Optimal SARS-CoV-2 vaccine allocation using real-time attack-rate estimates in Rhode Island and Massachusetts.

BACKGROUND: When three SARS-CoV-2 vaccines came to market in Europe and North America in the winter of 2020-2021, distribution networks were in a race against a major epidemiological wave of SARS-CoV-2 that began in autumn 2020. Rapid and optimized vaccine allocation was critical during this time. With 95% efficacy reported for two of the vaccines, near-term public health needs likely require that distribution is prioritized to the elderly, health care workers, teachers, essential workers, and individuals with comorbidities putting them at risk of severe clinical progression. METHODS: We evaluate various age-based vaccine distributions using a validated mathematical model based on current epidemic trends in Rhode Island and Massachusetts. We allow for varying waning efficacy of vaccine-induced immunity, as this has not yet been measured. We account for the fact that known COVID-positive cases may not have been included in the first round of vaccination. And, we account for age-specific immune patterns in both states at the time of the start of the vaccination program. Our analysis assumes that health systems during winter 2020-2021 had equal staffing and capacity to previous phases of the SARS-CoV-2 epidemic; we do not consider the effects of understaffed hospitals or unvaccinated medical staff. RESULTS: We find that allocating a substantial proportion (>75%) of vaccine supply to individuals over the age of 70 is optimal in terms of reducing total cumulative deaths through mid-2021. This result is robust to different profiles of waning vaccine efficacy and several different assumptions on age mixing during and after lockdown periods. As we do not explicitly model other high-mortality groups, our results on vaccine allocation apply to all groups at high risk of mortality if infected. A median of 327 to 340 deaths can be avoided in Rhode Island (3444 to 3647 in Massachusetts) by optimizing vaccine allocation and vaccinating the elderly first. The vaccination campaigns are expected to save a median of 639 to 664 lives in Rhode Island and 6278 to 6618 lives in Massachusetts in the first half of 2021 when compared to a scenario with no vaccine. A policy of vaccinating only seronegative individuals avoids redundancy in vaccine use on individuals that may already be immune, and would result in 0.5% to 1% reductions in cumulative hospitalizations and deaths by mid-2021. CONCLUSIONS: Assuming high vaccination coverage (>28%) and no major changes in distancing, masking, gathering size, hygiene guidelines, and virus transmissibility between 1 January 2021 and 1 July 2021 a combination of vaccination and population immunity may lead to low or near-zero transmission levels by the second quarter of 2021.

Effects of two centuries of global environmental variation on phenology and physiology of Arabidopsis thaliana.

Intraspecific trait variation is caused by genetic and plastic responses to environment. This intraspecific diversity is captured in immense natural history collections, giving us a window into trait variation across continents and through centuries of environmental shifts. Here we tested if hypotheses based on life history and the leaf economics spectrum explain intraspecific trait changes across global spatiotemporal environmental gradients. We measured phenotypes on a 216-year time series of Arabidopsis thaliana accessions from across its native range and applied spatially varying coefficient models to quantify region-specific trends in trait coordination and trait responses to climate gradients. All traits exhibited significant change across space or through time. For example, δ15 N decreased over time across much of the range and leaf C:N increased, consistent with predictions based on anthropogenic changes in land use and atmosphere. Plants were collected later in the growing season in more recent years in many regions, possibly because populations shifted toward more spring germination and summer flowering as opposed to fall germination and spring flowering. When climate variables were considered, collection dates were earlier in warmer years, while summer rainfall had opposing associations with collection date depending on regions. There was only a modest correlation among traits, indicating a lack of a single life history/physiology axis. Nevertheless, leaf C:N was low for summer- versus spring-collected plants, consistent with a life history-physiology axis from slow-growing winter annuals to fast-growing spring/summer annuals. Regional heterogeneity in phenotype trends indicates complex responses to spatiotemporal environmental gradients potentially due to geographic genetic variation and climate interactions with other aspects of environment. Our study demonstrates how natural history collections can be used to broadly characterize trait responses to environment, revealing heterogeneity in response to anthropogenic change.

A novel quantitative framework for riverscape genetics.

