Various geomatic measurement techniques can be efficiently combined for surveying glacier fronts. Aerial photographs and satellite images can be used to determine the position of the glacier terminus. If the glacier front is easily accessible, the classic surveys using theodolite or total station, GNSS (Global Navigation Satellite System) techniques, laser-scanner or close-range photogrammetry are possible. When the accessibility to the glacier front is difficult or impossible, close-range photogrammetry proves to be useful, inexpensive and fast. In this paper, a methodology combining photogrammetric methods and other techniques is applied to determine the calving front position of Johnsons Glacier. Images taken in 2013 with an inexpensive nonmetric digital camera are georeferenced to a global coordinate system by measuring, using GNSS techniques, support points in accessible areas close to the glacier front, from which control points in inaccessible points on the glacier surface near its calving front are determined with theodolite using the direct intersection method. The front position changes of Johnsons Glacier during the period 1957–2013, as well as those of the land-terminating fronts of Argentina, Las Palmas and Sally Rocks lobes of Hurd glacier, are determined from different geomatic techniques such as surface-based GNSS measurements, aerial photogrammetry and satellite optical imagery. This provides a set of frontal positions useful, e.g., for glacier dynamics modeling and mass balance studies.
Link to the data repository:
Hurd and Johnsons glaciers are located in Hurd Peninsula, Livingston Island,
the second largest island of the South Shetland Islands (SSI) archipelago
(Fig. 1). Johnsons is a tidewater glacier, calving small icebergs into the
proglacial bay known as Johnsons Dock, while the fronts of the various
tongues of Hurd Glacier (Argentina, Las Palmas and Sally Rocks) are
land-terminating. The three unnamed small sea-terminating glacier basins
draining to False Bay, to the south-east of Hurd Peninsula (U1, U2, U3 in
Fig. 1), which are very steep and heavily crevassed, are not covered in
the current study. Johnsons and Hurd glaciers are polythermal, though
Johnsons, as compared with Hurd, has a higher proportion of temperate ice
(Navarro et al., 2009). Typical velocities near the calving front of
Johnsons Glacier are about 50 m yr
Focusing on the studies dealing with geomatic techniques, Palà et al. (1999) did photogrammetric work in February 1999 focused on the terminal
zone of Johnsons Glacier. However, their study only dealt with cartography
of ash layers which originated from the eruptions in the neighboring Deception
Island (see inset of Fig. 1). They did not determine Johnsons calving
front position, since, due to the location of the observation points on the
top of Johnsons/Charrúa Peak (340 m a.s.l., see upper right corner of
Fig. 1), only a small fraction of the south-western part of the calving
front was visible. Calvet et al. (1999) did an interesting study of the
front position changes of the main basins of Livingston Island during the
period 1956–1996, based on the Directorate of Overseas Surveys (DOS, 1968a,
b)
The current study aims to make available to the scientific community the
whole set of front positions of Johnsons Glacier, and Argentina, las Palmas
and Sally Rocks fronts of Hurd Glacier, between 1957 and 2013. As discussed
above, only some front positions for 1957 and 2000 have been published
before (Molina et al., 2007; Navarro et al., 2013) while all others were
unpublished so far. There are two main reasons for the interest of these
front position changes: (1) the surface mass balance of both Hurd and
Johnsons glaciers has been monitored since 2001 as part of the World Glacier
Monitoring Service database (
The aerial and satellite photographs, and surface-based photogrammetric and
GNSS measurements which form the basis of the present study are summarized
below (Rodríguez et al., 2015):
DOCU 1: flight made by the British Antarctic Survey in December 1957. We
have selected a total of 5 frames (X26FID0052130, X26FID0052131,
X26FID0052132, X26FID0052160 and X26FID0052161) to study Hurd Peninsula
glacier fronts. DOCU 2: flight made by the United Kingdom Hydrographic Office (UKHO) in
January 1990. We have selected a total of 3 frames (0097, 0098 and 0099) for
our study. DOCU 3: photogrammetric survey, using metric camera, performed by Palà
et al. (1999) from the top of Johnsons Peak in 1999. DOCU 4: satellite image obtained by the Quickbird system in January 2010 for
Hurd Peninsula. DOCU 5: satellite image obtained by the Quickbird system in February 2007
for Hurd Peninsula. DOCU 6: inventory of data (2000–2012) by the Group of Numerical Simulation
in Science and Engineering (GSNCI, in its Spanish acronym) of Universidad
Politécnica de Madrid (several authors are affiliated to this group).
