We will be studying animal coat color miss-match by developing AI to process camera trap photographs. This will be off-campus overhead rate as work will be done at museum.

The overarching objective of our proposed work is to integrate heterogeneous satellite observations with ML-derived variables from existing networks of camera traps to better understand the composition of life on Earth through space and time. We will leverage the team's existing experience in species distribution modeling, camera traps, statistical data fusion and forecasting, and satellite image analysis to better characterize habitat features and their changes over space and time. Specifically, we will apply CV methods to camera trap images to quantify aspects of understory structure including vegetation density, greeness, tree stem density and size distribution, plant diversity, and downed woody debris. We will then develop relationships between these metrics and heterogeneous satellite data in order to estimate the understorey structure variables across the landscape. We have devised four example science questions meant to showcase the utility of these improved data across a variety of questions by evaluating the importance of our new understory characterization to: 1) as drivers of bird diversity, 2) improving species distribution models for squirrel species, 3) impact on deer demography and health, and 4) impact on wild boar population density. The core novelty of our proposed work is to use recent advances in computer vision to extract habitat variables from massive archives of camera trap imagery. Camera trap networks and their associated image archives are expansive but their primary use has been for recording the presence of animal species. However, the images typically contain views of the local habitat which can be extracted using emergent machine learning methods, including new tools that estimate the position and size of features. This type of image "by catch" represents an unexploited source of fine-scale habitat information that is critical in understanding the composition, abundance, and movement of animal species. Understorey structure measures at the camera trap sites will comprise a large sample of data with which to establish relationships with satellite imagery, including moderate resolution multispectral imagery and lidar (e.g. GEDI) and radar (e.g. Sentinel-1 and upcoming NISAR) that will ultimately permit estimating these variables at unobserved locations. This work will advance the state or remote sensing and biodiversity sciences by integrating in-situ and satellite observations in order to answer biodiversity questions and so is responsive to the stated goals of the solicitation. The impact of our proposed methodology has the potential to assist in answering many more than the four science questions we propose for demonstration. Moreover, by facilitating a better understanding of the composition of life on Earth our will make progress toward answering NASA Earth Science science questions: How is the global Earth system changing? What causes these changes in the Earth system? and How will the Earth system change in the future?

The sustainability of North American forests depends on seed production and seedling recruitment, neither of which are known at the biogeographic scales of greatest concern with climate change. Future forests will be dominated by the species that not only produce seed near and beyond current range limits. Seedling recruitment must succeed in the context of shifting climate and the predation pressure on seedlings that is likewise responding to changing climate. Understanding these recruitment responses requires a methodological shift from current monitoring of tiny plots to extensive sampling methods that can be implemented at community to biogeographic scales. This study combines continental scale fecundity estimates, newly available from the MASTIF project, with a new generation of monitoring and synthesis methods for integrating fecundity, seedling success, and consumers.

