Aaron Michael Grade

Aaron Michael Grade
U.S. Department of Energy · Office of Biological and Environmental Research

2021-2022 AAAS Science and Technology Policy Fellow at the U.S. Department of Energy, Office of Bio. and Env. Research


How we measure 'reads'
A 'read' is counted each time someone views a publication summary (such as the title, abstract, and list of authors), clicks on a figure, or views or downloads the full-text. Learn more
My research resides at the intersection of conservation, animal behavior and communication, landscape ecology, and community ecology. I want to know how wildlife responds to a changing landscape and human management, and how these responses are mediated by community interactions and the use of personal and social information. Utilizing experimental, comparative and modeling approaches, we can develop an understanding of the patterns and processes that underlie wildlife response to change.
Additional affiliations
September 2020 - present
Clark University
  • PostDoc Position
January 2018 - May 2018
University of Massachusetts Amherst
  • Research Assistant
  • Graduate Teaching Assistant: Ornithology, BIO 544L • Sole laboratory instructor, design lab course materials, slides, study guides, quizzes, practical exams, field trips, facilitate student research projects, grade all assignments.
September 2017 - January 2018
University of Massachusetts Amherst
  • Research Assistant
  • Graduate Teaching Assistant: Wildlife Habitat Management, NRC 564 • Organize labs and field trips, grade assignments, provide assistance to undergraduate research projects. • Coordinate multi-faculty remote camera mammal monitoring project in Amherst.
September 2015 - September 2020
University of Massachusetts Amherst
Field of study
  • Organismic and Evolutionary Biology
August 2013 - July 2015
University of Florida
Field of study
  • Interdisciplinary Ecology - Concentration in Wildlife Ecology and Conservation
August 2006 - May 2010
University of Connecticut
Field of study
  • Ecology and Evolutionary Biology


Publications (5)
Increased urbanization drives habitat loss, yet residential land-use represents significant habitat potential for mammals and could provide connectivity between patches of green spaces. Diverse mammal communities exist across urban gradients, but it is unclear how mammal community composition varies within residential lands. We conducted a camera t...
Full-text available
Predator fear effects influence reproductive outcomes in many species. In non‐urban systems, passerines often respond to predator cues by reducing parental investment, resulting in smaller and lighter nestlings. Since trophic interactions in urban areas are highly altered, it is unclear how passerines respond to fear effects in human‐altered landsc...
Full-text available
The urban development process results in the removal, alteration, and fragmentation of natural vegetation and environmental features, which has negatively impacted many wildlife species (McKinney 2002, Grimm et al. 2008). With the loss of large tracts of intact wildlands (e.g., forests, deserts, and grasslands), and the demise of specific habitat f...
Full-text available
Highway infrastructure and accompanying vehicle noise is associated with decreased wildlife populations in adjacent habitats. Noise masking of animal communication is an oft-cited potential mechanism underlying species loss in sound-polluted habitats. This study documents the disruption of between-species information transfer by anthropogenic noise...


