The analysis of hydroclimatic variability and predictability in Western Canada

An analysis of the role of several large-scale climate anomalies in forcing low-frequency variability in the precipitation and streamflow data of Western Canada was carried out using an ensemble of statistical techniques. Wavelet and wavelet coherence analysis of precipitation data from stations across the region shows that at both regional and local scales, precipitation anomalies exhibit significant interannual oscillations that occurred haphazardly. At many stations, the temporal locations of these haphazard oscillations were different from those of large-scale climate anomalies, leading to weak and inconsistent relationships with climate indices. On a seasonal basis, winter precipitation shows modest correlations at 0- to 3-season lead times with the El Niño/Southern Oscillation (ENSO) and the Pacific Decadal Oscillation (PDO) indices. Although there were fewer stations with statistically significant interdecadal oscillations, the relationship with interdecadal climate anomalies was much more consistent than at the interannual scale.;Based on the understanding derived from the diagnostic analysis, two seasonal ensemble streamflow forecasting models were developed and applied to the Bow and Castle rivers, both located in the headwaters of the South Saskatchewan River basin of Alberta. The first model is a statistical regression model based on the robust M-estimator and the nearest neighbor resampling algorithm, while the second one is based on the ensemble streamflow prediction (ESP) method. Results from the statistical model indicate that forecasts based on climate indices alone possess considerable skill (correlation of 0.65 and up for the Bow River) for forecasts issued early in the season and could thus extend the current forecast lead time by up to two months. In addition, ensemble forecasts were found to possess better skill than deterministic forecasts in terms of economic value because a wide range of forecast users with varying economic costs could potentially benefit from the probabilistic information contained in ensemble forecasts. While the ESP based forecasts also show promising results, they were generally found to be less skillful than the statistical forecasts partly because of hydrologic modeling uncertainties.;Since the influence of ENSO on Western Canada's precipitation appears to be stronger than the other climate anomalies, it was used as a basis to investigate patterns in streamflow response across the region. From statistical significance testing and cluster analysis, Western Canada was zoned into five spatially coherent streamflow response regions. For each of these regions, a detailed correlation analysis was carried out to examine the value of various climate indices for predicting seasonal streamflow anomalies. This analysis showed that some of the response regions show better correlations with ENSO indices while others exhibit higher correlations with PDO or other indices.