Airborne electromagnetic data levelling using principal component analysis based on flight line difference

Highlights

• New airborne data levelling method based on flight line difference and principal component analysis;

• Flight line difference enhances the features of levelling errors and improve the correlation between pseudo tie lines;

• Principal component analysis is introducted to extract levelling errors from the pseudo tie lines;

• Field data levelling examples to confirmed the reliability of the method.

Abstract

A novel technique is developed to level airborne geophysical data using principal component analysis based on flight line difference. In the paper, flight line difference is introduced to enhance the features of levelling error for airborne electromagnetic (AEM) data and improve the correlation between pseudo tie lines. Thus we conduct levelling to the flight line difference data instead of to the original AEM data directly. Pseudo tie lines are selected distributively cross profile direction, avoiding the anomalous regions. Since the levelling errors of selective pseudo tie lines show high correlations, principal component analysis is applied to extract the local levelling errors by low-order principal components reconstruction. Furthermore, we can obtain the levelling errors of original AEM data through inverse difference after spatial interpolation. This levelling method does not need to fly tie lines and design the levelling fitting function. The effectiveness of this method is demonstrated by the levelling results of survey data, comparing with the results from tie-line levelling and flight-line correlation levelling.

Introduction

Airborne electromagnetic survey has been widely applied in geological mapping, mineral exploration and groundwater search, while the proper levelling of AEM data remains a challenge and is still an active research area (Huang, 2008). Levelling errors can be easily recognized as the striping pattern along survey profile direction and significantly affect data quality. In the airborne survey, flight altitude variations (Huang, 2008; Beiki et al., 2010), flight direction changes (Huang and Fraser, 1999) and temperature variations (Valleau, 2000; Siemon, 2009) are main sources of levelling errors.

Airborne geophysical data levelling can be achieved in both tie-line direction (perpendicular to flight-line direction) and the flight-line direction. Early airborne geophysical data are levelled using tie lines which are flown cross profiles. Tie-line levelling deems the differences at the crossover points of the tie lines and the flight lines as levelling errors (Nelson, 1994). Unlike airborne magnetic data, AEM data are sensitive to the flight altitude. The fluctuation of flight altitude leads to residual corrugation in the tie-line levelling results of AEM data (Huang, 2008). Foster et al. (1970), Yarger et al. (1978) and Bandy et al. (1990) use the differences at the crossover points to fit and calculate the levelling errors, which improves the tie-line levelling method. Analysing the data features of levelling error, pseudo tie lines have also been used to level data without the need for tie lines. Huang and Fraser (1999) set the endpoints of the pseudo tie line in region without levelling errors and perform levelling through interpolation along the pseudo tie line. Davydenko and Grayver (2014) design a directional filter using principal component analysis (PCA) to levelling, considering that levelling error in the tie-line direction has a larger difference than that in the flight-line direction.

As for flight-line direction, Green (2003), Huang (2008), White and Beamish (2015) pre-set the error function to fit levelling errors using least-squared method. The levelling results are closely related to the selected error function. Beiki et al. (2010) design differential polynomial fitting levelling that avoids the selection of error function. However when the levelling errors of adjacent flight lines are similar, this algorithm cannot effectively remove the levelling errors (Davydenko and Grayver, 2014). Furthermore, median filtering (Huang and Fraser, 1999; Mauring et al., 2002; Mauring and Kihle, 2006), temporal filtering (Ishihara, 2015) and mean filtering (Li, 2007) have also been applied to level airborne data. These levelling methods generally need to configure the filter parameters.

This paper describes a method to level airborne geophysical data by analysing the characteristics of levelling error in tie-line direction and flight-line direction. Flight line difference is used to highlight the features of levelling error. Instead of levelling AEM data as usual, levelling is conducted to the difference data. Principal component analysis is applied to pseudo tie lines of the difference data to obtain the levelling error. To confirm the reliability of the method, we apply the method to airborne time-domain electromagnetic (TDEM) data and magnetic data acquired by Geotech Limited. Meanwhile, the levelled data are compared with results of tie-line levelling and flight-line correlation levelling for further analysis.