Field Report: GISP2 Summit Camp 1996 Boreal Summer Season

Bob Hawley, University of Washington


In the boreal summer of 1996, I accompanied Gary Clow of the USGS to the summit of the Greenland ice sheet. There, we carried out experiments as part of a project to further our understanding of paleoclimate changes, and how they may relate to the possibility of climate changes in the Holocene. When the air temperature changes above the surface, this changes the surface temperature of the ice, and the temperature change is propagated down into the ice. The result is that current temperatures in ice sheets can serve as a proxy for paleotemperatures. As an analogy, consider taking a frozen turkey and putting it in the oven. Ten minutes later the surface of the turkey has risen to the temperature of the inside of the oven, but the center of the turkey is still frozen. If you drill a hole into the turkey and measure the temperature of the center, you can tell something about the temperature of the freezer the turkey was in before it was put into the oven. This is an oversimplification of the ice sheet model, but the idea is the same. The time involved for these changes in an ice sheet is very large, so there is a very long paleotemperature record present in current polar ice sheet temperatures. By logging the current temperature in a borehole we can gain information to aid in the interpretation of climate signals contained in the chemistry of the ice core removed from the hole. The following is a summary of field activities in chronological order:

1. Install & log G - hole:

The G- hole is a 156M deep air-filled borehole a few kilometers away from the main borehole. It was drilled with a PICO 4 inch drill, leaving a hole about 5.625 inches in diameter. The chief difficulty of this experiment was how to use the existing hole to house a 1 inch ID fluid filled pipe for logging the temperatures in the hole. The pipe needed to be small to prevent convection from occuring in the fluid. Many different options for pipe materials were discussed. The final solution was as follows:

Installation:

We fed Driscopipe down the hole and filled it with diesel fuel (DFA) to ~150M. Since Driscopipe is manufactured in a continuous coil there are few couplings to make, but the pipe is naturally curved at the same radius as the storage coil. There is a risk of snagging the probe and losing it if the pipe in the hole is not straight. To make the pipe hang straight, we "splinted" and weighted the bottom of the Driscopipe with ~20ft of steel 1 1/2" sch 40 pipe. Also, before the installation, we stretched out the driscopipe on the snow and anchored it with deadman anchors at each end. After sitting for ~8-12 hours, the Driscopipe was relatively straight for most of its length, with bending restricted primarily to the ends. The bottom was of most concern to us, since it would have the least weight on it to keep it straight. The installation was effected using a 3/4" sheet of plywood with a hole cut in it for the borehole and a box to support the weight of the splinted pipe when the pipe was installed and hanging from the top of the hole. A tripod was constructed of 4x4" lumber and a sheave wheel attached at the apex (see figure #1). We attached a Dayton electric winch (not intended for vertical lifting) to one of the legs of the tripod and used a 3/32" steel multistrand cable with the winch to support the weight of the installation while lowering. We threaded the Driscopipe through the steel pipe for about 20 feet, and secured it at the bottom and the top with pipe clamps, to which we also attached the steel cable, using a nicropress cable-locking tool. We lowered the assembly into the hole with the winch. Because our hole was deeper than our pipe was long, we had to weld the Driscopipe about ~20m below the surface, and then again at the surface we welded on a drisco-to-steel pipe transition. We installed a reducer coupling to run the pipe up to 1 1/4" steel pipe, the size of the temperature logging well-head. Below the reducer coupling we mounted a steel plate and several washers to dissipate the weight of the system onto the box and thereby onto the plywood sheet (see figure #2). Once the system was installed (figure #3) , we filled the pipe with about 20 gallons of DFA. Each pour took up to a minute to work its way past all of the air bubbles, and they frequently "burped", occasionally spewing DFA. After the installation was completed, we removed the tripod, covered the hole and allowed its temperature to equilibrate for several hours. We backfilled the annulus around the pipe with snow to about 20m depth. We accomplished this by lowering a weather balloon attached to a pneumatic hose down the hole, and then inflating the balloon, occluding the annulus and allowing the snow to be bridged in a controlled manner.

Logging:

We logged this hole and all others during this season with the USGS Portable Logging system. For this particular log we utilized the short (600m) cable, and the temperature probe LT2. The log took about 3 hours to complete. We were aware that by installing the pipe and fluid we were disturbing the temperatures we wanted to record. In order to see the approach of thermal equilibrium between the pipe and the surrounding ice, we repeated this log several other times during the season.

