A multivariate technique for estimating New Zealand temperature normals
Multiple regression equations are developed to predict monthly and seasonal minimum, mean and maximum temperature normals in New Zealand. Three predictors (latitude, altitude and distance from the nearest coast) are used in the equations and explain between 66% and 96% of the variance in the temperature normals. The regression equations verify significantly with independent data. A seasonal pattern is apparent in the statistics describing the monthly minimum, mean and maximum temperature regression equations. These seasonal changes in the regression statistics are related to changes in the atmospheric circulation over the New Zealand region at different times of the year. Maximum temperature lapse rates, derived from the regression equations, show little seasonal variation and the annual value of 6.4 ºC per 1000m is close to the typical saturated adiabatic lapse rate for the middle troposphere. Minimum temperature lapse rates are lower (mean of 3.6 ºC per 1000m) and during winter are negative below 300m. This is a result of the formation of inversion layers in valley floors at this time. The regression equations are fully listed in the Appendix and should prove useful to ecologists, foresters and others attempting to interpolate temperature information for areas of New Zealand lacking instrumental climate data.