Canonical Correlation Analysis (CCA) is used to extract the relationships between the predictor (e.g., gridded observed Dec-Mar precipitation) and predictand (e.g, Apr-Sep river flow), with Empirical Orthogonal Function filtering applied prior to the CCA. A clear relationship between gridded regional winter precipitation and spring/summer river flows is obtained, with correlations exceeding 0.6 for several of the stations. This relationship can also be obtained using operationally available NCEP/NCAR Climate Data Assimilation System precipitation. The CCA results are regressed to large-scale wind and Sea Surface Temperature (SST) anomalies, to examine the relationship to large-scale climate variability. The winds show changes in the intensity of the westerly flow that impinges on the mountains of the region, consistent with the precipitation anomalies. These changes in regional winds and precipitation are associated with equatorial Pacific SST anomalies. The SST anomalies are similar to the El Nino pattern, but with greater strength in the central Pacific relative to the eastern Pacific: an SST pattern previously shown to affect the Central Asia climate. Moreover, this connection to Pacific SSTs suggests the potential to increase the lead of the river flow prediction even further, and the skill levels for forecasting spring and summer river flows based on SSTs in the previous autumn are also examined.