mass balance directly, variability in these processes can be observed and related to changing climatic conditions that do affect mass balance (Reference Fahnestock,, Bindschadler,, Kwok, and Jezed,Fahnestock and others, 1993). For example. Reference Forster,, Isacks and Das,Forster and others 1996, Reference Forster,, Smith, and Isacks,1997) identified seasonal trends and short-term weather effects, respectively, from SIR-C/X-SAR data acquired over Hielo Patagónico Sur (southern Patagonia icefield), providing a regional interpretation of the climatic regime over the icefield. Our own findings suggest that orbital SAR may be used to establish the timing of melt onset over glaciers. Support for this statement is provided by Reference Winebrenner,, Nelson,, Colony, and West,Winebrenner and others (1994), who found a close correlation between steep decreases in ERS-I SAR backscatter over sea ice and the ousel of 0– C (or higher) temperatures recorded by six drifting buoys in the Beaufort Sea.

In agreement with the results of Reference Rees,, Dowdeswell, and Diament,Rees and others (1995), we find that SAR may be used to monitor the position of the transient snowline on large glaciers. This may have value in efforts to monitor glacier mass balance. Equilibrium-line altitude (ELA) can in many cases be approximated as the position of the late-summer snowline (Reference Østream,Østrem. 1975; Reference Østrem, and Brugman,Østrem and Brugman, 1991). ELA is in turn related to glacier mass balance (Reference Østream,Østrem, 1975; Reference Braithwaite,Braithwaite, 1984), although some glaciers require knowledge of the transient snowline and the ELA at different times during the ablation season to determine mass balance (Reference Pelto,Pelto, 1987). For the study site, the maximum late-summer snowline elevations in 1992 and 1993 derived from ERS-1 SAR imagery were 1220 m and 1460 m, respectively. The ELA of the nearby Porcupine Glacier has previously been estimated at 1400 m (Reference Pelto,Pelto, 1987). The Closeness of these values suggests a potential use for SAR in complementing ELA-monitoring programs.

Comparison of air temperature and runoff measurements with patterns of M-zone development suggest that only a thin layer of surface melting is required to sharply decrease σ° over snow. The appearance of the M zone is correlated with the first arrival of above-freezing temperatures and precedes increased stream flow by around 2 weeks. ERS-1 SAR images acquired on 19 May, 8 and 25 June 1994 over a nearby non-glacierized snowpack show a diminishing area of low radar returns. Field measurements on this snowpack collected on 8 June indicate that by this time it was rapidly melting and 0° C throughout. Meltwater discharge measured in Bronson Creek increased steadily between 31 May and 10 June, suggesting that maximum runoff production occurs later than peak extent of low radar returns. The P2 zone may be correlated with high melt water production but more work is needed to determine the relationship between the evolution of radar glacier zones and meltwater discharge in streams.

Conclusion

35 ERS-1 SAR scenes of an icefield and two outlet glaciers reveal the temporal and spatial evolution of four zones related to surface wetness and textural properties. They appear, migrate up-elevation and disappear in response to spring and summer melting. Boundaries between these zones appear to represent: (1) the 0° C isotherm; (2) a diffuse transition between fresh wet snow and rough wet snow; and (3) snowline. After the first few weeks of melting each year, high backscatter returns (−6 to −8 dB) are found over first-year snow. A proposed mechanism for these high radar returns is multiple-path scattering from a wet. pitted snow surface. Meltwater volumes delivered to streams are sensitive to the occurrence of these zones. Increased stream flow begins about 2 weeks after the initiation of melting on the glacier surface.

Acknowledgements

Support for this work was provided by NASA through Earth Observing System (EOS) grant NAGW-2638, Spaceborne Imaging Radar (SIR-C) grant 958745 and a Graduate Student Researchers Program Fellowship NGT-51223. ERS-1 data were provided by the European Space Agenex and acquired by the NASA/Alaska SAR Facility. Hovercraft and helicopter transport was generously donated by Cominco Snip Operation and Pamicon Inc. The Denoth snow-wetness meter was provided by R. Davis at the U.S. Army Cold Regions Research and Engineering Laboratory. Assistance from T. Blodgett, S. Das, A. Krat, J. Liebeskind and I. Prcuss is gratefully acknowledged. Interpretation of P2 backscatter returns was aided by discussions with M. Miller (Glaciological and Arctic Sciences Institute. University of Idaho) and J. Dozier (School of Environmental Science and Management. University of California, Santa Barbara).

