HRHugh S. Robinson

Hugh S. Robinson

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Hugh S. Robinson has a total of 80 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Machine learning allows for large-scale habitat prediction of a wide-ranging carnivore across diverse ecoregions

    by W. Connor O’mAlley, L. Mark Elbroch, Katherine A. Zeller, Paul Beier, Meghan M. Beale, Richard A. Beausoleil, Brian Kertson, Kyle Knopff, Kryan Kunkel, Benjamin T. Maletzke, et al.

    Context: Resource selection functions are powerful tools for predicting habitat selection of animals. Recently, machine-learning methods such as random forest have gained popularity for predicting habitat selection due to their flexibility and strong predictive performance. Objectives: We tested two methods for predicting continental-scale, second-order habitat selection of a wide-ranging large carnivore, the mountain lion (Puma concolor), to support continent-wide conservation management, including estimating abundance, and to predict habitat suitability for recolonizing or reintroduced animals. Methods: We compared a generalized linear model (GLM) and a random forest model using GPS location data from 476 individuals across 20 study sites in the western USA and Canada and remotely-sensed landscape data. We internally validated models and examined their ability to correctly classify used and available points by calculating area under the receiver operating characteristics (AUC). We performed leave-one-out (LOO) out-of-sample tests of predictive strength on both models. Results: Both models suggested that mountain lions select for steeper slopes, areas closer to water, and with higher normalized difference vegetation index (NDVI), and against variables associated with human impact. The random forest model (AUC = 0.94) demonstrated that mountain lion habitat can be accurately predicted at continental scales, outperforming the traditional GLM model (AUC = 0.68). Our LOO validation provided similar results (x̄ = 0.93 for the random forest and x̄ = 0.65 for the GLM). Conclusions: We found that the added flexibility of the random forest model provided deeper insights into how individual covariates impacted habitat selection across diverse ecosystems. Our LOO analyses suggested that our model can predict mountain lion habitat selection in unoccupied areas or where local data are unavailable. Our model thus provides a tool to support discussions and analyses relevant to continent-wide mountain lion conservation and management including estimating metapopulation abundance.


  • Comparing abundance estimates of a cryptic carnivore in southern Patagonia using two experimental methods

    by L. M. Elbroch, S. H. Williams, O. Ohrens, K. Pilgrim, A. Moeller, S. Arroyo‐Arce, M. Parker, D. Goic, H. Robinson, M. K. Schwartz

    Determining the abundance of cryptic carnivores is central to building successful conservation management to mitigate conflicts and support coexistence strategies. For these reasons, there is considerable investment in developing reliable, cost‐effective tools for estimating the abundance of wildlife. Nevertheless, field‐based comparisons of abundance methods remain uncommon, even while essential to refining methods and coming to consensus around best practices. Here, we compare two approaches still being tested in real‐world application for an emblematic puma (Puma concolor) population in the Torres del Paine UNESCO Biosphere Reserve in southern Chile: (1) the unmarked estimator, space‐to‐event model (STE), which utilizes photographs gathered with camera traps, and (2) the genotype spatial partial identity model (gSPIM), which is an adaptation of the more established spatially explicit genetic capture‐recapture method (SECR) based on genetic data extracted from scats collected in systematic surveys. We show the tremendous variation in resulting STE estimates depending upon the start time of the analysis and length of the sampling window, and showcase a refined iterative sampling approach in a Bayesian framework to both utilize the full camera data and to stabilize density estimates for a given sampling window. Across all sampling, estimates from the STE model ranged from 3.19 (1.6–5.1 representing 10th and 90th percentile of credible intervals) to 7.38 (3.3–11.6) independent pumas 100 km−2. By comparison, our gSPIM model estimated 5.1 independent pumas 100 km−2 (excluding kittens) (with credible intervals of 2.2–10.3). Neither method was compared with any known density to determine their accuracy. Nevertheless, we provide initial density estimates to guide conservation strategies for wildlife agencies and local communities overseeing and hosting nascent puma tourism and livestock ranching, as well as guidelines for the use of these methods for any wildlife species.