These observations are made with GNSS techniques and theodolite and are
focused on Sally Rocks, Las Palmas and Argentina land-terminating fronts,
excluding the calving front of Johnsons Glacier. DOCU 7: photogrammetric survey (using non-metric camera) of the calving
front of Johnsons Glacier conducted in February 2013.
Location of Johnsons and Hurd glaciers in Hurd Peninsula.
Base map:
Photogrammetry is a well-known technique allowing one to obtain
three-dimensional information from photographs using stereoscopic vision
provided by two different points of view (Wolf, 1983). Its fundamental
principle is triangulation. By taking photographs from at least two
different locations, so-called “lines of sight” can be traced from each
camera to points on the object. These lines of sight (sometimes called rays
due to their optical nature) are mathematically intersected to produce the
three-dimensional coordinates of the points of interest. This requires a precise
knowledge of the position and orientation of the cameras. Resection is the
procedure used to determine the position and orientation (also called aiming
direction) of the camera, using ground control points appearing on the
images that have known coordinates. For a good resection, at least more than
10 well-distributed points in each photograph are needed (Kraus, 1993).
Each camera's position is defined by three coordinates, while three angles
are needed to define its orientation. The theoretical central projection can
be deformed by lens and film distortion. These influences can be accounted
for in a bundle block adjustment by introducing correction polynomials in
the observation equations, whose coefficients are determined in the
adjustment (Kraus, 2007). The distortion varies with the distance from each
point to the center of the optical axis. Distortion is often decomposed into
radial and tangential components (Brown, 1971), of which the radial is much
larger and the tangential is customarily ignored in practice. The radial
distortion
Photogrammetric restitution. By applying spatial
similarity transformations (Kraus, 2007), one can calculate the unknown
coordinates (
We took (February 2013) the photographs of Johnsons Glacier calving front
using a non-metric DSLR (Digital Single-Lens Reflex) camera Nikon D60. This
is a typical, inexpensive 10 MP digital camera without excessive loss of
accuracy. Obviously, its use for photogrammetric purposes requires a
photogrammetric calibration, aimed to determine with sufficient accuracy the
internal geometry (internal orientation) of the camera. This calibration
process involves the use of Eq. (1), to calculate the
Location of the various kinds of control points. Red triangles: location of the bases, measured using GNSS techniques, where theodolite was positioned (B1000, B2000, B3000). White circles: control points measured using GNSS techniques (P100, P200, P300, P400, P500, P600). Magenta points: other control points, measured using direct intersection method (I10, I20, I30, I40, I50, I60 points). Line 1 and Line 2 represent the zodiac boat tracks from which the photographs were taken. The scale of the figure does not match with the numerical map scale indicated. The latter corresponds to the full-size printed version of the figure. This comment also applies to Figs. 8–12.
