1. Introduction
Ice penetrating radar sounding is one of the most powerful and widely used geophysical techniques in glaciology (Dowdeswell and Evans, Reference Dowdeswell and Evans2004). Analysis of geometric information in radar sounding data has been used, for example, to analyze subglacial topography, ice-flow history and ice-shelf melt (Bingham and Siegert, Reference Bingham and Siegert2007). Radiometric analysis of the same data has also enabled observations of subglacial and englacial hydrology, ice-sheet temperature, and basal roughness among a wide variety of other conditions and processes that are critical for understanding and projecting future sea levels (Scambos and others, Reference Scambos2017; Schroeder and others, Reference Schroeder2020). Some of the earliest ice penetrating radar data were collected in the 1960s and 70s from airborne surveys and recorded on optical film (Bailey and others, Reference Bailey, Evans and Robin1964; Evans and Robin, Reference Evans and Robin1966; Crabtree and Doake, Reference Crabtree and Doake1986; Gudmandsen, Reference Gudmandsen1975). Recent work digitizing and analyzing archival radar film has demonstrated the potential to compare ice-sheet subsurface geometry in modern and archival radar sounding data over multiple decades (Schroeder and others, Reference Schroeder2019) However, due to a lack of radiometric calibration, these studies have been limited to visual interpretations of features such as subglacial lakes (Siegert and Ridley, Reference Siegert and Ridley1998), basal channels (Le Brocq and others, Reference Le Brocq2013), and ice shelf geometry (Khazendar and others, Reference Khazendar2016).
The Scott Polar Research Institute/National Science Foundation/Technical University of Denmark (SPRI/NSF/TUD) airborne radar sounding surveys of Antarctica in the 1960s and 70s (Drewry and Jordan, Reference Drewry and Jordan1980; Turchetti and others, Reference Turchetti, Dean, Naylor and Siegert2008) and the contemporaneous surveys of Greenland (Gudmandsen, Reference Gudmandsen1975; Kuiven, Reference Kuiven1975) recorded radar sounding data onto 35 mm optical film (Fig. 1). These profiles were the basis of the first subglacial bed maps (Drewry and Jordan, Reference Drewry and Jordan1980; Lythe and Vaughan, Reference Lythe and Vaughan2001) and early geological and glaciological interpretations (Dowdeswell and Evans, Reference Dowdeswell and Evans2004). However, in order to fit those profiles onto the finite dynamic range of optical film while preserving the visibility and interpretability of englacial layers and the bed, along-flight depth profiles called Z-scopes were processed using log-detection, fast-time differentiation and compressive limiting (Fig. 2). By contrast, these surveys also recorded A-scope records. A-scope signals, which passed through the same log-detection receivers as the Z-scope signals, but not the differentiator or limiter.
Fig. 1. Scans of 35 mm archival radar film include (a) dual (and single) frequency A-scope traces and (b) Z-scope profiles. The red vertical lines in (b) show the CBD counter interval where A-scope traces are recorded and (c) show an A-scope trace rotated 90 degrees and scaled (Gorman and Siegert, Reference Gorman and Siegert1999, after) to align with Z-scope traces. Intermittent ‘cal pips’ in both the (d) Z-scope and (e) A-scope records provide 2 microsecond fast-time calibration. (f) Bed (or other) echo strengths can be calibrated within the dynamic range (A scale) of A-scopes by measuring the vertical number of pixels between the saturated ‘main bang’ and the noise floor. (g) Z-scope signal strengths can be measured by as the difference between the opposite-signed peaks produced by differentiation within the receiver (see section 2.3). The film segments in this figure were selected to illustrate the types of film records and are not from a single or particular location.
Fig. 2. Flow chart of the radiometrically significant steps in producing the A-scope and Z-scope signals recorded on the 35 mm optical film (Christensen, Reference Christensen1970; Christensen and others, Reference Christensen, Gundestrup, Nilsson and Gudmandsen1970; Skou and Sondergaard, Reference Skou and Sondergaard1976).
