Michael M. Fuller

Research Focus

My research has focused on three important topics in ecology and resource management: space and scale, population dynamics, and computational methods. To link these topics into an organic whole, I use a combination of data collection and management, modeling and simulation, and the development of statistical and analytical tools.

My work on space and scale investigates the influence of spatial processes, such as dispersal, on species patterns including abundance, distribution, and association. I also study the interaction between environmental conditions, such as climate, and species traits, such as reproduction rate, and how this interaction influences the spatial structure of populations. An important part of my research involves developing and refining the computational techniques needed to analyze spatial patterns.

Recent Projects

In 2011, I worked with Dr. Marcy Litvak to develop methods for visualizing and managing high-throughput micro-meteorological (climate) data, at the University of New Mexico. Marcy and her colleagues have established eight climate towers in central New Mexico for the purpose of measuring changes in vegetation and ecosystem processes that are related to climate change.

This work revolves around the analysis of eddy-covariance data, which are measurements of the fluctuations in carbon dioxide, water, and energy at the ecosystem level.

To assist the analysis of large data files, I developed a visualization program in MATLAB that permits rapid assessment of the status of the many sensors that continuously monitor conditions at the towers. Each tower supports over 60 different sensors that record such variables as temperature, precipitation, wind speed and direction, and changes in carbon dioxide concentration over time.

Data are collected at 10Hz (10 times per second) and stored locally on data loggers. The software I developed downloads daily logger records and graphs the temporal sensor output according to data type, allowing the researchers to quickly identify faulty or inoperable sensors. This information is crucial to the task of assembling a complete, unbroken profile of conditions at each tower site.

The figure at right shows a typical plot window generated by the software. The figure contains four panels, each of which contains a plot of multiple sensors. The sensors are grouped by sensor type (e.g. "wind"). The lines show the output of the sensors over time (time period is set by the user). An empty plot, such as the lower right one, indicates that no data are available, either because a sensor is malfunctioning or there were simply no data for the period. The lower right plot is for certain error warnings, and there were no warnings for the plotted period.

Previous Research

Prior to working with Dr. Litvak, I was involved on several projects with Sean Thomas at the University of Toronto, Ontario, Canada. Our main area of research was forest dynamics and the effects of harvest systems on forest growth and yield, and ecosystem health. For example, we worked with several graduate students to determine how harvesting changes the structure and species composition of boreal and hardwood forests. Our studies required physically demanding field work, and made use of several types of computer models, including analystical models, statistical models, and individual-based models.

In addition to harvest impacts, I helped graduate student Rajit Patankar quantify the spatial structure of spindle gall mites in hardwood forests. Spindle gall mites are tiny arthropods that invade the leaves of sugar maples, inducing the leaves to grow spindle-shaped structures called galls. The mites live within the galls. Very little is known about the spatial distribution and natural history of the gall mites. And we are just beginning to understand the impacts the galls have on the growth and survival of maple trees. We are using Mantel tests to determine the strength of several spatially correlated environmental factors on gall density and distribution.

In other work, I developed harvest simulation software that interfaces with the forest succession model, SORTIE-ND. SORTIE-ND is an individual-based model developed by Stephen Pacala and Charles Canham. Our research aims to better understand harvesting effects and how they might be mitigated to improve forest regeneration and wildlife habitat value. We are also in the process of erecting a CO2 flux tower that will enable us to link forest community properties to changes in climate.

Research at The Institute for Environmental Modeling

My first post-doctoral appointment was with Louis Gross and Suzanne Lenhart, at The Institute for Environmental Modeling, and Department of Mathematics, University of Tennessee, Knoxville. I worked with Erika Asano and Andrew Whittle to develop a mathematical model for the spread of the Eurasian collared-dove, which has invaded North America. I was also involved in documenting the uncertainty of ecosystems models that were developed for the Florida Everglades Restoration Project.

Relative Assessment of Everglades Management Scenarios

In addition to the dove project, I worked with Dr. Louis Gross to document the modeling approach used in the the Florida Everglades Restoration project. In a recent paper (Ecological Applications 18:711-723) we describe an approach called relative assessment. We use this approach to analyze the robustness of habitat management decisions in the wetlands of Southern Florida.

The technique uses random variation in input data to generate empirically-based hypothetical climate scenarios. The effect of variation on the ranking of different management alternatives can then be compared using the criteria of interest.

For example, climate scenarios can be used as inputs to habitat quality models to project the effect of climate shifts on different species in the park. Because these species differ in their habitat needs, management scenarios must balance the positive and negative impacts of water management for species that may have conflicting needs. For example, increasing water flow to a particular area will generally favor wetlands over upland habitats.

The corresponding changes to vegetation and open freshwater can have divergent impacts on wildlife: the endangered Cape Sable seaside sparrow requires upland grassland whereas the endangered snail kite depends on wetlands (see figure at right). Relative assessment involves quantifying differences in the suitability of different management options according to a specific management-related criteria.

Computational Science and Natural Resource Management

The increased sophistication and application of models to ecological problems is just one example of how computers are changing the science of ecology. Increasingly, advances in miniaturization, computing power, remote sensing, and modeling are revolutionizing the field of natural resource management. But these advances also bring many challenges. The need for information management and communication, dynamic models, and real-time monitoring places increasing demands on legacy data structures and over-burdened networking infrastructures. To meet these demands, natural resource managers require access to high-performance computing tools and improvements in data storage, communication, and analysis.

