The spatial variability of soil moisture makes it difficult to represent watershed-scale soil moisture using traditional point-scale soil moisture sensors. In the temporal stability method, the spatial pattern of soil moisture is assumed to persist with time. Hence, measurements at a representative point can be used to represent the mean soil moisture. We investigated the factors that determine temporal stability and attempted to locate these representative points. Long-term simulated high-resolution soil moisture data, at a watershed scale, were used. Results showed that locations with a dominant vegetation cover and a low topographic wetness index (TI) can provide reasonable mean soil moisture estimates. Using vegetation cover and TI information, we minimized the number of the sampling locations needed for identifying the best estimate of the true watershed-scale mean. The sampling period duration is also a key factor. Using random combination tests, the minimum number of required sampling points and shortest sampling time were estimated. When 10 sampling points were used, a sampling period of 10 mo was required for accurately determining the representative point. Our study results will help apply the temporal stability method to the estimation of areal soil moisture and the calibration and validation of remote sensing data.