USGS Science⌕ Search
Home/Archive/USGS

USGS · 70229821

Diet analysis using generalized linear models derived from foraging processes using R package mvtweedie

James T. Thorson·Mayumi L. Arimitsu·Taal Levi·Gretchen Roffler
View original source ↗Export metadata ↓

Abstract

Diet analysis integrates a wide variety of visual, chemical, and biological identification of prey. Samples are often treated as compositional data, where each prey is analyzed as a continuous percentage of the total. However, analyzing compositional data results in analytical challenges, for example, highly parameterized models or prior transformation of data. Here, we present a novel approximation involving a Tweedie generalized linear model (GLM). We first review how this approximation emerges from considering predator foraging as a thinned and marked point process (with marks representing prey species and individual prey size). This derivation can motivate future theoretical and applied developments. We then provide a practical tutorial for the Tweedie GLM using new package mvtweedie that extends capabilities of widely used packages in R ( mgcv and ggplot2 ) by transforming output to calculate prey compositions. We demonstrate this approach and software using two examples. Tufted Puffins ( Fratercula cirrhata ) provisioning their chicks on a colony in the northern Gulf of Alaska show decadal prey switching among sand lance and prowfish (1980–2000) and then Pacific herring and capelin (2000–2020), while wolves ( Canis lupus ligoni ) in southeast Alaska forage on mountain goats and marmots in northern uplands and marine mammals in seaward island coastlines.

Explore related subjects

EcologyArticle

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

James T. Thorson, Mayumi L. Arimitsu, Taal Levi, Gretchen Roffler. 2022-03-16. Diet analysis using generalized linear models derived from foraging processes using R package mvtweedie. https://doi.org/10.1002/ecy.3637

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related USGS reports

USGS2026

DRIFteRS: A dataset of drift invertebrate densities in streams and rivers across western North America, 1997–2024

Prey availability is among the most influential and highly variable determinants of fish growth and freshwater habitat carrying capacity, yet it remains understudied compared to physical habitat variables (Ouellet et al., 2025; Rosenfeld et al., 2014; Weber et al., 2017). We often lack a clear understanding of how much food is available to fishes, how it varies spatially and temporally, and how it influences responses to restoration (Ouellet et al., 2025; Rossi et al., 2024; Wipfli et al., 2010). Drift invertebrates—the primary food source for juvenile salmonids and other drift-foraging fishes—play a pivotal role in these dynamics. To better understand the spatiotemporal variability of drift invertebrate abundance and biomass across the freshwater range of drift-feeding salmonids in western North America, we compiled the DRift Invertebrates For salmonids in River Systems (DRIFteRS) dataset. The dataset encompasses 6125 samples of drift invertebrates, and, for a subset of drift samples, associated benthic invertebrate density data, collected from 1360 reaches on 459 unique rivers and streams spanning 55 river basins considered hydrologically independent (i.e., not nested within the same larger watershed) across British Columbia, Canada, and the U.S. states of Alaska, Arizona, California, Colorado, Idaho, Nevada, New Mexico, Oregon, Utah, Washington, and Wyoming. Sample sites represent a diverse array of river and stream habitats (e.g., headwater, mainstem, side channel), in watersheds with diverse land uses (e.g., urban, wilderness, agricultural), and disturbance histories (e.g., fire, restoration). Collected between 1997 and 2024, the data span the full calendar year and capture daily and seasonal patterns in drift abundance and biomass densities. When paired with water quality and quantity data as well as remotely sensed environmental landscape data, such as land use/land cover, climate, and disturbance history, channel morphology, and riparian vegetation composition, the DRIFteRS dataset can aid in identifying key drivers of drift invertebrate densities and mean body size and support predictive modeling in unsampled locations and times. The dataset may also be used to analyze aquatic-terrestrial resource flows, derive prey-encounter rates and profitability (mean prey size), and inform broader investigations of sit-and-wait foraging ecology, especially when paired with data on drift-foraging predators. For salmonid-focused applications, the dataset can be integrated into habitat evaluation models, including bioenergetic (e.g., Naman et al., 2019) and life cycle models (e.g., Beechie et al., 2023), to improve estimates of habitat capacity and population dynamics for river- and stream-rearing salmonids. Understanding prey availability dynamics is increasingly important, because rising water temperatures increase salmonid metabolic demands (Crozier et al., 2010). Flow regime transitions (i.e., snow or glacier dominated to rain dominated; Beechie et al., 2013), wildfire frequency and intensity (Hessburg et al., 2021), as well as plant community and phenology shifts (Cleland et al., 2007; Franklin et al., 2016) are all predicted to change with rising temperatures and are potential drivers of terrestrial and aquatic invertebrate prey quality and availability in lotic systems. These insights can ultimately inform restoration prioritization and design, helping managers consider food resource implications when evaluating restoration priorities and match habitat improvement to food supply. The data are released under the Creative Commons Attribution 4.0 International license and may be reused with attribution.

