Abstract:

Because of airborne time-domain electromagnetic measurement data is large and the CDI imaging consumes plenty of time, it is still difficult to carry on the three-dimensional inversion. Facing with these problems, we propose multiple linear regression method to identify the conductivity and depth of anomalies. First, we use three-dimensional time-domain finite difference method to calculate the electromagnetic response of anomaly on multiple sets of conductivity and depth. Second, we extract seven attributes by analyzing three-dimensional anomaly profile curve. Third, we obtain the stepwise regression equation for the conductivity and depth of three-dimensional anomalies. By outliers’ rejection, collinearity diagnostics and significance testing, we get the optimal regression equations at last. According to the model analysis and accuracy analysis on regression equation, the optimal regression suggests as follows: when sampling time of profile curve is within the scope of 0.2-1.5 ms, the relative error of conductivity is less than 11.95% and the sum of squared for error (SSE) is 4 653.724. With the same sampling time, the relative error of depth is less than 13.19% and SSE is 10 873.645. Therefore, stepwise regression is entirely feasible to estimate conductivity and depth of anomalies. It is an efficient, accurate and rapid method, which lay a foundation for HEM flight measurement of mass data of rapid detection and identification in the open air.

Key words: airborne time-domain electromagnetic, three-dimensional, multiple linear regression, conductivity, depth