Introduction:

In the context of South Korea’s recurring drought challenges, this study explores season-ahead predictability of
hydrological drought, focusing on Bayesian autoregressive modeling incorporating stochastic volatility. To better
understand uncertainty, our approach integrates Bayesian inference and stochastic volatility, with comparisons
of conventional autoregressive (AR), autoregressive with exogenous variables (ARX), and autoregressive exog-enous with stochastic volatility (ARXSV) models. Our study explores the integration of climate-based information
with statistical modeling, presenting a comprehensive framework for water resource management. We aim to
predict June–July–August (JJA) streamflow and to assess the influence of large-scale climate predictors, such as
the sea surface temperature and sea level pressure associated with the North Atlantic Oscillation. The method-ology encompasses exploratory data analyses, the use of hydrological drought indices, and considerations related
to dam operation. The research provides tailored season-ahead streamflow forecasts specifically for the critical
monsoon season JJA, where hydrometeorological variability is paramount. By incorporating climate information
within a Bayesian inference framework and stochastic volatility, we show that the proposed approach out-performs conventional autoregressive models, providing decision-makers with an efficient tool to address climate
risk and support dam operation planning, ultimately contributing to improved water-resources management.

pdf : Fabian et al.(2026)