Abstract Choice of a predictive technique for land subsidence is based on the availability of appropriate field data. If only the depth and thickness of compressible beds can be estimated, a simple hand calculation is available as a predictive technique and for many purposes is adequate. An example of such a technique uses Schatz, Kasameyer, and Cheney's depth-porosity model. Their depth-porosity model for reservoir compaction is modified in the present paper and is combined with a modified form of Geertsma's nucleus of strain model to form a single predictive technique. The nucleus of strain model accounts for attenuation of vertical movement through the overburden. Based on this combined model, predictions of ultimate vertical subsidence are made at eleven specified locations. At five of these sites in California, Texas, and New Zealand, these predictions are compared to measured subsidence. Agreement between predicted and observed subsidence is good. If field measurements of water-level fluctuations and the resulting time-dependent compression and expansion of geologic strata are locally available, use of a more refined predictive technique is justified. An example of such a technique uses Tolman and Poland's aquitard-drainage model in conjunction with a field-based method of parameter evaluation (such as Riley's field-based method of estimating specific storage and vertical hydraulic conductivity). Tolman and Poland's conceptual model and Riley's parameter evaluation method have been combined with Helm's one-dimensional finite difference computational scheme to form a powerful time-dependent predictive technique. This technique has been field verified successfully at more than two dozen sites in California and Texas. Use of such a computer code allows prediction of observed time lags in nonlinear compression and expansion of layered sedimentary materials in response to arbitrary sequences of rising and falling water levels.