A-scope traces were recorded every 15 or 20 s (depending on the season) at each increment of the internal Coded Binary Decimal (CBD) counter corresponding to about every 1.5 or 2 km along-track. The CBD counter was used to cross-register the radar and positioning data and was recorded on the film along with the date, time, flight number and mode (Christensen, Reference Christensen1970; Christensen and others, Reference Christensen, Gundestrup, Nilsson and Gudmandsen1970; Skou and Sondergaard, Reference Skou and Sondergaard1976). In order to enable the radiometric comparison of subsurface conditions (such as englacial water storage, hydrology, attenuation and reflectivity) (Christianson and others, Reference Christianson2016; Chu and others, Reference Chu2016; Schroeder and others, Reference Schroeder, Seroussi, Chu and Young2016; Kendrick and others, Reference Kendrick2018) on similar timescales, in this paper, we develop and demonstrate an approach for the extraction of bed echo power in units of equivalent A-scope signal to noise ratio (SNR) and the production of relative reflectivity profiles from Z-scope records.
Z-scope records are also registered to the date, the time, mode, flight number, and CBD. Although these counters allow the film records to be geolocated, assuming positioning data are available, neither the Z-scope nor A-scope records have absolute power scales. The A-scope records are power traces. As such, utilizing their radiometric information is only a matter of calibrating their power scale. In contrast, the calibration of Z-scopes requires understanding and correcting for the conversion from A-scope signal to fast-time derivatives as well as the compression imposed by the differentiator, limiter and film itself.
There were early pioneering studies with A-scope records which utilized their radiometric information. These include single-trace analysis of subglacial lakes (Gorman and Siegert, Reference Gorman and Siegert1999), spatial analysis of echoes across the Ross Ice Shelf (Neal, Reference Neal1982) and studies of internal layering and ice fabric (Siegert and Fujita, Reference Siegert and Fujita2001; Siegert and Kwok, Reference Siegert and Kwok2000). Performing this type of analysis with modern scans of the film data requires: (1) calibrating the vertical scale of A-scopes (in microseconds), (2) identifying bed echoes and surface echoes in A-scope traces (using Z-scopes to aid in their identification), and (3) calibrating the power scale of A-scopes (in dB). Expanding this type of analysis to the more spatially continuous Z-scope records requires the additional steps of: (4) determining the relationship of A-scope signals (in dB) to Z-scope signals including the effects of Z-scope compression. In this paper, we develop and present a procedure for performing these calibration steps.
Our radiometric calibration of both A-scope and Z-scope records depends on the parameter A scale which describes the dynamic range of the A-scope SNR (in dB) and which we estimate using bench-top tests published in Christensen (Reference Christensen1970); Christensen and others (Reference Christensen, Gundestrup, Nilsson and Gudmandsen1970); Skou and Sondergaard (Reference Skou and Sondergaard1976) (see section 2.2). A scale is controlled by a portion of the SPRI/NSF/TUD receiver that is not user-adjustable (Christensen, Reference Christensen1970; Christensen and others, Reference Christensen, Gundestrup, Nilsson and Gudmandsen1970; Skou and Sondergaard, Reference Skou and Sondergaard1976) and is therefore expected to remain stable within and across field-seasons.
Our radiometric Z-scope calibration includes the additional empirical fitting of calibrated A-scope and uncalibrated Z-scope bed echoes (see section 2.4). This fitting is done using a logistic function which approximates the user-adjustable Z-scope receiver response. As a result, this calibration can change between, and within, survey lines. Therefore, investigators need to visually inspect Z-scope segments for signs of setting changes (e.g. step-changes in the saturation level or contrast of the Z-scope radargrams) and/or perform sub-segment or cross-segment tests to confirm that those parameters are stable. While this can be a labor-intensive (and at times subjective) sanity-checking process, the fact that these parameters change in discrete increments makes it tractable. Most significantly, our paper shows that within (and between) Z-scope profiles where the parameters describing receiver saturation are stable, the radiometric signature of relative reflectivity can be estimated and interpreted from Z-scope records.