Photogrammetric survey The fieldwork took place in February 2013, under local conditions of few
clouds and high visibility (more than 500 m). Several control points were
established making a network with permanent base stations near Johnsons
Glacier front (Fig. 3, B1000, B2000 and B3000). These bases were measured
using GNSS techniques, in particular, a Trimble 5700 GPS with horizontal and
vertical accuracies of Photograph shooting As mentioned, the camera used is a Nikon D60 DSLR 10 MP camera, with lens
55–200 mm AF-S DX. In our case, the only possibility for taking pictures
approximately perpendicular to the glacier calving front (normal
photogrammetry) is using a boat. We used a zodiac boat, taking photos from
lines 1 and 2 in Fig. 3. Line 1 is at approximately 400 m distance from
the glacier front, and we used a focal length of 95 mm and set the focus to
infinity. We used the same focal length and focus to infinity for the photos
taken from Line 2, located at an approximate distance of 700 m from the
glacier front. The overlap was higher than 95 %. To increase the number of
control points over the glacier, we took more pictures from a location near
P500. These photographs were taken with a focal length of 130 mm and focus
to infinity. In this case we mounted the camera on a tripod over the ground,
and took the photos using convergent photogrammetry. We thus obtained
coordinates for additional control points (about 20 new points). In these
photographs we could also observe three stakes from the net of velocity and
mass balance measurements, which are customarily measured using GNSS
techniques (Navarro et al., 2013). Camera calibration Calibration is made both before and after the photographic fieldwork, using
the same settings. We chose a building named “Mirador” for calibration,
which is located at Princesa de Eboli street in Madrid, which has an ideal
configuration for this project (Fig. 4). This building has a large open
space (square) at its front, allowing shot distances similar to those used
during fieldwork at Johnsons Glacier (400 m). This distance also makes it
easy to measure the corners of the windows to be used as calibration grid,
using a total station (theodolite plus laser distance meter). For
establishing the corners of the reticle, windows situated in a lower
horizontal line, an upper horizontal line, two central horizontal lines and
three vertical lines, defining the dot pattern shown in Fig. 4, were
chosen. Measuring the calibration points using total station was motivated
by the fact that we had no access to the architectural plans of the
building. On the other hand, there is no guarantee that the real elements of
the building coincide with those in the plans. We had to determine the
coordinates of the calibration points using angular measurements because the
total station did not allow measuring distances exceeding 100 m. We applied
the direct intersection method (Domínguez, 1993). Observations were
made from two stations. One of them had arbitrary local coordinates, and
provided those for the other station and for all calibration points. The
coordinates of the second station were obtained with an error of 12 mm. In
this process we used Leica Geo Office software to obtain the coordinates of
all calibration points with a root mean square (RMS) error of 53 mm,
eliminating the points with residuals exceeding this quantity. We did the
camera calibration for three focal lengths (85, 95 and 130 mm). We used the
coordinates of the calibration points to obtain the Image preparation One of the problems of fieldwork in extreme environments is that they often
do not allow the repetition of field observations. Consequently, it is
extremely important to collect as much data as possible during the
observations (in our case, taking a large number of photographs), so that
errors which are detected back in the office can be corrected or, at least,
minimized. To minimize errors, we did a selection of photographic peer
models with an overlap of 80 %. We also corrected distortion of all
photographs using internal orientation parameters and generated a new set of
distortion-corrected images (Fig. 5), which can be used by any
photogrammetric software. Calculation of ground coordinates Taking the radial distortion-free photographs as a starting point, we
applied collinearity conditions using the known control points to obtain the
parameters of different transformations (Helmert 3D transformations), with
which the ground coordinates of any point on the photographs can be
retrieved. We used a total of 10 photos to make 9 models, with estimated
errors Johnsons Glacier front has two main lines requiring photogrammetric
restitution. The upper one is the top of the cliff of the calving front and
the lower one is the waterline on the calving front. The calving front, as
usual, is heavily crevassed (Fig. 7b), which facilitates the use of
automatic correlation. An orthophoto corresponding to a total of 180 000
points produced by automatic correlation is shown in Fig. 7a, and the
restitution of the upper and lower lines of the glacier front, as well as
the main crevasses and/or fractures, are shown in Fig. 7b.
View of the facade of the building “Mirador”, where we did the calibration of the non-metric camera used for photographic shots of Johnsons Glacier front. In yellow, the points measured using a total station.
Coefficients of radial distortion and position for PPS. The two last columns show the RMS error and the maximum error for radial distortion respectively. We considered 102 points for the polynomial adjustment. The first column shows the values for the calibrated focal lengths (mm).
