Key points are not available for this paper at this time.
AbstractThe spatial variation in monthly temperature normals (1961-90) from southern Norway was studied. Linear regression as well as an approach combining a deterministic and a geostatistical model were applied (residual kriging). The deterministic component describes the large-scale trend in the temperature, and in this study is defined as the vertical temperature gradient. The spatial variability caused by differences in elevation is removed by reducing the temperatures to sea level. The resulting temperature fields are more suited for spatial statistical analysis. These temperatures are closer to fulfilling the assumptions most statistical interpolation methods require, i.e. stationarity and isotropy. The reduced temperatures were interpolated applying kriging. It is shown that residual kriging gives better estimates than kriging on station level temperatures. Residual kriging is also more credible than the linear regression approach, which regionally gives large estimation errors.
Tveito et al. (Wed,) studied this question.