Full-waveform inversion (FWI) has evolved to be the contemporary solution to resolve velocity models in areas of complex structure. Further, wide azimuth, long offset and rich low-frequency seismic data, resulting from broadband seismic acquisition, helps FWI update deeper with better convergence and stability. In this study from the South Mahakam area in offshore Indonesia, multiple layers of carbonate exist from shallow to deep with sharp velocity contrast. The target reservoir is down to 3.5 kilometers. However, for the acquired data with narrow azimuths (NAZ), short offsets (3 kilometers) and low signal to noise in the low frequencies, FWI encounters challenges of cycle skipping and unstable updates in the deeper targets that are beyond the diving-wave penetration depth. Time-lag FWI (TLFWI) (Zhang et al., 2018) uses time-shift differences between observed and modeled data as the cost function, and also makes better use of the low-frequency refraction and reflection energy. TLFWI gave good velocity updates in both the shallow and deep regions and, hence, gave an improved deep carbonate image. The anisotropic model is an important factor for the success of any FWI due to the coupling between velocity and anisotropy. In this paper, joint reflection and refraction tomography (Allemand et al., 2017) were applied in order to obtain stable anisotropy models for TLFWI. Following that, TLFWI with both refraction and reflection energy gives sensible velocity updates down to 3.5 kilometers. These updates to the model improve the seismic image and, importantly, reduce the depth uncertainties in this complex geological setting. The cumulative improvements increase interpretation confidence and can reduce future drilling risks. For the seismic processing community, the reprocessing of narrow azimuth, short-offset data with TLFWI, and associated technologies, offers great potential for generating improved and more reliable images from legacy, conventional, acquisition scenarios.