  • Open Access

    Correction to: Determining puma habitat suitability in the Eastern USA

    by Veronica Yovovich, Nathaniel Robinson, Hugh Robinson, Michael J. Manfredo, Shelby Perry, Jeremy T. Bruskotter, John A. Vucetich, Luis Aníbal Solórzano, Lydia A. Roe, Alison Lesure, et al.

  • Open Access

    Drivers of large carnivore density in non‐hunted, multi‐use landscapes

    by Allison L. Devlin, Jacqueline L. Frair, Peter G. Crawshaw, Luke T. B. Hunter, Fernando R. Tortato, Rafael Hoogesteijn, Nathaniel Robinson, Hugh S. Robinson, Howard B. Quigley

    Protected areas serve as population strongholds for many large carnivores, with multi‐use landscapes along their borders forming the front‐lines of wildlife conservation. Understanding large carnivore population dynamics within working landscapes is difficult where anthropogenic mortality is high and unregulated. This study focused on working ranches, where killing jaguars ( Panthera onca ) and their prey was prohibited, to gain insight into jaguar population potential across multi‐use landscapes. Faced with forest fragmentation, presence of domestic livestock, and dynamic land‐use practices, we expected jaguar populations in working landscapes to be predominantly male and transient, with low cub production, and inflated population densities in remnant forest patches, versus protected areas where we expected native forest habitat and stable jaguar territories. Using camera traps and spatial‐capture recapture analyses, we observed that male jaguars demonstrated larger‐scale movements and were more detectable than females (0.07 ± 0.01 SE vs. 0.02 ± 0.01 SE) in both working and protected landscapes. Female jaguars in ranches traveled farther than females in parks. Carnivore density increased with forest cover and wild prey activity, decreased with domestic prey activity, and was marginally higher in ranches (4.08 individuals/100 km 2 ± 0.73 SE) than in parks (3.59 individuals/100 km 2 ± 0.64 SE). Females outnumbered males in both landscapes (2.20–2.60 females/100 km 2 vs. ~1.60 males/100 km 2 ), although local male density reached up to 11.00 males/100 km 2 in ranches (vs. 3.50 males/100 km 2 in parks). While overall jaguar density was patchier in protected areas ( = 0.69 parks, 0.54 ranches), inter‐annual patchiness was higher within ranches (Moran's I = 0.49–0.60 ranches, 0.69–0.70 parks), reflecting changes in cattle management. Despite major habitat alterations, working landscapes can support carnivore densities equivalent to (or exceeding that of) unmodified forest habitat, provided that wildlife‐tolerant ranching practices are maintained.


  • Open Access

    Determining puma habitat suitability in the Eastern USA

    by Veronica Yovovich, Nathaniel Robinson, Hugh Robinson, Michael J. Manfredo, Shelby Perry, Jeremy T. Bruskotter, John A. Vucetich, Luis Aníbal Solórzano, Lydia A. Roe, Alison Lesure, et al.

    Pumas (Puma concolor) were eliminated from most of the eastern USA a century ago. In the past couple of decades, their recovery in the West has increased puma dispersal into the Midwest, with some individuals even traveling to the East Coast. We combined published expert opinion data and a habitat suitability index in an analysis that identified 17 areas in the Upper Midwest, Ozarks, Appalachia, and New England that could potentially host puma populations in the future. Thirteen of these were larger than 10,000 km2 and so likely to ensure a puma population’s long-term genetic health. Further, we quantified patch size, human density, livestock density, percent public land, and a sociocultural index reflecting wildlife values for comparing patches, as well as present a summary of current legislation relevant to puma management in the East. Our work may be useful in identifying suitable areas to restore pumas based not only on the quality of their biophysical habitat, but also on social values conducive to puma-human coexistence.


