New Bayesian and deep learning spatio-temporal models can reveal anomalies in sensor data more effectively.

Staying afloat in the sensor data deluge

To investigate the feasibility of the spatiotemporal filtering model in analysis of reported schistosomiasis cases, so as to provide insights into analysis of complicated data pertaining to schistosomiasis control. Demographic and epidemiological data of reported schistosomiasis cases in Anhui Province from 1997 to 2010 were collected from Anhui Provincial Center for Disease Control and Prevention, and the annual prevalence of Schistosoma japonicum human infections was calculated. The meteorological data were captured from meteorological stations in counties (cities, districts) of Anhui Province where schistosomiasis cases were reported from 1997 to 2010 at the National Meteorological Information Center, including monthly average air temperature and precipitation. Meteorological data were interpolated using the inverse-distance weighting method, and the annual average air temperature and annual precipitation were calculated in each county (city, district). The centroid of the county (city, district) where schistosomiasis cases were reported was extracted using the software ArcGIS 10.0, and the Euclidean distance from each centroid to the Yangtze River was calculated as the distance between that county (city, district) and the Yangtze River. The global Moran's I of the prevalence of S. japonicum human infections in Anhui Province for each year from 1997 to 2010 were calculated to analyze the spatial autocorrelation. A spatial weight matrix was constructed using Rook adjacency, and a first-order temporal weight matrix was built to quantify the relationship between disease changes over time. Subsequently, a spatiotemporal structure matrix was constructed. A negative binomial model was built based on the spatiotemporal structure matrix and data pertaining to reported schistosomiasis cases, and a linear model was created between the residual of the model and candidate set feature vectors to determine the optimal subset composition of the spatiotemporal filter through stepwise regression. Then, a spatio-temporal filtering model was constructed using the negative binomial model. Negative binomial models, Bayesian spatial models, and Bayesian spatiotemporal models were constructed and compared with the spatiotemporal filtering model to validate the performance of the spatiotemporal filtering model, and cross-validation was conducted for each model. The goodness of fit was evaluated using the deviance information criterion (DIC) and Watanabe-Akaike information criterion (WAIC), and the effectiveness of model validation was assessed using mean squared error (MSE), while the accuracy of assessment results was assessed using coefficients and their 95% confidence intervals (CI), and the computational efficiency was assessed based on the running time of the model. The four feature vectors with the largest Moran's I values were selected to identify regions with autocorrelation through their schematic diagrams to investigate the differences in spatiotemporal patterns of specific regions. Of all models created, the spatiotemporal filtering model exhibited the highest goodness of fit (DIC = 3 240.70, WAIC = 3 257.80), the best model validation effectiveness (MSE = 42 617.52), and the runtime was 3.18 s, exhibiting the optimal performance. Across all modeling results, the distance from the Yangtze River showed a negative correlation with the number of reported schistosomiasis cases (coefficient values = -4.93 to -3.78, none of the 95% CIs included 0), and annual average air temperature or average precipitation posed no significant effects on numbers of reported schistosomiasis cases (both of the 95% CIs included 0). Schematic diagrams of feature vectors showed that the transmission of schistosomiasis might be associated with water systems in Anhui Province, and localized clustering patterns were primarily concentrated in the northern and western parts of schistosomiasis-endemic areas in the province. The spatiotemporal filtering model is an effective spatiotemporal analysis characterized by simple modeling, user-friendly operation, accurate results and good flexibility, which may serve as an efficient alternative to conventional complex spatiotemporal models for data analysis in schistosomiasis researches.

