“…In the context of stochastic modeling of streamflows, a major limitation of the widely used periodic parametric models is their inability to simultaneously reproduce summary statistics and dependence structure at different temporal levels. To circumvent this, linear disaggregation models were developed [Harms and Campbell, 1967;Valencia and Schaake, 1973;Mejia and Rousselle, 1976;Lane, 1979;Stedinger, 1988, 1990;Santos and Salas, 1992]. However, these models are not parsimonious, and in addition they require empirical adjustments in order to restore summability of the disaggregated flows to the aggregate flows, in the event of normalizing transformations being applied.…”