Computer scientists are needed who can collaborate with natural resource managers and modelers to develop novel solutions. The figure at right is from my paper, written with colleagues at the Computer Science Dept. at the University of Tennessee (Computing in Science and Engineering 9:40-48), in which we highlighted several key problems in resource management that represent exciting opportunities for computer scientists and engineers in search of challenging practical problems.

Past Research Topics

My doctoral dissertation tested the predictions of the neutral theory of biodiversity and biogeography. The neutral theory emphasizes the role of random demographic change on species diversity and abundance. Proponents of the neutral theory have shown that simple neutral models, in which individuals have an equal probability of birth, death, and dispersal, can reproduce several observed community patterns, such as species relative abundance.

The above result suggests that random processes alone can explain the structure of communities. However, many empirical studies have shown that community patterns are also influenced by variation among species in their ability to survive and reproduce. To understand the extent to which neutral models can predict species distribution and abundance, I compared the variation inspecies abundances in natural communities with that predicted by neutral models. I developed a new technique for constructing "species association networks", which permit one to analyze the role of species differences on variation in species abundances. I also analyzed data on aquatic invertebrate communities (see below).

DISSERTATION: Species Association Networks of Tropical Trees

Tropical forest is often cited as a community that may be governed by neutral dynamics (e.g. Hubbell 2001). Trees are structurally and ecologically similar, and therefore may be more similar ecologically. In two projects, I collaborated with Brian Enquist and Andreas Wagner to analyze patterns of species association in a tropical forest (see our paper, Natural Resource Modeling 21:225-247). We used data on the geographic coordinates of over 19,000 tropical dry-forest trees (106 species) to determine whether species niche differences influence community structure. In an unprecedented approach, we used the principles of graph theory to analyze the effect of tree crown overlap and body size on community structure. We constructed networks representing the spatial association of species (see bottom figure in the frame to the right). This approach revealed how species interactions in local neighborhoods influence the structure of the community.

An important goal of the project was to determine the randomness of species distributions on the forest plot. We compared species networks constructed from the empirical distribution of species to those in which the geographic coordinates of individual trees had been randomized (see figure, above). We found that the networks which represented the empirical community often differed strikingly from those of randomized communities. However, our ability to detect the effect of niche differences on community structure was sensitive to how network complexity was measured. In the future, I want to examine the effect of intraspecific spatial autocorrelation on network structure in this forest.

DISSERTATION: Community-Metacommunity Dynamics of Aquatic Invertebr/ates

To analyze the invertebrate community patterns, I collaborated with Tamara Romanuk and Jurek Kolasa. We used a 9-year dataset collected by Jurek and his collaborators that represented 50 rock pool communities and 72 species (middle figure on the right). We used neutral models to predict the relative abundance of each species based on their metacommunity proportions. We then compared the predictions to empirical species proportions at the community (i.e. individual pool) and metacommunity scales.

We found that at the community scale, common species were far more variable in abundance than predicted by neutral models (lower figure at left). At the metacommunity scale, rare species were more common than predicted. In addition, variation in species diversity and abundance was strongly influenced by the relative density of predators. Trophic interactions influenced both community and metacommunity patterns. The findings of our metacommunity study are published in Community Ecology (Fuller et al. 2005, see references below) and were cited in a review article on the neutral theory, written by Brian McGill for the journal Ecology

Self-Organized Criticality in Ecological Systems

For my Master of Science research at the University of Oklahoma, I worked with Caryn Vaughn and the late Danish physicist and complexity theorist, Per Bak, along with his wife and colleague, Maya Paczuski, to test the theory of self-organized criticality (SOC), which Per co-developed. SOC is a theory from statistical physics which posits a mechanism for spontaneous self organization in complex systems. When applied to evolution and ecology, SOC asserts that extinction cascades and population correlations can arise as a consequence of intimate ecological relationships among species.

In the first empirical test of the ecological predictions of SOC, I collected time series data for 72 freshwater organisms found in vernal pools. I also monitored the physical and chemical conditions of the pools. My results neither refuted nor substantiated the predictions of SOC. Although the population trajectories were often correlated, the correlations corresponded with abiotic changes in the pools. I was therefore unable to differentiate the environmentally-driven population changes from those attributable to multivariate statistical approaches.

The above results illustrate that combining models with empirical data is a powerful approach for uncovering the influence of different factors on species patterns.

References Cited

Fuller, M.M., T.N. Romanuk, and J. Kolasa. 2005. Effects of predation and variation in species relative abundance on the parameters of neutral models. Community Ecology 6:229-240.

Fuller, M.M., D. Wang, L.J. Gross, and M.W. Berry. 2007. Current Problems and Future Directions in Computational Science for Natural Resource Management. IEEE Computing in Science and Engineering 9:40-48

Fuller, M.M., L.J. Gross, S.M. Duke-Sylvester, M. Palmer. 2008. Testing the robustness of management decisions: Everglades Restoration Scenarios. Ecological Applications 18:711-72

Fuller, M.M., B.J. Enquist, and A. Wagner. 2008. Using network analysis to characterize forest structure. Natural Resource Modeling 21:225-247

Hubbell, Stephen P. 2001. The Unified Theory of Biodiversity and Biogeography. Princeton, NJ. Princeton Univ. Press.

Contact Information
The Institute for Environmental Modeling