Results:

The data obtained from the logs at the G- hole had some expected and some unexpected results. First, the inflection points were identical to those observed in the deep hole logs, both in terms of curvature and depth. This was as expected. However, the data show the G-hole to be cooler than the main hole by about 0.095K. This is opposite from what was expected, since the DFA in the pipe had been warm when sitting on the surface.

One possible explanation could be that the fluid we put in the hole was colder than we expected. This seems unlikely, since the fluid was on the surface in blue containers, soaking up the warmth of the sun. Although the fluid was most likely at GISP2 over the winter, allowing it to cool significantly, the sun had been up for at least 3 months by the time we put the fluid in the hole. This would probably have given the fluid time to warm up. The other theory that has been advanced is the possibility that the 160m hole was acting as a cold sink over the several years between the time that the hole was drilled and the time we logged it. In the winter, although the hole was capped and covered by 1 m of snow, cold surface air could sink through the firn into the hole. Since cold air sinks, the hole would be a place the coldest air would accumulate, possibly cooling the surrounding ice. In the summers, the cold air would warm somewhat, but in all seasons the hole would be a sink for cold air. If this "cold sink" theory is correct, we would expect to see a greater difference at the bottom of the hole than at the top of the hole, since the top of the hole would see little effect from the cold sink at the bottom. This trend is not apparent in our logs, so there may be a combination of factors cooling the hole.

2. log main (deep) hole

The main hole at GISP2 is a large diameter borehole extending from the snow surface over 3000m to bedrock at the bottom of the ice sheet. It could be said that this hole is the reason that there is a field camp at the summit of the Greenland ice sheet. The large geodesic dome constructed over the hole is slowly being buried in the new and drifting snow. To keep the hole from closing due to the surrounding pressure, this hole is filled with N-butyl acetate, which has a similar density to glacier ice. This season several activities were undertaken in "The Dome".

Short cable log:

We logged the top ~600 m of the main hole with the same equipment that we used at the G-hole. The principle difference inside the dome was that we were optimistic about the temperature stability of the dome and therefore we decided to try to log without the logging tent. We placed more insulating material on and around the hot box, but we could not stabilize the temperature inside. This caused unstable readings on the multimeter and caused difficulty in completing that log. We decided to put up the tent for the main log.

Long cable log:

The main log in the deep hole was set up identical to the configuration used in years past. The electronics were set up in the logging tent in the dome, and the large 4km winch was used to lower the probe (LT1) to the bottom of the hole. Logging speed was ~11 feet/minute. The duration of the entire log, run continuously, was ~15 hours. The logging tent provided a much more stable temperature inside the hot box, and the log was monitored continuously.

Convection experiments:

In order to understand the convection cells believed to be present in the lower 1000 m of the hole, we ran several experiments by holding a continuously logging probe at the same depth for about 8 hours. The resulting data show that there indeed seem to be convection cells, but that they are turbulent convection cells and therefore not easy to characterize.

3. Log GRIP hole

GRIP is another field camp 28 km from GISP2. Like GISP2, GRIP is the site of a deep borehole. The camp is rarely visited anymore and is abandoned; most of the structures are almost completely buried in snow. The trip to log GRIP was scheduled near the end of the season. Shortly before the trip, we discovered a malfunction in our depth counter. At low speeds, the counter functioned normally, but as speed increased, the counter dropped counts increasingly until (at higher speeds) it failed entirely and would not count at all. For the most part this was a mere annoyance, since with the short cable we could log at slower than normal speeds and therefore be able to get an acurate log. However, in the GRIP hole, we planned to use the long cable and the big winch, and the big winch unfortunately did not go any slower than 10 fpm, just above the speed at which the depth counter began to fail. For this reason we were forced to abandon the possibility of logging GRIP with the big winch and the long cable. We felt lucky that the failure of the depth counter occurred after the log in the main GISP2 hole, the primary experiment of the grant.
We arrived at the GRIP hole and set up camp in the abandoned GRIP living dome. Our first task after setting up the logging system was to determine the level of fluid in the hole. We attempted to do this by means of a temperature probe. We reasoned that the temperature of the fluid, with its high specific heat relative to air, would be very close to the mean annual ice temperature, wheras the air in the hole would be close to the summer air temperature. We lowered the probe down the hole to where we thought the fluid would be, but saw no major temperature change as we expected when the probe hit the fluid. Several times the cable seemed to go slack, but after fishing more we felt that the tool was going downhole again. When we still had detected no large temperature change, even after 10 m more cable had been spooled out, we began to pull the tool out, to replace it with the fluid level detector. As the cable came out of the casing we discovered several tangles, suggesting that we had been simply spooling cable down the hole with the tool sitting on what we theorized was an ice plug. We then attached the fluid level detector and as much weight as we dared to see if we could break through the layer. We weren't sure if it was a layer of unconsolidated slush, or maybe a thick ice plug. To add to our worries, a crack that we discovered in the cablehead undermined the structural integrity of the threads that held the tool in place. We did not want to lose our tool. We sent the fluid level detector down and the cable went noticibly slack when we reached the fluid level recorded from last year. Nothing would get the tool through the plug. We eventually decided that the tool and the cablehead were too valuable to risk in an attempt to break the plug, so we went back to GISP2 to contact the Danish stewards of the GRIP hole and to equip ourselves with the tools needed to break through the plug. We got the OK from Denmark, and returned to GRIP with 20 feet of 1 1/2" steel pipe, (over 10 Kilos), and steel cable and swaging material for lowering the pipe. We would lower it and raise it by hand, with a snow machine ready to pull it up if need be in an emergency. Once the system was set up, we lowered the steel "battering ram" down the hole. At the same spot, it hung up. We raised and dropped the pipe several times before we finally broke through. After the initial breakthrough, we tried to break up the whole plug by holding the cable against the opposite side of the casing, but we had no way to verify that the proceedure was working. After several runs up and down with no apparent resistance, we pulled the "ram rod" out of the hole and put down the temperature probe. We never had another problem with the plug, although the cable and tool had slush from the plug on them when they came up. We then proceeded to log the hole with the longer of our 2 short cables. This cable was plagued by inconsistent speed and popping sounds (which cause interference in the electronics package) caused by improper spooling of the cable by the previous users. Nonetheless, the log was successful, and took about 7 hours.

