Abstract
In this paper, I use a number of examples of multiscale modeling in biology to argue that the primary challenge facing these modelers is not how to metaphysically interpret their models, but is instead using various idealizations to bring the available multiscale modeling techniques to bear on the phenomena of interest. This is particularly true when the aim of the biological modelers is to inform policy decisions concerning epidemics, climate change and conservation which involve a wide range of spatial and temporal scales. The ‘best-case scenario’ in these instances of multiscale modeling is when the dominant features of the system can be separated into distinct scales. When this occurs, scientists can effectively model the phenomenon by using modeling techniques designed for those particular scales (and type of processes). One example of this is attempting to use optimization techniques to model tradeoffs between the short-term and long-term adaptations of plants and animals to anthropogenic climate change. In these cases, biologists first model the adaptive strategies of individual plants and animals at shorter temporal scales (e.g. hours and days). They then model what would be adaptive for the overall population at longer temporal scales (e.g. generations). For example, when CO2 levels rise, in the short term, plants reduce their water use by reducing their stomatal conductance. However, at longer time scales, the models predict an increased water use due to increased photosynthetic capabilities, larger leaves, and deeper root depths. Finally, the optimal phenotypic trait (and the one we might want to use interventions to bring about) will be the one that best balances these short-term and long-term benefits. While scale separation is often useful for multiscale modelers, in many cases in the biological (and social) sciences, such a clean separation of scales is not possible. In these cases, biological modelers have begun borrowing various modeling techniques first deployed in physics. In particular modelers in spatial ecology have begun using 7 homogenization techniques to model plants and animals’ migration patterns across heterogeneous landscapes. In these cases, various idealizations are introduced to be able to model the system as a homogenous medium at the largest scale (e.g. the whole ecosystem) while taking into account the influences of variations at smaller scales (e.g. slower movement through mountains than fields). This results in a macroscale equation that encodes key features from smaller scales into its parameters and constants. This highly idealized modeling technique enables biologists to incorporate features from across a wide range of scales in much more computationally effective models that can more easily be used to inform policy decisions about the outcomes of specific interventions. What these two sets of cases show is the relationships between scales, the available modeling approaches, and scientists’ purposes for their models all influence which multiscale modeling technique is best suited to biologists’ goals and which idealizations can be justifiably used in order to deploy those modeling techniques.