A snapshot of the main screen of our in-house developed software for photogrammetric restitution. This software does not require artificial stereoscopic vision for the restitution. By clicking on an item in the frame to the left, it locates the corresponding point in the right frame using automatic correlation.
In addition to the positions of the land-terminating fronts of Sally Rocks, Las Palmas and Argentina lobes of Hurd Glacier determined from aerial photos (BAS flight of 1957 and UKHO flight of 1990) or satellite images (Quickbird images of 2007 and 2010), researchers from GSNCI have measured the positions of these fronts several times during the period 2000–2012. The measurements were done using GNSS techniques (with a Trimble 5700 system, with Data Controller TSC2), with estimated horizontal accuracy between 0.07 and 0.60 m, depending on the campaign. The measurements were done either in real-time kinematics or in fast static (post-processed) mode. In all cases the GNSS base station was located at the neighboring Juan Carlos I station.
ARCGIS shape files for the BAS photogrammetric flight of 1957. In the third column the RMS error is shown.
ARCGIS shape files for the UKHO photogrammetric flight of 1990. In the third column the RMS error is shown.
ARCGIS shape file corresponding to the satellite image of QUICKBIRD system program (2010). The image is corrected to sea level to obtain the correct planimetric position of Johnsons Glacier calving front. In the third column the RMS error is shown.
Here we further discuss the compilation and processing of the data from the various sources, with an emphasis on those not discussed so far (aerial photos and satellite images), and we also summarize the errors for each one.
The first set of data (DOCU 1) corresponds to the British photogrammetric flight of 26 December 1957, performed at a flight altitude of 13 500 feet, using a metric camera IX Eagle Mk I and a nominal focal length of 153.19 mm. Once restored using Digi3D software, it allowed us to obtain the position of the different glacier fronts, including Johnsons Glacier (Table 2). The estimated horizontal accuracies range within 0.60–1.0 m. The earlier work by Molina et al. (2007) also used these same photographs for 3-D restitution, but, as mentioned in the Introduction, they used paper-printed photos (implying additional distortion) while we used digitized versions from the original films. Moreover, in the current study we use the images only to get the planimetry at sea level, so we reach a better accuracy. Using the certificate of calibration for IX Eagle Mk I, we have rectified the photos and then georeferenced the photograms X26FID0052160 and X26FID0052131 using ARCGIS software with an 8-parameter transformation.
The second set of data (DOCU 2) corresponds to another British photogrammetric flight, done in January 1990. In this case, a helicopter based on the ship HMS Endurance was used as platform, at a flight altitude of 10 000 feet, using a metric camera RMK A 15/23 with a nominal focal length of 153 mm. Once restored the data using the same Digi3D software as before, we obtained the position of the different glacier fronts, including Johnsons Glacier (Table 3). The estimated horizontal accuracy is 2.0 m.
DOCU 3 corresponds to the photogrammetric survey of Johnsons Glacier front made in 1999 by researchers from the University of Barcelona using a metric camera. However, this has not been considered in the current study because they only determined the top of the calving front cliff and not the position of the waterline, which is the line of interest to us (it is what we are comparing for the various images of Johnsons Glacier calving front).
DOCU 4 corresponds to an image captured in January 2010 by the Quickbird
image system, which covers the entire work area. It is a raster file format
GEOTIFF UTM 20S on the ellipsoid WGS84. Its original name was
10JAN29132854-P2AS052832138010_01_P001.TIF and
it was obtained from
DOCU 5 is another Quickbird system image, taken in February 2007, which also
covers the entire work area (with the exception of an insignificant rock
outcrop to the northwest). It is a raster file format GEOTIFF UTM 20S on the
ellipsoid WGS84. Its original name was
07FEB03135449-M2AS-052422572010_01_P001.TIF
and it was obtained from
DOCU 6 corresponds to the surface-based GNSS measurements of the land-terminating fronts of Hurd Glacier discussed in Sect. 2.3 (Blewitt et al., 1997). The data compilation, with their estimated accuracies (ranging between 0.07 and 0.60 m), is shown in Table 6.
