A hierarchical aggregate data model with spatially correlated disease rates

The aggregate data study design (Prentice and Sheppard, 1995, Biometrika 82, 113-125) estimates individual-level exposure effects by regressing population-based disease rates on covariate data from survey samples in each population group. In this work, we further develop the aggregate data model to allow for residual spatial correlation among disease rates across populations. Geographical variation that is not explained by model predictors and has a spatial component often arises in studies of rare chronic diseases, such as breast cancer. We combine the aggregate and Bayesian disease-mapping models to provide an intuitive approach to the modeling of spatial effects while drawing correct inference regarding the exposure effect. Based on the results of simulation studies, we suggest guidelines for use of the proposed model.     

A spatio-temporal LSTM model to forecast across multiple temporal and spatial scales

This paper presents a novel spatio-temporal LSTM (SPATIAL) architecture for time series forecasting applied to environmental datasets. The framework was applied for three different ocean datasets: current speed, temperature, and dissolved oxygen. Network implementation proceeded in two directions that are nominally separated but connected as part of a natural environmental system – across the spatial (between individual sensors) and temporal dimensions of the sensor data. Data from twenty ocean sensors were used to train the model. Results were compared against four baseline models: two machine learning algorithms generated by robust autoML frameworks, and two deep neural networks based on CNN and LSTM, respectively. Results demonstrated ability to accurately replicate complex signals and provide comparable performance to state-of-the-art benchmarks. Learning from multiple sensors simultaneously increased robustness to missing data. This paper addresses two fundamental challenges related to environmental applications of machine learning: 1) data sparsity, particularly in a challenging ocean environment, and 2) environmental datasets are inherently connected in the spatial and temporal directions while classical ML approaches only consider one of these at a time. Furthermore, sharing of parameters across all input steps makes SPATIAL a fast, scalable, and easily-parameterized forecasting framework.     

Carbon balances of bioenergy systems using biomass from forests managed with long rotations: bridging the gap between stand and landscape assessments

Studies report different findings concerning the climate benefits of bioenergy, in part due to varying scope and use of different approaches to define spatial and temporal system boundaries. We quantify carbon balances for bioenergy systems that use biomass from forests managed with long rotations, employing different approaches and boundary conditions. Two approaches to represent landscapes and quantify their carbon balances – expanding vs. constant spatial boundaries – are compared. We show that for a conceptual forest landscape, constructed by combining a series of time‐shifted forest stands, the two approaches sometimes yield different results. We argue that the approach that uses constant spatial boundaries is preferable because it captures all carbon flows in the landscape throughout the accounting period. The approach that uses expanding system boundaries fails to accurately describe the carbon fluxes in the landscape due to incomplete coverage of carbon flows and influence of the stand‐level dynamics, which in turn arise from the way temporal system boundaries are defined on the stand level. Modelling of profit‐driven forest management using location‐specific forest data shows that the implications for carbon balance of management changes across the landscape (which are partly neglected when expanding system boundaries are used) depend on many factors such as forest structure and forest owners’ expectations of market development for bioenergy and other wood products. Assessments should not consider forest‐based bioenergy in isolation but should ideally consider all forest products and how forest management planning as a whole is affected by bioenergy incentives – and how this in turn affects carbon balances in forest landscapes and forest product pools. Due to uncertainties, we modelled several alternative scenarios for forest products markets. We recommend that future work consider alternative scenarios for other critical factors, such as policy options and energy technology pathways.     