  • Top Cited Papers
    Open Access

    Robust ecological analysis of camera trap data labelled by a machine learning model

    by Robin C. Whytock, JęDrzej Świeżewski, Joeri A. Zwerts, Tadeusz Bara‐Słupski, Aurélie Flore Koumba Pambo, Marek Rogala, Laila Bahaa‐El‐Din, Kelly Boekee, Stephanie Brittain, Anabelle W. Cardoso, et al.

    Ecological data are collected over vast geographic areas using digital sensors such as camera traps and bioacoustic recorders. Camera traps have become the standard method for surveying many terrestrial mammals and birds, but camera trap arrays often generate millions of images that are time‐consuming to label. This causes significant latency between data collection and subsequent inference, which impedes conservation at a time of ecological crisis. Machine learning algorithms have been developed to improve the speed of labelling camera trap data, but it is uncertain how the outputs of these models can be used in ecological analyses without secondary validation by a human. Here, we present our approach to developing, testing and applying a machine learning model to camera trap data for the purpose of achieving fully automated ecological analyses. As a case‐study, we built a model to classify 26 Central African forest mammal and bird species (or groups). The model generalizes to new spatially and temporally independent data ( n = 227 camera stations, n = 23,868 images), and outperforms humans in several respects (e.g. detecting ‘invisible’ animals). We demonstrate how ecologists can evaluate a machine learning model's precision and accuracy in an ecological context by comparing species richness, activity patterns ( n = 4 species tested) and occupancy ( n =  4 species tested) derived from machine learning labels with the same estimates derived from expert labels. Results show that fully automated species labels can be equivalent to expert labels when calculating species richness, activity patterns ( n = 4 species tested) and estimating occupancy ( n = 3 of 4 species tested) in a large, completely out‐of‐sample test dataset. Simple thresholding using the Softmax values (i.e. excluding ‘uncertain’ labels) improved the model's performance when calculating activity patterns and estimating occupancy but did not improve estimates of species richness. We conclude that, with adequate testing and evaluation in an ecological context, a machine learning model can generate labels for direct use in ecological analyses without the need for manual validation. We provide the user‐community with a multi‐platform, multi‐language graphical user interface that can be used to run our model offline.


  • Open Access

    High performance machine learning models can fully automate labeling of camera trap images for ecological analyses

    by Robin Whytock, Jędrzej Świeżewski, Joeri A. Zwerts, Tadeusz Bara-Słupski, Aurélie Flore Koumba Pambo, Marek Rogala, Laila Bahaa-El-Din, Kelly Boekee, Stephanie Brittain, Anabelle W. Cardoso, et al.

    Ecological data are increasingly collected over vast geographic areas using arrays of digital sensors. Camera trap arrays have become the ‘gold standard’ method for surveying many terrestrial mammals and birds, but these arrays often generate millions of images that are challenging to process. This causes significant latency between data collection and subsequent inference, which can impede conservation at a time of ecological crisis. Machine learning algorithms have been developed to improve camera trap data processing speeds, but these models are not considered accurate enough for fully automated labeling of images. Here, we present a new approach to building and testing a high performance machine learning model for fully automated labeling of camera trap images. As a case-study, the model classifies 26 Central African forest mammal and bird species (or groups). The model was trained on a relatively small dataset ( c .300,000 images) but generalizes to fully independent data and outperforms humans in several respects (e.g. detecting ‘invisible’ animals). We show how the model’s precision and accuracy can be evaluated in an ecological modeling context by comparing species richness, activity patterns ( n = 4 species tested) and occupancy ( n = 4 species tested) derived from machine learning labels with the same estimates derived from expert labels. Results show that fully automated labels can be equivalent to expert labels when calculating species richness, activity patterns ( n = 4 species tested) and estimating occupancy ( n = 3 of 4 species tested) in completely out-of-sample test data ( n = 227 camera stations, n = 23868 images). Simple thresholding (discarding uncertain labels) improved the model’s performance when calculating activity patterns and estimating occupancy, but did not improve estimates of species richness. We provide the user-community with a multi-platform, multi-language user interface for running the model offline, and conclude that high performance machine learning models can fully automate labeling of camera trap data.


  • Linking resource selection and mortality modeling for population estimation of mountain lions in Montana

    by Hugh S. Robinson, Toni Ruth, Justin A. Gude, David Choate, Rich DeSimone, Mark Hebblewhite, Kyran Kunkel, Marc R. Matchett, Michael S. Mitchell, Kerry Murphy, et al.


  • Mortality of Amur tigers: The more things change, the more they stay the same

    by Hugh S. Robinson, John M. Goodrich, Dale G. Miquelle, Clayton S. Miller, Ivan V. Seryodkin

    Poaching as well as loss of habitat and prey are identified as causes of tiger population declines. Although some studies have examined habitat requirements and prey availability, few studies have quantified cause‐specific mortality of tigers. We used cumulative incidence functions (CIFs) to quantify cause‐specific mortality rates of tigers, expanding and refining earlier studies to assess the potential impact of a newly emerging disease. To quantify changes in tiger mortality over time, we re‐examined data first collected by Goodrich et al. (; study period 1: 1992–2004) as well as new telemetry data collected since January 2005 (study period 2: 2005–2012) using a total of 57 tigers (27 males and 30 females) monitored for an average of 747 days (range 26–4718 days). Across the entire study period (1992 to 2012) we found an estimated average annual survival rate of 0.75 for all tigers combined. Poaching was the primary cause of mortality during both study periods, followed by suspected poaching, distemper and natural/unknown causes. Since 2005, poaching mortality has remained relatively constant and, if combined with suspected poaching, may account for a loss of 17–19% of the population each year. Canine distemper virus (CDV) may be an additive form of mortality to the population, currently accounting for an additional 5%. Despite this relatively new source of mortality, poaching remains the main threat to Amur tiger survival and, therefore, population growth.


  • A test of the compensatory mortality hypothesis in mountain lions: A management experiment in West‐Central Montana

    by Hugh S. Robinson, Richard Desimone, Cynthia Hartway, Justin A. Gude, Michael J. Thompson, Michael S. Mitchell, Mark Hebblewhite

    Mountain lions (Puma concolor) are widely hunted for recreation, population control, and to reduce conflict with humans, but much is still unknown regarding the effects of harvest on mountain lion population dynamics. Whether human hunting mortality on mountain lions is additive or compensatory is debated. Our primary objective was to investigate population effects of harvest on mountain lions. We addressed this objective with a management experiment of 3 years of intensive harvest followed by a 6‐year recovery period. In December 2000, after 3 years of hunting, approximately 66% of a single game management unit within the Blackfoot River watershed in Montana was closed to lion hunting, effectively creating a refuge representing approximately 12% (915 km2) of the total study area (7,908 km2). Hunting continued in the remainder of the study area, but harvest levels declined from approximately 9/1,000 km2in 2001 to 2/1,000 km2in 2006 as a result of the protected area and reduced quotas outside. We radiocollared 117 mountain lions from 1998 to 2006. We recorded known fates for 63 animals, and right‐censored the remainder. Although hunting directly reduced survival, parameters such as litter size, birth interval, maternity, age at dispersal, and age of first reproduction were not significantly affected. Sensitivity analysis showed that female survival and maternity were most influential on population growth. Life‐stage simulation analysis (LSA) demonstrated the effect of hunting on the population dynamics of mountain lions. In our non‐hunted population, reproduction (kitten survival and maternity) accounted for approximately 62% of the variation in growth rate, whereas adult female survival accounted for 30%. Hunting reversed this, increasing the reliance of population growth on adult female survival (45% of the variation in population growth), and away from reproduction (12%). Our research showed that harvest at the levels implemented in this study did not affect population productivity (i.e., maternity), but had an additive effect on mountain lion mortality, and therefore population growth. Through harvest, wildlife managers have the ability to control mountain lion populations. Published 2014. This article is a U.S. Government work and is in the public domain in the USA.