Benjamin Blonder, Professor

Closed (1) Niches and n-dimensional hypervolumes: moving toward inference

Closed. This professor is continuing with Spring 2021 apprentices on this project; no new apprentices needed for Fall 2021.

The n‐dimensional hypervolume was originally proposed to describe the fundamental niche of a species. These hypervolumes exist within a space defined by a set of n independent axes. The hypervolume is then defined as a subset of the space, i.e. an n‐dimensional geometrical shape. Because this geometrical concept is apparently simple and easily explained, it has been applied to other contexts where the shape does not represent a niche and the axes do not represent limiting conditions. That is, hypervolumes can be defined using many types of axes (limiting resource, competition parameters, climate, resource, functional trait) and for many systems (individuals, populations, species, communities, clades, regions) (Table 1). The hypervolume concept has therefore come to inspire a range of other research areas throughout ecology and evolutionary biology.

Our lab has recently developed software tools (the ‘hypervolume’ R package) to describe n-dimensional hypervolumes for multiple biological applications. However, these tools are currently only implemented for _descriptive_ statistics (e.g. ‘how big?’ ‘how much overlap?’). There is also an opportunity to build tools for _inferential_ statistics (e.g. ‘is A significantly larger than B?’, ‘what is a confidence interval for the size of C given the sampled data?’). Such tools would enable researchers to ask much more complex and useful statistical questions about ecology and evolution.


The student will develop tools for inferential statistics using hypervolumes. These tools will include confidence intervals for volumes and overlaps based on data resampling approaches, as well as tools for comparing within/across groups similar to ANOVAs. The outcome of the work will be improved software algorithms implemented in our widely used software package. The student will also be invited to write a short peer-reviewed paper for a scientific journal highlighting this methods work. The project would be idea for a student interested in contributing mathematical, statistical, or computational tools to ecology, and who would like to learn more about interfaces between these disciplines.

The Macrosystems Ecology Laboratory is small but growing, giving members the opportunity to work closely together to develop a research project. The PI prefers to work with people who have a good sense of curiosity, an enthusiasm for natural history, and a desire to grow into independent scientists. The lab also provides regular mentoring support via regular meetings, an open-door policy, 360° feedback sessions, and a strong learning community bringing together undergraduate students, graduate students, and postdoctoral researchers. Development of soft skills, collaborations, and community linkages is strongly encouraged.
The PI is half Chinese, from a second-generation immigrant family on my mother's side, and feels strongly that we should be building inclusive communities that allow ecology to be open to everyone. The lab is a safe and welcoming community for all members.


Day-to-day supervisor for this project: Courtenay Ray

Qualifications: Required: prior experience writing code in any language (eg Python, MATLAB, R, C) Desirable: prior experience writing code in R, or coursework in theoretical probability or statistics

Weekly Hours: 6-8 hrs

Related website: http://www.benjaminblonder.org
Related website: https://besjournals.onlinelibrary.wiley.com/doi/10.1111/2041-210X.12865

Closed (2) Multiscale linkages in leaf venation network form and function

Closed. This professor is continuing with Spring 2021 apprentices on this project; no new apprentices needed for Fall 2021.

Leaf network form is highly variable among plant species (e.g. single, branching or reticulate veins) and across spatial scales (e.g. one or multiple vein orders). Such multiscale variation in network architecture might reflect different evolved solutions to simultaneously maximize leaf function (e.g. efficiency in water transportation along the veins) and minimize leaf construction costs (e.g. lignified tissue comprising veins is costly to construct). For instance, while lower vein orders may contribute more to cost, due to their disproportionate mass allocation, higher vein orders may contribute more to efficiency, due to their disproportionate impact on hydraulic resistance. Due to the difficulty of collecting network architectural data for whole-leaves, prior studies have mostly focused on those categorical descriptors of spatial scale (higher vs. lower vein orders). Here, we will apply an innovative approach (using machine learning algorithms) to fully describe multiscale network properties on whole-leaves of a phylogenetic broad set of plant species (collected from USA Botanical Gardens). Then, we will link those architectural properties to whole-leaf efficiency and cost. In this way, we will be able to investigate the rules linking network form and function across the whole spatial scale. Deciphering such rules would advance theory for the evolution of transportation efficiency, not only in leaves, but also in others types of spatial networks (e.g. informational, traffic and electrical systems).