References

Benson,, C S. 1962. Stratigraphic studies in the snow and firn of the Greenland ice sheet, SIPRE Res. Rep. 70.Google Scholar
Bicknell,, T. 1992. User’s guide to products, Alaska SAR facility SAR processor system. Version 1.0. Pasadena, CA, Jet Propulsion Laboratory. (Report JPL D- 9362.)Google Scholar
Bindschadler,, R.A. and Vornberger,, P.L. 1992. Interpretatio of SAR imagery of the Greenland ice sheet using coregistered TM imagery. Remote Sensing Environ., 42(3), 167–175.Google Scholar
Braithwaite,, R.J. 1984. Cand the mass balance of a glacier be estimated from its equilibrium-line altitude. J. Glaciol., 30(106), 364–168.Google Scholar
Denoth,, A. 1989. Snow dielectric measurements. Adv. Space Res., 9(1), 233–243.Google Scholar
Donald,, J.R., Seglenieks,, F.R., Soulis,, E.D., Kouwen,, N. and Mullins,, D.W. 1993. Mapping partial snowcover during the melt season using C-band SAR imagery. Can. J. Remote Sensing., 19(1), 68–76.Google Scholar
Dowdeswell,, J.A., Rees,, W.G. and Diament,, A.D. 1994. ERS-1 SAR investigations of snow and ice facies on ice caps in te Ruropean High Arctic. In Space at the Service of our Environment. Procedings of the Second ERS-1 Symposium, 11–40 October 1993, Hnburg, Germany. Vol. 2. Paris. European Space Agencyh, 1171–1176. (ESA SP-361).Google Scholar
Fahnestock,, M., Bindschadler,, R., Kwok,, R. and Jezed,, K. 1993. Greenland ice sheet surface properties and ice dynamics fromERS-1 SAR imagery. Science., 262 (5139), 153–1534.Google Scholar
Forster,, R.R., Isacks, B.L., and Das,, S. 1996. Shuttle imaging radar (SIR-C/S-SAR) reveals near surafce properties of the South Patagonian Ice-field. J. Geophys. Res., 101(E10), 23,169–23,180.Google Scholar
Forster,, R.R., Smith,, L.C. and Isacks,, B.L. 1997. Effects of weather events on X-SAR returns from ice fields: case-study of Hiclo Patagonico Sur, South America. Ann. Glaciol., 24, 367–374.Google Scholar
Fung,, A.K., Li,, Z. and Chen,, K.S. 1922. Backscatter from a randomly rough diclectric surface. IEEE Trans. Geasci,. Remote SEnsing., GE-30(2), 356–369.Google Scholar
Haeberli,, W., Müller,, P. Aleand,, P. and Büsch., H. 1989. Glacier changes folowing the Little Ice Age —a survey of the international data basis and its perspeectives. In. Ocrlemans,, J., ed. Glacier fluctuations and climatic change. Dordrecht, etc., Kluwer Academic Publishers, 77–101.Google Scholar
Haefner,, H., Holecz,, F., Meier,, W., Ntesch,, D. and Piesbergen,, J. 1994. Capabilities and limitation of ERS-1SAR data for snowcover determination in mountainous regions. In Space at the Service fo our Environment, Proceeding of the Second ERS-1 Symposium, 11–14 October 1993, Hamburg, Germany. Vol. 2. Paris, European Space Agency, 971–967. (ESA SP-361)Google Scholar
Hall,, D.K., Williams,, R.S. Jr., and Sigursson,, O. 1995. Glaciological observations of Brúarjükull, Iceland, using synthetic aperture radar and thematic mapper satellite data. Ann. Glaciol., 21, 271–276.Google Scholar
Hallikainen,, M., Jääskeläinen,, V. Kurvonen,, L., Koskinen,, J., Herland, E.-A. and Preälä., J. 1933. Application of ERS-1 SAR data to snow mapping. In. Kaldeich,, B., ed. Proceedings of the First ERS-1 Symposium Space at the service of our Environment, 4– November 1992, Cannes, France. Vol. 2. Paris, European Space Agency, 589–594. (ESA Special Publication SP-359).Google Scholar
Jezek,, K.C., Drinkwater,, M.R., Crawford,, J.P., Bindschadler,, R. and Kwok,, R., 1993. Analysis of synthetic aperture radar data collected over the southwestern Greenland ice sheet. J.Glaciol., 39(131), 119–132.Google Scholar
Kattelmann,, R.C. 1995. Water movement and ripening processes in snowpacks of the Sierra Nevada. (Ph.D. thesis, University of California, Santa Barbara.)Google Scholar
Maxfield,, A.W. 1994. Radar satellite snowmelt detection in the Canadian Rocky Mountains. In. Stein,, T.I., ed. IGARSS ’94. Surface and Atmospheric Remote Sensing: Technologies, Data Analysis and Interpretation, Pasadena, California Institute of Technology, California, USA, August 8–12, 1944. Proceedings. Vol. 4. Pisacataway, NJ, Institute of Electrical and Electronics Engineers, 2074–2077.Google Scholar
Meier,, M.F. 1984. Contribution of small glaciers to global sea level. Science., 226(4681), 1418–1421.Google Scholar
O’Connor,, J.E. and Costa,, J.E. 1993. Geologic and hydrologtic hazards in glacierized basins in North America resulting from 19th and 20th century global warming. Natural Hazards., 8(2), 121–140.Google Scholar
