This paper examines the scale dependence of surface climate interannual variability over six regions of the world using a gridded observed dataset of precipitation and surface air temperature for the period 1961--1990. The scales analysed cover the range of 0.5 to 3.5 degrees in longitude/latitude. The interannual variability of precipitation exhibits a strong sensitivity to spatial scale, especially at the finer scales analysed, while the interannual variability of temperature shows a weak sensitivity to spatial scale. The scale sensitivity of interannual variability is maximum in warm climate regimes, such as over tropical and sub-tropical regions and in the summer season. These results have important implications for climate predictability (e.g., at seasonal and longer temporal scales) and for the identification of climatic trends and changes. They indicate that the prediction and detection of precipitation changes becomes a substantially more difficult task as the scale of interest is increasingly refined. Conversely, for temperature, since the variability is characterized by broader spatial scales, scale refinement is a less important factor.