Detection and Attribution of Climate Change and Its Impacts

Abstract

The concept of change builds upon the assumption that some kind of constancy or repeatability naturally exists in the system of interest, and that change is a negation of such constancy. For example, one may compare some characteristic of temperature (e.g. its average, at a location of interest, regionally, or globally), for two different longer time periods, e.g. 30-year climatological standard normals. When detecting a significant difference in the distribution of temperature between the two periods, one might conclude that this temperature has differed between the two periods. This, in turn, would lead to the conclusion that something in the system has changed. Usually, the nature of the change is of interest. For example, one might observe a trend as a continued change that occurs over time. This trend might be viewed either as a manifestation of a time-dependent deterministic component (possibly with a known underlying mechanism), or simply as a tendency in the statistical properties of the process. Detection is the act of extraction of particular information from a larger stream of information (e.g. determination of presence or absence of a useful signal in telecommunication). It is the process of becoming aware that a change has occurred. The process of detection is germane to the work of any detective attempting to reconstruct a sequence of past events, based on whatever information is available and considered relevant. Detection of change in a time series of observations (e.g. related to climate and its impacts) means demonstrating that a system has changed in some statistical sense, i.e. that an observed change is unusual, significantly different from what can be explained by natural internal variability. Detection itself does not identify a cause for the change. Detectability, i.e. the possibility of detecting a change depends on signal-to-noise ratio, and the relative size of the trend versus any natural variability (amplitude and duration of change). It may not be possible to detect a weak signal amidst a strong natural variability. Usually trends of simple shape (linear, low-order polynomial, piecewise linear, i.e. broken line, exponential, etc.) are considered. Different trend shapes are possible, including steeper trends similar to abrupt step-like changes. There is a continuum of cases and, in practice, the terms “trend” and “change” can be almost interchangeable. One can also speak of trends in a non-parametric, comparative sense; e.g. an increasing