Author(s): E. N. Tsykin
Linked Author(s):
Keywords: No Keywords
Abstract: Rainfall-runoff regression models are generally designed on the basis of multiple linear regress ions deve loped in linear or logar ithmic space. This places restrictions on their st ructure. Consequently, they often cannot be ext rapolated beyond the cal ibrat ion range because of lack of correspondence between the curvatures of the actual process and the model. Various rainfal l-runoff models consist ing of an Intercept and a number of non linear members have been deve loped. Each nonl inear member compr ises rainfall averaged over a selected period and a system of ant ecendent soil moisture indices. A mathematical procedure to fit these models has been descr ibed. Two kinds of models for runoff s imulation are presented. The first uses rainfall excess over evaporat ion as the independent variable and the second. just rainfal. The average R2 of calibrations in the six demonstrated models was 0.96 and the average R~2 of nine extrapolations to periods before and after the calibration was 0.86, even though in some models the max imum runoff used for calibration was less than half of the ext rapolated max imum.
Year: 1985