A Random Block-Based Nonparametric Approach for Temporal Disaggregation of Net Basin Supplies in the Lake Champlain–Richelieu River Basin

Stochastic simulations have been widely applied in water-related risk management, particularly for estimating annual Net Basin Supplies (NBS) in the Lake Champlain–Richelieu River (LCRR) basin. Following the unprecedented flood event in 2011, simulated NBS datasets were required to re-evaluate existing flood-protection infrastructure and to support the development of future mitigation strategies within the basin. Because water-resources operation and flood-management planning are typically conducted at monthly or quarter-monthly resolutions, the simulated annual NBS data must be disaggregated to finer temporal scales. In this study, several existing disaggregation approaches were applied to the simulated annual NBS series, with the objective of reproducing the key statistical characteristics associated with the 2011 flood event in the LCRR basin. The 2011 flood was characterized by its persistence over multiple months, indicating that an appropriate disaggregation framework must be able to maintain both interannual dependence and month-to-month temporal relationships in the resulting monthly series. The analysis shows that currently available parametric and nonparametric disaggregation models exhibit clear limitations, particularly in their ability to preserve sufficient temporal dependence. To address these deficiencies, this study proposes a new random block-based nonparametric disaggregation (RB-NPD) model. In addition, the proposed framework is further enhanced by incorporating a Genetic Algorithm–based mixture scheme to improve the representation of lagged correlations. The results demonstrate that the RB-NPD model provides a viable alternative to existing methods, and that its enhanced version is well suited for disaggregating annual NBS data in the LCRR basin.