Riverscape genetics, which applies concepts in landscape genetics to riverine ecosystems, lack appropriate quantitative methods that address the spatial autocorrelation structure of linear stream networks and account for bidirectional geneflow. To address these challenges, we present a general framework for the design and analysis of riverscape genetic studies. Our framework starts with the estimation of pairwise genetic distance at sample sites and the development of a spatially structured ecological network (SSEN) on which riverscape covariates are measured. We then introduce the novel bidirectional geneflow in riverscapes (BGR) model that uses principles of isolation-by-resistance to quantify the effects of environmental covariates on genetic connectivity, with spatial covariance defined using simultaneous autoregressive models on the SSEN and the generalized Wishart distribution to model pairwise distance matrices arising through a random walk model of geneflow. We highlight the utility of this framework in an analysis of riverscape genetics for brook trout (Salvelinus fontinalis) in north central Pennsylvania, USA. Using the fixation index (FST ) as the measure of genetic distance, we estimated the effects of 12 riverscape covariates on geneflow by evaluating the relative support of eight competing BGR models. We then compared the performance of the top-ranked BGR model to results obtained from comparable analyses using multiple regression on distance matrices (MRM) and the program STRUCTURE. We found that the BGR model had more power to detect covariate effects, particularly for variables that were only partial barriers to geneflow and/or uncommon in the riverscape, making it more informative for assessing patterns of population connectivity and identifying threats to species conservation. This case study highlights the utility of our modeling framework over other quantitative methods in riverscape genetics, particularly the ability to rigorously test hypotheses about factors that influence geneflow and probabilistically estimate the effect of riverscape covariates, including stream flow direction. This framework is flexible across taxa and riverine networks, is easily executable, and provides intuitive results that can be used to investigate the likely outcomes of current and future management scenarios.

Ecological prediction at macroscales using big data: Does sampling design matter?

Although ecosystems respond to global change at regional to continental scales (i.e., macroscales), model predictions of ecosystem responses often rely on data from targeted monitoring of a small proportion of sampled ecosystems within a particular geographic area. In this study, we examined how the sampling strategy used to collect data for such models influences predictive performance. We subsampled a large and spatially extensive data set to investigate how macroscale sampling strategy affects prediction of ecosystem characteristics in 6,784 lakes across a 1.8-million-km2 area. We estimated model predictive performance for different subsets of the data set to mimic three common sampling strategies for collecting observations of ecosystem characteristics: random sampling design, stratified random sampling design, and targeted sampling. We found that sampling strategy influenced model predictive performance such that (1) stratified random sampling designs did not improve predictive performance compared to simple random sampling designs and (2) although one of the scenarios that mimicked targeted (non-random) sampling had the poorest performing predictive models, the other targeted sampling scenarios resulted in models with similar predictive performance to that of the random sampling scenarios. Our results suggest that although potential biases in data sets from some forms of targeted sampling may limit predictive performance, compiling existing spatially extensive data sets can result in models with good predictive performance that may inform a wide range of science questions and policy goals related to global change.

Animal movement constraints improve resource selection inference in the presence of telemetry error.

Multiple factors complicate the analysis of animal telemetry location data. Recent advancements address issues such as temporal autocorrelation and telemetry measurement error, but additional challenges remain. Difficulties introduced by complicated error structures or barriers to animal movement can weaken inference. We propose an approach for obtaining resource selection inference from animal location data that accounts for complicated error structures, movement constraints, and temporally autocorrelated observations. We specify a model for telemetry data observed with error conditional on unobserved true locations that reflects prior knowledge about constraints in the animal movement process. The observed telemetry data are modeled using a flexible distribution that accommodates extreme errors and complicated error structures. Although constraints to movement are often viewed as a nuisance, we use constraints to simultaneously estimate and account for telemetry error. We apply the model to simulated data, showing that it outperforms common ad hoc approaches used when confronted with measurement error and movement constraints. We then apply our framework to an Argos satellite telemetry data set on harbor seals (Phoca vitulina) in the Gulf of Alaska, a species that is constrained to move within the marine environment and adjacent coastlines.

Reconciling resource utilization and resource selection functions.