4. Dante (pressure/temperature experiment)

The purpose of this experiment was to determine the effects on temperatures in an open borehole that small changes in barometric pressure could exert. If the air above the firn is significantly colder than the firn itself, an increase in barometric pressure could force cold air down into the porous firn, affecting temperature readings there. The same possibility exists if the air is warmer than the firn, or if there is a temperature gradient within the firn. Since our work is dependent on characterizing the temperature in the ice and the heat flow in and out, this measurement could prove important. This effect could also be magnified in an air-filled borehole, as there is a passage for air to freely flow in. Because of the equipment failures described below, we were unable to execute this experiment.

Setra differential pressure transducers function by measuring small differences in pressure between a reference pressure (usually attached to a sealed volume held at as constant temperature as possible) and the pressure at a signal port (usually open to the outside through a sintered porous filter). The signal port and reference port on the transducer lead to opposite sides of a thin metal membrane dividing the chamber in the body of the transducer. Since the membrane is one plate of a capacitor, a differential pressure causes the membrane to deform slightly, changing the capacitance. The resulting change is output as a voltage.

Dante is a unit built by Brian Peterka for the 1994-95 antarctic summer field season. Its main function is to ease installation and use of up to 4 Setra differential pressure transducers. Dante consists of a set of pneumatic tubes and wires which allow the transducers to be buried about 1m in the snow, where temperatures are more stable than on the surface. Setra pressure transducers are very sensitive to temperature, so to get a reliable pressure reading it is necessary to bury them. In addition, they sometimes drift, making it necessary to "zero" them by creating a short circuit in the plumbing between the reference and signal ports on the transducer. This is acomplished by means of several valves located at Dante's "head". For a more complete discussion of Dante's initial setup, see Brian Peterka's "Weather Station Manual". I made several modifications to Dante, both in the lab and in the field, to adapt him to better fill my needs.

