The purpose of this paper is to outline the principles and use of a method for sampling factors—such as the amount of rain reaching the ground under a forest during a series of storms—which exhibit variations in both space and time. Solutions of two somewhat different sampling problems are discussed. One is designed simply to provide a sample estimate (average or sum) for any desired factor over a period of time. The other, with little loss of efficiency, provides a sample estimate for each of a series of short time units making up the period, as well as an estimate for the period as a whole. It is shown that these sampling methods are likely to supply a substantial gain in efficiency as compared to methods that have been commonly employed.