2. Methods
2.1 Geometric calibration
Although Z-scopes are more challenging than A-scopes to calibrate radiometrically, they are better suited for initial visual interpretation and identification of bed and surface interfaces. Z-scope traces can be associated with the individual A-scope measurements using the CBD counter. The vertical red lines in Figure 1b show the locations and spacing of CBD markers where A-scope traces are recorded along a continuous Z-scope profile.
In terms of fast-time calibration, both record types include calibration pips or ‘cal pips’ recorded every 2 microseconds. In Z-scope records, they appear as horizontal lines along a vertical ruler (Fig. 1d), and in A-scope records, they appear as vertical lines superimposed on the A-scope trace itself (Fig. 1e). The presence of these marks on both types of records enables the cross-calibration of the vertical scale of a Z-scope to the horizontal time axis of A-scopes and both to physical units of time (in microseconds).
Picking the Z-scope bed and surface follows the common visual interpretation process used in modern radar sounding data and does not require radiometric calibration to be performed. There are automatic picking algorithms for radar sounding data (Panton and Karlsson, Reference Panton and Karlsson2015, e.g.) but the most common approach is semi-automated picking in which a human interpreter picks two bounds, one above and one below the bed or surface interface and then an automatic re-picker selects the highest power echo between those bounds. However, because Z-scopes are produced by taking the fast-time derivative of A-scope echo power, a radar echo will show up as two sequential opposite-signed peaks in the Z-scope (Fig. 1g). As a result, the automated portion of picking Z-scope data involves selecting the largest Z-scope signal (or positive-to-negative difference) rather than the largest peak. This includes first measuring the number of pixels between positive and negative peaks in a particular scanned Z-scope image (3, e.g., in Fig. 1g) and then identifying the largest difference (within the manual picking bounds) across that number of vertical pixels.
Once Z-scope picks have been recorded, the delay between the bed and surface can be used to identify bed echoes in A-scope traces. In order to do so, the surface return must first be identified in the A-scope record. The left side of an A-scope record (e.g. Fig. 1f) corresponds to the shortest fast-time delay after transmission and the first strong signal is usually the ‘main bang’ caused by leakage between the transmitter and receiver (or another artifact of the transmission event itself). This is discernible from the surface return because it has a fixed delay and strength across many observations, as is evident in both A-scope and Z-scope records. The surface return is usually followed by a decay in echo power until the bed echo. The bed echo is usually followed by the relatively constant noise floor power level. Once the surface return is identified, the delay at each CBD can be used to identify the bed echo in the A-scope traces.
2.2 A-scope signal, saturation and calibration
Although A-scopes directly record log-detected received power, they do not contain a radiometric calibration reference. Fortunately, the dynamic range of A-scope records can be used to radiometrically calibrate echoes within A-scope traces. Figure 3 is the result of bench-top loop-back testing of the receiver with an in-line variable attenuator (Christensen, Reference Christensen1970; Christensen and others, Reference Christensen, Gundestrup, Nilsson and Gudmandsen1970; Skou and Sondergaard, Reference Skou and Sondergaard1976) and shows a typical amplifier response starting with a noise floor at low input levels, increasing linearly, and saturating at high input levels (Pozar, Reference Pozar2011). We adapted Figure 3 from Christensen (Reference Christensen1970); Christensen and others (Reference Christensen, Gundestrup, Nilsson and Gudmandsen1970); Skou and Sondergaard (Reference Skou and Sondergaard1976) by assuming that the highest A-scope values measured in the test were at saturation and that the lowest values were at the noise floor. As a result, the vertical pixel range in A-scopes (e.g. Fig. 1f) can be converted into A-scope SNR values by picking the noise floor and setting the SNR value of that vertical pixel to be 0 dB, picking the ‘main bang’ and setting the SNR value of that vertical pixel to A scale = 70 dB (Fig. 3) (Christensen, Reference Christensen1970; Christensen and others, Reference Christensen, Gundestrup, Nilsson and Gudmandsen1970; Skou and Sondergaard, Reference Skou and Sondergaard1976), and linearly interpolating (in dB) between these values.
