GPR was used to assess the potential for subsurface features of archaeological and morphological interest occurring in the 100 m stretch of open space in between the summer and winter settlements (Fig. 4). GPR relies on reflected energy from the propagation of electromagnetic pulses generated at the surface. In air, the velocity of a radar wave approaches the speed of light in a vacuum. In the ground, however, this velocity is greatly reduced and the wave propagates at velocities governed by the electrical properties of the medium, with the primary controlling properties being electrical conductivity and relative permittivity (dielectric constant; Davis & Annan 1989; Conyers 2004; Daniels 2004; Jol 2009). At interfaces where electrical properties vary, a backscattering of reflected energy will occur. This reflected energy is the primary information sought with a GPR survey. The two-way travel time of the observed signal can then be used to estimate the depth to an embedded object or interface in the ground (Davis & Annan 1989). Due to the very short wavelengths involved, GPR is capable of resolving finer detail than most other geophysical methods, though it may not work well where the ground is very conductive or insufficient contrast exists between the electrical properties of the target and the host medium (Conyers 2004). With appropriate processing, GPR data can be presented as profiles, planar images (time-depth slices), three-dimensional volumes or combinations thereof.

The GPR survey was undertaken using the Noggin GPR system (Sensors and Software, Mississauga, ON, Canada), with a 250 MHz centre-frequency antenna. The instrument was deployed on a sled configuration that glided smoothly across the lichen and moss covered ground. In this setting, such a mode of GPR deployment may be preferable to the more typical cart set-ups due to reduced packing space, an issue for small float planes, and to ease of manoeuvrability: a sled moves more easily across this type of surface that a wheeled cart likely would. Transects were spaced at 0.25 m and in-line distances were recorded with both a survey tape and an instrument odometer. The subsurface environment offered minimal attenuation of the radar signal, with only a weak gain being necessary to view the data in real time. While this may seem surprising for a once saline environment, it was clear that hydrologic processes have acted to remove any remnant salts, which would have increased the electrical conductivity, thereby attenuating the GPR signal.

A number of typical signal processing protocols were applied to the radar data to enhance desired features. The most common procedure where radar data are concerned is the application of a gain to amplify a signal as it weakens with depth. This was achieved by implementing a moderate spherical and exponential compensation gain. Subsurface velocities were determined with a hyperbola fitting procedure during post-processing, and average velocity was shown to be 0.1 m/ns. A signal-saturation removal procedure (dewow) was implemented as a high-pass filter to remove unwanted low frequency trends from shallow induction currents (generated by the instrument itself or other sources), along with a background removal filter which subtracted the average value of all traces in the collected grid for an additional high-pass effect. To eliminate the unwanted tails of diffraction hyperbolas, a common GPR data artefact, a velocity migration procedure was implemented using the velocity previously determined with hyperbola fitting. This procedure resulted in a more spatially accurate rendering of embedded features. Finally, an enveloping procedure was applied to eliminate the oscillatory nature of radar traces so that similar reflections in close spatial proximity could be packaged across multiple traces for more accurate spatial rendering. Two-way travel times were converted to depths using the previously determined velocity, and individual profiles were grouped for time-slicing and three-dimensional rendering. A more detailed discussion of the processing techniques described here can be found in Annan (2004), Conyers (2004), Daniels (2004) and Jol (2009).

In addition to GPR, magnetic gradiometry was used to obtain information about subsurface features on the site and to supplement the GPR results. Magnetometers are sensitive to remnant magnetization (permanent magnetization locked into the structure of a material) and induced magnetization (caused by Earth's magnetic field), both of which vary significantly across material types. Since the instrument used for this survey measures the total-field response (total flux density without a directional component), it is difficult to differentiate between induced and remnant contributions, though this can be possible through comparison with electromagnetic in-phase data (Won & Huang 2004).

In this setting, a vertical magnetic gradient was preferable to a total field survey. While Earth's magnetic field fluctuates temporally throughout the course of the survey, the differential measurement between two sensors of varying elevations (vertical gradient) is often used as a relative measurement of variations in the local magnetic environment, eliminating the need for additional diurnal corrections to the data. This also enhances the response from shallow magnetic features by minimizing low frequency trends, acting as a type of high-pass filter (Silliman et al. 2000). Use of the vertical gradient is especially beneficial in northern latitudes, such as Cape Krusenstern, that experience more rapid and intense fluctuation in the magnetosphere throughout the day, thus making diurnal corrections impractical. This is especially true in the auroral zone during the summer months (Urban et al. 2012), the effects of which were responsible for the failure of J. Louis Giddings’ early attempt at magnetic surveying in this region (Urban 2012).

A G-858 cesium vapour magnetometer (Geometrics, San Jose, CA, USA) deployed in gradiometry configuration (vertical gradient survey) was used for this project. With the sensors positioned at 0.25 and 1.25 m above grade, transects were collected with a 0.50 m interval. In-line distance was recorded with a survey tape. Limiting factors to the magnetic survey included the presence of modern ferrous metals on the surface and the natural effects of the northern magnetosphere. With the former issue, little could be done to mitigate the problem. The offending objects included a permanent steel datum set by the National Park Service (NPS), as well as steel cans and stakes left behind from previous archaeological work that were well hidden by surface vegetation. The latter issue of Arctic diurnal effects was mitigated by using the relative measurement of vertical gradient only rather than attempting diurnal corrections, which would have required a continuous base station, as noted by previous studies in the region (Urban et al. 2012).