Østream,, G. 1975. ERTS data in glacioloty—an effort to monitor glacier mass balance from satellite imagery. J. Glaciol., 15(73), 403–145.Google Scholar
Østrem,, G. and Brugman,, M. 1991. Glacier mass-balance meaurements. A manual for field and office work. Saskatoon, Sask., Environment Canada. National Hydrology Research Institute. (NHRI Science Report 4.)Google Scholar
Pelto,, M.S. 1987. Mass balance of south-east Aaska and north-west British Columbia glaciers form 1976 to 1984: methods and results. Ann. Glaciol., 9, 189–194.Google Scholar
Pelto,, M.S. 1989. Satellite determination of Coast Range, Alaskan glacier mass balance related to atmospheric circulation. Intrnational Association of Hydrological Science Publication. 186 (Symposium at Baltimore 1989—Remote Sensng and Large-Scale global Processes), 127–137.Google Scholar
Rees,, W.G., Dowdeswell,, J.A. and Diament,, A.D. 1995. Analysis of ERS-1 synthetic aperture radar data from Nrdaustlandet, Svalbard. Int. J. Remote Sensing., 16(5), 905–924.Google Scholar
Rignot,, W. 1995. Backscatter model for the unusualradar properties of the Greenland ice sheet. J. Geophys. Res., 100(Es), 9389–9400.Google Scholar
Rott,, H. 1994. Thematic studies in alpine areas by means of polarimetric SAR and optical imagery. Adv. Space Res., 14(3), 217–226.Google Scholar
Rott,, H. and Davis,, R.E. 1993. Multifrequency and polarimetric SAR obervations on alpine glaciers. Ann. Glaciol., 17, 98–104.Google Scholar
Rott,, H. and Môtzler., C. 1987. Possibilities and limits of synthetic aperture radar for snow and glacier surveyng. Ann. Glaciol., 9, 195–199.CrossRefGoogle Scholar
Rott,, H. and Nagler,, T. 1993. Snow and glacier nvestigations by ERS-1 SAR — first result. In. Kaldeich,, B., ed. Proceedings of the First ERS-1-Symposium Space at the Service of our Environment, 4–6 November 1992, Cannes, France. Vol. 2. Paris, European Space Agancy, 577–582. (ESA Special Publication SP-359.)Google Scholar
Rott,, H. and Nagler,, T. 1994. Capabilities of ERS-1 SAR for snow and glacier monitoring in alpine areas. In Spce at the service of our Environment. Proceedings of the Second ERS-1 Symposium, 11–14 October 1993, Hamburg, Germany. 1992. Paris, European Space Agency, 965–970. (ESA SP-361.)Google Scholar
Jiancheng, Shi and Dozier,, J. 1992. Radar backscattering response to wet snow. In. Williamson,, R., ed. IGARSS’22. International Space Year: Space Remote Sensing, Houston, Texas, May 26–29, 1992. Proceedings. Vol 2. New York, Institute of Electrical and Electronics Engineers, 927–929.Google Scholar
Jiancheng, Shi and Dozier,, J. 1993. Measurements of snow-and glacier-covered areas with single-polarization SAR. Ann, Glaciol., 17, 72–76.Google Scholar
Jiancheng, Shi ad Dozier,, J. 1995. Inferring snow wetness using C-band data from SIR-C’s polarimetric sthetic aperture radar. IEEE Trans. Geosci. Remote Sensing., GE-33(4), 905–914. (Correction: GE-33(6), 1995, 1340).Google Scholar
Jiancheng, Shi, Dozier,, J. and Rott,, H. 1993. Deriving snow liquid wate content using C-band polarimetric SAR. In IGARSS’93. Better Understanding of Earth Environment, Tokyo, Japan, 18–21 August 1993. Proceedings. Vol. 3. New York, Institute of Electrical and Electronies Engineers, 1038–1041.Google Scholar
Tangborn,, W.V. 1984. Prediction of glacier derived runoff for hydroelectric development. Geogr. Ann., 66A(3), 257–265.Google Scholar
Ulaby,, F.T., Moore,, R.K. and Fung,, A. K., eds. 1982. Radar remote sensing and surface scattering and emission theory. vol.2. Reading, MA, Addison-Wesley Publishing Co.Google Scholar
Ulaby,, F.T., Moore,, R.K. and Fung,, A.K. 1986. Microwae remote sensing: active and passive. Vol.3. Dedham, MA, Artech House.Google Scholar
Van Zyl,, J., Chapman,, B., Dubois, P and WShi., P 1993. The effect of topography on SAR calibration. IEEE Trans. Geosci. Remote Sensing. GE-31(5), 1036–1043.Google Scholar
Winebrenner,, D.P., Nelson,, E.D., Colony,, R. and West,, R.D. 1994. Observation of melt onset on mutiyear Arctie sea ice usig the ERS 1 synthetic aperture radar. J. Geophys. Res., 99(C11), 22,425–22,441.Google Scholar
Yarnal,, b. 1984. Relationships between synoptic-scale atmospheric circulation and glacier mass balance in south-western Canada during the International Hydrological decade, 1965–74. J. Glaciol., 30(105), 188–198.Google Scholar
Young,, G.J., ed. 1985. Technique for prediction of runoff from glacierized areas. Wallingfore, Oxon., International Association of Hydrological Sciences. (IAHS Publication 149.)Google Scholar
Figure 0

