Development and performance evaluation of SCS-CN based hybrid model

In this study, a hybrid approach has been used to increase the predictive efficiency of the SCS-CN model. A recently proposed Ajmal model (developed after randomized configuration) that ignored initial abstraction and maximum potential retention has been given the conceptual framework of the SCS-CN model and a new outcome-based hybrid model (Miv) was formulated. A total of 78 watersheds (7817 events) were used for calibration and the remaining 36 watersheds (3967 events) for validation to develop this hybrid model. The numerical value of hybrid model parameters Lc, λ and S were calibrated using calibration dataset and a simple non-linear one-parameter model has been developed. The performance of the Ajmal (Miii) and hybrid model (Miv) was compared with the original SCS-CN method (λ = 0.2 as Mi and λ = 0.05 as Mii). The performance of models was compared by using four statistical error indices i.e. RMSE, NSE, PBIAS, and n(t) and applying ranking and grading system (RGS). The mean RMSE, NSE, PBIAS, and n(t) values were found superior for Miv (5.60 mm, 0.71, 6.97%, 1.15) model followed by Miii (5.98 mm, 0.65, 16.52%, 1.01), Mii (6.27 mm, 0.61, 20%, 0.90) and Mi (6.98 mm, 0.46, 24.2%, 0.72) model for tested watersheds. The hybrid model (Miv) exhibited consistently well performance for all size watersheds. On the basis of the agreement between watershed runoff coefficient (C) and calibrated model parameter (Lc or CN), R2 value was found relatively higher for hybrid model (Miv) than other models. HIGHLIGHTS The Ajmal model, which was tested on South Korean watersheds, has been investigated in a large set of US watersheds having different sizes.; The Ajmal model has been given the conceptual framework of SCS-CN model by merging both Ajmal and SCS-CN models and a hybrid model having three parameters (Lc, λ and S) has been evolved.; The three-parameter model was calibrated and a simplified version of the one-parameter hybrid model has been developed.; The performance of the hybrid model was found superior than other models.;

Abstract licence: CC BY-ND

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.