<p>With the plethora of open data and computational resources available, environmental data science research and applications have accelerated rapidly. Therefore, there is an opportunity for community-driven initiatives compiling and classifying open-source research and applications across environmental systems (polar, oceans, forests, agriculture, etc). Building upon the Pangeo Gallery, we propose <em>The Environmental Data Science book</em> (https://the-environmental-ds-book.netlify.app), a community-driven online resource showcasing and supporting the publication of data, research and open-source developments in environmental sciences. The target audience and early adopters are i) anyone interested in open-source tools for environmental science; and ii) anyone interested in reproducibility, inclusive, shareable and collaborative AI and data science for environmental applications. Following FAIR principles, the resource provides multiple features such as guidelines, templates, persistent URLs and Binder to facilitate a fully documented, shareable and reproducible notebooks. The quality of the published content is ensured by a transparent reviewing process supported by GitHub related technologies. To date, the community has successfully published five python-based notebooks: two forest-, two wildfires/savanna- and one polar-related research. The notebooks consume common Pangeo stack e.g. intake, iris, xarray, hvplot for interactive visualisation and modelling from Environmental sensor data. In addition to constant feature enhancements of the GitHub repository https://github.com/alan-turing-institute/environmental-ds-book, we expect to increase inclusivity (multiple languages), diversity (multiple backgrounds) and activity (collaboration and coworking sessions) towards improving scientific software practises in the environmental science community.</p>

Opioid-related overdose mortality has remained at crisis levels across the United States, increasing 5-fold and worsened during the COVID-19 pandemic. The ability to provide forecasts of opioid-related mortality at granular geographical and temporal scales may help guide preemptive public health responses. Current forecasting models focus on prediction on a large geographical scale, such as states or counties, lacking the spatial granularity that local public health officials desire to guide policy decisions and resource allocation. The overarching objective of our study was to develop Bayesian spatiotemporal dynamic models to predict opioid-related mortality counts and rates at temporally and geographically granular scales (ie, ZIP Code Tabulation Areas [ZCTAs]) for Massachusetts. We obtained decedent data from the Massachusetts Registry of Vital Records and Statistics for 2005 through 2019. We developed Bayesian spatiotemporal dynamic models to predict opioid-related mortality across Massachusetts' 537 ZCTAs. We evaluated the prediction performance of our models using the one-year ahead approach. We investigated the potential improvement of prediction accuracy by incorporating ZCTA-level demographic and socioeconomic determinants. We identified ZCTAs with the highest predicted opioid-related mortality in terms of rates and counts and stratified them by rural and urban areas. Bayesian dynamic models with the full spatial and temporal dependency performed best. Inclusion of the ZCTA-level demographic and socioeconomic variables as predictors improved the prediction accuracy, but only in the model that did not account for the neighborhood-level spatial dependency of the ZCTAs. Predictions were better for urban areas than for rural areas, which were more sparsely populated. Using the best performing model and the Massachusetts opioid-related mortality data from 2005 through 2019, our models suggested a stabilizing pattern in opioid-related overdose mortality in 2020 and 2021 if there were no disruptive changes to the trends observed for 2005-2019. Our Bayesian spatiotemporal models focused on opioid-related overdose mortality data facilitated prediction approaches that can inform preemptive public health decision-making and resource allocation. While sparse data from rural and less populated locales typically pose special challenges in small area predictions, our dynamic Bayesian models, which maximized information borrowing across geographic areas and time points, were used to provide more accurate predictions for small areas. Such approaches can be replicated in other jurisdictions and at varying temporal and geographical levels. We encourage the formation of a modeling consortium for fatal opioid-related overdose predictions, where different modeling techniques could be ensembled to inform public health policy.

BackgroundRoad traffic fatalities are a major global health burden, with climate factors such as temperature and rainfall potentially influencing crash risks. However, the impact of these factors on road traffic...