Processing of the magnetic data required a standard set of procedures including despiking, drop-out removal and gridding with the Kriging interpolation algorithm. Some less routine procedures were also applied. For example, in order to enhance certain subsurface features an upward continuation was implemented. The upward continuation is a potential field transformation procedure that develops synthetic magnetic observations for a specified plane above the actual sensor (Blakely 1995). By simulating a magnetic surface at a higher elevation in this fashion, a pseudo-gradient was developed with the originally observed magnetic data and the upward continued data. This procedure was implemented as a noise reduction strategy that acted to suppress local noise at the survey surface—in essence acting as a low-pass filter—resulting in a smoother image. This was done specifically to reduce noise generated at the low sensor by the shallow ferrous debris previously described. Though the continuation was over a short distance (50 cm), it was effective in enhancing general trends that were obscured by the debris, and without sacrificing significant detail where such trends occurred.

GPR revealed a complex dendritic morphology of the Old Whaling beach ridge's subsurface (Fig. 5). This morphology between the two known settlements (summer and winter) appears to be the result of natural morphologic processes rather than tent pads as the Darwents hypothesized (Darwent & Darwent 2005:151), and offers a glimpse into the past drainage and cryogenic processes of the site area. Most of the anomalous patterns detected with GPR are likely the perimeters of buried permafrost polygons and associated shallow ponds generated by the seasonal freeze–thaw cycle. Similar features are visible on the modern surface near the coast. Those detected in the GPR results may have been surficial features when the coast was closer to the Old Whaling beach ridge.

To further evaluate if these were natural features, several test excavations were conducted to investigate if the slight depressions were possible cultural features since many are roughly the same size and shape as prehistoric house features from the region. This could also explain the Darwents’ interpretation, although they noted fire-cracked rock and debitage on the surface at one of these locations. A 50×50 cm test excavation at that same location revealed only eight pieces of chert debitage at or near the surface, as well as infilling episodes related to natural processes (Fig. 6). No indication of an interior hearth feature is present in the area and our magnetic survey revealed no buried external hearth features. Depth estimates of a similar feature derived from GPR and test excavations support the argument that the feature was a pond generated by a low-centred frost polygon that appears to be contemporaneous to the documented settlements (based on depth of the feature). In addition to the dendritic patterning of the frost polygons, a relatively deep channel cutting through the middle of the survey area appears in the GPR survey results. This represents a present topographic feature related to site drainage rather than a subsurface feature. This feature appears to bisect several of the deeper subsurface features.

In areas between the two settlements, the drainage patterns detected with GPR were also evident with gradiometry (Fig. 7). These manifested more prominently in the west, toward the summer settlement, and are not nearly as well defined as with the GPR results; however, the perimeters of frost-polygons are apparent. Some of these appear as magnetic lows, while others appear as magnetic highs. There are a number of possible explanations for this, including magnetization and demagnetization of sediments from hydrologic processes, variations in the density of the sediments (i.e., these may be fissures infilled with various material including sediment, water and organics), and variations in respiration with associated changes in microbial activity. These observed disparities are likely related to whether a given polygon is low-centred or high-centred as the fore fosters the accumulation of water (i.e., pooling) while the latter fosters draining. With the magnetic results alone, several of these features could easily be mistaken for house sites on the basis of scale and layout. However, when compared with the radar results, it is clear that these are part of the broader dendritic pattern across the site. That is not to say that all of these features are without cultural relevance. Indeed, some of these may have served as good locations for tent pads or other features as the Darwents suggested. It is evident, however, that houses of the type and scale seen in the summer and winter settlements are not apparent across the rest of the site.

The underlying foundational structure of the beach ridge is also evident with both survey methods (Fig. 8). Comparison of upward continued magnetic data with some of the deeper GPR time-slices reveal a consistent, underlying trend striking across the full length of the site. This may be related to the early formation of the beach ridge.

The footprints of the previously excavated houses in the summer settlement were easily visible in the magnetic results, which are likely a result not only of the features themselves, but also of topographic variations, disturbance from previous excavation, and additional anomalies clearly related to modern ferrous trash, such as cans and stakes (Fig. 9). A metal rebar datum placed at the site by the NPS (which we were unable to remove) also generated a prominent anomaly. The additional deeply buried structure identified by the Darwents (Darwent & Darwent 2005) was not detected with gradiometry. Weak randomly distributed dipoles were visible throughout the settlement area, which could be the result of heat-exposed rocks. Interestingly excavation in the area of the possible buried structure (Fig. 10), revealed a small amount of animal bone (marine, terrestrial and possibly bird), wood fragments, a projectile point base fragment, and a few pieces of chert debitage at a depth of approximately 47–51 cm below surface, slightly above where the Darwents identified possible house timbers. None of these artefacts are diagnostic so there affiliation remains undetermined. They may relate to the nearby house structure (House 204) excavated by Giddings & Anderson (1986) and possible midden deposits.