Correcting for SPM effects in airborne EM

Abstract

INTRODUCTION Superparamagnetism(SPM) effects in airborne electromagnetic (AEM) surveys can be a source of needless expense in exploration if not identified, due to unnecessary drilling or further exploration. AEM surveys are commonly used for conductor mapping. Figure 4 shows, for example, the EMFlow apparent conductance (surface to 600 m depth) for a VTEM survey in the Mwese area in Africa. There are many high conductance anomalies on the map that would appear at first glance to be suitable drilling candidates in this dataset. With recent reductions in AEM noise levels, SPM effects (explained in the next section) has become an issue in lowamplitude, late-delay time data, and many anomalies have been drilled based on mistaken interpretation (Mutton, 2012). The challenge lies in determining which of these anomalies indicate basement conductor responses, and which indicate SPM responses. METHOD AND RESULTSFor a distribution of SPM decays, the signal from SPM appears as proportional to 1/t, in a dB/dt receiver. We use a 1/t basis function, together with exponential decays, to decompose AEM data. Previous work has described basis function decomposition of AEM data for EM response characterisation. This approach has recently been extended to airborne IP effect detection through least-squares fitting of both EM and IP basis functions as described in Kratzer and Macnae (2012). This last paper provides a comprehensive description of basis function fitting methodology, which we will briefly summarise below. To fit for SPM effects we add a single basis function to the equation representing an inverse delay-time:where t k and t k+1 are the start and end of the sample windows with respect to the primary field turn-off. We then construct our least-squares problem:where R is our time-series data, AEM are our exponential EM basis functions, A IP our IP basis functions, A SPM is our SPM basis function from equation 4, Λ is the smoothing parameter bidiagonal matrix of -λ and λ, and a EM , a IP and a SPM are the EM, IP and SPM amplitudes. AIP effects in AEM were accurately modelled with AIP as well as AEM basis functions as described by Kratzer and Macnae (2012). In the current work, we found that several modifications were necessary to the AIP fitting process, in order to reliably detect SPM. For example, we found that if any significant weight was given to minimising errors in fitting the early channels (<0.5ms delay) in VTEM, the SPM basis function was not used, and so these early channels were not used in the processing. We have found that normalisation of the EM decay basis functions is necessary to prevent the large dynamic range of data leading to unstable fitting. This process may have the effect of allowing unstable fits of the very long and very short decays by the fitting algorithm. In order to reduce the incentive of the least-squares fitting algorithm to fit very long EM time constants to SPM signals, we also normalise the EM smoothing matrix Λ (defined above) (along with the EM basis functions). Figure 1 shows an e...