1. Analyses based on utilization distributions (UDs) have been ubiquitous in animal space use studies, largely because they are computationally straightforward and relatively easy to employ. Conventional applications of resource utilization functions (RUFs) suggest that estimates of UDs can be used as response variables in a regression involving spatial covariates of interest. 2. It has been claimed that contemporary implementations of RUFs can yield inference about resource selection, although to our knowledge, an explicit connection has not been described. 3. We explore the relationships between RUFs and resource selection functions from a hueristic and simulation perspective. We investigate several sources of potential bias in the estimation of resource selection coefficients using RUFs (e.g. the spatial covariance modelling that is often used in RUF analyses). 4. Our findings illustrate that RUFs can, in fact, serve as approximations to RSFs and are capable of providing inference about resource selection, but only with some modification and under specific circumstances. 5. Using real telemetry data as an example, we provide guidance on which methods for estimating resource selection may be more appropriate and in which situations. In general, if telemetry data are assumed to arise as a point process, then RSF methods may be preferable to RUFs; however, modified RUFs may provide less biased parameter estimates when the data are subject to location error.

A combination of annual and nonannual forces drive respiratory disease in the tropics.

Background: It is well known that influenza and other respiratory viruses are wintertime-seasonal in temperate regions. However, respiratory disease seasonality in the tropics remains elusive. In this study, we aimed to characterize the seasonality of influenza-like illness (ILI) and influenza virus in Ho Chi Minh City (HCMC), Vietnam. Methods: We monitored the daily number of ILI patients in 89 outpatient clinics from January 2010 to December 2019. We collected nasal swabs and tested for influenza from a subset of clinics from May 2012 to December 2019. We used spectral analysis to describe the periodicities in the system. We evaluated the contribution of these periodicities to predicting ILI and influenza patterns through lognormal and gamma hurdle models. Findings: During ten years of community surveillance, 66,799 ILI reports were collected covering 2.9 million patient visits; 2604 nasal swabs were collected 559 of which were PCR-positive for influenza virus. Both annual and nonannual cycles were detected in the ILI time series, with the annual cycle showing 8.9% lower ILI activity (95% CI: 8.8%-9.0%) from February 24 to May 15. Nonannual cycles had substantial explanatory power for ILI trends (ΔAIC = 183) compared to all annual covariates (ΔAIC = 263). Near-annual signals were observed for PCR-confirmed influenza but were not consistent along in time or across influenza (sub)types. Interpretation: Our study reveals a unique pattern of respiratory disease dynamics in a tropical setting influenced by both annual and nonannual drivers. Timing of vaccination campaigns and hospital capacity planning may require a complex forecasting approach.

A combination of annual and nonannual forces drive respiratory disease in the tropics.

INTRODUCTION: It is well known that influenza and other respiratory viruses are wintertime-seasonal in temperate regions. However, respiratory disease seasonality in the tropics is less well understood. In this study, we aimed to characterise the seasonality of influenza-like illness (ILI) and influenza virus in Ho Chi Minh City, Vietnam. METHODS: We monitored the daily number of ILI patients in 89 outpatient clinics from January 2010 to December 2019. We collected nasal swabs and tested for influenza from a subset of clinics from May 2012 to December 2019. We used spectral analysis to describe the periodic signals in the system. We evaluated the contribution of these periodic signals to predicting ILI and influenza patterns through lognormal and gamma hurdle models. RESULTS: During 10 years of community surveillance, 66 799 ILI reports were collected covering 2.9 million patient visits; 2604 nasal swabs were collected, 559 of which were PCR-positive for influenza virus. Both annual and nonannual cycles were detected in the ILI time series, with the annual cycle showing 8.9% lower ILI activity (95% CI 8.8% to 9.0%) from February 24 to May 15. Nonannual cycles had substantial explanatory power for ILI trends (ΔAIC=183) compared with all annual covariates (ΔAIC=263) in lognormal regression. Near-annual signals were observed for PCR-confirmed influenza but were not consistent over time or across influenza (sub)types. The explanatory power of climate factors for ILI and influenza virus trends was weak. CONCLUSION: Our study reveals a unique pattern of respiratory disease dynamics in a tropical setting influenced by both annual and nonannual drivers, with influenza dynamics showing near-annual periodicities. Timing of vaccination campaigns and hospital capacity planning may require a complex forecasting approach.

A Mechanistic Model of Annual Sulfate Concentrations in the United States.