Fig. 1. Stikine icefields, British Columbia, Canada. Locations of study sites and ground temperature and discharge stations are also shown.

Figure 1

Table 1.Description of the four radar glacier zones seen in the ERS-1 SAR time series. “Label” is the zone name, as annotated in Fig. 2

Figure 2

Fig. 2.Development and upslape migration of four radar glacier zones revealed by the ERS-1 SAR. These zones are interpreted as: (1) cold winter snow with no liquid water present (dry); (2) initial snowmelt (M), with an upper limit around the 0° C isotherm; (3) rapidly melting, metamorphosed and roughened snow cover (P2); and (4) bar ice. SAR illumination is from the top of each image.

Figure 3

Fig. 3.(a) Hand-held oblique air photograph of the study ice-field and outlet glaciers acquired within minutes of the ERS-1 SAR acquisition shown n (b). Mottled appearance at left is from cloud shadows. Arrow ponts to nunnatak for reference in (b), (b) ERS-1 SAR image of the area photographed in (a), revealing an nterface between a low-return initial melt (M) and a developing P2 radar glacier zone. This interace cannot be seen in the air photograph (note reference arrow at nunatak). Transect used to derive backscatter profiles in Fig. 4 is also shown. SAR illumination is from the bottom of the image.

Figure 4

Fig. 4.Time series of ERS-1 SAR σ° transects between an outlet-glacier terminus (0 km) and the icefield interior (21 km). Transect location is shown in Fig. 3b. Inset contains the transect topographic profile and a sample σ° transect showing the bare. P2 and M radar glacier zones.

Figure 5

Fig. 5.Time series for the 1992 and 1993 melt seasons of: (1) boundary elevations between the dry, M, P2 and bare radar glacier zones (each point is derived from a single ERS-1 SAR image); (2) daily elevations of the 0° isotherm calculate from ground observations of air temperature; and (3) daily mean meltwater discharge measured in Forrest Kerr Creek.

Figure 6

Table 2.Results of snow-pit measurements collected from a non-glacier snowpack within hours of an 8 June 1994 ERS-1-SAR acquisition. Surface roughness near the snow pit is characterized by the standard deviation of the mean surface height(σ°) and the correlation length (lc). Snow depths are based on five probe measurements made in the general areas of each snow pit. Volumetric snow wetness (Wv) is calculated from a capacitance wetness meter reading and snow density at upper (25 cm) and lower (26–34 cm) snow depths. Snow wetness calculations for pits 6–9 use the mean density of pits 1–5. Error bounds on mean snow depth and grain-size are one standard deviation. Snow temperatures were measured between 26 and 34 cm in depth

Figure 7

Fig. 6.Multi-temporal ERS-1 SAR images of the study snowpack (see arrow) and surrounding area. The extent of low radar returns is greatest on 19 May 1991, intermediate on 8 June and diminished by 25 June. Field measurements of snowpack properties given in Table 2 were collected within hours of the 8 June image. Location of meltwater-discharge measurements made in Bronson Creek is also shown. Radar illumination is from the top of each image.

Figure 8

Fig. 7.Simulated C-VV σ° curves for three wet snowpacks (liquid water contents: 6, 10 and 15%), using a geometric optics model of radar backscattering. Increased backscatter is associated with increased surface roughness and liquid-water content. The ranges of surface roughness measured on the ground on 8 June 1994 (A) and observed σ° recorded hours later (B) are also shown. Model value of σ° simulated using average density (0.52 Mg m−3) and incidence angle (35°) measured at the test site is shown at (C).