Publication Abstract

The Natural History of Time Series

R.R. Dickson

In a volume that describes the collection, analysis, and interpretation of the key time series by which environmental variation and its effects are understood-all of them screened, calibrated, and corrected with considerable thought and care-it is perhaps appropriate to remind ourselves that the business of achieving a homogeneous series is neither easy nor obvious. When collected over decades, time series are almost bound to be affected by subtle changes in the site and method, which are important to environmental change only insofar as they obscure it; and even in the case of shorter series, records may still suffer from instrumental errors and malfunctions of greater or lesser subtlety.

The reason we cannot simply screen these errors out is usually twofold: either the errors and assumptions build so insidiously that we do not know they are there; or, if we do spot them, we are uncertain as to whether an unexpected result looks unusual because it is wrong or because a variable has merely changed outside the range of our past experience.

The best we can do is to pass on the sense that errors are possible on a wide range of fronts and to illustrate this point with a sufficiently broad selection of cases to show how. In doing so, we concentrate and therefore exaggerate the problems; but if it seems quirky to use a "natural history" analogue to unify a rather disparate set of cases, the important point is that all the examples described are real, and, although these specific examples may not recur, their generic equivalents almost certainly will.

Reference:

R.R. Dickson, 1995. pp 70-98 In: T.M. Powell and J. H. Steele, (Eds.) Ecological Time-Series. Chapman and Hall, NY., 491 pp.