The Yellowstone to Yukon Corridor (Y2Y) is North America's largest nature corridor and connectivity project for wildlife. The 2,000-mile swath of land between Yellowstone National Park and the Yukon of Canada is one of the last remaining intact mountain ecosystems on Earth, and home to many endangered and at-risk species. The Y2Y is a mosaic of protected and unprotected land including Canadian and US national parks, federally managed wildland and national forests, state/province/territory parks, Indigenous territories, and privately managed conservation easements. Our project will focus on developing a common animal movement archive for the Y2Y, and research tools to study the effectiveness of protected areas, and migration and movement connectivity that will be applied by our end users throughout the Y2Y. Animal tracking data in the Arctic Animal Movement Archive, compiled during a NASA-funded ABoVE project, will provide the starting core of the Y2Y archive. Additional tracking data will be provided by our end users, their collaborators, and other researchers that study movement ecology throughout the Y2Y. This dynamic ensemble of studies will collectively contain millions of occurrences in the region, allowing our end users to document baseline movements and behavior of many individuals and species over past decades and up to the present. Our Movebank-based archive will provide a uniform platform and QA protocol for conducting large-scale, long-term, and multi-species analyses addressing critical research questions in support of wildlife management efforts in the region, contribute to biodiversity assessments related to climate and other regional and global changes, and provide early signals of local or large-scale ecosystem changes. Furthermore, we will develop a set of data access, preparation, and analysis tools within the MoveStore platform, a GUI-based environment for data flow and analysis tools. The data access and preparation tools will allow users to compile movement data meeting specific criteria from the Y2Y archive and link these data with contextual environmental data from remote sensing and weather resources, including NASA������������������s LANDSAT NDVI, GPM precipitation, human population density, the MERRA and NARR weather reanalyses, and local environmental data layers. We will provide tools to downscale data, resample, model null presence, and prepare movement tracks and corresponding environmental data for statistical analyses in R. Finally, we will provide tools to quantify utilization distributions and connectivity surfaces by fitting spatial and spatial-temporal point-process models. The animal movement data archive will provide a resource for the Y2Y region, and it will be expanded over time. The preparation and analysis tools will be applicable worldwide, using movement data stored on Movebank. We will develop and test tools on a set of applications catered to the needs of our end users, addressing questions such as: How are protected lands utilized by mammals throughout the Y2Y? How is connectivity between conservation areas influenced by current and predicted future environmental and anthropogenic characteristics? What characteristics are most critical for stopover locations for migrating raptors? The archive and the tools we will develop will be hosted on Movebank.org, an international database and movement analysis portal. Movebank is supported and run by the Max Planck Institute of Animal Behavior. The German government and the Max Planck Institute have committed to guaranteeing the long-term operability and accessibility of the archive and tools we will develop. Our end users are a coalition of state/province/territory wildlife management agencies, and US and Canadian NGOs. Each will commit weeks of in-kind labor to contribute movement data to the archive and to help us design, test, and implement the tools we develop.

We will use fecal DNA to estimate the density of coyotes in two barrier island National Seashores. We will also use GPS collars to track the movement of coyotes there. This will fund one MS student for 2.5 years and pay for field work.

This project will create biodiversity models for the abundance of organisms using data from the NEON system and the GJAM statistical models developed by colleagues at Duke University. We will focus on mammal and insect systems expected to be driven by variation in seed mast by trees. We will combine existing NEON data on plants, insects, and small mammals with new data on seed production and medium-large mammals.

We have assembled a multidisciplinary team of faculty experts and students from NC State, Duke, and Peru to implement a prospective study on risk factors for cutaneous leishmaniasis (CL) in humans and animals in San Martin, Peru (Figure 1). We will investigate the hypothesis that agriculturally driven deforestation increases the risk for CL exposure in humans and dogs by reducing the social and physical distances to natural reservoirs of CL. In early 2016 (T1), 350 households (HH) will be enrolled. Up to two adults and two domestic dogs per HH will be tested for CL. Individuals (and their dogs) testing negative at T1 will be retested approximately 9 months later (T2). Participants will be administered risk factor surveys and 9-month life histories. Between T1 and T2, small wild mammals will be trapped in communities, near-to homes and in the fields of participants and tested for CL. To our knowledge, the proposed research will be the first to study the role of agricultural related deforestation on CL incidence anywhere in the world. As such, this project has the potential to open up new opportunities for funding through the NIAID and CONSYTEC (Peruvian funding agency)

Studying the bear populations of the park.

This grant funds the development of a new research tool that will be integrated into the Movebank animal tracking database and website. Movebank.org is an archive of animal tracking data created by Dr. Kays and Dr. Martin Wikelski (Max Plank Institute). This new NASA funded tool will allow Movebank users to obtain data from a variety of satellite derived data sets (e.g. weather, land use, sea surface temperature etc?) for each location on earth visited by their tracked animal. Participants in the NASA funded workshops will develop this tool and then use it to ask biological questions from existing data. The tool will also be available for other scientists to use within Movebank.