A model of the extent and distribution of woody linear features in rural Great Britain

Hedges and lines of trees (woody linear features) are important boundaries that connect and enclose habitats, buffer the effects of land management, and enhance biodiversity in increasingly impoverished landscapes. Despite their acknowledged importance in the wider countryside, they are usually not considered in models of landscape function due to their linear nature and the difficulties of acquiring relevant data about their character, extent, and location. We present a model which uses national datasets to describe the distribution of woody linear features along boundaries in Great Britain. The method can be applied for other boundary types and in other locations around the world across a range of spatial scales where different types of linear feature can be separated using characteristics such as height or width. Satellite‐derived Land Cover Map 2007 (LCM2007) provided the spatial framework for locating linear features and was used to screen out areas unsuitable for their occurrence, that is, offshore, urban, and forest areas. Similarly, Ordnance Survey Land‐Form PANORAMA®, a digital terrain model, was used to screen out where they do not occur. The presence of woody linear features on boundaries was modelled using attributes from a canopy height dataset obtained by subtracting a digital terrain map (DTM) from a digital surface model (DSM). The performance of the model was evaluated against existing woody linear feature data in Countryside Survey across a range of scales. The results indicate that, despite some underestimation, this simple approach may provide valuable information on the extents and locations of woody linear features in the countryside at both local and national scales.     

Spatial Distribution of a Large Herbivore Community at Waterholes: An Assessment of Its Stability over Years in Hwange National Park,Zimbabwe

The spatial structuring of populations or communities is an important driver of their functioning and their influence on ecosystems. Identifying the (in)stability of the spatial structure of populations is a first step towards understanding the underlying causes of these structures. Here we studied the relative importance of spatial vs. interannual variability in explaining the patterns of abundance of a large herbivore community (8 species) at waterholes in Hwange National Park (Zimbabwe). We analyzed census data collected over 13 years using multivariate methods. Our results showed that variability in the census data was mostly explained by the spatial structure of the community, as some waterholes had consistently greater herbivore abundance than others. Some temporal variability probably linked to Park-scale migration dependent on annual rainfall was noticeable, however. Once this was accounted for, little temporal variability remained to be explained, suggesting that other factors affecting herbivore abundance over time had a negligible effect at the scale of the study. The extent of spatial and temporal variability in census data was also measured for each species. This study could help in projecting the consequences of surface water management, and more generally presents a methodological framework to simultaneously address the relative importance of spatial vs. temporal effects in driving the distribution of organisms across landscapes.     

Long-term monitoring data meet freshwater species distribution models: Lessons from an LTER-site

Long-term monitoring datasets provide a solid framework for ecological research. Such a dataset from the German long-term ecological research (LTER) site Rhine-Main-Observatory was used to set up a species distribution model (SDM) for the Kinzig catchment. The extensive knowledge on the monitoring data provided by the LTER-site framework allowed to calibrate a robust model for 175 taxa of stream macroinvertebrates and to project their distributions on the Kinzig River stream network using bioclimatic, topographical, hydrological, land use and geological predictors. On average, model performance was good, with a TSS of 0.83 (±0.09 SD) and a ROC of 0.95 (±0.03 SD). The model delivered valuable insights on three sources of bias that plague SDMs in general: (a) level of taxonomic identification of the modeled organisms, (b) the spatial arrangement of sampling sites, and (c) the sampling intensity at each sampling site. Taxonomic resolution did not affect SDM performance. The distribution of high predicted probabilities of occurrence in the stream network coincided with those segments in the stream network most densely and frequently sampled, indicating both a spatial and temporal sampling bias. Species richness curves confirmed the temporal sampling bias. Next to spatial bias, sampling frequency also plays an important role in data collection, affecting further analysis and modeling procedures. Results indicate an underrepresentation of low order streams, an important aspect that should be addressed by both monitoring schemes and modeling approaches.     

Mapping Cancer Risk in Southwestern Ontario with Changing Census Boundaries

Summary Mapping disease risk often involves working with data that have been spatially aggregated to census regions or postal regions, either for administrative reasons or confidentiality. When studying rare diseases, data must be collected over a long time period in order to accumulate a meaningful number of cases. These long time periods can result in spatial boundaries of the census regions changing over time, as is the case with the motivating example of exploring the spatial structure of mesothelioma lung cancer risk in Lambton County and Middlesex County of southwestern Ontario, Canada. This article presents a local‐EM kernel smoothing algorithm that allows for the combining of data from different spatial maps, being capable of modeling risk for spatially aggregated data with time‐varying boundaries. Inference and uncertainty estimates are carried out with parametric bootstrap procedures, and cross‐validation is used for bandwidth selection. Results for the lung cancer study are shown and discussed.     

On gene ranking using replicated microarray time course data

Summary .  Consider the ranking of genes using data from replicated microarray time course experiments, where there are multiple biological conditions, and the genes of interest are those whose temporal profiles differ across conditions. We derive a multisample multivariate empirical Bayes' statistic for ranking genes in the order of differential expression, from both longitudinal and cross-sectional replicated developmental microarray time course data. Our longitudinal multisample model assumes that time course replicates are independent and identically distributed multivariate normal vectors. On the other hand, we construct a cross-sectional model using a normal regression framework with any appropriate basis for the design matrices. In both cases, we use natural conjugate priors in our empirical Bayes' setting which guarantee closed form solutions for the posterior odds. The simulations and two case studies using published worm and mouse microarray time course datasets indicate that the proposed approaches perform satisfactorily.     

Object-oriented satellite image time series analysis using a graph-based representation

Nowadays, remote sensing technologies produce huge amounts of satellite images that can be helpful to monitor geographical areas over time. A satellite image time series (SITS) usually contains spatio-temporal phenomena that are complex and difficult to understand. Conceiving new data mining tools for SITS analysis is challenging since we need to simultaneously manage the spatial and the temporal dimensions at the same time. In this work, we propose a new clustering framework specifically designed for SITS data. Our method firstly detects spatio-temporal entities, then it characterizes their evolutions by mean of a graph-based representation, and finally it produces clusters of spatio-temporal entities sharing similar temporal behaviors. Unlike previous approaches, which mainly work at pixel-level, our framework exploits a purely object-based representation to perform the clustering task. Object-based analysis involves a segmentation step where segments (objects) are extracted from an image and constitute the element of analysis. We experimentally validate our method on two real world SITS datasets by comparing it with standard techniques employed in remote sensing analysis. We also use a qualitative analysis to highlight the interpretability of the results obtained.     

Disentangling the paradox of insect phenology: are temporal trends reflecting the response to warming?

The strength and direction of phenological responses to changes in climate have been shown to vary significantly both among species and among populations of a species, with the overall patterns not fully resolved. Here, we studied the temporal and spatial variability associated with the response of several insect species to recent global warming. We use hierarchical models within a model comparison framework to analyze phenological data gathered over 40 years by the Japan Meteorological Agency on the emergence dates of 14 insect species at sites across Japan. Contrary to what has been predicted with global warming, temporal trends of annual emergence showed a later emergence day for some species and sites over time, even though temperatures are warming. However, when emergence data were analyzed as a function of temperature and precipitation, the overall response pointed out an earlier emergence day with warmer conditions. The apparent contradiction between the response to temperature and trends over time indicates that other factors, such as declining populations, may be affecting the date phenological events are being recorded. Overall, the responses by insects were weaker than those found for plants in previous work over the same time period in these ecosystems, suggesting the potential for ecological mismatches with deleterious effects for both suites of species. And although temperature may be the major driver of species phenology, we should be cautious when analyzing phenological datasets as many other factors may also be contributing to the variability in phenology.     

Modeling Trends from North American Breeding Bird Survey Data: A Spatially Explicit Approach

Population trends, defined as interval-specific proportional changes in population size, are often used to help identify species of conservation interest. Efficient modeling of such trends depends on the consideration of the correlation of population changes with key spatial and environmental covariates. This can provide insights into causal mechanisms and allow spatially explicit summaries at scales that are of interest to management agencies. We expand the hierarchical modeling framework used in the North American Breeding Bird Survey (BBS) by developing a spatially explicit model of temporal trend using a conditional autoregressive (CAR) model. By adopting a formal spatial model for abundance, we produce spatially explicit abundance and trend estimates. Analyses based on large-scale geographic strata such as Bird Conservation Regions (BCR) can suffer from basic imbalances in spatial sampling. Our approach addresses this issue by providing an explicit weighting based on the fundamental sample allocation unit of the BBS. We applied the spatial model to three species from the BBS. Species have been chosen based upon their well-known population change patterns, which allows us to evaluate the quality of our model and the biological meaning of our estimates. We also compare our results with the ones obtained for BCRs using a nonspatial hierarchical model (Sauer and Link 2011). Globally, estimates for mean trends are consistent between the two approaches but spatial estimates provide much more precise trend estimates in regions on the edges of species ranges that were poorly estimated in non-spatial analyses. Incorporating a spatial component in the analysis not only allows us to obtain relevant and biologically meaningful estimates for population trends, but also enables us to provide a flexible framework in order to obtain trend estimates for any area.     

Monitoring large herbivore diversity at different scales: comparing direct and indirect methods

Monitoring of large herbivores is central to research and management activities in many protected areas. Monitoring programs were originally developed to estimate (trends in) population sizes of individual species. However, emphasis is shifting increasingly towards conservation of diversity and communities instead of individual species, as a growing literature shows the importance of herbivore diversity for ecosystem functioning. We argue that the design of monitoring programs has not yet been adapted well to this new conservation paradigm. Using large herbivore census data from Hluhluwe-iMfolozi Park, South Africa, we studied how monitoring methodology (observational counts vs. dung counts) and spatial scale interact in influencing estimates of large herbivore species richness and diversity. Dung counts resulted in higher herbivore species richness and diversity estimates than direct observational counts, especially at finer monitoring resolutions (grid cells smaller than 25 km²). At monitoring resolutions coarser than 25 km² both methods gave comparable diversity estimates. The methods also yielded different spatial diversity estimates, especially at finer resolutions. Grid cells with high diversity according to the dung count data did not necessarily have high diversity according to the observational counts, as shown by low correlation of grid cell values of both methods. We discuss these results in the light of estimates of the sampling effort of each method and, hence, suggest new monitoring designs that are more suitable for tracking temporal and spatial trends in large herbivore diversity and community composition.     

Including species interactions in risk assessments for global change

Most ecological risk assessments for global change are restricted to the effects of trends in climate or atmospheric carbon dioxide. In order to move beyond investigation of the effects of climate alone, the climex ^™ model was extended to investigate the effects of species interactions, in the same or different trophic levels, along environmental gradients on a geographical scale. Specific needs that were revealed during the investigations include: better treatment of the effects of temporal and spatial climatic variation; elucidation of the nature of boundaries of species ranges; data to quantify the role of species traits in interspecies interactions; integrated observational, experimental, and modelling studies on mechanisms of species interactions along environmental gradients; and high‐resolution global environmental datasets. Greater acknowledgement of the shared limitations of simplified models and experimental studies is also needed. Above all, use of the scientific method to understand representative species ranges is essential. This requires the use of mechanistic approaches capable of progressive enhancement.     

Path analysis of family resemblance with temporal trends: applications to height, weight, and Quetelet index in northeastern Brazil.

A multifactorial model incorporating temporal trends in its parameters is discussed. The model is a generalization of the tau model of Rice et al. in which the parameters are assumed to be specific functions of time. A special case of this model is fit to data on height, weight, and Quetelet index in 1,067 nuclear families to demonstrate the utility of the approach. The results indicate that there is considerable temporal variation in family resemblance over time for all three traits. For height and Quetelet index, both the transmissibility, comparable to heritability, and residual sibling environmental correlation show temporal changes, while for weight, only the latter exhibits significant trends. Trends were not found in the marital correlation for any of the traits, and only limited evidence was found for trends in the maternal transmission parameter for height. This provides an objective method for evaluating the nature and sources of temporal trends in family resemblance, which can easily be incorporated into the framework of any model-based approach.     

Cross-Checking Different Sources of Mobility Information

The pervasive use of new mobile devices has allowed a better characterization in space and time of human concentrations and mobility in general. Besides its theoretical interest, describing mobility is of great importance for a number of practical applications ranging from the forecast of disease spreading to the design of new spaces in urban environments. While classical data sources, such as surveys or census, have a limited level of geographical resolution (e.g., districts, municipalities, counties are typically used) or are restricted to generic workdays or weekends, the data coming from mobile devices can be precisely located both in time and space. Most previous works have used a single data source to study human mobility patterns. Here we perform instead a cross-check analysis by comparing results obtained with data collected from three different sources: Twitter, census, and cell phones. The analysis is focused on the urban areas of Barcelona and Madrid, for which data of the three types is available. We assess the correlation between the datasets on different aspects: the spatial distribution of people concentration, the temporal evolution of people density, and the mobility patterns of individuals. Our results show that the three data sources are providing comparable information. Even though the representativeness of Twitter geolocated data is lower than that of mobile phone and census data, the correlations between the population density profiles and mobility patterns detected by the three datasets are close to one in a grid with cells of 2×2 and 1×1 square kilometers. This level of correlation supports the feasibility of interchanging the three data sources at the spatio-temporal scales considered.     

Assessing metacommunity processes through signatures in spatiotemporal turnover of community composition

Although metacommunity ecology has been a major field of research in the last decades, with both conceptual and empirical outputs, the analysis of the temporal dynamics of metacommunities has only emerged recently and consists mostly of repeated static analyses. Here we propose a novel analytical framework to assess metacommunity processes using path analyses of spatial and temporal diversity turnovers. We detail the principles and practical aspects of this framework and apply it to simulated datasets to illustrate its ability to decipher the respective contributions of entangled drivers of metacommunity dynamics. We then apply it to four empirical datasets. Empirical results support the view that metacommunity dynamics may be generally shaped by multiple ecological processes acting in concert, with environmental filtering being variable across both space and time. These results reinforce our call to go beyond static analyses of metacommunities that are blind to the temporal part of environmental variability.     

Double‐wavelet transform for multisubject task‐induced functional magnetic resonance imaging data

The goal of this article is to model multisubject task‐induced functional magnetic resonance imaging (fMRI) response among predefined regions of interest (ROIs) of the human brain. Conventional approaches to fMRI analysis only take into account temporal correlations, but do not rigorously model the underlying spatial correlation due to the complexity of estimating and inverting the high dimensional spatio‐temporal covariance matrix. Other spatio‐temporal model approaches estimate the covariance matrix with the assumption of stationary time series, which is not always feasible. To address these limitations, we propose a double‐wavelet approach for modeling the spatio‐temporal brain process. Working with wavelet coefficients simplifies temporal and spatial covariance structure because under regularity conditions, wavelet coefficients are approximately uncorrelated. Different wavelet functions were used to capture different correlation structures in the spatio‐temporal model. The main advantages of the wavelet approach are that it is scalable and that it deals with nonstationarity in brain signals. Simulation studies showed that our method could reduce false‐positive and false‐negative rates by taking into account spatial and temporal correlations simultaneously. We also applied our method to fMRI data to study activation in prespecified ROIs in the prefontal cortex. Data analysis showed that the result using the double‐wavelet approach was more consistent than the conventional approach when sample size decreased.     