Finally, DOCU 7 corresponds to the photogrammetric survey of Johnsons Glacier front done by the authors in February 2013 with a non-metric camera, discussed in Sect. 2.2. The corresponding metadata are given in Table 7. The estimated horizontal accuracy is 0.70 m.
ARCGIS shape file corresponding to the satellite image of QUICKBIRD system program (2007). The image is corrected to sea level to obtain the correct planimetric position of the glacier fronts. In the third column the RMS error is shown.
ARCGIS shape files corresponding to the GSNCI data inventory for Argentina, Las Palmas and Sally Rocks fronts of Hurd Glacier, determined using GNSS techniques. In the third column the RMS error is shown.
ARCGIS shape file corresponding to the photogrammetric restitution of the Johnsons Glacier calving front in February 2013. The photo was obtained with a non-metric camera. In the third column the RMS error is shown.
The position of Johnsons Glacier calving front (the waterline position) at the beginning and end of 1957 is shown in Fig. 8, while its position for various years within the period 1957–2013 is shown in Fig. 9. As Johnsons is a tidewater glacier, its front position changes, though influenced by climate, are mostly driven by its internal dynamics, including feedback mechanisms involving the balance of forces, the flow and calving. We observe that the calving front advanced 74 m in its central part (segment A) between 1957 and 1990. Then the glacier front retreated 171 m (sum of the segments A, L and F) between 1990 and 2007, to remain stable until 2010 and then re-advance by 31 m in its central part (segment L) between 2010 and 2013. In spite of this latter re-advance of the central part, the southern part of the front retreated ca. 57 m (segment J) during the same period 2010–2013. Also note that, in the northern part, the glacier front has retreated over 97 m between 1957 and 2013 (E segment). The position of the front in the neighbourhood of segment E was estimated using the photo X26FID0093015 obtained at the beginning of 1957 (Rodríguez, 2014) from an incomplete BAS photogrammetric flight, shown in Fig. 8a.
The darker area seen on Fig. 8, in the southwestern area of the terminal part of the glacier, corresponds to an accumulation, due to intense folding and faulting in this part of Johnsons Glacier, of volcanic ash from tephra layers stemming from the volcanic eruptions at the neighboring Deception Island (Ximenis et al., 2000).
Calving front position changes of Johnsons Glacier. The oldest front position available corresponds to 1957 (in light blue) and the most recent to 2013 (in red). The base map corresponds to the BAS aerial photograph taken on 19 January 1957. The scale of the figure does not match with the numerical map scale indicated.
In this case, the front position changes are shown in Fig. 10. Hurd is a
land-terminating glacier, which implies that the front position changes are
mainly driven by climate-induced variations (e.g., increased melting),
coupled with the (slow) dynamic response to the climate variations (due,
e.g.,
to geometry changes associated to accumulation-ablation changes or variation
in basal lubrication because of changes in the amount of meltwater reaching
the glacier bed). We observe that the glacier front glacier retreated by 116 m
in its central area (segment A) between 1957 and 1990. This was followed
by a further retreat by 60 m (segment B) between 1990 and 2000, and yet
further retreats by 47 m (C segment) during 2000–2006, and by 36 m (D
segment) during 2006–2009. From 2009, the front position has remained stable
until 2012. These changes are consistent with the overall warming trend in
the South Shetland Islands, where the decade 1995–2006 has been the warmest
since the 1960s, while the decade 2006–2015 has been colder than the
previous one (by 0.5
Terminus position changes of Sally Rocks front of Hurd Glacier. The oldest front position available corresponds to 1957 (in purple) and the latest to 2012 (in cyan). The base map corresponds to the BAS aerial photograph of 1957. The scale bar shown in the figure does not match with the map scale indicated.
