Understanding and predicting surface movement is important both technically and for social reasons. The shallow processes contributing to subsidence include construction works, peat oxidation, clay compaction, and groundwater withdrawal; deep causes are hydrocarbon and salt production. We describe an inversion procedure we have devised to disentangle the deep and shallow causes of surface movement. It employs a Bayesian inversion scheme, using forward models and other ‘a priori’ information about shallow and deep compaction. Parameter estimation thus takes place at two different depths, thereby disentangling the deep and shallow compaction processes responsible for surface movement. The uncertainty in the surface measurements and ‘a priori’ estimates is naturally incorporated. Furthermore, spatial and temporal correlations can be taken into account through inclusion of the covariance matrix. The inversion scheme is demonstrated for two synthetic cases. The first combines a compacting gas field and a compacting shallow peat layer. We demonstrate that assumptions on the shape of the subsidence bowl are not necessary. We also show how neglecting either deep or shallow causes of subsidence can produce spurious results. The advantage of using the ‘a priori’ estimates of the compaction and the covariance matrix obtained by Monte Carlo simulations is demonstrated with a second synthetic example involving two polders and different depths of their water table. A robust solution is obtained for each polder unit, while a simpler (and faster) ‘a priori’ estimate based on the expected average clay thickness fails to reproduce the actual compaction. Monte Carlo simulations can also be applied to compaction in depleting gas reservoirs. Information on spatial correlations is often available, even when the absolute values of the ‘a priori’ compaction data are quite uncertain. Explicitly incorporating such ‘a priori’ known spatial correlations improves the result significantly.