This special issue consists of a collection of articles that describe innovations in spatio-temporal methodology. Spatio-temporal statistics has been developing at a rapid pace over the past 25 years. The topic is briefly covered in Cressie's (1993) comprehensive book on spatial statistics. Subsequently, Cressie and Wikle (2011) provided the first comprehensive book on spatio-temporal statistics, but it focused primarily on linear, univariate, and Gaussian methods. More recently, there have been numerous significant advancements in non-linear, multivariate, and non-Gaussian spatio-temporal methods, particularly those suited for ‘big’ data problems. Yet, each of these topic areas is relatively underdeveloped, and there are still many research challenges. Historically, spatio-temporal statistics methodology has developed more as an extension of spatial statistics rather than time series. That is, the emphasis has been on specification of models through their second-order structure, typically within a Gaussian process framework. Alternatively, dynamic spatio-temporal models (DSTMs) have been used to model processes for which it is more realistic to think of spatial processes evolving through time, that is, when it is more reasonable to think of the process conditionally (in time) rather than marginally (as with the Gaussian process framework). DSTMs often have model forms as in classical multivariate time series models, but they present unique challenges in that the types of relationships between space and time are often driven by mechanistic processes, and the associated statistical models must attempt to accommodate these relationships. In addition, the dimensionality of the spatial components of these models often prohibits the use of classical multivariate time series methods. Although much of the development of spatio-temporal methodology has been driven by environmental, epidemiological, and ecological applications (e.g. pollution monitoring, weather forecasting, climate, oceanography, disease mapping, invasive species, animal movement, etc.), there is an increasing number of novel methodologies that are motivated by the biological sciences (e.g. brain science), federal statistics (e.g. multivariate surveys, non-Gaussian change of support), sociological statistics (e.g. crime analysis), and econometric (e.g. multivariate panel) applications. Indeed, these processes are often multivariate, non-linear, and/or non-Gaussian. The articles in this special issue provide a representative snapshot of innovative work in these areas. Specifically, the articles selected for this issue highlight non-Gaussian spatio-temporal models, computational efficiency, and/or improving parameterizations of DSTMs. In their article ‘Scalable Inference for Space-Time Gaussian Cox Processes’, Shirota and Banerjee show how one can use an efficient data augmentation approach in conjunction with nearest-neighbor log-Gaussian processes to efficiently model count (crime event) data via spatio-temporal Gaussian Cox processes. In ‘Estimating Spatial Changes Over Time of Artic Sea Ice Using Hidden 2 × 2 Tables’, Zhang and Cressie consider non-Gaussian (Bernoulli) data on the presence/absence of sea ice. The model is cast in an efficient manner by using dimension-reduced spatio-temporal processes in an EM algorithm, cleverly describing the process as time-varying 2 × 2 tables. Tagle, Castruccio, Crippa, and Genton, in their article ‘A Non-Gaussian Spatio-Temporal Model for Daily Wind Speeds Based on a Multivariate Skew-T Distribution’, show how to build an efficient and realistic ‘weather generator’ for daily wind speeds that is able to simulate the type of skewed distributions that are often found in weather variables. In their article ‘On a Semiparametric Data-Driven Nonlinear Model with Penalized Spatio-Temporal Lag Interactions’, Al-Sulami, Jiang, Lu, and Zhu consider a non-linear (semiparametric) regression spatio-temporal model that also includes a dynamic interaction term and consider an adaptive lasso approach for selecting the space–time lag interactions that are most useful (in their example, to model US housing price data). Their methodology does not rely on a Gaussian error assumption. In the spirit of efficient modeling of the spatio-temporal terms beyond a trend and seasonality, Gao and Tsay present an efficient structured factor approach for multivariate time series and spatio-temporal data (e.g. PM2.5 pollutant observations) in their article ‘A Structural-Factor Approach to Modeling High-Dimensional Time Series and Space-Time Data’. Finally, in their article ‘Spatio-Temporal Models for Big Multinomial Data Using the Conditional Multivariate Logit-Beta Distribution’, Bradley, Wikle and Holan develop a novel efficient conjugate Bayesian algorithm to model high-dimensional spatio-temporal multinomial data, which includes a spatio-temporal dynamic process. They apply this model to public-use Quarterly Workforce Indicators from the Longitudinal Employer Household Dynamics program of the US Census Bureau. These articles provide excellent examples of how spatio-temporal statistical models can be applied to address complex non-Gaussian, non-linear, and multivariate data with efficient computational algorithms. These problems are by no means ‘solved’, but it is our hope that the readers of the Journal of Time Series Analysis may see connections to their own work that can be applied to this rich set of problems and this growing and important area of statistics. Finally, we would like to end this Editorial by thanking Prof. A. M. R. Taylor for encouraging us to edit this special issue.