Questions (6)
I am new with occupancy models, and I am working on selecting a model-fitting function for an occupancy model to use for a camera trap study that incorporates abundance and multiple seasons into detection. I hope to use unmarked in R, though I am open to other options. The Royle and Nichols 2003 function (occuRN) seems to be best for using abundance models, while the MacKenzie et al. 2003 (colext) function seems to be used for multi-seasons. My study design is as follows:
  • 3 summers of data collection, camera traps deployed in backyards
  • Cameras deployed 2X per season in each backyard for 7 days at a time
  • 2 cameras in each yard per deployment (at the same time)
I realize that I will have to account for the spatial and temporal autocorrelation, and that they can't be treated as independent samples, but having 2 cameras per deployment may give me a unique advantage to modeling detection. They were always deployed in roughly the same areas of the yard with roughly the same amount of cover and other habitat metrics. Should I be utilizing the multiple season functions, the abundance-informed functions, or another type of function?
Thank you for your assistance!
I am working on a linear mixed-effects model for bird nestling mass. Each individual nestling is within a nest that is within a site (residential backyard) across an urbanization gradient. Urban gradient is one of the potential predictors, as well as an experimental treatment. There can be multiple nests within a site (across years).
I generated a few models and compared them with AICc:
Fixed effects only:
mass = urban*treatment
Nested random effects:
mass = urban*treatment + (Random: nest ID in site ID)
Non-nested random effects:
mass = urban*treatment + (Random: nest ID)
mass = urban*treatment + (Random: site ID)
All three random effects model structures had delta AICc < 2. Should I go for the simplest one (just nest ID) or the one that is safer given the sampling design (nest ID in site ID)? Code for nlme in r is below.
Thank you!
Form1 <- formula(Mass ~ P_Type*urbanindex
, data = nle.12)
M.gls <- gls(Form1, data = nle.12)
M1.lme <- lme(Form1, random=(~1|SID/NID), #Nested random effects with NID %in% SID
method = "REML", data = nle.12)
M2.lme <- lme(Form1, random=(~1|SID), #Random effect just SID
method = "REML", data = nle.12)
M3.lme <- lme(Form1, random=(~1|NID), #Random effect just NID
method = "REML", data = nle.12)
summary(M.gls) #Model with no random effects
AICc(M.gls) #AICc = 333.91
summary(M1.lme) #Model with nested random effects
AICc(M1.lme) #AICc = 307.13 better fit with nested random effects
AICc(M2.lme) #AICc = 306.37 no difference with M1
AICc(M3.lme) #AICc = 304.86 no difference with M1 or M2
I ran a zero-inflated negative binomial mixed model (ZINB) and now have a statistically significant factor (recording type, either NOCA, BCCH, or Human).
To assess the differences between these three possibilities in the factor, could I rely on a post-hoc tukey HSD? Would this post-hoc test be appropriate for assessing pairwise difference of a factor in a ZINB model?
Also, would I be able to just run separate ZINB mixed models for each pairwise comparison on their own? For example, run one model with a subset of the data that's just NOCA and BCCH, one model that's just BCCH and Human, and one that's just NOCA and Human?
We are looking to develop an abundance scale to look at a sample of mites taken via berlese funnel out of House Wren nests as a part of a larger bird nest ectoparasite project. The mites are preserved in ethanol.
Our idea so far is to use a grid of cells of a certain size on a petri dish, then to have the researcher count the number of mites in 5 randomly selected cells under a dissecting microscope.
Does anyone have any opinions on this method, have an alternative/easier method, or have any papers to share that have done something similar? We are having the toughest time finding a paper that uses this method. Thank you!
I plan on conducting playback surveys at points for Owls along an urban-to-rural gradient in Western Massachusetts. Some sites are pretty rural, but others need to be in more urban areas (e.g., just outside of forested parks). In a perfect world, Owl surveys should be conducted from 1 hour after sunset until about 3 a.m., but this presents unique safety challenges in urban areas (along with permitting issues). How would one go about surveying for Owls in these urban sites with undergraduate field technician safety in mind? The owls I will be surveying for are Northern Saw-whet, Eastern Screech, Barn, Barred, Great Horned. Thanks much!
I have a small sample size, large effect size generalized linear statistical model that I am building in SPSS. The response variable is binary (i.e., response or no response), and my main independent variables of interest are the interaction between level of noise (dBA, continuous) and distance from highway.
I have "ecologically relevant" covariates (both continuous and categorical) that have the potential for interactions. For example, vegetation structure principle components, presence of other birds, etc. I also have so called "nuisance variables", most of which have no realistic potential for interactions with other covariates (e.g., observer, playback file number), and some which might (e.g., Julian Date).
Should I do separate logistic regressions for each of these, and if they are insignificant, throw them out? Model selection with all of these? How do others "deal" with nuisance covariates in their analyses? Do you treat them separately from ecologically relevant covariates? We obviously can't throw 30+ covariates in a model with a small sample size.
Thanks for the help!


Cited By


Projects (3)
Humans in residential areas provide a variety of resources that birds and other animals may use, such as bird feeders, pollinator gardens, and water features. These additional food and water subsidies could alter or even stabilize species compositions and abundances, especially in regions with low or seasonally fluctuating resource densities. Availability of human-provided resources, in contrast to naturally occurring resources, is often consistent across time and decoupled from environmental stressors such as temperature. We suggest that human-provided food and water in residential neighborhoods may buffer the effects of temperature for urban-dwelling birds in a desert landscape. We are investigating the interacting effects of seasonality, temperature and human-provided resource availability on abundances of focal bird species that occupy residential neighborhoods in the Central Arizona-Phoenix Long-Term Ecological Research (CAP LTER) study system. We developed a series of species distribution models using 2015 – 2016 bird abundance data generated by Phoenix Area Social Survey (PASS) bird censuses, as well as data on food and water as self-reported by residents in the 2017 PASS survey and collected in the 2015 Ecological Survey of Central Arizona (ESCA). We predicted that if there is a resource buffer effect on a species, then there will be a greater influence of human-provided resources on species distributions in the spring than in the winter within the same neighborhood, and in neighborhoods with higher temperatures than in neighborhoods with lower temperatures. We plan to apply these methods to additional species with varying life history traits, and we predict that species life history traits will influence the presence of a buffer effect. The results of this study will provide insights into how human-provided resources mediate effects of temperature and natural resource fluctuations on birds in residential neighborhoods.
To determine if sample selection biases associated with socioeconomic composition influence the use of community science data to determine metrics of community structure and biodiversity.
We are conducting an experimental and comparative study to see how House Wrens respond to the perception of predator threats differently across an urban-to-rural gradient in Western MA.