First Attempt of Applying Factor Analysis in Moving Base Gravimetry

Abstract:

For gravimetric observation systems on mobile platforms (land/sea/airborne), the Low Signal to Noise Ratio (SNR) issue is the main barrier to achieving an accurate, high resolution gravity signal. Normally, low-pass filters (Childers et al 1999, Forsberg et al 2000, Kwon and Jekeli 2000, Hwang et al 2006) are applied to smooth or remove the high frequency “noise” - even though some of the high frequency component is not necessarily noise. This is especially true for aerogravity surveys such as those from the Gravity for the Redefinition of the American Vertical Datum (GRAV-D) project. These gravity survey flights have a spatial resolution of 10 km between tracks but higher resolution along track. The along track resolution is improved due to the lower flight height (6.1 km), equipment sensitivity, and improved modeling of potential errors. Additionally, these surveys suffer from a loss of signal power due to the increased flight elevation. Hence, application of a low-pass filter removes possible signal sensed in the along-track direction that might otherwise prove useful for various geophysical and geodetic applications. Some cutting-edge developments in Wavelets and Artificial Neural Networks had been successfully applied for obtaining improved results (Li 2008 and 2011, Liang and Liu 2013). However, a clearer and fundamental understanding of the error characteristics will further improve the quality of the gravity estimates out of these gravimetric systems. Here, instead of using any predefined basis function or any a priori model, the idea of Factor Analysis is first employed to try to extract the underlying factors of the noises in the systems. Real data sets collected by both land vehicle and aircraft will be processed as the examples.