Day-to-day supervisor for this project: Ilaíne Silveira Matos , Post-Doc

Qualifications: We will hire up to two undergraduate students to specifically work on: i. obtaining whole-leaf network images (clearing, staining, imaging and segmenting leaves); ii. extracting multiscale network architecture from leaf images (machine learning algorithms); iii. measuring functional traits related to leaf efficiency (leaf hydraulic conductance) and cost (specific leaf area and gram of glucose per gram of dry mass); and iv. analyzing the dataset to investigate multiscale trade-offs between leaf network form, function and cost. Students will obtain an in-depth knowledge on laboratorial work and will develop skills in machine learning, imaging processing (ImageJ, GIMP) and computational data analysis with R/MATLAB. This will also be an excellent opportunity for students to improve their presentation (for e.g. presenting their results at a national meeting/conference), teamwork and independent thinking skills. Students will also be welcomed to engaged in weekly lab meeting activities and social events organized by the Macrosystem Ecology Laboratory. Candidates are required to be detailed oriented, to show a strong motivation to work in a laboratory space and to have a STEM background. Biology major are preferred. Prior experience in acquiring/processing images or in data analysis with R or in laboratory work is desirable, but not required. Sophomore, Junior or Senior are preferred.

Weekly Hours: 9-11 hrs

Related website: http://www.benjaminblonder.org

Closed (3) Does seed dispersal reflect plant community composition in an alpine environment?

Closed. This professor is continuing with Spring 2021 apprentices on this project; no new apprentices needed for Fall 2021.

Seeds can move in many ways including by being consumed by animals, carried by water, or blown by wind. For every species, there is a range of suitable habitats that a seed may be fortunate enough to reach. We know from previous research that in environmentally stressful environments, like the alpine, it is often beneficial to have close neighbors. One reason for this is that some plants modify surrounding environmental conditions making growing conditions more tolerable. However, in our previous research in an alpine community in SW Colorado we did not find evidence for improved plant performance with more neighbors. As an alternative hypothesis, we are now testing if the clustering of plants that we observe in the field could be driven by seed dispersal processes. To collect a baseline of the quantity and diversity of seeds dispersed in our field site, in summer 2019 we deployed 40 seed traps. Once these seed trap collections are sorted and identified the student can test 1) if seed species composition is homogenous across the field site, 2) whether seed trap contents reflect the abundance and diversity of adult plants at our site.

Day-to-day supervisor for this project: Erin Carroll (PhD student, NSF graduate research fellow), Graduate Student

Qualifications: Two undergraduate students will assist an ESPM graduate student in processing seed trap samples collected from our alpine field site in Colorado. Students will learn lab etiquette and safety, how to sort and identify seeds, data management techniques, and statistical analyses in R. Students would be encouraged to develop their work into a research project for presentation at a College of Natural Resources undergraduate poster session. The student should be interested in ecology and plants, but no research experience is required.

Weekly Hours: 9-11 hrs

Related website: http://www.benjaminblonder.org
Related website: https://courtenayray.com/

Closed (4) Mapping quaking aspen distribution across Colorado using remote sensing and deep learning

Closed. This professor is continuing with Spring 2021 apprentices on this project; no new apprentices needed for Fall 2021.

Mortality in many tree species is accelerating due to climate change, though with intraspecific variation in mortality response within and across landscapes. This spatial variation in mortality under similar environmental conditions is driven by intraspecific genetic and phenotypic differences in ecophysiology. Mapping intraspecific genetic variation across landscapes is key to understanding differential vulnerability to mortality in a changing climate. Once impractical due to costly molecular methods, advances in remote sensing and machine learning now present the opportunity to continuously map genetic variation within species from canopy spectra. This approach has been proven over small extents using high-resolution and hyperspectral imagery but has not yet been scaled to regional levels. This study will map ploidy level variation in quaking aspen (P. tremuloides) across all of Colorado using moderate resolution (20m), multispectral satellite imagery from the Sentinel-2 mission. Ploidy level in the species has been shown to drive differences in water use efficiency, suggesting that cytotype may play a role in the observed spatial variation in recent mass mortality events linked to drought. Large scale, continuous maps of ploidy level in quaking aspen will allow us to test the hypothesis that drought mortality risk in the species is an interaction between ploidy and environment, potentially equipping land managers to identify climate-resilient seed sources and prioritize high and low-risk areas for management.

In order to map ploidy level within quaking aspen populations, we must first create a map of aspen distribution. The first step of this project will be to construct and train a convolutional neural network (CNN) to predict aspen presence or absence from multispectral, 20m resolution imagery from the Sentinel-2 satellite mission. The student will contribute to the construction and training of the CNN, with the exact role to be determined by the student’s interest and experience in applying geospatial and/or deep learning methods to ecology. For example, the student may contribute to the development of the training dataset by using higher resolution satellite imagery (1m) and GIS to create polygons capturing aspen presence and absence. The student may also contribute to the design, construction, and deployment of the CNN itself. The student will be encouraged to design an independent research project complementing their contribution to the larger project.

Qualifications: Previous experience in GIS and/or deep learning is required. No previous research experience necessary.

Weekly Hours: 9-11 hrs

Related website: http://www.benjaminblonder.org