The time series of annually averaged global surface temperature anomalies f
or the years 1856-1998 is studied through nonlinear time series analysis wi
th the aim of estimating the predictability time. Detection of chaotic beha
viour in the data indicates that there is some internal structure in the da
ta, the data may be considered to be governed by a deterministic process an
d some predictability is expected. Several tests are performed on the serie
s, with results indicating possible chaotic behaviour. (C) 2001 Elsevier Sc
ience B.V. All rights reserved.