Non-parametric inference on trends in temperature data

Abstract

This paper contributes to the existing literature on the topic of trend estimation in temperature series by applying a non-standard estimation procedure to data from cities all over Europe. It seems that statements about the important topic of global warming are oftentimes based on linear trend estimation, where a straight line is fitted to the temperature data. It is argued in this paper that this approach of linear trend estimation is too restrictive in the case of temperature data and that a different approach, allowing for a more flexible form of the trend, yields more accurate results. For instance, regressing the European temperature data on a linear trend provides evidence of no global warming for all series under consideration, whereas the non-parametric approach advocated in this paper shows a recent upward trend for all series underlining the importance of taking the complexity of temperature data into account.