The black spots on the glacier surface seen in Figure 10 are mostly ash from tephra layers, but these are often mixed with subglacial sediments taken to the glacier surface through fractures associated to thrust faults in the glacier snout resulting from the compressional regime. The compressional regime is mainly a consequence of fact that the glacier snout is frozen to bed (Molina et al., 2007; Navarro et al., 2009; Molina, 2014). This mixture of volcanic ash and subglacial sediments often accumulates at the surface in the form of pinnacles. According to Molina (2014, p. 126–127), most tephra shown in Fig. 10 likely correspond to eruptions well before 1829 (what is denoted as “oldest layers” in Ximenis et al., 2000, for Johnsons Glacier). We say likely because Molina did not analyze 1957 photos but images taken during 2004–2009, and did the geochemical analysis (X-ray fluorescence) with samples taken within this period. In any case, as Molina (2014) has pointed out, the tephra deposits in these terminal zones should not be used for purposes of dating structures, because these spots in the terminal part of Hurd Glacier fronts are the result of accumulation of tephra flushed out, upon melting, from tephra layers corresponding to various eruptions. Consequently, the ash layers in this zone cannot be used to reliably infer past mass balance rates.
The front position changes for Las Palmas Glacier are shown in Fig. 11. We can see that the glacier front retreated 11 m in its central part (segment A) between 1957 and 1990. Then, the front retreated 24 m (segment F) between 1990 and 2000. This was followed by a further retreat by 17 m (segment B) during 2000–2005, another 14 m (D segment) during 2005–2007, and finally another 10 m during 2007–2009, to remain stable thereafter. For Las Palmas front, the retreat during the period 1957–1990 was significantly lower than that for Sally Rocks front. This is likely due to the fact that, in 1957, Las Palmas front ended at sea, and thus had a comparatively larger thickness at its terminal zone as compared with Sally Rocks front.
Terminus position changes of the front of Las Palmas side lobe of Hurd Glacier. The oldest front position available corresponds to 1957 (in green) and the latest to 2012 (blue points). The base map corresponds to the BAS aerial photograph of 1957. The scale bar shown in the figure does not match with the map scale indicated.
In this case, shown in Fig. 12, the glacier front seems to have advanced
by a small amount, ca. 5 m, in its central area (segment A) between 1957 and
1990. Then the front retreated by 70 m (segment A
Terminus position changes of the front of Argentina side lobe of Hurd Glacier. The oldest front position available corresponds to 1957 (in purple) and the latest for the year 2012 (in cyan). The base map corresponds to the BAS aerial photograph of 1957. The outline marked on the “diffuse” 2007 Quickbird image corresponds to ground-based GNSS measurements. The scale bar shown in the figure does not match with the map scale indicated.
The following conclusions may be drawn from our study:
Close-range photogrammetry with non-metric cameras is very suitable for
determining the position of glacier fronts and, in particular, the calving
fronts of tidewater glaciers, for which other techniques, such as
surface-based GNSS measurements, are not possible. Moreover, it compares
favorably in terms of costs with other methods such as photogrammetry with
metric cameras or laser scanner systems. For the measurement of terminus
position of land-terminating glaciers, surface-based GNSS techniques are
fast and reliable. These surface-based techniques are effectively
complemented, whenever available, with remote-sensing images, from either
aerial photogrammetric flights or satellite optical imagery. The calving front of Johnsons Glacier has shown, during the period
1957–2013, several episodes of advance and retreat (and likely many more
happened during long intermediate periods lacking observations). This is a
tidewater glacier, so its front position changes, though can be influenced
by climate, are mostly driven by the internal glacier dynamics, including
various feedback mechanisms. The land-terminating fronts of Hurd Glacier (Sally Rocks, Las Palmas and
Argentina) have experienced during the period 1957–2009 an overall
retreating trend, remaining nearly stationary during 2009–2012. Being
land-terminating, the front position changes are mostly driven by
climate-related processes such as accumulation and melting variations, and
their associated glacier dynamic response. The observed front variations are
consistent with the regional climate changes. In particular, with an
extremely warm 1995–2006 decade, followed by a colder 2006–2015 decade.
This work was supported by grant CTM2014-56473-R from the Spanish National Plan of R&D.Edited by: D. Carlson Reviewed by: two anonymous referees