This award is funded under the American Recovery and Reinvestment Act of 2009 (Public Law 111-5).
Long-term studies of mammals are precious resources for scientists and the public, but rarely are such data sets fully exploited. This is largely because, with volumes of data stored in many formats (e.g. spreadsheets, text, image files), biological inquiry is often limited to manual approaches and traditional statistical analyses. To improve analytical capability, we have developed a data warehouse for the most comprehensive long-term dolphin dataset collected to date. This warehouse, containing 25+ years of detailed observational data,, including >14,000 sighting records on >1200 dolphins, focal follows on 214 individuals (detailed individual behavioral data), genetic, ecological (habitat, prey, predators), and extensive demographic data. Biologists working with this data have used traditional statistical and linear (cause-effect) approaches for analysis. Even with large longitudinal data sets, scientists must choose just a few variables, and can easily select less-important or even the "wrong" feature. To promote more interactive data exploration, we propose developing a comprehensive visual graph inquiry platform that contains a query language for dynamic graphs, graph mining algorithms, and an intuitive visual mapping for community and individual animals that will allow biologists to explore broader data patterns before they commit to a particular set of analyses. While new research is surfacing in graph databases, graph mining, and graph visualization, an integrated, holistic approach for querying, exploring, and visualizing dynamic graphs does not exist.
Such computational tools offer biologists access to more dynamic multi-dimensional approaches to help answer biological questions related to (1) spatio-temporal and dynamic dimensions of network structure; (2) socio-cultural transmission of behavior; and (3) social, ecological, and demographic factors influencing female reproduction. By exploiting both biological and computational approaches, our new tools will help unveil the properties of large, dynamic, heterogeneous data sets and their underlying social complexity. This project will serve as a template for observational large scale animal studies in the areas of data collection, data integration, data management, visual data exploration, and data analysis. Very few longitudinal datasets on mammals are widely available. Some of the tools we develop will enable scientists to visually explore patterns rather than download a set of variables in table form. This will enable scientists to explore data properties and is expected to yield more insights than traditional data analysis.
