Stories from the Field: Climate Change at the Arctic’s Edge
PRINCIPLES OF GROUND-PENETRATING RADAR
A ground-penetrating radar (GPR) system generates electromagnetic signals and detects the interaction of the electromagnetic field with the surrounding material. GPR utilizes an antenna (containing a transmitter and receiver a small fixed distance apart) to send electromagnetic waves into the subsurface. The antenna is moved over the surface of the medium to be inspected. The transmitter sends a diverging beam of energy pulses into the subsurface and the receiver collects the energy reflected from interfaces between materials of differing electromagnetic properties. The reflected energy is recorded as a pattern on radargrams that are displayed in real time. The radargrams, which look similar to seismic sections, constitute the raw GPR data.
The GPR methodology is a very high-frequency electromagnetic wave technique used to produce high-resolution images from the subsurface. GPR is used, for example, to characterize the subsurface stratigraphy, water table, permafrost depth, and/or geology. When used to survey over time at the same site, GPR is also an effective method for monitoring changes in the subsurface. Electromagnetic energy from the antenna can propagate at frequencies ranging from 10 MHz to 3 GHZ. The peak power of this antenna is 20 to 100 times less than the wattage of a cellular phone, and the energy is directed into the ground (and not at the operator) by means of shielding on the top side of the antenna.
The GPR signal is reflected back to the antenna by materials with contrasting electrical impedance, which is primarily determined by the dielectric and conductivity properties of the material, its magnetic permeability, and its physical properties.
A material's dielectric properties are primarily determined by its mineralogy and water content. The greater the contrast in the real dielectric permittivity of two materials, the greater the reflection amplitude. Typically, high-amplitude reflections occur at lithologic or mineralogic changes, or where there is a sudden change in water content (water table or depth of ice).
Relevant for our area here in Churchill, we expect ground-level reflection (air to ground) to be positioned at depth = 0 meters and to appear as a "hard" kick, either a positive or negative loop depending on the signal polarity and polarity convention. Then the permafrost loop is probably a "soft" kick (if we assume water on ice) according to the relative permittivities and so should have opposite polarity than the ground- level reflection.
Reflections observed on GPR records are nonunique, meaning that a similar reflection can be caused by many different objects or combinations of layers.
GPR is measured as a function of time. To relate to the depth below surface, a velocity estimate is required. The velocity differs between materials with different electrical properties, and a signal passed through two materials with different electrical properties over the same distance will arrive at different times. The interval of time that it takes for the wave to travel from the transmit antenna to the receive antenna is simply called the travel time. The basic unit of electromagnetic wave travel time is the nanosecond (ns), where 1 ns = 10-9 seconds. The velocity of an electromagnetic wave in air is 3 x 108 m/s (0.3 m/ns). Air is the fastest medium. The travel time of a wave in a material other than air is always smaller than 3 x 108 m/s (0.3 m/ns). For time to depth conversion, an average velocity of 0.1 m/ns is taken for our Churchill environment. Note that in reality, within a vertical section, the velocities can vary considerably (water, 0.033 m/ns; sand or clay, 0.06–0.09 m/ns; ice, 0.16 m/ns), leading to extreme variations in depth sections per time unit (time to depth conversion) if we vary from water-saturated rock to ice. This means that the time position and loop character of an interface reflection at, say, 3 meters will vary dramatically (the time scale can vary by a factor of 10), depending on whether the overburden is air, water, or ice filled.
Peter Kershaw briefed us on the first day at 7:45 A.M. in the large, unheated laboratory. Here we were introduced to the GPR. We all knew we would grow to love this instrument over the next 11 days. We spent about an hour in a detailed lecture on the systematic setup and operation of this very data acquisition tool.
We had three minutes to get ready to depart for a first survey on a tundra site, a long drive over very rough, potholed roads. We saw a mother bear and her cub on our drive, which was good as we could all tick this one off our list now. We set up two 50-meter tapes perpendicular to each other and sampled the permafrost underground every 25 cm. The instrument has an all-plastic construction with a transmitter mounted on a heavy plastic ski approximately 1 meter long with tall plastic handles for holding the instrument in position. This is powered by two 6-volt motorbike-size batteries. The detector is identical in construction and is operated with a separation of exactly 1 meter with both tracks held parallel. The transmitter and detector are connected together by optical fiber, which is then connected to a signal processor, which must be kept more than 5 meters away from the measuring instruments to reduce electrical "noise." The processor is powered by a 12-volt bike battery and is connected to an HP Palm to store the data. Operators with steel-toed boots are excused from taking their turn operating the GPR. The unit is set to send pulses of radar at 100 MHz, which relates to a target depth of 5 meters (maximum depth of 15 meters). Every 25 cm, the heavy unit is lifted and moved for a new reading , which is recorded on a second data logger. When the readings are taken near a steel object that cannot be removed, such as a leg of an environmental monitoring station or other steel pegs, this must be manually noted in the data logger. This is a slow, tedious process but one that should provide relevant data on the depth of permafrost that can be periodically compared.
