Students will work as a member of a research team on one of the projects described below. In the Letter of Interest required for the EISI application, students should give preference for two of the projects and explain how you will contribute to and benefit from each project.
We have three goals this summer. One is to continue our work on species distribution models for multiple species. We have assembled a data set of the presence/absence of 422 species of moths measured at 256 sites in the HJ Andrews Forest. We seek to create a statistical/machine learning model that predicts which moth species will be present/absent at all locations throughout the forest. The model will take as input variables describing a site (e.g., vegetation, elevation, etc.) and predict the species that will be found there. We are particularly interested in discovering which species co-occur and which are mutually exclusive. We have found that different machine learning algorithms produce models with very different levels of accuracy, and we want to understand why this is so.
Our current data lack important information about the site variables, so the field component of this study will be to measure a variety of variables at as many of the sites as possible. The field surveys will focus on quantifying the presence and abundance of a set of host plants where moth caterpillars are known to feed; many of these host plants are species-specific. An important factor influencing moth populations is the difference between conifer/evergreen and broadleaf/deciduous vegetation. The results of this field survey should allow us to make our model much more accurate. Also, moths are important food sources for birds in the Andrews Forest landscape. Hence, these field surveys will also help with other modeling efforts at the HJA, especially models of bird behavior.
The second goal is to model what conditions can predict the "flight time" of moths--that is, the date when they first emerge and start flying. Our main hypothesis is that flight time is determined by the integrated energy (e.g., temperature degree days) received by the forest. When the total energy exceeds some (unknown) threshold, we predict the moths will fly. We want to test this and several other hypotheses, because of the implications for how moths will respond to global warming. We have similar data for bird migration, and we would like to fit similar models to predict spring arrival dates and autumn departure dates for migratory species. What factors determine these dates? How much do they vary across species? Do the dates reflect competition for food? nesting sites? mating opportunities? This work will build on existing datasets at the Andrews Forest, such as climate variables and geographical information (like elevation, vegetation cover, temperature, precipitation, wind, humidity) but will also use field data collected about moth host plants.
The third goal is to build mathematical models to understand how the dispersal behaviors of moths and birds affect the data that we collect on presence/absence. When we place an observer at location x and a bird has a nest at location y, what is the probability that the observer will detect the bird? How does this vary with the distance between x and y? Analogous questions arise for moths. When we measure site variables at x rather than at y, what problems does this cause for fitting species distribution models? Mathematical models also can address the question of how moths (or birds) select habitat locations. These models will use applied probability approaches.
Velocity distributions and fish use of engineered log jams
Project mentors: Desiree Tullos (Biological and Ecological Engineering), Eugene Zhang (Computer science), and Matt Cox (Biological and Ecological Engineering)
Increases in human population and industrial, commercial, and residential development have placed heavy demands on stream corridors at the expense of healthy streams and riverine ecosystems. Traditional engineering responses to accommodate development (i.e. flood control) have failed to incorporate natural river processes, riparian functions or aesthetic value and consequently have led to degraded streams and aquatic ecosystems. In an effort to restore streams and improve declining fish populations, stream restoration has taken center stage for fisheries management. Streambank stabilization is often implemented at river bends where natural vegetation has been removed and the streambank is eroding, causing both instream habitat degradation and land use concerns. Streambank stabilization projects are increasingly required to use natural materials (e.g. wood) as opposed to traditional engineering methods (riprap) because natural materials are thought to provide more natural habitat for fish and aquatic organisms, have esthetic appeal, are more easily permitted, and are often available locally. However, questions remain regarding the effectiveness of these Engineered Log Jams (ELJs) in both providing habitat and stabilizing streambanks, primarily due to lack of direct observations of fluid mechanics around and fish use of these engineered structures.
Snorkel surveys in Quartz Creek Velocity measurements in Quartz Creek
Summer 2013 activities