Research: Melanie Moses

Research Courses Publications Postdocs/Students CV (PDF) Links Home

Overview

I study complex biological and information systems, the scaling properties of networks, and the general rules governing the acquisition of energy and information in complex adaptive systems. My focus is on the efficiency of growth and information exchange in biological and computational networks, and how the size and topology of networks determine emergent system behavior. I draw insights, tools and approaches from different disciplines in an effort to find unifying principles in the natural world.

Human Societies

Just as individual organisms construct arteries and veins to distribute energy to cells throughout their bodies, societies construct distribution systems to transport energy and materials to individuals. In ant colonies, social infrastructure includes foraging trails, pheromonal communication systems, and nest structure. In modern human societies, these distribution networks take the form of highways and airline routes, oil pipelines and electric grids, and numerous other physical and virtual systems that enable modern society to function.

I have used metabolic scaling and life history theories as a lens through which to view fertility patterns in contemporary societies (Moses and Brown, 2003). Fertility declines as per capita energy consumption (including consumption of fossil fuels) increases with a scaling exponent of -1/3 as predicted by scaling theory. The pattern extends what is already known about large mammalslarger organisms with high metabolic rates have proportionally less energy to allocate to reproduction. In large, consumptive societies, individuals also appear to have a smaller percentage of their total energy available to allocate to reproduction.  We hypothesize that this reflects that very consumptive societies spend a higher proportion of their energy in transporting, processing and acquiring more energy.

This study of human societies indicates that the nonlinear relationship between resource use and size may extend to social organisms, as well as individual organisms. This suggests that the networks that constitute social infrastructure are characterized by the same diminishing returns as biological networks: larger networks deliver proportionally less energy in both organisms and societies. As societies grow and become more complex, more and more work goes into sustaining the society (acquiring, transporting and defending resources) rather than directly delivering services and resources to the population. This analysis also suggests that societies use fractal resource acquisition networks which mimic the fractal networks in organisms. We are currently studying the scaling properties of road networks and the information networks described below.

Foraging in Ant Colonies

Large ant colonies with large territories face a disadvantage due to the geometry of resource distribution networks that is very similar to large organisms: ants in large colonies spend more time transporting resources from distant locations in the territory. However, large colonies also have an advantage: they use their foraging network to gain information about the location of rich resource patches. Large colonies exploit that information to reduce the time it takes to search for resources and to allow a higher density of foragers in the territory.

I developed a model based on scaling principles to predict how territory area, the density of ants in the territory and the time each ant spends on a foraging trip vary as nonlinear functions of the size of the ant colony. My field research shows that foraging in Pogonomyrmex (desert seed harvesting ants) is consistent with the model predictions. The territory area of a colony is proportional to the colony population raised to the ¾ power, the density of ants in the territory are proportional to population size to the ¼ power, and the transport distances are proportional to population raised to the ¼ power. The practical effect of these relationships is that a 30-fold increase in colony size means a 10-fold increase in territory area and a doubling of density and distances to collect seeds.

The study also highlighted an interesting departure from scaling predictions. Even though larger colonies had greater transport distances, the complete time of a foraging trip is equivalent in the largest and smallest colonies. Once ants in large colonies reach their destination, they find seeds much more quickly than ants in small colonies. Experimental seed additions showed that large colonies locate rich food patches and recruit other workers to collect seeds much faster than small colonies.

Large colonies appear to be less efficient at transporting resources because they gather them from distant locations, but they are more efficient at searching for resources because additional workers gather and communicate information more quickly. The tradeoff between a transportation cost and an information benefit in large social groups may apply to other social systems. I intend to extend a simple agent based model of foraging behavior to provide a more general test of how the number of cooperating agents affects the rate at which those agents gather information about their environment.

This work is described in Chapter 4 of my dissertation.

Scaling Theory for Information Networks

Computer networks and biological organisms are both complex systems. In both cases higher-level processes, such as the rate at which an organism consumes energy or the rate at which a computer processes information, emerge from the integrated operation of many smaller components. I explore how properties of information networks scale as a function of their size in collaboration with Stephanie Forrest and James Brown. We use metabolic scaling as a framework to examine similarities between energy processing in biological systems and information processing in computer systems. We are studying the scaling properties of computer chips which contain hundreds of millions of transistors networked in a few square millimeters of surface area, and the Internet which connects hundreds of millions of hosts and spans the 5 x107 km2 surface of the earth.

Scaling in Immunology and Epidemic Spread

Epidemiological models, such as the Susceptible-Infected-Recovered (SIR) model, must be parameterized by a variety of demographic and immunological rates (e.g. rates of birth, death, infection probability, and recovery). While many demographic rates and some rates associated with pathogenesis decrease with body size (Cable, Enquist and Moses, 2007), our analyses (work in prep with Soumya Banerjee) suggest that immunological response times are largely independent of body size. Horacio Samaniego and I have developed an SIR model based on the premise that the different scaling of immune and demographic rates strongly influences the spread of multi-host pathogens such as West Nile Virus (WNV) and Influenza. Our model combines laboratory measured immunological response times, bird community compositions reported in the Breeding Bird Survey, and theoretical scaling predictions in order to create a computationally tractable model that allows for spatial variation in community structure and taxonomic variation in epidemiological parameters (Samaniego and Moses, in prep). The initial results of this work are promising: the locations that we predict for WNV persistence are strongly correlated with observations of US bird populations. Ultimately, I endeavor to incorporate spatial networks of bird interactions during epidemic spread and to explain the type and quantity of experimental data that are most useful to parameterize predictive multi-host epidemiological models.

Last modified: November 19, 2007