In formulating a response to I the risk of future human influence on the climate, it is important to understand what has happened in the recent past. Has the climate actually been changing? And if so, how do we know if we are causing it? These two simple questions raise some of the most contentious and complicated topics in the whole climate change debate: the detection and attribution of change. To introduce this complex topic, a look at a more familiar problem may be useful.
Suppose you step onto your bathroom scale some morning and look down to find the number is one pound higher than you expected. You then call out, "Honey; I'm gaining weight. I've been eating too much!" What does it take to reach such a conclusion with any degree of confidence?
First, how do you know your weight is actually higher? You may suspect that the accuracy of your bathroom scale is not that great: you paid only $19.95 for it. You have also observed that readings on your health club scale differ from those at home. So there is error in the measurement itself: you can detect your weight increase only with some error. Thus informed, you might say, "There is a 90 percent probability that my weight is between 0.8 and 1.2 pounds higher than the last time I looked!"
Second, does this measurement indicate a change that might have some underlying cause, like eating too much? Here another complexity enters: your body weight may go up and down by amounts comparable to your observed gain this day regardless of what you eat, in response to changes in temperature and humidity or your psychological state--perhaps your boss is on vacation for the month. If the magnitude of this natural variability is similar to the change you are seeing on the bathroom scale, you should be cautious when saying that you have detected something significant. You have a problem of sorting the "signal" of a significant change in your body, perhaps attributable to food intake, from the "noise" of its natural fluctuations. So a still more accurate statement might be, "There is a 70 percent probability that my weight has gone up between 0.8 and 1.2 pounds for some reason other than natural variability."
The third question that arises is, why the apparent change? Weight gain does have a basis in physiology; and this knowledge may be bolstered by past experience and observation of others. But the relationship cannot be stated precisely because it depends on many factors, such as adjustments in metabolism and amount of physical activity, that are poorly understood.
This gives rise to a fourth question, how accurately have you recorded or recalled your food intake, physical activity level, and other factors that you know affect your weight? With this uncertainty about the relationship of food intake to weight, the most that can be said with scientific accuracy about attribution may be something like, "There is a 90 percent chance that at least one-half of the increase shown on this scale is due to my eating too much." Or, where formal analysis is missing, "Honey, the preponderance of the evidence suggests that I'm eating too much."
To summarize, detection of weight gain and its attribution to increased food intake involves four elements:
* An estimate of the weight change and the potential for error in measuring it,
* Knowledge of the natural variability in your body weight,
* An understanding of the mechanisms by which body weight responds to food intake and other factors, and a model of the relationships, and
* A record of food intake and understanding of its accuracy.
Detection and Attribution
Now consider the global climate system. We have only imperfect measurements of how climate has changed over past decades, and our understanding of patterns of natural climate variability is limited. Moreover, our models of the interacting chemistry; physics, and biology of the global system are as yet incomplete, and we do not have an accurate inventory of past changes in factors, human and natural, that could have altered the Earth's radiation balance, the so-called radiative forcing. …