The Chevron-operated Jansz field, discovered in 2000 with the Jansz-1 well, is situated in the North West Shelf of Australia at a distance of 250km off the coast. The water depths range from 1.1-1.4km with lowest known gas in the field at 2970 mSS TVD at the Io-2 well. The closure of the Upper Jurassic age gas-filled trap has an aerial extent of more than 2000km2 and a thickness of up to 65m.
A narrow-azimuth towed-streamer (NATS) seismic survey was shot in 2004 and the XWITM proof-of-concept subset consists of 9 sail lines taken from the core area of the field. The velocity cube with dimensions 26km (IL direction) by 2.8km (XL direction) by 4km (depth) contains the main updip wells Jansz-3 and Io-1.
The study input data consists of 5186 shots minimally processed beyond bandpass filtering and is utilised within XWITM to iteratively optimise a velocity model to go from a smooth to a highly-resolved fully-quantitative characterisation of the subsurface.
XWITM final velocity model
This is the final FWI after AWITM velocity model obtained from 36 passes through the shots at frequencies up to 22Hz. Velocity extractions from the model at the well locations highlight the significant change that occurs between start (dashed line) and final (continuous line) results.
The wells intersecting the displayed section are Jansz-3 (left) and Io-1 (right) with the logs given in this integrated field study AAPG paper. In the two wells, the depths of the Base Cretaceous Unconformity (BCU) regional marker are found at 2776 and 2811 mTVDSS and align with a high-velocity contrast at the corresponding locations in the XWITM result. The study states that the BCU corresponds to a cemented package explaining the velocity kick. Impedance logs published in this IPTC paper also confirm the presence of the hard interface at these depths.
Directly underlying the BCU kick, the model optimisation uncovers a thin low-velocity layer which begins to delineate the high-porosity interval of the field. By moving to a finer model grid and predicting data to greater frequencies, the layer is captured with greater detail and resolution.
Model validation: PSDM before and after XWITM
Pre-stack depth-migrated stacks are generated to QC the velocity update. The starting model stack has a clear pull-up at the Cretaceous wedge due to the missing velocity heterogeneity in the overburden section of this model.
The XWITM velocity update results in a downward shift of 120m and 135m at the wells (blue to green dashed line) and the continuous negative amplitude event tracked across the section lands immediately below the BCU markers, in close proximity to the top porosity marker in the two wells.
3D narrow-azimuth streamer survey
The survey is a narrow-azimuth marine streamer acquisition with 8 receiver cables of length 5.5 km. With the limited offsets, the model updates need to penetrate below the reach of the diving waves for broadband velocity recovery at the target zone.
The raw shot gathers that are input into the process have all free-surface effects retained. No pre-processing is applied to the traces beyond bandpass filtering for progressively widening the frequency range through blocks of iterations.
Model validation: predicted and observed data alignment
The panels show predicted data on the left and the corresponding observed data being matched on the right for a randomly selected shot gather. The data is inverted at widening bandwidths with both diving wave and pre-critical reflection energy converging through cost function reduction.
XWITM model evolution
The successful XWITM model evolution employs the AWITM objective function for the long to intermediate length scale updates before switching to the FWI objective function for final refinements to the detail.
FWI alone applied from the same starting model and same lowest frequency range leads to waveform inversion misconvergence (bottom right panel).
Try S-Cube Cloud with your Seismic Data
AWS-hosted Virtual Private Cloud for industry-leading Full Waveform Inversion.
Launch a cloud-optimised XWITM cluster in minutes - no need for a supercomputer.