Evaluation of interpolated daily temperature data for high elevation areas in New Zealand
Accurate estimates of daily maximum and minimum temperature at locations where there are no actual observations are extremely useful for many purposes. The daily gridded Virtual Climate Station (VCS) estimates of these two variables (plus nine others), which are based on a thin plate smoothing spline model, have become a well-used dataset. However, there is some concern about the usefulness of the VCS temperature estimates in high elevations due to the possibility that the values are often too high (warm). This study presents two alternate daily temperature interpolation methods, both based on the use of fixed lapse rates, and evaluates the accuracy of all three methods against two independent datasets; the first a set of mostly low elevation stations and the second a single high elevation site. Results of the comparison show that while the choice of interpolation method makes little difference in low elevation areas, the ‘Norton’ fixed lapse rate method is clearly the most accurate for the high
elevation site. This study has confirmed previous concerns about the significant warm bias (mostly in summer) of the operational VCS daily temperature estimates in high elevation locations. It also presents an improved methodology, which reduces this bias and is unlikely to have any negative impact on the accuracy of temperature estimates in low elevation locations. As a result, the operational VCS temperature interpolation methodology will be changed forthwith.