The NRCS abstraction method is based on two assumptions. The first is
that the ratio of actual water retention after ponding to maximum
potential retention after ponding is equal to the ratio of actual
surface runoff to potential surface runoff. The second assumption is
that the initial abstraction for the watershed is twenty percent of the
maximum potential retention. This study shows that both assumptions
violate continuity principles and proposes a modification that renders
an elementary relationship accounting for all abstraction forms by
dividing them into a variable and constant components. Consequently, the
surface runoff computation becomes dependent on the soil initial
moisture content and implicitly influenced by the initial abstraction,
while retaining the advantage of the subjective selection of curve
number from extensive database from which the NRCS method has gained
popularity. A new time of concentration model is also proposed to extend
the computation for flood hydrograph generation.