The water balance climatology of the Mackenzie basin with reference to the 1994/95 water year

This project attempted to develop an accurate baseline climatology for the Mackenzie basin suitable for modelling and other evaluation applications in the Mackenzie GEWEX (Global Energy and Water Cycle Experiment) Study. Particular attention and effort were applied to adjusting the observed station precipitation data in order to minimize systematic measurement errors. Various geostatistical methods were applied to interpolate the station data onto a uniform grid. This resulted in datasets of 50-km grid square values for mean monthly temperature and monthly total precipitation for the years 1950 to 1997 inclusive. An independent estimate of basin evapotranspiration was derived using an empirical model (Morton, 1983) to generate similar grid square values of monthly total evapotranspiration for the years 1953 to 1996 inclusive. From the grid square datasets, we computed the mean annual basin temperature (T) to be -3.4degreesC, the mean annual basin precipitation (P) to be 421 mm and the mean annual basin evapotranspiration (E) to be 277 mm. A simple water balance was applied to test the consistency of the P and E fields with the observed basin discharge (Q). For the 24-year period 1972 to 1995, the mean annual residual (P-E-Q) for the water balance was -28.4 mm. This residual is a combination of errors in the three water balance components and the assumption of zero annual storage. It is within the estimation errors associated with the measurement and analysis methods used. Further analysis on the correlation of Q with the total basin net surface moisture supply (P-E) showed that P and (P-E) are most strongly correlated with Q with a 3-month lag, i.e., a discharge water-year of October-September corresponds best with a (P-E) year of July-June. In examining the seasonal correlation of T and E with Q, we found that summer T was significantly correlated with annual Q but there was no significant correlation between any seasonal E or annual Q. This suggests that although the Morton model estimates of E provided a reasonable magnitude for the long-term annual basin water balance, it cannot be considered reliable for year-to-year or shorter-term estimates of the basin evapotranspiration. In examining the 1994/95 water year, it was found to be the lowest discharge year (October-September) in the observed record. This is consistent with the 3-month lagged climate data year (July-June) which for 1994/95 has the largest (P-E) anomaly; it is also the driest year since 1950 having the warmest summer months on record and the third lowest precipitation on record. To demonstrate a practical application of the climate datasets generated by this study, we constructed a multilinear regression model for annual (October-September) discharge based on annual (July-June) basin precipitation and summer temperature. The resulting linear model estimated the annual discharge remarkably well, explaining 61% of the variance for this period. Using this model we were able to reconstruct an estimate of annual discharge back to 1950.