California's Central Valley region has been called the “bread‐basket” of the United States. The region is home to one of the most productive agricultural systems on the planet. Such high levels of agricultural productivity require large amounts of fresh water for irrigation. However, the long‐term availability of water required to sustain high levels of agricultural production is being called into question following the latest drought in California. In this paper, we use Bayesian multilevel spatiotemporal modeling techniques to examine the influence of the structure of surface water rights in the Central Valley on agricultural production during the recent drought. California is an important place to study these dynamics as it is the only state to recognize the two dominant approaches to surface water management in the United States: riparian and appropriative rights. In this study, Bayesian spatiotemporal modeling is employed to account for spatial processes that have the potential to influence the effects of water right structures on agricultural production. Results suggest that, after accounting for spatiotemporal dependencies in the data, seniority in surface water access significantly improves crop health and productivity on cultivated lands but does not independently affect the ability to maintain cultivated extent. In addition, agricultural productivity in watersheds with more junior surface water rights shows less sensitivity to cumulative drought exposure than other watersheds, however the extent of cultivation in these same watersheds is relatively more sensitive to cumulative drought exposure.

Modelling drivers of trawl fisheries discards using Bayesian spatio-temporal models

BackgroundMalaria is a significant public health problem, particularly among children aged 6–59 months who bear the greatest burden of the disease. Malaria transmission is high and more pronounced in poor tropical and subtropical areas of the world. Climate change is positively correlated with the geographical distribution of malaria vectors. There is substantial evidence of spatial and temporal differences in under-five malaria risk. Thus, the study aimed to create intelligent maps of smooth relative risk of malaria in children under-5 years in Ghana that highlight high and low malaria burden in space and time to support malaria prevention, control, and elimination efforts.MethodThe study extracted and merged data on malaria among children aged 6–59 months from the 2014 Ghana Demographic and Health Surveys (GDHS), 2016 and 2019 Ghana Malaria Indicator Surveys (GMIS). The outcome variable of interest was the count of children aged 6–59 months with a positive test on the rapid diagnostic test (RDT) result. Bayesian Hierarchical spatiotemporal models were specified to estimate and map spatiotemporal variations in the relative risk of malaria. The existence of local clustering was assessed using the local indicator of spatial association (LISA), and the points were mapped to display significant local clusters, hotpot, and cold spot communities.ResultsThe number of positive malaria cases in children aged 6–59 months decreased marginally from 946.7 (36.4%) in 2014 to 603.6 (22.9) in 2019 DHS survey periods. Smooth relative risk of malaria among children aged 6–59 months has consistently increased in the Northern and Eastern regions between 2014 and 2019. Socioeconomic and climatic factors such as household size [Posterior Mean: -0.198 (95% CrI: 3.52, 80.95)], rural area [Posterior Mean: 1.739 (95% CrI: 0.581, 2.867)], rainfall [Posterior Mean: 0.003 (95% CrI: 0.001, 0.005)], and maximum temperature [Posterior Mean: -1.069 (95% CrI: -2.135, -0.009)] were all shown as statistically significant predictors of malaria risk in children aged 6–59 months. Hot spot DHS clusters (enumeration areas) with a significantly high relative risk of malaria among children aged 6–59 months were repeatedly detected in the Ashanti region between 2014 and 2019.ConclusionThe findings of the study would provide policymakers with practical and insightful information for the equitable distribution of scarce health resources targeted at reducing the burden of malaria and its associated mortality among children under five years.