Understanding how individual pollution sources contribute to ambient sulfate pollution is critical for assessing past and future air quality regulations. Since attribution to specific sources is typically not encoded in spatial air pollution data, we develop a mechanistic model which we use to estimate, with uncertainty, the contribution of ambient sulfate concentrations attributable specifically to sulfur dioxide (SO2) emissions from individual coal-fired power plants in the central United States. We propose a multivariate Ornstein-Uhlenbeck (OU) process approximation to the dynamics of the underlying space-time chemical transport process, and its distributional properties are leveraged to specify novel probability models for spatial data that are viewed as either a snapshot or time-averaged observation of the OU process. Using US EPA SO2 emissions data from 193 power plants and state-of-the-art estimates of ground-level annual mean sulfate concentrations, we estimate that in 2011 - a time of active power plant regulatory action - existing flue-gas desulfurization (FGD) technologies at 66 power plants reduced population-weighted exposure to ambient sulfate by 1.97 µg/m3 (95% CI: 1.80 - 2.15). Furthermore, we anticipate future regulatory benefits by estimating that installing FGD technologies at the five largest SO2-emitting facilities would reduce human exposure to ambient sulfate by an additional 0.45 µg/m3 (95% CI: 0.33 - 0.54).

SARS-CoV-2 Attack Rate and Population Immunity in Southern New England, March 2020 to May 2021.

Importance: In emergency epidemic and pandemic settings, public health agencies need to be able to measure the population-level attack rate, defined as the total percentage of the population infected thus far. During vaccination campaigns in such settings, public health agencies need to be able to assess how much the vaccination campaign is contributing to population immunity; specifically, the proportion of vaccines being administered to individuals who are already seropositive must be estimated. Objective: To estimate population-level immunity to SARS-CoV-2 through May 31, 2021, in Rhode Island, Massachusetts, and Connecticut. Design, Setting, and Participants: This observational case series assessed cases, hospitalizations, intensive care unit occupancy, ventilator occupancy, and deaths from March 1, 2020, to May 31, 2021, in Rhode Island, Massachusetts, and Connecticut. Data were analyzed from July 2021 to November 2021. Exposures: COVID-19-positive test result reported to state department of health. Main Outcomes and Measures: The main outcomes were statistical estimates, from a bayesian inference framework, of the percentage of individuals as of May 31, 2021, who were (1) previously infected and vaccinated, (2) previously uninfected and vaccinated, and (3) previously infected but not vaccinated. Results: At the state level, there were a total of 1 160 435 confirmed COVID-19 cases in Rhode Island, Massachusetts, and Connecticut. The median age among individuals with confirmed COVID-19 was 38 years. In autumn 2020, SARS-CoV-2 population immunity (equal to the attack rate at that point) in these states was less than 15%, setting the stage for a large epidemic wave during winter 2020 to 2021. Population immunity estimates for May 31, 2021, were 73.4% (95% credible interval [CrI], 72.9%-74.1%) for Rhode Island, 64.1% (95% CrI, 64.0%-64.4%) for Connecticut, and 66.3% (95% CrI, 65.9%-66.9%) for Massachusetts, indicating that more than 33% of residents in these states were fully susceptible to infection when the Delta variant began spreading in July 2021. Despite high vaccine coverage in these states, population immunity in summer 2021 was lower than planned owing to an estimated 34.1% (95% CrI, 32.9%-35.2%) of vaccines in Rhode Island, 24.6% (95% CrI, 24.3%-25.1%) of vaccines in Connecticut, and 27.6% (95% CrI, 26.8%-28.6%) of vaccines in Massachusetts being distributed to individuals who were already seropositive. Conclusions and Relevance: These findings suggest that future emergency-setting vaccination planning may have to prioritize high vaccine coverage over optimized vaccine distribution to ensure that sufficient levels of population immunity are reached during the course of an ongoing epidemic or pandemic.

SARS-CoV-2 epidemic after social and economic reopening in three U.S. states reveals shifts in age structure and clinical characteristics.

State-level reopenings in late spring 2020 facilitated the resurgence of severe acute respiratory syndrome coronavirus 2 transmission. Here, we analyze age-structured case, hospitalization, and death time series from three states-Rhode Island, Massachusetts, and Pennsylvania-that had successful reopenings in May 2020 without summer waves of infection. Using 11 daily data streams, we show that from spring to summer, the epidemic shifted from an older to a younger age profile and that elderly individuals were less able to reduce contacts during the lockdown period when compared to younger individuals. Clinical case management improved from spring to summer, resulting in fewer critical care admissions and lower infection fatality rate. Attack rate estimates through 31 August 2020 are 6.2% [95% credible interval (CI), 5.7 to 6.8%] of the total population infected for Rhode Island, 6.7% (95% CI, 5.4 to 7.6%) in Massachusetts, and 2.7% (95% CI, 2.5 to 3.1%) in Pennsylvania.

Reconciling multiple data sources to improve accuracy of large-scale prediction of forest disease incidence.

Ecological spatial data often come from multiple sources, varying in extent and accuracy. We describe a general approach to reconciling such data sets through the use of the Bayesian hierarchical framework. This approach provides a way for the data sets to borrow strength from one another while allowing for inference on the underlying ecological process. We apply this approach to study the incidence of eastern spruce dwarf mistletoe (Arceuthobium pusillum) in Minnesota black spruce (Picea mariana). A Minnesota Department of Natural Resources operational inventory of black spruce stands in northern Minnesota found mistletoe in 11% of surveyed stands, while a small, specific-pest survey found mistletoe in 56% of the surveyed stands. We reconcile these two surveys within a Bayesian hierarchical framework and predict that 35-59% of black spruce stands in northern Minnesota are infested with dwarf mistletoe.

Social fluidity mobilizes contagion in human and animal populations.

Humans and other group-living animals tend to distribute their social effort disproportionately. Individuals predominantly interact with a small number of close companions while maintaining weaker social bonds with less familiar group members. By incorporating this behavior into a mathematical model, we find that a single parameter, which we refer to as social fluidity, controls the rate of social mixing within the group. Large values of social fluidity correspond to gregarious behavior, whereas small values signify the existence of persistent bonds between individuals. We compare the social fluidity of 13 species by applying the model to empirical human and animal social interaction data. To investigate how social behavior influences the likelihood of an epidemic outbreak, we derive an analytical expression of the relationship between social fluidity and the basic reproductive number of an infectious disease. For species that form more stable social bonds, the model describes frequency-dependent transmission that is sensitive to changes in social fluidity. As social fluidity increases, animal-disease systems become increasingly density-dependent. Finally, we demonstrate that social fluidity is a stronger predictor of disease outcomes than both group size and connectivity, and it provides an integrated framework for both density-dependent and frequency-dependent transmission.

Optimal SARS-CoV-2 vaccine allocation using real-time seroprevalence estimates in Rhode Island and Massachusetts.

As three SARS-CoV-2 vaccines come to market in Europe and North America in the winter of 2020-2021, distribution networks will be in a race against a major epidemiological wave of SARS-CoV-2 that began in autumn 2020. Rapid and optimized vaccine allocation is critical during this time. With 95% efficacy reported for two of the vaccines, near-term public health needs require that distribution is prioritized to the elderly, health-care workers, teachers, essential workers, and individuals with co-morbidities putting them at risk of severe clinical progression. Here, we evaluate various age-based vaccine distributions using a validated mathematical model based on current epidemic trends in Rhode Island and Massachusetts. We allow for varying waning efficacy of vaccine-induced immunity, as this has not yet been measured. We account for the fact that known COVID-positive cases may not be included in the first round of vaccination. And, we account for current age-specific immune patterns in both states. We find that allocating a substantial proportion ( > 75%) of vaccine supply to individuals over the age of 70 is optimal in terms of reducing total cumulative deaths through mid-2021. As we do not explicitly model other high mortality groups, this result on vaccine allocation applies to all groups at high risk of mortality if infected. Our analysis confirms that for an easily transmissible respiratory virus, allocating a large majority of vaccinations to groups with the highest mortality risk is optimal. Our analysis assumes that health systems during winter 2020-2021 have equal staffing and capacity to previous phases of the SARS-CoV-2 epidemic; we do not consider the effects of understaffed hospitals or unvaccinated medical staff. Vaccinating only seronegative individuals avoids redundancy in vaccine use on individuals that may already be immune, and will result in 1% to 2% reductions in cumulative hospitalizations and deaths by mid-2021. Assuming high vaccination coverage ( > 28%) and no major relaxations in distancing, masking, gathering size, or hygiene guidelines between now and spring 2021, our model predicts that a combination of vaccination and population immunity will lead to low or near-zero transmission levels by the second quarter of 2021.

