Extensive quality control was performed on the observed climate data. Daily records of precipitation, and maximum and minimum temperature were examined for accuracy in geographic position and uncertainty in the data due to missing and estimated values, trace precipitation amounts, and precipitation accumulation over multiple days. These efforts substantially increased the number of available good quality station-days.
The interpolation was performed in automatic and optimised fashion by minimising the generalised cross validation. The fitted tri-variate smoothing splines incorporated a spatially varying dependence on elevation and were able to adapt automatically to the large variation in station density over Canada. Mean absolute errors in daily maximum and minimum temperature for withheld data averaged over all years were 1.1 ºC and 1.6 ºC respectively. Daily temperature extremes were also well matched.
Daily precipitation is particularly challenging due to often short correlation length scales, the prevalence of zeros, the large positive skew in positive precipitation data and measurement errors of snow in particular. A two-stage approach was applied in which precipitation occurrence was first estimated and then used in conjunction with a surface of positive precipitation values to arrive at the final daily estimates. For withheld data the occurrence/non-occurrence of daily precipitation was correctly predicted 83% of the time and mean percent absolute errors in seasonal and annual precipitation totals were 14% and 9% respectively.
The withheld data tests agreed with the statistical diagnostics of the final models that used all available data. The final models can be used with confidence in applications that depend on daily temperature and accumulated seasonal and annual precipitation. They should be used with care in applications that depend more critically on daily precipitation extremes.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner