An important parameter in land surface hydrology is the fractional area of a model grid receiving rainfall when rain is observed. Using NEXRAD (Next Generation Weather Radar) hourly rainfall estimates for the southern Plains, we examined the temporal variability in this quantity and tested the important assumptions of the threshold method. The following conclusions are reached. First, seasonal/diurnal variations explain less than 18% of the variance at all spatial scales examined; an efficient method of computing fractional coverage must consider the large event-scale variability. Second, stationarity in conditional distributions of spatial rainfall, a key assumption in the threshold method, is not warranted at the event scale for small grids or at the seasonal/diurnal scale for large grids. Third, biases are introduced by the high correlation between the parameter and the independent variable of the threshold method. Fourth, the spatial conditional mean rain rate is higher than its temporal counterpart for all cases examined; the ergodicity assumption overpredicts the fractional coverage. Two simple methods were proposed to solve the above problems. Results of an intercomparison showed that they can modestly improve the threshold method. ¿ American Geophysical Union 1996