SARS-CoV-2 attack rate and population immunity in southern New England, March 2020 - May 2021.

Estimating an infectious disease attack rate requires inference on the number of reported symptomatic cases of a disease, the number of unreported symptomatic cases, and the number of asymptomatic infections. Population-level immunity can then be estimated as the attack rate plus the number of vaccine recipients who had not been previously infected; this requires an estimate of the fraction of vaccines that were distributed to seropositive individuals. To estimate attack rates and population immunity in southern New England, we fit a validated dynamic epidemiological model to case, clinical, and death data streams reported by Rhode Island, Massachusetts, and Connecticut for the first 15 months of the COVID-19 pandemic, from March 1 2020 to May 31 2021. This period includes the initial spring 2020 wave, the major winter wave of 2020-2021, and the lagging wave of lineage B.1.1.7(Alpha) infections during March-April 2021. In autumn 2020, SARS-CoV-2 population immunity (equal to the attack rate at that point) in southern New England was still below 15%, setting the stage for a large winter wave. After the roll-out of vaccines in early 2021, population immunity in many states was expected to approach 70% by spring 2021, with more than half of this immune population coming from vaccinations. Our population immunity estimates for May 31 2021 are 73.4% (95% CrI: 72.9% - 74.1%) for Rhode Island, 64.1% (95% CrI: 64.0% - 64.4%) for Connecticut, and 66.3% (95% CrI: 65.9% - 66.9%) for Massachusetts, indicating that >33% of southern Englanders were still susceptible to infection when the Delta variant began spreading in July 2021. Despite high vaccine coverage in these states, population immunity in summer 2021 was lower than planned due to 34% (Rhode Island), 25% (Connecticut), and 28% (Massachusetts) of vaccine distribution going to seropositive individuals. Future emergency-setting vaccination planning will likely have to consider over-vaccination as a strategy to ensure that high levels of population immunity are reached during the course of an ongoing epidemic.

SARS-CoV-2 epidemic after social and economic reopening in three US states reveals shifts in age structure and clinical characteristics.

In the United States, state-level re-openings in spring 2020 presented an opportunity for the resurgence of SARS-CoV-2 transmission. One important question during this time was whether human contact and mixing patterns could increase gradually without increasing viral transmission, the rationale being that new mixing patterns would likely be associated with improved distancing, masking, and hygiene practices. A second key question to follow during this time was whether clinical characteristics of the epidemic would improve after the initial surge of cases. Here, we analyze age-structured case, hospitalization, and death time series from three states - Rhode Island, Massachusetts, and Pennsylvania - that had successful re-openings in May 2020 without summer waves of infection. Using a Bayesian inference framework on eleven daily data streams and flexible daily population contact parameters, we show that population-average mixing rates dropped by >50% during the lockdown period in March/April, and that the correlation between overall population mobility and transmission-capable mobility was broken in May as these states partially re-opened. We estimate the reporting rates (fraction of symptomatic cases reporting to health system) at 96.0% (RI), 72.1% (MA), and 75.5% (PA); in Rhode Island, when accounting for cases caught through general-population screening programs, the reporting rate estimate is 94.5%. We show that elderly individuals were less able to reduce contacts during the lockdown period when compared to younger individuals. Attack rate estimates through August 31 2020 are 6.4% (95% CI: 5.8% ‒ 7.3%) of the total population infected for Rhode Island, 5.7% (95% CI: 5.0% ‒ 6.8%) in Massachusetts, and 3.7% (95% CI: 3.1% ‒ 4.5%) in Pennsylvania, with some validation available through published seroprevalence studies. Infection fatality rates (IFR) estimates for the spring epidemic are higher in our analysis (>2%) than previously reported values, likely resulting from the epidemics in these three states affecting the most vulnerable sub-populations, especially the most vulnerable of the ≥80 age group.

Machine learning for modeling animal movement.

Animal movement drives important ecological processes such as migration and the spread of infectious disease. Current approaches to modeling animal tracking data focus on parametric models used to understand environmental effects on movement behavior and to fill in missing tracking data. Machine Learning and Deep learning algorithms are powerful and flexible predictive modeling tools but have rarely been applied to animal movement data. In this study we present a general framework for predicting animal movement that is a combination of two steps: first predicting movement behavioral states and second predicting the animal's velocity. We specify this framework at the individual level as well as for collective movement. We use Random Forests, Neural and Recurrent Neural Networks to compare performance predicting one step ahead as well as long range simulations. We compare results against a custom constructed Stochastic Differential Equation (SDE) model. We apply this approach to high resolution ant movement data. We found that the individual level Machine Learning and Deep Learning methods outperformed the SDE model for one step ahead prediction. The SDE model did comparatively better at simulating long range movement behaviour. Of the Machine Learning and Deep Learning models the Long Short Term Memory (LSTM) individual level model did best at long range simulations. We also applied the Random Forest and LSTM individual level models to model gull migratory movement to demonstrate the generalizability of this framework. Machine Learning and deep learning models are easier to specify compared to traditional parametric movement models which can have restrictive assumptions. However, machine learning and deep learning models are less interpretable than parametric movement models. The type of model used should be determined by the goal of the study, if the goal is prediction, our study provides evidence that machine learning and deep learning models could be useful tools.

Identifying and characterizing extrapolation in multivariate response data.

Faced with limitations in data availability, funding, and time constraints, ecologists are often tasked with making predictions beyond the range of their data. In ecological studies, it is not always obvious when and where extrapolation occurs because of the multivariate nature of the data. Previous work on identifying extrapolation has focused on univariate response data, but these methods are not directly applicable to multivariate response data, which are common in ecological investigations. In this paper, we extend previous work that identified extrapolation by applying the predictive variance from the univariate setting to the multivariate case. We propose using the trace or determinant of the predictive variance matrix to obtain a scalar value measure that, when paired with a selected cutoff value, allows for delineation between prediction and extrapolation. We illustrate our approach through an analysis of jointly modeled lake nutrients and indicators of algal biomass and water clarity in over 7000 inland lakes from across the Northeast and Mid-west US. In addition, we outline novel exploratory approaches for identifying regions of covariate space where extrapolation is more likely to occur using classification and regression trees. The use of our Multivariate Predictive Variance (MVPV) measures and multiple cutoff values when exploring the validity of predictions made from multivariate statistical models can help guide ecological inferences.

Confronting models with data: the challenges of estimating disease spillover.

For pathogens known to transmit across host species, strategic investment in disease control requires knowledge about where and when spillover transmission is likely. One approach to estimating spillover is to directly correlate observed spillover events with covariates. An alternative is to mechanistically combine information on host density, distribution and pathogen prevalence to predict where and when spillover events are expected to occur. We use several case studies at the wildlife-livestock disease interface to highlight the challenges, and potential solutions, to estimating spatio-temporal variation in spillover risk. Datasets on multiple host species often do not align in space, time or resolution, and may have no estimates of observation error. Linking these datasets requires they be related to a common spatial and temporal resolution and appropriately propagating errors in predictions can be difficult. Hierarchical models are one potential solution, but for fine-resolution predictions at broad spatial scales, many models become computationally challenging. Despite these limitations, the confrontation of mechanistic predictions with observed events is an important avenue for developing a better understanding of pathogen spillover. Systems where data have been collected at all levels in the spillover process are rare, or non-existent, and require investment and sustained effort across disciplines. This article is part of the theme issue 'Dynamic and integrative approaches to understanding pathogen spillover'.

Ant colonies maintain social homeostasis in the face of decreased density.

Interactions lie at the heart of social organization, particularly in ant societies. Interaction rates are presumed to increase with density, but there is little empirical evidence for this. We manipulated density within carpenter ant colonies of the species Camponotus pennsylvanicus by quadrupling nest space and by manually tracking 6.9 million ant locations and over 3200 interactions to study the relationship between density, spatial organization and interaction rates. Colonies divided into distinct spatial regions on the basis of their underlying spatial organization and changed their movement patterns accordingly. Despite a reduction in both overall and local density, we did not find the expected concomitant reduction in interaction rates across all colonies. Instead, we found divergent effects across colonies. Our results highlight the remarkable organizational resilience of ant colonies to changes in density, which allows them to sustain two key basic colony life functions, that is food and information exchange, during environmental change.