Notes

^1.

A description of the CSERA can be found in Statistics Canada, 1997.

^2.

HYDEX is a relational database that contains inventory information on the various streamflow, water level, basin area, and sediment stations (both active and discontinued) in Canada. HYDAT is a relational database that contains the actual computed data for the stations listed in HYDEX. These data include: daily and monthly means of flow, water levels and sediment concentrations (Environment Canada, 2007).

^3.

Albers Conic Equal Area projection is well suited for projecting large landmasses with an east-west orientation while at the same time preserving areas accurately. This study used this projection because of Canada's large east-west extent and the need to accurately preserve areas for calculating the values of water yield.

^4.

The bull's-eye effect occurs when certain observations unduly influence the interpolation, resulting in localized spots of extreme value in the surface centered over the observation, which decreases radially with distance forming what looks like a target, or "bull's-eye".

^5.

For a more detailed description of the kriging system along with its matrix notation see Borga and Vizzaccaro, (1997) or Webster and Oliver, (2000).

^6.

The 300 km distance was determined by evaluating several different distances (larger and smaller) and visually assessing the predicted surfaces along their zone divides. Distances smaller than 300 km did not result in a smooth transition between zones while distances larger than 300 km did not correspond well to the semivariogram model built with the observations within the zone (for example, the semivariogram model produced in zone 1 is not suitable for the observations found in zone 4).

^7.

This rule was used because the lag increment can strongly influence the semivariogram, an increment smaller than the minimum distance between the data observations may lead to too few observation pairs, resulting in an erratic semivariogram (Webster and Oliver, 2001).

^8.

Artefacts are defined here as breaks in the continuous surface caused by abrupt changes in values where it becomes visually clear the interpolation does not perform well. This usually occurs when the interpolation is extrapolating to areas beyond the convex hull formed by the observations.