Fig. 3. (a) 60 MHz and (b) 300 MHz A-scope saturation curves from benchtop loop-back tests with variable attenuation (Christensen, Reference Christensen1970; Christensen and others, Reference Christensen, Gundestrup, Nilsson and Gudmandsen1970; Skou and Sondergaard, Reference Skou and Sondergaard1976).
With the y-axis of A-scopes calibrated in term of A-scope SNR (in dB), A-scope records can be used to perform the same kind of relative, along-profile echo-power, attenuation and reflectivity analysis common in modern digital radar sounding data (Schroeder and others, Reference Schroeder2020). However, A-scope traces are recorded only every 1.5 or 2 km and can therefore only be used to observe reflectivity patterns that would not be aliased by such sparse spatial sampling. Given the significant variability between individual radar traces on scales of 2 km and below, A-scope reflectivity patterns should be interpreted either at the scale of profiles, catchments or ice shelves (Neal, Reference Neal1982, e.g.) or, with great care, as single traces (Gorman and Siegert, Reference Gorman and Siegert1999, e.g.). However, despite these caveats, the pioneering A-scope based analysis of relative reflectivity and echo character of the Ross Ice Shelf by Neal (Reference Neal1982) is just now being matched by modern data (Tinto and others, Reference Tinto2019).
2.3 Z-scope signal
Z-scope records are produced by differentiating log-detected A-scope signals in fast-time resulting in sequential opposite-signed peaks for a single echo (Fig. 1g). The vast majority of the SPRI/NSF/TUD survey (and surveys of Greenland from the same period) were collected in a short-pulse (rather than chirped) mode so that the pulse-length (rather than the waveform) determined the system bandwidth and range resolution (Christensen, Reference Christensen1970; Christensen and others, Reference Christensen, Gundestrup, Nilsson and Gudmandsen1970; Kuiven, Reference Kuiven1975). In this configuration, Z-scope signals can be treated as the fast-time derivatives of a square-wave with a ‘height’ equal to the A-scope SNR (in log-detected A-scope units of dB) resulting in Z-scope signals with units of dB/ns. To illustrate this, we simulate Z-scope signal strength as a function of A-scope SNR using the A-scope geometry in Figure 4a and calculate the Z-scope signal (Fig. 4b) as its fast-time derivative. Figure 4c illustrates the 1 to 1 mapping of A-scope SNR to Z-scope signal before compression. This intuitive relationship is the result of the fast-time differentiation calculating a slope for which the A-scope SNR is the vertical component.
Fig. 4. (a) Simulated A-scope echo with red arrows showing the A-scope SNR (in dB) and the fast-time differentiation window (in ns) used to calculate the (b) simulated Z-scope signal (in dB/ns). (c) These simulations show that, without compression effects from the actual differential amplifier, limiter, and film, Z-scope signals vary linearly with A-scope SNR. The vertical scale in (c) is normalized to the range of SNR values used in the simulation.
