I work to understand how the abiotic environment has shaped ectotherms (mostly phytoplankton). I am interested in understanding the strategies they use to survive in a variable and dynamic world, and how this will affect their survival and competition as the world changes.
These are the major themes of my research:
These are the major themes of my research:
Predicted change in potential phytoplankton diversity as a result of ocean warming this century (Thomas et al. 2012)
1) Temperature-driven variation in ecophysiology: implications for competition, survival and evolution
Rising temperatures will alter ecosystems and food webs. But we have a fairly patchy understanding of how the temperature influences ecological patterns and processes on different scales. We understand reasonably well how temperature affects metabolic processes, but we are missing large pieces of the picture when it comes to understanding how temperature shapes populations, communities, and ecosystem processes, not to mention how evolution alters short-term responses.
I've worked to understand how phytoplankton species respond to temperature, mostly using data from lab experiments. I've shown that from pole to pole, in both lakes and oceans, they are adapted to local temperatures. Even now, temperatures in the tropics have risen sufficiently that we should be seeing either evolutionary adaptation or local extinctions (parallel findings in other marine & terrestrial taxa point towards tropical environments as being particularly sensitive to warming). I've also worked on the implications of warming for competition between species.
Goals I am still working towards:
- understanding the implications of predicted population-level changes for whole communities and ecosystems.
- evaluating how well we can predict species occurrence patterns in nature from lab-measured temperature responses.
Collaborators: Colin Kremer, Elena Litchman, Chris Klausmeier
Papers:
Rising temperatures will alter ecosystems and food webs. But we have a fairly patchy understanding of how the temperature influences ecological patterns and processes on different scales. We understand reasonably well how temperature affects metabolic processes, but we are missing large pieces of the picture when it comes to understanding how temperature shapes populations, communities, and ecosystem processes, not to mention how evolution alters short-term responses.
I've worked to understand how phytoplankton species respond to temperature, mostly using data from lab experiments. I've shown that from pole to pole, in both lakes and oceans, they are adapted to local temperatures. Even now, temperatures in the tropics have risen sufficiently that we should be seeing either evolutionary adaptation or local extinctions (parallel findings in other marine & terrestrial taxa point towards tropical environments as being particularly sensitive to warming). I've also worked on the implications of warming for competition between species.
Goals I am still working towards:
- understanding the implications of predicted population-level changes for whole communities and ecosystems.
- evaluating how well we can predict species occurrence patterns in nature from lab-measured temperature responses.
Collaborators: Colin Kremer, Elena Litchman, Chris Klausmeier
Papers:
- A global pattern of thermal adaptation in marine phytoplankton [pdf]
- Environment and evolutionary history determine the global biogeography of phytoplankton temperature traits [pdf]
- Marine phytoplankton temperature versus growth responses from polar to tropical waters – outcome of a scientific community-wide study [pdf]
- Temperature- and size-scaling of phytoplankton population growth rates: Reconciling the Eppley curve and the metabolic theory of ecology [pdf]
- Swift thermal reaction norm evolution in a key coccolithophore species [pdf]
Population growth rate across temperature and nutrient gradients, and shifts in important traits (Thomas et al. 2017)
2) How interacting environmental drivers shapes fitness
Multiple environmental dimensions are changing simultaneously because of human activity - temperature, pH, rainfall, nutrients, light, and more. Studying how temperature alone affects species and communities cannot tell us enough to predict the future, because these environmental drivers interact in complex ways that are ignored in most experiments and models. Till a few years ago, our mathematical understanding of these interactions and their implications was limited.
My collaborators and I developed a model and did experiments to show how temperature and nutrients interact to influence growth and explored the consequences of this interaction. I've since worked to extend this budding framework and understand the implications for our communities and ecosystems.
Goals I am still working towards:
- understanding how interactions affect community- and ecosystem-level processes.
- developing a framework that can integrate additional drivers more easily.
Collaborators: Colin Kremer, Elena Litchman, Chris Klausmeier, Kyle Edwards, Anita Narwani
Papers:
Multiple environmental dimensions are changing simultaneously because of human activity - temperature, pH, rainfall, nutrients, light, and more. Studying how temperature alone affects species and communities cannot tell us enough to predict the future, because these environmental drivers interact in complex ways that are ignored in most experiments and models. Till a few years ago, our mathematical understanding of these interactions and their implications was limited.
My collaborators and I developed a model and did experiments to show how temperature and nutrients interact to influence growth and explored the consequences of this interaction. I've since worked to extend this budding framework and understand the implications for our communities and ecosystems.
Goals I am still working towards:
- understanding how interactions affect community- and ecosystem-level processes.
- developing a framework that can integrate additional drivers more easily.
Collaborators: Colin Kremer, Elena Litchman, Chris Klausmeier, Kyle Edwards, Anita Narwani
Papers:
- Phytoplankton growth and the interaction of light and temperature: A synthesis at the species and community level. [pdf]
- Temperature-nutrient interactions exacerbate sensitivity to warming in phytoplankton [pdf]
- Temperature‐dependence of minimum resource requirements alters competitive hierarchies in phytoplankton [pdf]
- Are we underestimating the ecological and evolutionary effects of warming? Interactions with other environmental drivers may increase species vulnerability to high temperatures [pdf]
There are many possible experimental designs one can use when studying multiple drivers. The first two are the most common, and also the worst (Thomas & Ranjan 2024).
