Recipient Organization
UNIVERSITY OF MAINE
(N/A)
ORONO,ME 04469
Performing Department
Department of Wildlife, Fisheries and Conservation Biology
Non Technical Summary
In aiming to advance biodiversity conservation, my proposed research will expand our understanding the drivers of wildlife distributions and abundances and develop tools to assist with management of species of concern. Focusing primarily on forest-dependent wildlife, this research will combine remote-sensing techniques with field surveys to investigate the relationships between forest structure, disturbance history, landscape pattern and wildlife habitat. The research will entail 1) developing computer algorithms to map forest habitat quality with satellite imagery; 2) understanding the spatial and temporal dynamics of the unisexual salamander complex; and 3) other opportunities to advance our understanding of the landscape ecology of sensitive wildlife species. This work will be carried out in collaboration with state and federal agencies, nonprofit organizations, multiple academic institutions and other institutions to ensure broad applicability and utilization of the results.
Animal Health Component
34%
Research Effort Categories
Basic
33%
Applied
34%
Developmental
33%
Goals / Objectives
I am interested in developing our understanding of how forest structure and configuration across the landscape influences wildlife species of conservation concern, with the aim of advancing tools and recommendations for conserving these species. My initial research projects will include: 1) developing and implementing techniques for mapping forest structure from satellite imagery; 2) advancing our understanding of the spatial dynamics of the unisexual salamander complex, and 3) related opportunities that explore biodiversity conservation and the relationships between human activities and wildlife dynamics on the landscape.
Project Methods
My approach will involve a combination of field, laboratory, and computational work. For mapping forest structure (Objective 1), my team will develop machine-learning algorithms, such as Convoluted Neural Networks (CNN), that can identify and segment individual forest elements within high-resolution ortho-imagery (Weinstein et al., 2019). We will focus on forest characteristics that indicate particular types of wildlife habitat, that indicate elements of the disturbance history, and other features relevant to conservation. We will train the CNN networks to perform image segmentation by manually labelling orthoimagery, beginning within small spatial extents where we can also hand-collect detailed ground data for validation. In both training and validating, orthoimagery will be cut into ecologically-relevant tile sizes, such as 30 m by 30 m, each of which will serve as an independent observation. Once trained within small regions, we will expand to regional extents using the same tiling technique. At the larger extents, we will make use of regional ground-based observation networks such as the U. S. Forest Service Forest Inventory and Analysis data for validation. Once sufficiently optimized, we will use the algorithms to produce multi-layer raster-based map data, aggregating over appropriate resolutions across broad spatial extents, depending on data availability and computational resources. In future work, we intend to partner with organizations to use these data in guiding conservation decision-making and to understand fundamental relationships between human land use, forest structure, and wildlife.To expand our understanding of the unisexual salamander complex (Objective 2), my team will begin by advancing techniques for distinguishing different mitochondrial lineages that occur within the complex in New England - Jefferson salamander, blue-spotted salamander, and unisexual Salamander. We will work with morphometric techniques, genetic techniques based on tissue samples, and environmentally sampled genetic techniques (Charney et al., 2019, 2014; King et al., 2015). We will use previously collected tissue samples and new sampling at sites where we have previously documented occurrence data to test protocol effectiveness. To optimize eDNA protocols, we will collect samples from a small set of ponds weekly throughout the vernal pool season to assess how detectability varies with time. We will also collect one-time samples from additional ponds that represent the different genotypic components of the unisexual complex found in New England. Once we have sufficiently developed these techniques, we will use them to map distributions of the lineages in wetlands from Massachusetts to Maine. We will also use genetic markers to explore connectivity and historical patterns of dispersal across the landscape. In future years, we will link these field observations with GIS land-use layers and individual-based simulations of salamander movements to understand spatiotemporal dynamics within a heterogenous landscape.I will pursue related opportunities as they arise, including landscape ecological work on species of conservation concern, species distributions, host-parasite dynamics, expanding natural history knowledge, sexual evolution, species-habitat relationships, impacts of urbanization and climate change on wildlife, and other topics germane to biodiversity conservation using both observational and manipulative approaches.