Improving bioregional frameworks for conservation by including mammal distributions

Large identifiable landscape units, such as ecoregions, are used to prioritize global and continental conservation efforts, particularly where biodiversity knowledge is inadequate. Setting biodiversity representation targets using coarse large‐scale biogeographic boundaries, can be inefficient and under‐representative. Even when using fine‐scale biodiversity data, representation deficiencies can occur through misalignment of target distributions with such prioritization frameworks. While this pattern has been recognized, quantitative approaches highlighting misalignments have been lacking, particularly for assemblages of mammal species. We tested the efficacy of Australia's bioregions as a spatial prioritization framework for representing mammal species, within protected areas, in New South Wales. We produced an approach based on mammal assemblages and assessed its performance in representing mammal distributions. Substantial spatial misalignment between New South Wales's bioregions and mammal assemblages was revealed, reflecting deficiencies in the representation of more than half of identified mammal assemblages. Using a systematic approach driven by fine‐scale mammalian data, we compared the efficacy of these two frameworks in securing mammalian representation within protected areas. Of the 61 species, 38 were better represented by the mammalian framework, with remaining species only marginally better represented when guided by bioregions. Overall, the rate at which mammal species were incorporated into the protected area network was higher (5.1% ± 0.6 sd) when guided by mammal assemblages. Guided by bioregions, systematic conservation planning of protected areas may be constrained in realizing its full potential in securing representation for all of Australia's biodiversity. Adapting the boundaries of prioritization frameworks by incorporating amassed information from a broad range of taxa should be of conservation significance.     

Granger mediation analysis of multiple time series with an application to functional magnetic resonance imaging

This paper presents Granger mediation analysis, a new framework for causal mediation analysis of multiple time series. This framework is motivated by a functional magnetic resonance imaging (fMRI) experiment where we are interested in estimating the mediation effects between a randomized stimulus time series and brain activity time series from two brain regions. The independent observation assumption is thus unrealistic for this type of time‐series data. To address this challenge, our framework integrates two types of models: causal mediation analysis across the mediation variables, and vector autoregressive (VAR) models across the temporal observations. We use “Granger” to refer to VAR correlations modeled in this paper. We further extend this framework to handle multilevel data, in order to model individual variability and correlated errors between the mediator and the outcome variables. Using Rubin's potential outcome framework, we show that the causal mediation effects are identifiable under our time‐series model. We further develop computationally efficient algorithms to maximize our likelihood‐based estimation criteria. Simulation studies show that our method reduces the estimation bias and improves statistical power, compared with existing approaches. On a real fMRI data set, our approach quantifies the causal effects through a brain pathway, while capturing the dynamic dependence between two brain regions.     

Understanding ecological change across large spatial,temporal and taxonomic scales: integrating data and methods in light of theory

The difficulty of integrating multiple theories, data and methods has slowed progress towards making unified inferences of ecological change generalizable across large spatial, temporal and taxonomic scales. However, recent progress towards a theoretical synthesis now provides a guiding framework for organizing and integrating all primary data and methods for spatiotemporal assemblage‐level inference in ecology. In this paper, we describe how recent theoretical developments can provide an organizing paradigm for linking advances in data collection and methodological frameworks across disparate ecological sub‐disciplines and across large spatial and temporal scales. First, we summarize the set of fundamental processes that determine change in multispecies assemblages across spatial and temporal scales by reviewing recent theoretical syntheses of community ecology. Second, we review recent advances in data and methods across the main sub‐disciplines concerned with ecological inference across large spatial, temporal and taxonomic scales, and organize them based on the primary fundamental processes they include, rather than the spatiotemporal scale of their inferences. Finally, we highlight how iteratively focusing on only one fundamental process at a time, but combining all relevant spatiotemporal data and methods, may reduce the conceptual challenges to integration among ecological sub‐disciplines. Moreover, we discuss a number of avenues for decreasing the practical barriers to integration among data and methods. We aim to reconcile the recent convergence of decades of thinking in community ecology and macroecology theory with the rapid progress in spatiotemporal approaches for assemblage‐level inference, at a time where a robust understanding of spatiotemporal change in ecological assemblages is more crucial than ever to conserve biodiversity.