Observed climate change and variability (for definitions, see the IPCC Glossary, Appendix I) are considered in this chapter by addressing seven commonly asked questions related to the detection of climate change and sensitivity of the climate to anthropogenic activity. The questions are:
How much is the world warming?
Is the recent warming unusual?
How rapidly did climate change in the distant past?
Have precipitation and atmospheric moisture changed?
Are the atmospheric/oceanic circulations changing?
Has climate variability, or have climate extremes, changed?
Are the observed trends internally consistent?
This chapter emphasises change against a background of variability. The certainty of conclusions that can be drawn about climate from observations depends critically on the availability of accurate, complete and consistent series of observations. For many variables important in documenting, detecting, and attributing climate change, Karl et al. (1995a) demonstrate that the data are still not good enough for really firm conclusions to be reached, as noted in the IPCC WGI Second Assessment Report (IPCC, 1996) (hereafter SAR). This especially applies to global trends in variables that have large regional variations, such as precipitation, whereas conclusions about temperature changes are often considerably more firmly based. The recently designated Global Climate Observing System (GCOS) upper air network (Wallis, 1998) and a GCOS surface network (Peterson et al., 1997), maintained and reporting to higher standards, may have had a limited positive impact on the quality and availability of some of our results. New data sets e.g., on surface humidity, sea-ice thickness and sub-surface ocean temperature, have widened the range of conclusions than can be drawn since the SAR, albeit tentatively. However, a wider range of analytical techniques and tests of the data have increased our confidence in areas such as surface temperature changes.
Throughout the chapter we try to consistently indicate the degree of our confidence in trends and other results. Sometimes we provide quantitative estimates of uncertainty, as far as possible the value of twice the standard error, or we estimate statistical significance at the 0.05 (5%) level. This is the appropriate terminology and implies that what we see is very unusual, given the null hypothesis. We use the word "trend" to designate a generally progressive change in the level of a variable. Where numerical values are given, they are equivalent linear trends, though more complex changes in the variable will often be clear from the description. We use the word "consistent" to imply similarity between results or data sets that are expected to be related on physical grounds. Where this is not possible, we use the following words to indicate judgmental estimates of confidence: virtually certain (>99% chance that a result is true); very likely (90% but 99% chance); likely (>66% but <90% chance); medium likelihood (>33% but 66% chance), unlikely (>10% but 33% chance); very unlikely (1% but 10% chance) and exceptionally unlikely (<1% chance).
Other reports in this collection