With technological advances, the sources of available information have become more and more diverse. Recently, a new source of information has gained growing importance: sensor data. Sensors are devices sensing their environment in various ways and reporting in general a numeric result. A sensor continuously reports values, thus the flow of information is also continuous, like a stream. As the field has developed, the usage paradigm has shifted from stand-alone sensors to interconnected sensors, or sensor networks. Sensors became more complex, generating larger quantities of data and having wireless communication modules for transmitting their data. Initially, data from sensor networks was first stored, and then processed. Thus, classical database technologies could be used. However, the focus has soon shifted towards reacting to sensor data in real time. A user query reacting in real time to a stream of data is called a continuous query, and to answer such a query requires that it is continuously processed, as new values appear in the sensor stream. As sensor networks and sensor based applications become more popular, users identified the need to query sensor data pertaining to different sensor networks. This setting, of interconnected sensor networks, consists of more powerful computational devices, connected with a wired communication, which can process and relay sensor data. Users can launch queries at any node to query sensor events coming from any part of the interconnected network. In this setting, the number of data sources (sensors) is orders of magnitude smaller than the number of user queries, which themselves are orders of magnitude smaller than the full content of the (sensor) data streams, and the communication becomes by far the greatest communication bottleneck. In this thesis, we present our research for reducing communication cost generated by applications accessing large scale interconnected sensor networks. Our first contribution is a probabilistic algorithm for detecting and exploiting subsumption of queries over correlated data sources. This technique reduces the communication traffic generated by query forwarding in an interconnected sensor network, by filtering out queries subsumed by a set of existing queries. In addition, this reduces the number of results that need to be transmitted. We propose an efficient forwarding algorithm of the elements of the result sets, by employing a publish/subscribe data dissemination. To support the general setting of distributed data sources in an interconnected sensor network, we propose a Filter-Split-Forward approach that adapts set subsumption to the case of join queries over distributed data sources. We base our approach on the concept of filter-split-forward phases for efficient query filtering and placement inside the network, and an efficient, publish/subscribe forwarding of matching events. We also propose distributed adaptations of state of the art solutions for continuous query processing over multiple data sources. We adapt these techniques to require only local interactions between nodes, without relying on global knowledge or a centralized server. We show how our approach achieves lower traffic through query subsumption and efficient event dissemination. In many applications using sensor data, users are only interested in the most relevant events. To that end, we present our solutions for processing top-k queries over distributed sensor data streams in the presence of query subsumption. We analyze the impact of query subsumption on top-k processing. We propose different strategies for incorporating query subsumption into top-k processing, in order to obtain sufficiently accurate result sets, while keeping network traffic low. We show that the best tradeoff is achieved by updating throughout the network the values of k for the queries resulting from splitting a query between nodes and also for the set of queries subsuming a query. By this work we contribute a framework for increasing the efficiency of continuous query processing over distributed data sources for a wide range of applications, such as environmental and living spaces monitoring, network and traffic monitoring, and in general for all sensor enhanced monitoring applications.

BackgroundCOVID-19 brought enormous challenges to public health surveillance and underscored the importance of developing and maintaining robust systems for accurate surveillance. As public health data collection efforts expand, there is a critical need for infectious disease modeling researchers to continue to develop prospective surveillance metrics and statistical models to accommodate the modeling of large disease counts and variability. This paper evaluated different likelihoods for the disease count model and various spatiotemporal mean models for prospective surveillance.MethodsWe evaluated Bayesian spatiotemporal models, which are the foundation for model-based infectious disease surveillance metrics. Bayesian spatiotemporal mean models based on the Poisson and the negative binomial likelihoods were evaluated with the different lengths of past data usage. We compared their goodness of fit and short-term prediction performance with both simulated epidemic data and real data from the COVID-19 pandemic.ResultsThe simulation results show that the negative binomial likelihood-based models show better goodness of fit results than Poisson likelihood-based models as deemed by smaller deviance information criteria (DIC) values. However, Poisson models yield smaller mean square error (MSE) and mean absolute one-step prediction error (MAOSPE) results when we use a shorter length of the past data such as 7 and 3 time periods. Real COVID-19 data analysis of New Jersey and South Carolina shows similar results for the goodness of fit and short-term prediction results. Negative binomial-based mean models showed better performance when we used the past data of 52 time periods. Poisson-based mean models showed comparable goodness of fit performance and smaller MSE and MAOSPE results when we used the past data of 7 and 3 time periods.ConclusionWe evaluate these models and provide future infectious disease outbreak modeling guidelines for Bayesian spatiotemporal analysis. Our choice of the likelihood and spatiotemporal mean models was influenced by both historical data length and variability. With a longer length of past data usage and more over-dispersed data, the negative binomial likelihood shows a better model fit than the Poisson likelihood. However, as we use a shorter length of the past data for our surveillance analysis, the difference between the Poisson and the negative binomial models becomes smaller. In this case, the Poisson likelihood shows robust posterior mean estimate and short-term prediction results.

