Probabilistic/ensemble forecasting: a case study using hydrologic response distributions associated with El Niño/Southern Oscillation (ENSO)

Abstract

Due to the non-linear processes and interactions of the hydroclimatic system, a given hydroclimatic event such as the El Niño/Southern Oscillation (ENSO) can lead to a range of possible hydrologic system responses described by a probability distribution. This probability distribution changes in space and time reflecting the non-stationary behavior of the hydroclimatic system. An initial approach in quantifying the evolving probability distributions of hydrologic system response utilizes a physically based hemispheric hydrologic model, PBHHM, that incorporates the salient physics of the hydroclimatic system for the midlatitudes of the Northern Hemisphere. The state variables of the model include atmospheric temperature, atmospheric water content, quasi-geostrophic potential vorticity, land hydrologic water storage, and land/sea surface temperature. The model is structured in such a way that characteristics (e.g. sea surface temperature, geopotential anomalies, etc.) of a hydroclimatic event such as ENSO can be incorporated into the model as a forcing event. The hydrologic system response probability distribution is quantified, via the land hydrologic water storage state variable. As a case study, the hydrologic system response probability distributions of the western continental United States to both the El Niño and La Niña phases of ENSO have been simulated. One hundred realizations were run for each phase using random initial conditions for the state variables in order to reflect differing hydroclimatic conditions during the initiation and evolution of the forcing event. The probability distributions of hydrologic system response and their evolution in space and time are described using relative frequency histograms, cumulative distribution functions, and contour plots of frequency histogram categories. Simulation results of the hydrologic system response probability distribution associated with each phase of the ENSO phenomenon are presented which show a distinct response that varies in space and time. The influence of the number of realizations upon these distributions will be discussed along with a means of incorporating the distributions into a water resources planning scheme.