2.4 Z-scope saturation and calibration
Z-scope records also pass through a differential amplifier, limiter and are recorded on film, the net effect of which is to compress the Z-scope signal (Christensen, Reference Christensen1970; Christensen and others, Reference Christensen, Gundestrup, Nilsson and Gudmandsen1970; Kuiven, Reference Kuiven1975). In order to calibrate that compression, we utilize a segment of radar sounding data collected in the Gamburtsev Mountain region of interior East Antarctica, which is an ideal area for calibration because the significant relief in ice thickness produces a wide range of bed echo strengths across the dynamic range of both the A-scope and Z-scope records. To empirically constrain and correct for Z-scope compression, we plot the calibrated A-scope SNR (as described in section 2.2) and Z-scope signal for bed echoes along this profile as a scatter plot (Fig. 5b) and approximate the receiver response (Pozar, Reference Pozar2011) including compression due to the differential amplifier and limiter (see Supplementary Materials) as a logistic function (Fig. 5a) given by
where Z is the Z-scope signal strength, A SNR is the calibrated A-scope SNR, A is the max value, B is the growth rate, and C is the sigmoid mid point. Fitting the curve in Figure 5b yields values of A = 0.378, B = −0.212 and C = −7.78 for Equation (1). Since A-scope traces are acquired much more sparsely than Z-scope signals (Fig. 1b), Z-scope signal strengths are compared to interpolated A-scope SNR values (as shown in Fig. 6). As a result, the spread in Figure 5b is due to the along-track variability in Z-scope bed echo power (due to noise, speckle and clutter as well as changing bed reflectivity and roughness) and the coarse spatial sampling of A-scope traces rather than the receiver compression itself (which is a smooth, deterministic function with unknown parameter values). As a result, the impact of the scatter in Figure 5b is increased uncertainty in the estimations of the parameters (A, B and C) in the saturation function rather than the introduction of a stochastic character to the receiver itself. With A, B and C estimated, Equation (1) can be used to convert the more spatially continuous Z-scope records into equivalent A-scope SNR values.
Fig. 5. (a) Generic response of a receiver (adapted from Pozar, Reference Pozar2011), showing the output power (P out) as a function of input power (P in) including the noise floor, linear and saturation regions of that response. The dotted lines illustrate the one realization of a logistic function, which we use to approximate that response. (b) Observed Z-scope signal strength as a function of A-scope SNR for the profile in Figure 6. The black line shows the logistic curve used to estimate the effect of compression for the Z-scope signal along this profile.
Fig. 6. (a) A Z-scope profile from the Gamburtsev Mountain region of East Antarctica (f) along with (b) extracted ice thickness, (c) Z-scope signal, (d) calibrated A-scope SNR (in dB) and (e) calculated Z-scope relative reflectivity profiles. The red lines correspond to the span of the zoomed-in Z-scope of a single subglacial lake (g) and its corresponding reflectivity signature (h).
3. Results
3.1 Z-scope reflectivity signatures of subglacial lakes
Even though the deep interiors of sheets are relatively stable and unlikely to exhibit large radiometric changes between radar film records and modern data, they are compelling regions for initial demonstration of Z-scope signal extraction, calibration and analysis. As mentioned above, they exhibit significant ice-thickness relief, which enables the empirical calibration of and correction for attenuation (e.g. Schroeder and others, Reference Schroeder, Seroussi, Chu and Young2016) and host many stable subglacial ‘radar lakes’ (MacKie and others, Reference MacKie, Schroeder, Caers, Siegfried and Scheidt2020). These lakes are among the brightest, earliest and least ambiguous subglacial radiometric signatures in radar sounding data (Oswald and Robin, Reference Oswald and Robin1973; Siegert and others, Reference Siegert, Carter, Tabacco, Popov and Blankenship2005; Wolovick and others, Reference Wolovick, Bell, Creyts and Frearson2013; Rutishauser and others, Reference Rutishauser2018).
For this analysis, we use the same flight segment (Fig. 6) from the Gamburtsev Mountains region described above. Visually, the Z-scope radargram in Figure 6a shows the characteristically flat, bright signatures of subglacial lakes (Carter and others, Reference Carter2007) in the mountain valleys (Wolovick and others, Reference Wolovick, Bell, Creyts and Frearson2013, consistent with). We use Equation (1) and the estimated saturation parameters for this profile to convert the Z-scope signal (Fig. 6c) to its equivalent A-scope SNR values and used these values, along with thickness (Fig. 6b) to calculate the attenuation rate. We follow common radioglaciological practice (Jacobel and others, Reference Jacobel, Welch, Osterhouse, Pettersson and MacGregor2009; Schroeder and others, Reference Schroeder, Seroussi, Chu and Young2016) and calculate the attenuation rate (in dB/km) as a slope fit to bed power vsthickness. For the profiles in Figure 6, this leads to an estimated one-way depth-averaged attenuation rate 〈N〉 = 4.7 dB/km, which is consistent with modeled values for the region (Matsuoka and others, Reference Matsuoka, MacGregor and Pattyn2012). Using this attenuation rate, we correct for the total attenuation and calculate the relative (mean adjusted) along-track reflectivity (Fig. 6e). This results in a (99 percentile) range of reflectivity values of 28.9 dB, which is compatible with those theoretically and empirically observed in modern sounding data (Christianson and others, Reference Christianson2016). The highest reflectivity values in Figure 6 and the portions of the radargrams to which they correspond are consistent with the canonical signatures of subglacial ‘radar lakes’ (Siegert and Ridley, Reference Siegert and Ridley1998; Carter and others, Reference Carter2007; Wolovick and others, Reference Wolovick, Bell, Creyts and Frearson2013; MacKie and others, Reference MacKie, Schroeder, Caers, Siegfried and Scheidt2020)
In order to explore the effect of uncertainty in A, B and C on the relative reflectivity signal (e.g. Fig. 6e), we introduce a range of errors (from 0 to 100%) to each of the parameters and calculate the resulting relative reflectivity. We then plot the sensitivity of the calculated relative reflectivity to errors in each of the fitting parameters (Fig. 7) in terms of the ‘reflectivity sensitivity’ or the root mean squared (RMS) difference between the perturbed and unperturbed relative reflectivity profiles. The modest values in Figure 7 indicate that the relative reflectivity signal (Fig. 6e) can be larger than the uncertainty introduced by even significant (e.g. 100%) errors in the estimated logistic parameters with larger reflectivity errors occurring at higher Z-scope values due to shape of the logistic function itself (Fig. 5b).
The variety of system configurations, surveys, and targets as well as film processing, preservation and digitization across corpus of archival radar film (Gudmandsen, Reference Gudmandsen1975; Drewry and Jordan, Reference Drewry and Jordan1980; Björnsson and Pálsson, Reference Björnsson and Pálsson2020) means that realized uncertainty in any radiometric analysis will depend on the particular features and film under consideration. However, individual researchers can evaluate that uncertainty for their analysis by comparing calibration/fitting-based uncertainties as presented above (e.g. Fig. 7) with the uncertainty/variability/noise in extracted Z-scope signals (as expressed in Fig. 6c and the vertical spread in Fig. 5). In practice, as with modern digital data (Chu and others, Reference Chu2021), this latter target/environment-based uncertainty is likely to be the dominant source of uncertainty in relative signals for most studies.
3.2 Z-scope reflectivity signatures of an ice shelf
While the ice-sheet interior is ideal for demonstrating the radiometric extraction and calibration because of significant ice-thickness relief and easily identifiable features like subglacial lakes, the interior is less likely to exhibit change on the scale of decades spanned by modern and archival data. Such multi-decadal subsurface evolution is more likely nearer to the coast where active subglacial hydrology (Chu and others, Reference Chu2016), grounding zones (Christianson and others, Reference Christianson2016) and ice shelves (Khazendar and others, Reference Khazendar2016) are undergoing relatively rapid change. Some ice shelves were observed and investigated in the original SPRI/NSF/TUD survey (Neal, Reference Neal1982; Crabtree and Doake, Reference Crabtree and Doake1986) and fortunately, there was a near repeat observation of a 1977 SPRI/NSF/TUD flight by the University of Texas Institute for Geophysics (UTIG) High Capability Airborne Radar Sounder (HiCARS) system (Young et al., 2011) in 2011 (Fig. 8a). The archival (Fig. 8b) and modern (Fig. 8d) radargrams are visually and geometrically similar, with the same general bed, ice-shelf and crevasse generated clutter geometries.
