Unlocking the potential of stochastic simulation through Bluecat: Enhancing runoff predictions in arid and high‐altitude regions

Abstract

Runoff prediction is crucial for effective water resource management and risk mitigation. However, predicting these catchment responses is challenging due to their unique characteristics and the randomness of hydrological processes. This manuscript explores two different types of modelling frameworks (deterministic and stochastic) and aims to answer questions regarding the reliance on stochastic simulation based on deterministic simulations, the suitability of simple deterministic models, and the influence of catchment characteristics on the results. A simple deterministic rainfall‐runoff model (with only one model parameter) was used to feed the Brisk Local Uncertainty Estimator for Hydrological Simulations and Predictions (Bluecat) framework, exploring the whole range of values of the model parameter. Our findings showed that Bluecat enhanced the Kling‐Gupta Efficiency (KGE) outcomes in arid and semi‐arid regions as well as high‐altitude catchments. Additionally, using the mean of the confidence band of the stochastic simulation as the simulated discharge, rather than the median, resulted in improved KGE values for all catchments. Hysteresis between S‐KGE (stochastic KGE) and D‐KGE (deterministic KGE) was observed, indicating a non‐monotonic relationship between the two variables, and therefore, S‐KGE optimisation can be achieved even when D‐KGE is not optimal. Bluecat showed exceptional performance extended to arid and semi‐arid regions, as well as high‐altitude areas, making it a promising alternative for rainfall‐runoff simulations in these challenging locations.