Back to list of UTIG presentations at SEG 2002

Lopamudra Roy, Thomas Richter, Donald Blankenship, Mrinal Sen, John Holt and Paul Stoffa.

Since 1994 researchers at the University of Texas Institute for Geophysics (UTIG) have utilized a light twin engine aircraft in Antarctica to make tightly integrated geophysical measurements including gravimetry. Over 200,000 line-km have been surveyed during seven field deployments, each lasting 3-4 months. Below we describe the acquisition and analysis procedures for our airborne gravity data and study of the propagation of error for a recent experiment over subglacial lake Vostok, East Antarctica.

The UTIG airborne platform is a contracted DeHavilland Twin Otter instrumented with ice-penetrating radar, laser altimeter, magnetometer, and a gravimeter. Navigation systems include both uncorrected and differential CA code GPS and GLONASS. After the field season, positioning is determined via differential carrier-phase GPS, providing accuracy of about 0.1 m. Aircraft attitudes and 3-axis accelerations are measured with a laser-gyro inertial navigation system (INS). Measurements (other than GPS) are linked in a single data acquisition system with accurate time stamping provided by a GPS synchronized counter-timer. A ground instrumentation suite mirroring the airborne suite provides reference observations for the GPS and magnetometer data. The gravimeter utilized is a Bell Aerospace BGM-3 marine system, modified for airborne use. The BGM-3 provides measurements of vertical accelerations at 1 Hz, with verticality of the sensor maintained by a gyro-stabilized platform. Dual-frequency carrier phase GPS observations are made by three independent GPS receivers (both airborne and ground references) operating at 1 or 2 Hz. In order to maximize GPS data quality, GPS receivers of different design are used simultaneously. The most recent suite of GPS receivers consists of an AOA Turborogue, an Ashtech Z-12, and an Ashtech Z-Surveyor. All GPS data sets are reduced using two different software packages -- K&RS and GIPSY-OASIS. K&RS produces the most accurate positions but is inappropriate for long baselines. While GIPSY-OASIS yields positions in circumstances unfavorable to K&RS (i.e., long baselines and lines without closure), it is about half as accurate as K&RS and is insufficient for achieving the desired accuracy of 1-2 mGal in the reduced gravity data. Up to 21 GPS solutions are available for each line. Selection is made through correlation of the high-frequency accelerations recorded by the gravity meter and those derived from the GPS positions. In some cases, this selection is guided by the data reduction process used for the laser altimetry. After inertial and other corrections are applied, a moving average filter with a tri-weight kernel width of 15 km (amplitude of 0.46 at 7.5 km) is applied in order to remove residual high-frequency noise. For most cases, this produces an agreement between repeated lines of <1.5 mGal rms difference and preserves small-scale gravity features.

In response to a National Science Foundation proposal from Lamont Doherty Earth Observatory (R. Bell and M. Studinger) to study subglacial Lake Vostok in East Antarctica, a team from the University of Texas Institute for Geophysics conducted the first comprehensive aerogeophysical survey of Lake Vostok during the 2000/01 austral summer. The survey block was 165 x 330 km (line spacing 7.5 km with 11.25 km and 22.5 km ties), augmented by 12 regional lines extending 180 to 440 km outward from the primary grid. Four km of ice cover, high altitude, and extreme cold presented significant technical challenges, but the survey was completed successfully and has resulted in excellent geophysical data sets.

The gravity measurements over lake Vostok were gridded to produce a map of the free-air gravity field which shows many correlations to the subglacial topography determined by radar sounding of the overlying ice sheet. Presently, we are using the radar determined ice thickness as a priori information to invert the gravity profiles for mapping the bedrock topography where radar can not penetrate. We know well that these analysis are highly non-unique with uncertainties arising from several sources, viz., noise in the data, lack of a priori information, simplification of the forward model. Our forward model is composed of known ice thickness over water and we are trying to determine unknown water and sediment thickness lying over crystalline bedrock. So, the objective of the present work is (i) to visualize how error in the data propagates into the interpreted results and (ii) to understand how the distribution of a priori information can help to improve the results. Instead of providing a single solution with acceptable fitness, we generate a large number of acceptable solutions which can be condensed into an interactive solution space. To accomplish this, we use very fast simulated annealing (VFSA) - a global optimization technique to invert the gravity data. Statistical analysis is performed over 10,000 inverted models to obtain a weighted mean model with variance at each parameter point. The correlation of each parameter with others is presented to visualize the uncertainty holistically. Initially, simulated data is analyzed to quantify the effect of data error, as well as, wrong a priori information. A similar error analysis is performed for gravity data along profiles over Lake Vostok with ‘single point’ information of water and sediment thickness obtained from seismic measurements.