Fig. 8. (a) Location of near-repeat surveys of an East Antarctic ice shelf along the Sabrina Coast, collected by the NSF/SPRI/TUD survey in 1977 (Drewry and Jordan, Reference Drewry and Jordan1980; Schroeder and others, Reference Schroeder2020) and the UTIG HiCARS system in 2011 (Young and others, Reference Young2011). (b) Z-scope radargram and (c) Z-scope reflectivity from an NSF/SPRI/TUD survey in 1977 (Drewry and Jordan, Reference Drewry and Jordan1980; Schroeder and others, Reference Schroeder2020) and (d) radargram and (e) reflectivity from a 2011 HiCARS survey (Young and others, Reference Young2011).
We follow the same calibration approach described in 2.4 to confirm that the values for A, B, and C did not change significantly for this survey segment and used these values to calibrate Z-scope signal strength to A-scope equivalent SNR values. We again followed the same attenuation-fitting approach (Jacobel and others, Reference Jacobel, Welch, Osterhouse, Pettersson and MacGregor2009, e.g.) described in section 3.1 to calculate attenuation rates for both the Z-scope (Fig. 8b) and HiCARS (Fig. 8d) profiles. We used the mean value of 〈N〉 (11.6 dB/km), which is also consistent with modeled values (Matsuoka and others, Reference Matsuoka, MacGregor and Pattyn2012) for the region, to calculate both relative reflectivity profiles (Figs 8c,e). Notably, our results show that the pattern and strength of the archival (Fig. 8c) and modern (Fig. 8e) reflectivity signals are similar, with the floating ice shelf about 10–20 dB higher than the grounded ice (Christianson and others, Reference Christianson2016, consistent with) and rising toward the center of the shelf. Because this is not an exact reflight, some of the finer-scale differences may be due to temporal change, spatial variationor a combination of the two. However, what is clear is that the general pattern and magnitude of basal reflectivity is consistent across two systems with different signal and processing parameters (albeit with the same center frequency) and recorded on different media (film vs digital records) more than three decades apart.
4. Discussion
The results presented above provide an initial demonstration and a basic understanding of how radiometric information in A-scope and Z-scope records was captured in the NSF/SPRI/TUD archival film record. By leveraging that understanding, we were able to extract, calibrate and interpret geometric and radiometric signals between A-scope and Z-scope records and convert compressed Z-scope signals into equivalent of A-scope SNR values. By converting fast-time measurements of delay into thickness (assuming an index of refraction for ice) and using the A-scope equivalent SNR values extracted from Z-scope records, we were then able to produce interpretable relative reflectivity profiles. In the deep East Antarctic interior, these relative reflectivity values reproduced the empirical strengths, forms and characters of subglacial lakes in modern data (i.e. ~10 dB jumps in the relative reflectivity and increased power variability over visually flat basal interfaces) (Carter and others, Reference Carter2007; Wolovick and others, Reference Wolovick, Bell, Creyts and Frearson2013; Jordan and others, Reference Jordan2018). Additionally, across an East Antarctic Ice Shelf, the Z-scope based relative reflectivity matched both the general modern radiometric signature of ice shelves (Greenbaum and others, Reference Greenbaum2015; Khazendar and others, Reference Khazendar2016) and the specific pattern of a modern near-reflight of the archival film line (Drewry and Jordan, Reference Drewry and Jordan1980; Schroeder and others, Reference Schroeder2019; Young and others, Reference Young2011). Overall, we believe that these results provide a promising proof-of-concept demonstration of the radiometric information richness and interpretability of Z-scope signals and hope that this provides the motivation and basis of significant future work. In addition to greater scientific exploration and utilization of the NSF/SPRI/TUD radar film archive, we believe that there is also compelling follow-on work in the areas of survey-wide calibration, targeted reflights, and the application of these approaches to archival radar film from Greenland (Gudmandsen, Reference Gudmandsen1975; Kuiven, Reference Kuiven1975).
