RUI: Temporal analysis of reproduction distributed in space to characterize plant populations' responses to environmental variation

A species' biology limits their geographic range and is a fundamental question of ecology. However, answering such questions can take years of research. Species distribution models (SDMs) have become widely used for understanding how the environment limits a species' range, by comparing soil, weather, or other environmental factors in places where a species is present to the same factors in places where it is absent. For example, some species, like the common dandelion, have taken root on every continent where humans have brought their seeds. Others, like the Mount Gleason paintbrush, are confined to a single mountain range. This project will develop a method that improves conventional SDMs by modeling not just whether a plant species is present, but the health of its populations. The method uses a form of artificial intelligence, specifically, machine learning models, trained on observations of plants' flowering and fruit production, in a pipeline the researchers call temporal analysis of reproduction distributed in space, or TARDIS. TARDIS has already been used to study the population health of Joshua trees, to aid in conservation planning, and this project will build TARDIS into a user-friendly software package that can work with data from any plant species. The research team will use the TARDIS package in a series of focused studies of plants whose geographic ranges have recently changed, such as invasive purple loosestrife or endangered lady's slipper orchids. They will also organize a course-based undergraduate research experience, or CURE, in which students learn data science methods while developing independent projects using TARDIS. Compiling results across all these projects will develop a clearer understanding of how different plant species' life histories interact with environmental variation to determine their geographic distributions. TARDIS improves conventional presence-based SDMs by modeling the reproductive niche, and it draws on data that is readily available for many species through the iNaturalist crowdsourcing program. It also takes advantage of Bayesian additive regression trees (BART) model, a powerful and flexible machine learning method that is emerging as broadly useful for ecological applications. The completed TARDIS package will provide functions for data management, training and interpreting multiple types of BART model, and analysis of predicted flowering activity projected into places and times that lack direct observations — a key TARDIS result is to reconstruct the frequency and intensity of past flowering events, and identify trends in flowering activity as environmental conditions change. The researchers expect to identify how environmental variation has facilitated range expansion, or driven range contractions, in species selected for focused study. Researchers will also synthesize student project results from the CURE in a phylogenetic regression analysis, to examine how plants' mating systems, dispersal modes, and growth forms interact with their reproductive niches to define realized geographic distributions. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria. NSF Award ID: 2536007 | Program: 01002627DB NSF RESEARCH & RELATED ACTIVIT | Principal Investigator: Jeremy Yoder | Institution: The University Corporation, Northridge, NORTHRIDGE, CA | Award Amount: $814,012 View on NSF Award Search: https://www.nsf.gov/awardsearch/show-award/?AWD_ID=2536007 View on Research.gov: https://www.research.gov/awardapi-service/v1/awards/2536007.html