Reaping the rewards of the big data revolution in ecology

Abstract

The study of ecology has been irreversibly transformed by the ongoing big data revolution. Technological advancements in data collection and a cultural shift toward data sharing mean that ecological data are more available than ever before. However, availability is not equivalent to accessibility; analytical methods in ecology are typically highly customized for specific datasets, and their resulting complexity makes it difficult to adapt them to new analyses. Consequently, ecologists spend a lot of time developing their own analytical tools rather than building on the work of others. To take full advantage of ecology’s big data revolution we need tools that are easy to implement and adapt to new use cases. The first part of this dissertation presents a series of such tools designed for the data collected in mapped forest stands; areas of the forest where the precise locations of all trees are recorded. In chapter one, I develop a new regularized regression model of tree growth and compare it to the most popular tree growth model. I find that my model reaches the same conclusions regarding important ecological hypotheses as the classical model, but does so in minutes on a personal laptop whereas the classical model required hundreds of hours on a high-performance computing cluster. In chapter two, I present a new R package, ForestPlotR, which provides a toolkit for the exploration and analysis of mapped forest stand data, including an implementation of the regularized regression tree growth model developed in chapter one. Together, the products of these first two chapters have the potential to increase the accessibility of mapped forest stand data, and will hopefully encourage exciting new research directions in this field. Despite the power of big data, it is important to recognize that big ecological datasets typically arise from the combining of many contributed datasets based on a diversity of study systems. In this age of big data, we must remember that each of these contributed datasets represents the result of a well-designed experiment, and such experiments can only be developed by researchers with an intimate knowledge of their system obtained through detailed observation. In chapter three, I demonstrate this critical axis of contemporary ecology by investigating how the climate change-mediated range shifts of a coniferous tree may be moderated by the distributions of their mutualistic fungi. Surprisingly, I find some evidence that high dependence on a mutualism may not slow the rate of range shift. Overall, this dissertation highlights some important considerations that we ecologists must make as our field adapts to its increasing preponderance of data.