In terms of calibration, this paper demonstrates the possibility and process of calibrating individual sounding profiles on a segment-by-segment basis where radar system settings were documented to be stable across significant portions of flights or even across campaigns (Christensen, Reference Christensen1970; Christensen and others, Reference Christensen, Gundestrup, Nilsson and Gudmandsen1970; Skou and Sondergaard, Reference Skou and Sondergaard1976). Although it is currently necessary for investigators to check that A, B, and C are stable across the collection of profiles used in a given study, it should be possible to detect setting changes as step-functions in calibration values along flights and across surveys using a larger portion of the radar film archive. This would likely require leveraging optimization approaches. However, it has the potential to provide better-constrained and survey-wide calibration values.
This paper also provides a first demonstration of quantitative radiometric cross-interpretability between archival radar film and modern data. Quantitative analysis of direct reflights are rare, even in modern data(Chu and others, Reference Chu2016; Khazendar and others, Reference Khazendar2016; Schroeder and others, Reference Schroeder, Hilger, Paden, Young and Corr2018), but they provide a unique window into the temporal evolution of subsurface conditions. With portions of SPRI/NSF/TUD surveys collected 50 years ago (Schroeder and others, Reference Schroeder2019), contemporary reflights of these lines (especially in more rapidly changing regions) could provide a window into subsurface processes at the half-century timescale. Given the criticality of decadal-scale predictions for climate adaptation and the widespread use of a century as a planning baseline in reports like the International Panel on Climate Change (IPCC) (Solomon, Reference Solomon, Manning, Marquis and Qin200; Scambos and others, Reference Scambos2017; Kennicutt and others, Reference Kennicutt2019), the ability to observe subsurface conditions over five decades could enable deeper understanding of critical processes and inform numerical modeling used in sea-level projections.
Finally, the successful extraction and analysis of radiometric signals from Z-scope profiles stand to enable the application of modern radiometric analysis techniques to archival film data. For example, in dynamic areas of the Greenland Ice Sheet, interaction between supraglacial and englacial hydrology plays a critical, but poorly understood, role in the dynamic response of Greenland ice flow to surface melt (Moon and others, Reference Moon2014). Other work has demonstrated that the synthesis of ice-sheet modeling and radar sounding data can estimate water storage and firn aquifers thickness from airborne surveys (Chu and others, Reference Chu, Schroeder and Siegfried2018). Applying these analysis techniques to the radar sounding film archive of data collected in Greenland in the 1960s and 70s (Gudmandsen, Reference Gudmandsen1975; Kuiven, Reference Kuiven1975) could enable a multi-decadal window into the evolution of englacial hydrology across the Greenland Ice Sheet.
5. Conclusion
In this paper, we demonstrated geometric fast-time and slow-time calibration of the A-scope and Z-scope record types in archival radar sounding film and radiometrically cross-calibrate Z-scope signal strength with A-scope SNR values. These calibrations can be performed on any of the archival NSF/SPRI/TUD film for which both A-scope and Z-scope records are available. We demonstrated the successful extraction, calibration and calculation of relative reflectivity values for a profile in the Gamburtsev Mountain region of East Antarctic, producing canonical radiometric signatures of subglacial lakes. In the same manner, we also demonstrated the successful extraction and calculation of a Z-scope reflectivity profile across an East Antarctic ice shelf with similar character and strength to a near-repeat profile from a modern digital radar sounding collected 30 years later. In combination, these results highlight the potential (and provide the technical foundation) for radiometric analysis of Z-scope records in archival radar sounding film and motivate future reflights of archival data to enable radiometric analysis of multi-decadal subsurface evolution.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/jog.2021.130
Acknowledgements
We thank N. Holschuh and M. Siegert for their thorough, thoughtful, and constructive reviews. This work was supported, in part, by NSF award 1745137, Heising-Simons Foudation Grant 2020-1771, the Stanford Summer Undergraduate Research in Geoscience and Engineering Program, the Stanford Earth Summer Undergraduate Research Program, and the National Defense Science and Engineering Graduate Fellowship Program. The authors would also like to acknowledge Julian Dowdeswell, Lucy Martin and Naomi Boneham of the Scott Polar Research Institute at the University of Cambridge for their partnership in scanning the radar film archive.
Open access