First, I completely rewired and replumbed Dante to make it simpler as a system. All 3 Setras are connected to the same surface port, and there is 1 valve to zero the reference volume. I later added a second valve to allow the setra reference port to be opened to the atmosphere while the reference volume was equilibrating(see diagram, figure #5 ). I got rid of all of the db type connectors, which I think may have been hard to crimp, causing loose connections. I soldered the setra wires directly into a length of 9-conductor seiscord, with the power leads going to a separate length of 3-conductor seiscord(see wiring diagram, figure #6 ), so that I would be sure of all connections. The only non-solder connection in the system was the connection at the CR-10 itself. I therefore feel that I have eliminated faulty connections as a source of error.
Once this was done, I wrote a CR-10 program (gisp296.doc) to measure the setras, a thermocouple at the panel, a thermocouple between 1 and 3 meters, and wind speed and direction. The execution interval was 5 seconds, and every minute the instantaneous readings and minute- averages were output. It worked flawlessly in the weatherport lab, but when I buried the sensors, they "spun their heads around" and output ~14 V. Since the setras are specified to have a maximum output voltage of +-2.5V, this was more than a little disturbing. There seemed to be nothing to do but have breakfast, and after an hour or so the voltages were dropping, roughly 1V/hr. Eventually the readings were in the range for which the setras were designed. I then decided to try to get a feel for how "zero" looked and set up the valves accordingly. The results were less than desirable. The output of the setras varied over the entire range of output expected of them, appearing to be a huge signal. Since I was trying to measure "zero", I was not convinced. I had theorized that the erroneous readings were not a result of faulty wiring or damaged units, not only because I had rewired them specifically to prevent this, but also because the outputs of the 2 working setras(one of the wires connecting the other had broken) closely paralleled one another throughout all of the spurious voltage spikes(see figure #7 ). I thought that possibly something in the turning of the valves themselves had caused some signal, and removed them from the system. The final configuration was greatly simplified(see diagram, figure #9 ). Even with every possible simplification made, the outputs were such as to make measurements impossible, since the noise of the system was the same order of magnitude as the expected signal. The field season was drawing to a close, and the experiment had to be abandoned. As might have been expected, the zero output did not change by more than +-25mV overnight in the lab back at UW in Seattle (see figure #8 ). I had measured many parameters in the field and could find no correlating temperature, windspeed, battery voltage, or any other corresponding spikes that might help explain the readings. Perhaps the most plausible explanation for the erroneous readings is the possibility of a small ice blockage in one of the tubes leading to or from the setras. If this were the case, some thermal gradient in the snowpack could have caused a pressure differential across the setras even though they were "short circuited", since an ice blockage would effectively cut off the short circuit, allowing the 2 sides to drift to different pressures. Unfortunately this could not be tested for after the unit was brought back from the field, and dwindling field time made it impossible to make a second attempt at installation.

The possible ice blockage could have occurred at any location along the tube leading from the reference port to the signal port. Figure #9 shows the two most extreme cases. In the first case the blockage is at the surface, so any thermal gradient would have the minimum effects on the "zero" reading. The second case shows the maximum limit of the effect where the entire 2m of tubing going to the surface and back can be warmed or cooled by a possible temperature change in the snowpack, which will not be constant over the 1m depth. When it is sealed, the pressure fluctuates in proportion to the change in temperature.

We can make simple calculations to check this theory. In the "worst case scenario", the blockage might have occurred in the tube exactly at the reference port, leaving the 2M of tubing going to and from the surface to cause an erroneous signal. It is sealed from the outside, so the amount of air inside and the volume remain constant in the Ideal gas equation,

PV=nRT.

For the location in question, with V and n held constant using air at a pressure of roughly .57 atmospheres due to elevation and latitude and an average temperature of -20 degrees C, this equation reduces to:

P =.00225 T

Where P is the pressure in atmospheres and T is the temperature in kelvins. The Setra transducers fluctuated on the order of 1000mV, corresponding to ~2.5mb, or .00245 atm. To create this pressure difference in the tube, the temperature must have fluctuated on the order of 1 degree C. Fig 10.1 of "The Physics of Glaciers", by W.S.B. Paterson shows that midsummer temperature gradients in central greenland can be ~4 degrees C over the first meter. This is a reasonable expectation at GISP2 in mid june. The fact that no discernable pattern with respect to wind, air temperature, time of day, or any other data I collected might be explained by the fact that after the assembly was installed a Scott tent was erected over it, causing the surface temperatures in the 3m^2 around the device to vary with no predictible pattern. My presence inside the tent, at odd times, would have an effect. The warming inside the tent due to solar energy would be a factor, though not regular enough to define a pattern. Unfortunately, we did not have the time in the field to fix the problems that were occuring.

In retrospect, I believe that in my effort to simplify Dante as much as possible, I made the device more vulnerable to this particular problem; if each setra had been left with its own surface port and its own tubing, as it was initially constructed, an ice blockage in just 1 tube would have knocked out only 1 setra, not all 3, and we would have been able to go forward with the experiment using the remaining functional pressure transducers. In this case, redundancy would have proved successful, where simplicity failed.

5. Air hole drilled and logged near the G-hole

Near the G-hole, we drilled a 10m air hole with the PICO hand auger. This hole was logged by Gary using the Portable Logging System and the air probe.

6. Problems encountered in the field season:


The majority of the 1996 field season was a great success. We were able to obtain data for our main objectives, and the equipment failures that did happen caused only the loss of "add-on" data. Much has been learned about temperatures in ice sheets, and their relation to past temperatures.
Robert Hawley
Email -- bo@geophys.washington.edu
Web -- http://www.geophys.washington.edu/People/Students/bo/