3. Experimental design
I've come to believe that progress in ecology - especially the study of global change - is hampered by poor philosophy of science, filtered through a blinkered use of statistics. This shows up most clearly in the still-prevalent use of p-values as a measure of knowledge gain, which has led to over-replicated ANOVA experimental designs applied to continuous experimental factors/drivers. For the most part, ANOVA designs and analyses answer a question ("Are these two levels different?") that does not improve our scientific understanding or ability to predict anything useful. Regression designs (or response surface designs, if manipulating >1 driver) can tell us a great deal more.
But there are still a number of choices involved in setting up a regression or response surface design, and even those of us who use them rely on rules of thumb or intuition to make experimental choices. I've therefore become interested in optimising experimental design, a remarkably rich and interesting field that is unfortunately mostly unknown in ecology, oceanography, and limnology. With collaborators, I am working to apply these ideas to the experiments we need in ecology, especially for multiple driver studies.
Goals I am still working towards:
- Generating new rules of thumb for how to optimise design for single-driver regression experiments, and tools to enable anyone to do this.
Collaborators: Sinéad Collins, Ravi Ranjan
Papers:
I've come to believe that progress in ecology - especially the study of global change - is hampered by poor philosophy of science, filtered through a blinkered use of statistics. This shows up most clearly in the still-prevalent use of p-values as a measure of knowledge gain, which has led to over-replicated ANOVA experimental designs applied to continuous experimental factors/drivers. For the most part, ANOVA designs and analyses answer a question ("Are these two levels different?") that does not improve our scientific understanding or ability to predict anything useful. Regression designs (or response surface designs, if manipulating >1 driver) can tell us a great deal more.
But there are still a number of choices involved in setting up a regression or response surface design, and even those of us who use them rely on rules of thumb or intuition to make experimental choices. I've therefore become interested in optimising experimental design, a remarkably rich and interesting field that is unfortunately mostly unknown in ecology, oceanography, and limnology. With collaborators, I am working to apply these ideas to the experiments we need in ecology, especially for multiple driver studies.
Goals I am still working towards:
- Generating new rules of thumb for how to optimise design for single-driver regression experiments, and tools to enable anyone to do this.
Collaborators: Sinéad Collins, Ravi Ranjan
Papers:
Older projects
The predictability of natural phytoplankton communities
How predictable are natural communities such as phytoplankton? Somewhat embarrassingly, we don't really know. Phytoplankton species may experience a new generation every day, under the right conditions. But when we study communities in natural systems, we miss a great deal of the picture because we measure them once every 2-4 weeks, roughly comparable to measuring forests every 200-500 years. If we are to improve earth systems models (that help forecast climate change) by including more realistic biology, we need to make reasonable forecasts on the range of decades to centuries.
I worked with Francesco Pomati at Eawag in Switzerland as a postdoc to address this. We used high-frequency scanning flow cytometry data and long-term datasets to quantify predictability over timescales of hours to a decade. I found that machine learning offers a good path towards improving our ability to assess this. We found that (i) phytoplankton communities are highly predictable, (ii) despite simultaneous changes in several environmental dimensions, we are able to disentangle how each is independently (to use the term loosely) affecting the phytoplankton, and (iii) these environmental dependencies point towards clear trade-offs that will help us develop better mechanistic models to forecast the future.
As part of this work, I developed semi-automated machine learning approaches to clean and analyse flow cytometry datasets. We have made resources for these methods are publicly available on Github and Zenodo (Example dataset, R code). In the future, we hope to apply these methods in real time to monitor natural communities in the field.
Collaborators: Francesco Pomati, Simone Fontana, Marta Reyes, Michael Kehoe
Papers:
How predictable are natural communities such as phytoplankton? Somewhat embarrassingly, we don't really know. Phytoplankton species may experience a new generation every day, under the right conditions. But when we study communities in natural systems, we miss a great deal of the picture because we measure them once every 2-4 weeks, roughly comparable to measuring forests every 200-500 years. If we are to improve earth systems models (that help forecast climate change) by including more realistic biology, we need to make reasonable forecasts on the range of decades to centuries.
I worked with Francesco Pomati at Eawag in Switzerland as a postdoc to address this. We used high-frequency scanning flow cytometry data and long-term datasets to quantify predictability over timescales of hours to a decade. I found that machine learning offers a good path towards improving our ability to assess this. We found that (i) phytoplankton communities are highly predictable, (ii) despite simultaneous changes in several environmental dimensions, we are able to disentangle how each is independently (to use the term loosely) affecting the phytoplankton, and (iii) these environmental dependencies point towards clear trade-offs that will help us develop better mechanistic models to forecast the future.
As part of this work, I developed semi-automated machine learning approaches to clean and analyse flow cytometry datasets. We have made resources for these methods are publicly available on Github and Zenodo (Example dataset, R code). In the future, we hope to apply these methods in real time to monitor natural communities in the field.
Collaborators: Francesco Pomati, Simone Fontana, Marta Reyes, Michael Kehoe
Papers:
