Research

Quantifying large-scale, long-term controls on vegetation composition and ecosystem function

A large portion of my dissertation research involves understanding what controls the composition of vegetation, and subsequent ecosystem function over large spatial domains and long temporal domains in the Midwest, US. First, I advised an undergraduate thesis project, which quantified the relationship between taxon- and ecosystem-level vegetation occurrences and commonly-hypothesized environmental drivers of vegetation occurrences across Illinois and Indiana prior to European settlement. We used reconstructions of the taxa occurring across this region using the Public Land Survey record, a historical record of tree occurrences across central and western US prior to widespread European settlement. We found that patterns in vegetation occurrences were broadly explained by environmental covariates, including climate, soil, and topographic variables. However, mesoscale patterns in occurrences, such as when forest and savanna vegetation existed in a mosaic, were not predicted by the environment, but instead, could be predicted by knowing the relationship between taxa or ecosystems. We hypothesize that these residual relationships are indicative of self-regulating feedbacks between ecosystem states in this region, such as the effects of fire and herbivory. These feedbacks may give rise to multiple ecosystem states, where multiple states can exist in the same environmental niche space. We are excited to submit a manuscript on this research to Ecography in the near future!

Next, I was interested in understanding the consequences of the existence multiple ecosystem states in the Midwest on the predictability of vegetation responses to changing environmental conditions. To investigate this, I used popular joint species distribution models (random forest, genearlized linear models) to quantify the relationship between vegetation and environmental characteristics prior to European settlement across the Upper Midwest, US. Then, I used the fitted relationship to predict vegetation occurrences on the modern landscape. My hypothesis is that predictions in the modern period will poorly characterize the observed modern landscape because of the shift from open ecosystem states (prairie and savanna) historically to a closed ecosystem state (forest) in the modern period.

Finally, I extended the concept of the environment-vegetation relationship in time by using reconstructions of climate, vegetation composition, and aboveground biomass over the last 2,000 years of the pre-Industrial Holocene. Again, I found that broad-scale patterns in vegetation composition and biomass were related to patterns in mean climate and climate varaibility. Additionally, we found more evidence that climate conditions fail to completely explain patterns in vegetation composition, suggesting that other drivers such as fire and herbivory were important, even at regional and centennial spatial and temporal scales.

This research was supported by an NSF Macrosystems grant for the PalEON Project, which I co-wrote with my advisor, Jason McLachlan, as a graduate student. In addition, I was awarded an NSF Graduate Research Fellowship to support these research ideas.

Model-data fusion to understand forest ecosystem processes

As an additional part of my dissertation research, I investigated different methods of combining data and a forest gap model to improve process understanding of forest functioning at the decadal temporal scale. I worked with collaborators in the PalEON project to estimate taxon-level aboveground biomass from tree ring chronologies at a set of forest sites across the Midwest and Northeast US. Then, we used a previously-developed data assimilation algorithm to iteratively assimilate aboveground biomass into the LINKAGES forest gap model. We found that LINKAGES systematically underpredicted aboveground biomass even with annual data assimilation, highlighting the need for further model-data fusion to understand limitations of process-based ecosystem models more broadly.

To continue understanding the predictive limitations of ecosystem models, I conducted a parameter optimization procedure to choose taxon-specific model parameter values for each site. The hypothesis is that parameter optimization may allow for more flexible parameter combinations to improve estimates of aboveground biomass, but that this flexibility comes at the cost of having trait combinations that are ecologically impossible. Preliminarily, we have found that our parameter optimization algorithm also struggles to find parameter values for which aboveground biomass matches data, with systematic underprediction.

Ecological forecasting

Ecological forecasting is the methodological lens through which I view my research. Iterative out-of-sample prediction with uncertainty quantification is a powerful tool for testing theory because it requires the quantification of hypotheses and confrontation of hypotheses with data. I have been an active participant in the Ecological Forecasting Initiative, where I have co-led the Ecological Forecasting Initiative Student Association, served on the Steering Committee, and collaborated with researchers across disciplines in the Education, Diversity, Equity, and Inclusion, and Theory Working Groups. Through these collaborations, I have been able to apply methodology from the discipline of ecological forecasting to long-term proceses in paleo- and historical ecology. Additionally, through my participation in the Education and Diversity, Edquity, and Inclusion Working Groups, I became passionate about identifying opportunities for providing more inclusive education to undergraduate students interested in quantitative sciences. One chapter of my dissertation is dedicated to understanding the availability and accessibility of educational tools related to quantitative ecology and ecological forecasting, which was a collaboration with members of the Ecological Forecasting Initiative.

Contributions of changing climate and functional traits on tropical trees and lianas

I began my PhD doing rotations, one of which was in the Medvigy Lab. With Dr. Medvigy, I investigated functional trait differences between tropical trees and lianas with the purpose of identifying functional differences between trees and lianas to incorporate a liana functional type into Earth system models. I found that lianas on average have more acquisitive hydraulic traits than tropical trees, the axis by which trees and lianas differ most. I then developed a model of competition between an individual tree-liana pair based on previous work by Dr. Anna Trugman to understand the consequences of the difference in hydraulic functional traits on tree-liana competition and carbon assimilation. My key finding was that, under future scenarios of drying hydroclimatic conditions in the Neotropics, lianas may surpass a survival threshold, wherein the acquisitive hydraulic architecture of lianas is unable to maintain photosynthesis under severly dry atmospheric conditions.