Columbia River Streamflow Forecasting Based on ENSO and PDO Climate Signals

Abstract

A simple method has been devised to incorporate the El Niño Southern Oscillation (ENSO) and Pacific Decadal Oscillation (PDO) climate signals into the well-known extended streamflow prediction forecasting approach. Forecasts of ENSO are currently available up to a year or more in advance, which facilitates forecasting of the streamflow response to this climate signal at interannual forecast lead times. The biomodal phase of the PDO can be identified in real time using a combination of assumed persistence of the existing phase and the tracking of extreme events to identify transitions. The technique makes use of a gridded meteorological data set to drive a macroscale hydrology model at 1° spatial resolution over the Columbia River Basin above The Dalles. A streamflow forecast ensemble is created by resampling from the historical meteorological data according to six predefined PDO/ENSO categories. Given a forecast of the ENSO climate signal for the coming water year and the existing phase of the PDO, these meteorological ensembles are then used to drive the hydrology model based on the initial soil and snow conditions as of the forecast date. To evaluate the technique, a retrospective forecast of the historic record was prepared (1989–1998), using October–September as the forecast period, as well as an ensemble forecast for water years 1999 and 2000 that were prepared on June 1, 1998 and May 10, 1999, respectively. The results demonstrate the increase in lead time and forecast specificity over climatology that can be achieved by using PDO and ENSO climate information to condition the forecast ensembles.