Hand, foot, and mouth disease (HFMD) shows spatiotemporal heterogeneity in China. A spatiotemporal filtering model was constructed and applied to HFMD data to explore the underlying spatiotemporal structure of the disease and determine the impact of different spatiotemporal weight matrices on the results. HFMD cases and covariate data in East China were collected between 2009 and 2015. The different spatiotemporal weight matrices formed by Rook, K-nearest neighbour (KNN; K = 1), distance, and second-order spatial weight matrices (SO-SWM) with first-order temporal weight matrices in contemporaneous and lagged forms were decomposed, and spatiotemporal filtering model was constructed by selecting eigenvectors according to MC and the AIC. We used MI, standard deviation of the regression coefficients, and five indices (AIC, BIC, DIC, R2, and MSE) to compare the spatiotemporal filtering model with a Bayesian spatiotemporal model. The eigenvectors effectively removed spatial correlation in the model residuals (Moran’s I < 0.2, p > 0.05). The Bayesian spatiotemporal model’s Rook weight matrix outperformed others. The spatiotemporal filtering model with SO-SWM was superior, as shown by lower AIC (92,029.60), BIC (92,681.20), and MSE (418,022.7) values, and higher R2 (0.56) value. All spatiotemporal contemporaneous structures outperformed the lagged structures. Additionally, eigenvector maps from the Rook and SO-SWM closely resembled incidence patterns of HFMD.

Advancements in Bayesian spatial and spatio-temporal modelling have been observed in recent years. Despite this, there are unresolved issues about the choice of appropriate spatial unit and adjacency matrix in disease mapping. There is limited systematic review evidence on this topic. This review aimed to address these problems. We searched seven databases to find published articles on this topic. A modified quality assessment tool was used to assess the quality of studies. A total of 52 studies were included, of which 26 (50.0%) were on infectious diseases, 10 (19.2%) on chronic diseases, 8 (15.5%) on maternal and child health, and 8 (15.5%) on other health-related outcomes. Only 6 studies reported the reasons for using the specified spatial unit, 8 (15.3%) studies conducted sensitivity analysis for prior selection, and 39 (75%) of the studies used Queen contiguity adjacency. This review highlights existing variation and limitations in the specification of Bayesian spatial and spatio-temporal models used in health research. We found that majority of the studies failed to report the rationale for the choice of spatial units, perform sensitivity analyses on the priors, or evaluate the choice of neighbourhood adjacency, all of which can potentially affect findings in their studies.

BackgroundMeasles-containing vaccine (MCV) has been effective in controlling the spread of measles. Some countries have declared measles elimination. But recently years, the number of cases worldwide has increased, posing a challenge to the global goal of measles eradication. This study estimated the relationship between meteorological factors and measles using spatiotemporal Bayesian model, aiming to provide scientific evidence for public health policy to eliminate measles.MethodsDescriptive statistical analysis was performed on monthly data of measles and meteorological variables in 136 counties of Shandong Province from 2009 to 2017. Spatiotemporal Bayesian model was used to estimate the effects of meteorological factors on measles, and to evaluate measles risk areas at county level. Case population was divided into multiple subgroups according to gender, age and occupation. The effects of meteorological factors on measles in subgroups were compared.ResultsSpecific meteorological conditions increased the risk of measles, including lower relative humidity, temperature, and atmospheric pressure; higher wind velocity, sunshine duration, and diurnal temperature variation. Taking lowest value (Q1) as reference, RR (95%CI) for higher temperatures (Q2–Q4) were 0.79 (0.69–0.91), 0.54 (0.44–0.65), and 0.48 (0.38–0.61), respectively; RR (95%CI) for higher relative humidity (Q2–Q4) were 0.76 (0.66–0.88), 0.56 (0.47–0.67), and 0.49 (0.38–0.63), respectively; RR (95%CI) for higher wind velocity (Q2–Q4) were 1.43 (1.25–1.64), 1.85 (1.57–2.18), 2.00 (1.59–2.52), respectively. 22 medium-to-high risk counties were identified, mainly in northwestern, southwestern and central Shandong Province. The trend was basically same in the effects of meteorological factors on measles in subgroups, but the magnitude of the effects was different.ConclusionsMeteorological factors have an important impact on measles. It is crucial to integrate these factors into public health policies for measles prevention and control in China.

Evaluating consumer and clinical sleep technologies: an American Academy of Sleep Medicine update