Measurements of the radioactive content of environmental samples are potentially very costly, especially when these are made ex situ in a laboratory. A less expensive alternative is to acquire in situ measurements in the field. Both measurement types are subject to uncertainties, some of which arise from different sources depending on the measurement method used. Surveys on radioactively contaminated land found that in situ measurements produced results that were as useful in satisfying the typical objectives of such surveys as ex situ measurements. The random component of analytical uncertainty estimated from duplicated in situ measurements was 2-4 times higher than would have been expected from Poisson statistics, however the sampling uncertainty (0-10 %) was found to be much lower than that for ex situ measurements (44-73 %). This resulted from the combined effects of high heterogeneity of the target radionuclide (137Cs) in the ground, and the comparatively large primary sample mass associated with in situ measurements of gamma-emitting radionuclides. A large sampling mass also means that in situ measurements have an advantage in finding small hotspots of activity, although they may not provide sufficient resolution for spatially mapping lateral distributions of contaminants for remediation purposes. The degree of resolution can be readily changed in the field, however, by the simple expedient of changing the detector height. Experiments with an in situ detector close to the ground surface enabled the position of a small hotspot to be determined to within a few centimetres. To evaluate activity concentrations in the soil, assumptions need to be made about the dimensions of the measured sample, and the distributions of activity within it. This requires some information that might be best obtained from ex situ measurements of excavated samples. However, well planned in situ surveys have the potential to significantly reduce the requirement for these expensive laboratory measurements. A new method of optimising the design of in situ surveys has been developed, based on a generic model for predicting the detector response to small particles of activity at different positions relative to the detector. The new mathematical model used by this method compares well with field measurements, and also with predictions made using a commercially available calibration program.