Investigating the probability of detection of typical cavity shapes through modelling and comparison of geophysical techniques

With a growing need for housing in the U.K., the government has proposed increased development of brownfield sites. However, old mine workings and natural cavities represent a potential hazard before, during and after construction on such sites, and add further complication to subsurface parameters. Cavities are hence a limitation to certain redevelopment and their detection is an ever important consideration. The current standard technique for cavity detection is a borehole grid, which is intrusive, non-continuous, slow and expensive. A new robust investigation standard in the detection of cavities is sought and geophysical techniques offer an attractive alternative. Geophysical techniques have previously been utilised successfully in the detection of cavities in various geologies, but still has an uncertain reputation in the engineering industry. Engineers are unsure of the techniques and are inclined to rely on well known techniques than utilise new technologies. Bad experiences with geophysics are commonly due to the indiscriminate choice of particular techniques. It is imperative that a geophysical survey is designed with the specific site and target in mind at all times, and the ability and judgement to rule out some, or all, techniques. To this author's knowledge no comparative software exists to aid technique choice. Also, previous modelling software limit the shapes of bodies and hence typical cavity shapes are not represented. Here, we introduce 3D modelling software (Matlab) which computes and compares the response to various cavity targets from a range of techniques (gravity, gravity gradient, magnetic, magnetic gradient and GPR). Typical near surface cavity shapes are modelled including shafts, bellpits, various lining and capping materials, and migrating voids. The probability of cavity detection is assessed in typical subsurface and noise conditions across a range of survey parameters. Techniques can be compared and the limits of detection distance assessed. The density of survey points required to achieve a required probability of detection can be calculated. The software aids discriminate choice of technique, improves survey design, and increases the likelihood of survey success; all factors sought in the engineering industry. As a simple example, the response from magnetometry, gravimetry, and gravity gradient techniques above an example 3m deep, 1m cube air cavity in limestone across a 15m grid was calculated. The maximum responses above the cavity are small (amplitudes of 0.018nT, 0.0013mGal, 8.3eotvos respectively), but at typical site noise levels the detection reliability is over 50% for the gradient gravity method on a single survey line. Increasing the number of survey points across the site increases the reliability of detection of the anomaly by the addition of probabilities. We can calculate the probability of detection at different profile spacings to assess the best possible survey design. At 1m spacing the overall probability of by the gradient gravity method is over 90%, and over 60% for magnetometry (at 3m spacing the probability drops to 32%). The use of modelling in near surface surveys is a useful tool to assess the feasibility of a range of techniques to detect subtle signals. Future work will integrate this work with borehole measured parameters.