Description:
This group aims to be forward-thinking in the development and incorporation of new and cutting-edge methods and models to maximize the information gained from ecological data. The group works in direct support of the taxon experts and other modeling efforts in the broader lab and addresses a diverse set of research questions ranging from technical/specific computational questions about methods and data to theoretical questions about spatial modeling. Our projects and team members are advancing our capabilities in describing ecological processes from spatiotemporal species distribution models to individual niche dynamics. Many of our approaches are based around connecting datasets from disparate, multi-modal sources by building new models that allow us to gain deeper insights on complex ecological processes.
Core members:
Kevin Winner (lead), Jeremy Cohen, Lukas Gabor, Fabiola Iannarilli, Muyang Lu, Jussi Mäkinen, Shubhi Sharma, David Shen
Guest members:
Kate Ingenloff, Diego Ellis Soto, Gabriel Reygondeau (Alumni), Cory Merow (Honorary alumni), Erica Stuber (Honorary alumni)
Projects:
(primary contact in bold, other team members listed optionally)
MOLSDM
(Kevin Winner, Kate Ingenloff, Charles Marsh, Cory Merow) — 1km species distribution modeling workflow combining primary occurrence data and expert range predictions via Maxent and (Merow, Wilson, and Jetz; 2016).
Borrowing Strength for Data-Poor Species
(Kevin Winner, Shubhi Sharma) — Accurately characterizing the environmental niche and geographic distribution of species for which we have very little primary observational data by leveraging data from closely related data-rich species.
Anthropogenic Impacts on Species Occurrence at the Range-Wide Scale
(Fabiola Iannarilli) — Determining how anthropogenic drivers affect mammals and birds trends in species occurrence at the range-level scale by collating information from multiple camera-trap projects available on Wildlife Insights.
Camera traps as a new lens on biodiversity
(Fabiola Iannarilli, co-led with Ruth Oliver) — Exploring whether, where, and for which taxa, camera-trap data has the greatest potential to fill spatio-temporal gaps in data coverage of terrestrial vertebrates.
Developing automated analytics for camera-trap data
(Fabiola Iannarilli, in collaboration with the Wildlife Insights Team) — Facilitating the analysis of camera-trap data by creating an automated analytical workflow accessible through a user-friendly GUI interface soon available in the WIldlife Insights web platform.
Combining Species Survey Data Sets in Joint Models
(Jussi Mäkinen) — Integrating species observational data from presence-only datasets, presence-absence datasets, abundance datasets, and other available data to better calibrate predictions of species distributions.
Seasonal Niche Tracking
(Jeremy Cohen) — Modeling the trait and phylogenetic drivers of seasonal niche tracking in birds.
Positional Uncertainty in SDMs
(Lukas Gabor) —Investigation of the sensitivity of SDMs to various spatial scales and spatial data quality. Recently focused mainly on the effect of positional uncertainty in species occurrences to SDMs performance and ecological interpretability.
Multivariate-Normal Hypervolume Niche Framework
(Muyang Lu, Kevin Winner) — An efficient framework for niche hypervolumes based on a multivariate-normal (MVN) distribution and corresponding measures of niche breadth and similarity based on partitioning the MVN distribution.
Scale Dependence of Species Environmental Niche and Climate Change Vulnerability
(Muyang Lu) — Understanding how estimates of species environmental niche and climate change vulnerability to climate change depend on spatial grain.
Phylogenetic SDMs
(Shubhi Sharma, Kevin Winner, Jussi Mäkinen) — Multi-species distribution modeling approach connecting species in niche space based on phylogenetic information.
Active Learning for Ecology
(Kevin Winner, Charles Marsh, Diego Ellis Soto, Fabiola Iannarilli, Jussi Mäkinen, Muyang Lu, Shubhi Sharma) — Consideration of active learning strategies and an online-learning approach to data collection and sampling design for a variety of common ecological modeling problems including SDMs, camera-traps, and bio-logging.
Linking Bio-Logging and Climatological Data
(Diego Ellis Soto, Jussi Mäkinen, Kevin Winner) — Hierarchical models to connect hyperlocal data from animal-borne sensors with remote sensing data.
Modeling Species Niches with Inverse Reinforcement Learning
(Kevin Winner, Scott Yanco, Ben Carlson) — By treating individual animals as efficient solvers of an unknown Markov decision process, we can use IRL to estimate the species' niche as the reward function of that MDP.