Abstract: Airship-based ground-airborne time domain electromagnetic system enjoys high depth of prospecting and spatial resolution, as well as outstanding detection efficiency and easy flight controlling. However, due to low-altitude flying is markedly affected by wind direction, air flow, landform, and the difference of temperature filed around ground surface, the front-fixed receiving coil would cut earth magnetic field, which results from changes of the altitude, track, and gesture of the airship, and causes electromagnetic and moving noises, baseline drift as well. The drawbacks mentioned above could lead to inferior resistivity image formation of electromagnetic data. Consequently, to investigate methods of removing noises of electromagnetic data is of vital importance to inversion explanation. With simultaneously occurred frequency band of valid electromagnetic data and part of the noise, traditional filter technique or noise-removing ways have their own limitations. Therefore, this study proposes a combined de-noising method, which adopts sym8 wavelet basis according to the characteristic of the electromagnetic data, corrects the baseline drift using estimation of high-level approximation by applying wavelet multi-resolution analysis, and suppresses other noises by five-level wavelet decomposition with MiniMaxi threshold and Hard shrinking function by applying wavelet threshold shrinkage theory. The combined de-noising method is validated by adopting the theoretical response and wild measured data based on the Anomaly Loop Model, respectively. The final results confirm that the proposed method eliminates noises effectively, which indicates it is applicable and effectual to ground-airborne time domain electromagnetic data.