Empirical Evidence Shows That Curve Numbers Vary from Event to Event

The Natural Resources Conservation Service’s (NRCS) curve number method is a popular approach for estimating discharges from rainfall data. It makes use of a dimensionless parameter—called the curve number—that is obtained from soil, antecedent moisture, and land use conditions. This parameter is traditionally taken to be a constant watershed characteristic; however, in this paper, the validity of this assumption is questioned by generating event curve numbers from a data set of 1,393 rainfall-runoff events from 61 watersheds in Texas, with each watershed having at least 10 rainfall-runoff events. For a given event, the accumulated precipitation was obtained from the rainfall data, and the accumulated depth of runoff was obtained by dividing the volume of direct runoff by the area of the watershed. From the accumulated precipitation and runoff depth, the curve number was obtained using the NRCS’s curve number method. The standard deviations of the curve numbers for each watershed were calculated, and it was found that 24 watersheds had standard deviations of the calculated curve numbers greater than 10; 35 watersheds, between 5 and 10; and only two watersheds had standard deviations of less than five. Moreover, the maximum, minimum, and average standard deviations considering all watersheds were 18, 4, and 10, respectively. The fact that these values are not small indicates that the curve number is not a constant watershed characteristic; rather, it varies from event to event—although the subject of what standard deviation is considered small may be subject to debate.