Sunday, 7 January 2018
Exhibit Hall 5 (ACC) (Austin, Texas)
Future streamflow is often inferred by precipitation projections, but this can cause uncertainty due to the difficult nature of parameterizing precipitation in models. There is a need for another method that infers future daily streamflow dynamics and does not rely on precipitation as the means to attain it. This project uses climate modes as a covariate in estimating the parameters of a stochastic streamflow generator. We apply the climate-influenced streamflow analysis framework to a test case on the Lower Susquehanna River. To complete the objective of this project, streamflow generator parameters are determined through maximum likelihood estimation as nonstationary functions of North Atlantic Oscillation (NAO). Parameters’ probability distributions are modeled using systematically varied combinations of nonstationary parameters and forms of NAO index. Models’ performance is compared using the Bayesian Information Criterion (BIC) for models of daily streamflow states persistence and Akaike Information Criterion (AIC) for flood frequency and magnitude. Although many models that include NAO as a covariate are outperformed by the baseline model without external covariates, several models conditioned on lagged-DJFM (Dec.-Mar.) NAO non-stationary models are competitive against the stationary model. Of the well-performing models, the fully non-stationary parameter estimates of the neutral lagged-DJFM NAO models showed positive relationships to flood frequency and dry spell persistence. This modeling framework can be used to investigate the connection between different large-scale climate patterns in the stochastic streamflow generator. This project contributes to the calibration of the streamflow generator which can be applied to other water management decisions outside of the Susquehanna River Basin.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner