skip to main content

Title: Estimating the movements of terrestrial animal populations using broad-scale occurrence data

As human and automated sensor networks collect increasingly massive volumes of animal observations, new opportunities have arisen to use these data to infer or track species movements. Sources of broad scale occurrence datasets include crowdsourced databases, such as eBird and iNaturalist, weather surveillance radars, and passive automated sensors including acoustic monitoring units and camera trap networks. Such data resources represent static observations, typically at the species level, at a given location. Nonetheless, by combining multiple observations across many locations and times it is possible to infer spatially continuous population-level movements. Population-level movement characterizes the aggregated movement of individuals comprising a population, such as range contractions, expansions, climate tracking, or migration, that can result from physical, behavioral, or demographic processes. A desire to model population movements from such forms of occurrence data has led to an evolving field that has created new analytical and statistical approaches that can account for spatial and temporal sampling bias in the observations. The insights generated from the growth of population-level movement research can complement the insights from focal tracking studies, and elucidate mechanisms driving changes in population distributions at potentially larger spatial and temporal scales. This review will summarize current broad-scale occurrence datasets, discuss more » the latest approaches for utilizing them in population-level movement analyses, and highlight studies where such analyses have provided ecological insights. We outline the conceptual approaches and common methodological steps to infer movements from spatially distributed occurrence data that currently exist for terrestrial animals, though similar approaches may be applicable to plants, freshwater, or marine organisms.

« less
; ; ;
Award ID(s):
1915909 1915913
Publication Date:
Journal Name:
Movement Ecology
Springer Science + Business Media
Sponsoring Org:
National Science Foundation
More Like this
  1. Abstract

    Data availability limits phenological research at broad temporal and spatial extents. Butterflies are among the few taxa with broad-scale occurrence data, from both incidental reports and formal surveys. Incidental reports have biases that are challenging to address, but structured surveys are often limited seasonally and may not span full flight phenologies. Thus, how these data source compare in phenological analyses is unclear. We modeled butterfly phenology in relation to traits and climate using parallel analyses of incidental and survey data, to explore their shared utility and potential for analytical integration. One workflow aggregated “Pollard” surveys, where sites are visited multiple times per year; the other aggregated incidental data from online portals: iNaturalist and eButterfly. For 40 species, we estimated early (10%) and mid (50%) flight period metrics, and compared the spatiotemporal patterns and drivers of phenology across species and between datasets. For both datasets, inter-annual variability was best explained by temperature, and seasonal emergence was earlier for resident species overwintering at more advanced stages. Other traits related to habitat, feeding, dispersal, and voltinism had mixed or no impacts. Our results suggest that data integration can improve phenological research, and leveraging traits may predict phenology in poorly studied species.

  2. Abstract

    The immune system is the primary barrier to parasite infection, replication, and transmission following exposure, and variation in immunity can accordingly manifest in heterogeneity in traits that govern population-level infectious disease dynamics. While much work in ecoimmunology has focused on individual-level determinants of host immune defense (e.g., reproductive status and body condition), an ongoing challenge remains to understand the broader evolutionary and ecological contexts of this variation (e.g., phylogenetic relatedness and landscape heterogeneity) and to connect these differences into epidemiological frameworks. Ultimately, such efforts could illuminate general principles about the drivers of host defense and improve predictions and control of infectious disease. Here, we highlight recent work that synthesizes the complex drivers of immunological variation across biological scales of organization and scales these within-host differences to population-level infection outcomes. Such studies note the limitations involved in making species-level comparisons of immune phenotypes, stress the importance of spatial scale for immunology research, showcase several statistical tools for translating within-host data into epidemiological parameters, and provide theoretical frameworks for linking within- and between-host scales of infection processes. Building from these studies, we highlight several promising avenues for continued work, including the application of machine learning tools and phylogenetically controlled meta-analyses tomore »immunology data and quantifying the joint spatial and temporal dependencies in immune defense using range expansions as model systems. We also emphasize the use of organismal traits (e.g., host tolerance, competence, and resistance) as a way to interlink various scales of analysis. Such continued collaboration and disciplinary cross-talk among ecoimmunology, disease ecology, and mathematical modeling will facilitate an improved understanding of the multi-scale drivers and consequences of variation in host defense.

    « less
  3. We are developing a system for long term Semi-Automated Rehabilitation At the Home (SARAH) that relies on low-cost and unobtrusive video-based sensing. We present a cyber-human methodology used by the SARAH system for automated assessment of upper extremity stroke rehabilitation at the home. We propose a hierarchical model for automatically segmenting stroke survivor's movements and generating training task performance assessment scores during rehabilitation. The hierarchical model fuses expert therapist knowledge-based approaches with data-driven techniques. The expert knowledge is more observable in the higher layers of the hierarchy (task and segment) and therefore more accessible to algorithms incorporating high level constraints relating to activity structure (i.e., type and order of segments per task). We utilize an HMM and a Decision Tree model to connect these high level priors to data driven analysis. The lower layers (RGB images and raw kinematics) need to be addressed primarily through data driven techniques. We use a transformer based architecture operating on low-level action features (tracking of individual body joints and objects) and a Multi-Stage Temporal Convolutional Network(MS-TCN) operating on raw RGB images. We develop a sequence combining these complimentary algorithms effectively, thus encoding the information from different layers of the movement hierarchy. Through this combination,more »we produce a robust segmentation and task assessment results on noisy, variable and limited data, which is characteristic of low cost video capture of rehabilitation at the home. Our proposed approach achieves 85% accuracy in per-frame labeling, 99% accuracy in segment classification and 93% accuracy in task completion assessment. Although the methodology proposed in this paper applies to upper extremity rehabilitation using the SARAH system, it can potentially be used, with minor alterations, to assist automation in many other movement rehabilitation contexts (i.e., lower extremity training for neurological accidents).« less
  4. Tracking studies for invasive lionfish ( Pterois volitans and P. miles ) in the Western Atlantic can provide key information on habitat use to inform population control, but to date have likely underestimated home range size and movement due to constrained spatial and temporal scales. We tracked 35 acoustically tagged lionfish for >1 yr (March 2018-May 2019) within a 35 km 2 acoustic array in Buck Island Reef National Monument, St. Croix, US Virgin Islands (an area 10× larger than previous studies). Tracking lionfish at this scale revealed that home range size is 3-20 times larger than previously estimated and varies more than 8-fold across individuals (~48000-379000 m 2 ; average: 101000 m 2 ), with estimates insensitive to assumptions about potential mortality for low-movement individuals. Lionfish move far greater distances than previously reported, with 37% of fish traveling >1 km from the initial tagging site toward deeper habitats, and 1 individual moving ~10 km during a 10 d period. Movement rates, home range size, and maximum distance traveled were not related to lionfish size (18-35 cm total length) or lunar phase. Lionfish movement was lowest at night and greatest during crepuscular periods, with fish acceleration (m s -2 )more »increasing with water temperature during these times. Our results help reconcile observed patterns of rapid recolonization following lionfish removal, and suggest complex drivers likely result in highly variable patterns of movement for similarly sized fish occupying the same habitat. Culling areas ≥ the average lionfish home range size identified here (i.e. ~10 ha) or habitat patches isolated by ≥ ~180 m (radius of average home range) may minimize subsequent recolonization. If the shallow-deep long-distance movements observed here are unidirectional, mesophotic habitats may require culling at relatively greater frequencies to counteract ongoing migration.« less
  5. Low precision deep neural network (DNN) training is one of the most effective techniques for boosting DNNs’ training efficiency, as it trims down the training cost from the finest bit level. While existing works mostly fix the model precision during the whole training process, a few pioneering works have shown that dynamic precision schedules help NNs converge to a better accuracy while leading to a lower training cost than their static precision training counterparts. However, existing dynamic low precision training methods rely on manually designed precision schedules to achieve advantageous efficiency and accuracy trade-offs, limiting their more comprehensive practical applications and achievable performance. To this end, we propose LDP, a Learnable Dynamic Precision DNN training framework that can automatically learn a temporally and spatially dynamic precision schedule during training towards optimal accuracy and efficiency trade-offs. It is worth noting that LDP-trained DNNs are by nature efficient during inference. Further more, we visualize the resulting temporal and spatial precision schedule and distribution of LDP trained DNNs on different tasks to better understand the corresponding DNNs’ characteristics at different training stages and DNN layers both during and after training, drawing insights for promoting further innovations. Extensive experiments and ablation studies (seven networks,more »five datasets, and three tasks) show that the proposed LDP consistently outperforms state-of-the-art (SOTA) low precision DNN training techniques in terms of training efficiency and achieved accuracy trade-offs. For example, in addition to having the advantage of being automated, our LDP achieves a 0.31% higher accuracy with a 39.1% lower computational cost when training ResNet-20 on CIFAR-10 as compared with the best SOTA method.« less