Alaska, Arizona, British Columbia, California, Col↗
USGS2026

Combining individual and close-kin mark–recapture to design an effective wildlife population survey

Close-kin mark–recapture (CKMR) is a promising approach for assessing population size of species that have been difficult to survey using more traditional methods. Here, we combine individual and close-kin mark–recapture in a single modeling framework (ICKMR) and provide an example of study design using this approach for Pacific walrus ( Odobenus rosmarus divergens ). We develop the ICKMR model and test it using simulated datasets, then use properties of the pseudo-likelihood to investigate the expected precision in estimates of abundance with different proposed survey designs. Our motivating example, the Pacific walrus, is an ice-associated marine mammal found in the Bering and Chukchi seas, where it is an important resource for Indigenous peoples. Pacific walrus abundance declined in the late 20th century, and it is currently a species of conservation concern due to potential impacts of climate change, particularly the loss of sea ice. To reduce uncertainty in population size estimates, researchers undertook a genetic mark–recapture sampling campaign from 2013 to 2017 and collected tissue samples from over 8000 individuals. Another campaign of a similar scale is ongoing (2023–2028). While sample collection was designed for individual mark–recapture, advances in CKMR methods and associated molecular techniques mean that these samples could also be suitable for CKMR. The advantages of CKMR over mark–recapture include an increased effective sample size (because each individual tags itself and its parents, siblings, and offspring) and additional insights into demographic quantities of interest. To make best use of genetic samples, we combine individual mark–recapture (IMR) with CKMR (ICKMR) and investigate whether different sampling strategies can increase precision in estimates of abundance. Our modeling approach includes special considerations for walrus life history, including a multi-year inter-birth interval. We found that expected coefficients of variation (CVs) of the ICKMR estimates of abundance, adult female survival, juvenile female survival, and proportion of breeding females are lower than those expected from IMR alone, and with ICKMR, fewer years of sampling can be conducted to obtain sufficient precision in estimates of abundance. This work demonstrates the utility of ICKMR and could be applicable across a variety of taxa.

Ecology↗
USGS2026

Environmental and demographic mechanisms underlying population dynamics provide relative stability in an island songbird

Understanding demographic and ecological mechanisms underlying population dynamics is a key goal in population ecology and can inform effective conservation and management. Species may be well adapted to persist under normal ranges of environmental conditions, but increasingly novel conditions due to climate change may strain their demographic buffering mechanisms. Small isolated populations on islands are expected to be especially vulnerable to declines caused by novel environmental conditions, but the ecological drivers of population dynamics on islands are often unknown. We used an integrated population model to (1) examine ecological drivers of population growth and (2) assess global population trends of the Island Scrub-Jay ( Aphelocoma insularis ) from 2009 to 2019. Our results suggest that population size increased slightly over this interval despite declines during severe drought. We also found evidence that density dependence, precipitation, and food availability affected fecundity and the survival of nonbreeding individuals. Breeder survival was relatively stable and had a weak effect on population growth as expected for long-lived species. Overall, our results provide an optimistic snapshot for this species by demonstrating resilience to contemporary drought but also emphasize the species' potential vulnerability due to its small population size.

California↗