Attribution (1)More and more I hear the following – there is no doubt that the Earth is warming, but there is still not agreement about whether or not humans are the cause. This introduces the problem of attribution; in general, if we have a set of observations of the Earth and we observe a trend, how can we decide whether or not the trend is human caused or associated with variability that is natural? Natural? I will avoid the philosophical arguments of man being part of nature ... natural changes would be those that occur in the absence of things that we, collectively, humans have done.
Climate change is not the first big environmental problem where we have faced the problem of attribution. Two that are familiar are acid rain and ozone depletion. In the case of acid rain there were measurements of the increasing acidity of precipitation and lakes. The source of the increasing acidity was attributed, in part, to the use of high-sulfur coal at electrical power plants. Compared with the global warming due to increasing greenhouse gases, the acid rain problem is regionally localized. This is because the sulfur release and the conversion to sulfuric acid and its rainout takes place in a small number of days. Models, which are an important part of attribution, were able to trace the acid rain, more or less directly, to regional locations of emission. Observations were able to distinguish different types of coal based on chemical fingerprints. And, in the case of acid rain, we are able to measure a reduction of acid rain when low sulfur coal is used to generate electricity.
The ozone depletion problem is more nearly a direct analogue to climate change. Most of the chemicals that are initiators of human-caused ozone depletion are long-lived gases, chlorofluorocarbons, that become well mixed in the atmosphere. They are emitted into the atmosphere as a result of human commerce and industry; hence, their emission is linked to the more affluent nations of the world.
How then is any observed depletion in ozone linked to the industrial chemicals?
The persistent problem that is faced by atmospheric scientists is that they do not often have the ability to do experiments. In the acid rain problem there was some ability for experimentation because the emission of sulfur compounds and the acid rain were localized and the time scales short. In the ozone problem we have the atmosphere distinctly absent of ozone-depleting chlorofluorocarbons prior to some date followed by measured increases. We have a measure of the natural ozone variability prior to the release of the chemicals. Or do we? Can we state that the variability that has been observed does, in fact, represent all natural variability? If we start to observe a change of ozone that is outside of the realm of natural variability after the release of the chlorofluorocarbons can we say that it was caused by chlorofluorocarbons? Or, is it some unobserved mode of natural variability?
One way this attribution problem is approached is through the use of models. The great strength of models, and some would argue a weakness, is that the modeler can prescribe exactly the “world” of the model. Hence in the case of ozone depletion, two model experiments are easy to conceive. In one experiment the “natural” atmosphere is run without the chlorofluorocarbons. In the other the chlorofluorocarbons are added to experiment. This then provides two expectations, one with only natural causes of ozone variability and one with the industrial chemical-based depletion added on top on the natural variability.
Models are a critical part of the attribution process. Observations are also a critical part. First the model of the natural environment must be evaluated through comparison with the observations. If we do not determine that the model represents natural variability with some accuracy, then we have no foundation for attribution or prediction. We also use models in prediction experiments to determine the ability or inability for the model to make predictions. Then we take the observations of the environment in the presence of the human-made perturbation. For the case of ozone, this would be in the presence of the chlorofluorocarbons. We compare these observations with the model predictions and determine, first, if there is a formally significant difference between the observations and what would be expected in the “natural” atmosphere. At this point, the “natural” atmosphere is defined by a model prediction.
I do not understand those who maintain that climate models have not been compared with and evaluated by formal comparison with observations. Billions, literally billions, of observations have been used in the evaluation of climate models. There are aspects of the atmosphere or the climate that are modeled very well, and there are aspects that are not. The errors in model simulations are used to define the development paths for next generation models and the strategies for new observing systems.
Trying to stay with the ozone problem --- as long as the changes in ozone were small relative to the natural variability it was difficult to determine whether or not there was human-induced ozone depletion. There were arguments exactly parallel to the arguments that we see today with climate change. But in the mid 1980s the ozone hole appeared. It was a shock. The ozone hole proved to be the “smoking gun” for ozone depletion. This led to observing missions which proved without doubt the role of chlorine chemistry in destroying ozone. It also led to the improvement of models, specifically the inclusion of nitric acid and water clouds in stratospheric chemistry models.
This “smoking gun” laid the arguments to rest, with the exception of a few people who searched for possible alternative explanations. Most of these searches ended at dead ends or required processes that are not observable. When one has an observation-supported, testable theory, and an alternative that relies on unobserved processes, then the rationalist is pushed towards the observation-based theory.
That is the introduction. What about climate change? Is there a smoking gun?
r (jet lagged from ECMWF)
WU Ozone Hole FAQ WU Blog on Models and Attribution An ozone picture