Mathematical models can be developed and used for different purposes. Whereas in heuristic modeling model development and use predominantly serves heuristic purposes such as the investigation and better understanding of atmospheric and climate processes, predictive modeling aims at the production of predictive knowledge such as future projections of climate change. These different scopes of heuristic and predictive modeling are important, because they involve different research priorities, practices and strategies. Climate models and data need to be developed and adjusted specifically for this scope.[1]
The first generation of climate modelers comprised predominantly theoretical meteorologists, who were interested in investigating atmospheric processes and improving its physical understanding. They pursued heuristic modeling and did not take an interest in producing predictions of future climate.
The path to predictive modeling was paved by concerns about climate warming due to rising CO2 concentrations. In a time of social unrest and rising environmentalism around 1970, a new generation of climate scientists such as William W. Kellogg and Stephen H. Schneider gave climate modeling a fundamentally new interpretation. NCAR Atmospheric Division Director William Welch Kellogg wrote in 1971 that
"there is the haunting realization that man may be able to change the climate of the planet Earth. This, I believe, is one of the most important questions of our time, and it must certainly rank near the top of the priority list in atmospheric science." (Kellogg 1971, p. 123)
Kellogg and Schneider concluded that climate models should be developed and used in order to simulate projections of future climate. Even though they were aware that climate models were still very simple and had to be improved, they strongly demanded to develop and pursue predictive climate modeling. Schneider described the dilemma of using imperfect models for prediction 1956 in a popular book as follows:
"The real problem is: If we choose to wait for more certainty before actions are initiated, then can our models be improved in time to prevent an irreversible drift toward a future calamity? (…) This dilemma rests, metaphorically, in our need to gaze into a very dirty crystal ball; but the tough judgment to be made here is precisely how long we should clean the glass before acting on what we believe we see inside" (Schneider 1976, p. 149)
[1] The distinction of heuristic and predictive climate modeling is not always sharp. The aim of predictive modeling requires heuristic forms of research for the development and testing of appropriate predictive models. And – like all physical theory – heuristic modeling involves the calculation of simulation results to be compared to observational data, a type of computation which some scientists also refer to as prediction, as well.
References:
Kellogg, WilliamW. 1971. Predicting the climate, in: Matthews, William H., William W. Kellogg, W. D. Robinson (eds.), Man’s Impact on the Climate (Cambridge, MA: MIT Press): 123-132.
Schneider, S.H. 1976. The Genesis Strategy: Climate and Global Survival (New York: Plenum Press).