GPR data from the field are processed using the pulseEKKO program (Sensors & Software Inc.) provided with the GPR tools (a bit outdated, from 1996, but still functioning appropriately). Key processing steps include input data selection, topographical correction, scaling, display parameter selection, and plot file creation. Some processing issues and tests are described below.
Various sections had to be drift corrected (Timezero Adjust), as the first reflection was at times below 2 meters depth, was drifting away (changing from left to right), or appeared to have opposite polarity. Timezero Adjust shifts the tracks so that the first reflection (assumed to be ground level) is set at time zero (according to the manual, although it looks closer to depth = 0). I tested and found the most stable correction using a threshold of +2,000.
The polarity of produced GPR sections is not clear. Is a positive number (black loop) a "hard" kick (increase of permittivity) or a "soft" kick (decrease of permittivity)? In several instances, zero depth, which should in principle represent ground level (air to ground interface should be a "hard" kick), appears as a black loop. Sometimes it is a white loop. Note that while recording, polarity depends on the system setup, as the electric unit orientation controls the recording polarity (always have the pulseEKKO name on the electronic units pointed in the direction of antenna movement).
Appropriate amplitude scaling needs to be selected to see the relevant features. With no scaling, it appears that typically below 50–80 ns (3–4 meters), the formation is immediately very transparent. Automatic Gain Control (AGC), which scales amplitudes to common level (within specified window), allows that a weak signal (low amplitude) can still be "seen." Applying AGC here, however, shows that the data below 60–100 ns (3–4 meters) are not coherent and thus not meaningful.
All recorded sections of October 2008 have been processed according to this protocol. The October 2006 data have also been reprocessed following the same workflow in order to attempt to compare them.
FURTHER STEPS AND INTERPRETATION
Calibration and Validation—Ground Truthing
The GPR section shows loops that represent reflections at interfaces of various lithologies (with relative permittivity or electromagnetic "hardness" variation). Loops could represent many interfaces, such as sand/peat, air/water, or water/ice.
Calibration and validation with "hard" measurements such as soil coring or soil profiling is required to answer which interface is the interface with ice (top permafrost). The least would be to use a permafrost probe. It is suggested to do such calibration and validation at various points of the GPR survey simultaneously with the acquisition (e.g., every 5 meters, depending on the expected variability).
After calibration, once it is clear which loop represents the depth of permafrost, the GPR section can be interpreted for depth of top permafrost all along the profile.
A repeatability test would analyze how well we can repeat a survey (without any changes in the subsurface) and thus which differences are attributable to the repeatability of the technique (tool warming, tool settings, air temperature, operators, acquisition protocol, etc.) and which to variations in the subsurface. A simple repeatability test could be done by coming back the next day or week and having different operators repeat a few meters of the survey.
Time Lapse Comparison
Finally, time lapse analysis and interpretation would compare the same sections shot in different seasons or years and analyze the differences in terms of changes in permafrost depth. Under favorable conditions, visual comparison could be sufficient to interpret changes in permafrost depth. Shifting the time lapse sections to common datum (Timezero Adjust) is essential. In more complex cases, more detailed time elapse analysis and processing may be required, looking at time shifts and amplitude differences independently.
At first glance, it may look comparable, but time scale- and frequency-wise, it is quite different. A change of water table as well as frost depth (causing considerable velocity changes) could possibly explain such differences. Although we interpret the first negative loop (white) in 2008 as top permafrost (as calibrated by 2008 permafrost probing), it is not clear how to interpret the 2006 section. It may be that the first white loop is also the top permafrost, but that is not certain as the character and shape are quite different. Permafrost probing in 2006 would have been needed to validate this. A series of time lapse plots comparing October 2008 with October 2006 is made for most surveys.
Learn more about
Climate Change at the Arctic’s Edge