Routine amplitude‐versus‐offset or amplitude‐versus‐angle (AVA) analysis is founded on the predicted reflectivity of a planar boundary between isotropic homogeneous half‐spaces, yet many real boundaries of interest to hydrocarbon exploration may violate these assumptions. Here, we evaluate consequences of boundary roughness, i.e. small‐scale, sub‐resolution topography, on the reflecting boundary, to find whether sub‐resolution topography can deceive routine AVA analysis and produce misleading hydrocarbon indicators. We use noise‐free synthetic CMP and stacked‐section seismograms, generated with a Kirchhoff‐integral method, for offsets up to 4000 m. The reflecting boundary was at 2000 m depth below a single overburden layer, and comprised isolated mounds or channels (sinusoidal cross‐section, 5–40 m high and 100–750 m wide), and more complex roughness (simulated with seven 1.5–2 km long bathymetric profiles across modern‐day river‐beds, with dune and bar features up to approximately 20 m or 20 ms TWT vertical relief, and 10–100 m or more lateral extent). Flat‐boundary responses were taken as the ‘reference’ against which to compare waveforms and amplitudes through semblance analysis, AVA intercept A and slope B cross‐plots, and inferred Poisson's ratios. Physical properties of the media were based on shales and brine‐ or oil‐sands from an offshore UK oilfield.
For the CMP centred on the topographic feature, isolated mounds and channels produced correlated excursions of A and B from the flat‐boundary response (by approximately 35–200%), simulating the familiar ‘background trend’ but sometimes oblique to it. Inferred Poisson's ratios were around 0.3, but for some channels often fell as low as 0.2, potentially interpretable as gas sand. For complex boundary topographies on a shale/brine‐sand model, AVA parameters were extracted for 29 CMPs, 100 m apart along each of seven profiles. On a cross‐plot, they spanned the flat‐boundary response on a linear trend B = (0.01 ± 0.02) – (1.72 ± 0.09)A, similar to reported real data and a realistic ‘mudrock trend’ but with outliers both above and below it that could be interpreted as ‘real’ weak anomalies. Poisson's ratios were 0.30 ± 0.01, between the expected values for brine‐ and oil‐sand. This suggests that boundary roughness may contribute to observed trends on cross‐plots and possibly small, but potentially laterally extensive, false AVA anomalies may also be induced.