Bayesian Maximum Entropy space/time estimation of surface water chloride in Maryland using river distances

A review of geospatial exposure models and approaches for health data integration

BACKGROUND

Geospatial methods are common in environmental exposure assessments and increasingly integrated with health data to generate comprehensive models of environmental impacts on public health.

OBJECTIVE

Our objective is to review geospatial exposure models and approaches for health data integration in environmental health applications.

METHODS

We conduct a literature review and synthesis.

RESULTS

First, we discuss key concepts and terminology for geospatial exposure data and models. Second, we provide an overview of workflows in geospatial exposure model development and health data integration. Third, we review modeling approaches, including proximity-based, statistical, and mechanistic approaches, across diverse exposure types, such as air quality, water quality, climate, and socioeconomic factors. For each model type, we provide descriptions, general equations, and example applications for environmental exposure assessment. Fourth, we discuss the approaches used to integrate geospatial exposure data and health data, such as methods to link data sources with disparate spatial and temporal scales. Fifth, we describe the landscape of open-source tools supporting these workflows.

Evaluating the performance of Bayesian geostatistical prediction with physical barriers in the Chesapeake Bay

The Chesapeake Bay is one of the most widely studied bodies of water in the United States and around the world. Routine monitoring of water quality indicators (e.g., salinity) relies on fixed sampling stations throughout the Bay. Utilizing this rich monitoring data, various methods produce surface predictions of water quality indicators to further characterize the health of the Bay as well as to support wildlife and human health research studies. Bayesian approaches for geostatistical modelling are becoming increasingly popular and can be preferred over frequentist approaches because full and exact inference can be computed, along with more accurate characterization of uncertainty. Traditional geostatistical prediction methods assume a Euclidean distance between two points when characterizing spatial dependence as a function of distance. However, Euclidean approaches may not be appropriate in estuarine environments when water-land boundaries are crossed during the modelling process. In this study, we compare stationary and barrier INLA geostatistical models with a classic kriging geostatistical model to predict salinity in the Chesapeake Bay during 4 months in 2019. Cross-validation is conducted for each approach to evaluate model performance based on prediction accuracy and precision. The results provide evidence that the two Bayesian-based models outperformed ordinary kriging, especially when examining prediction accuracy (most notably in the tributaries). We also suggest that the non-Euclidean model accounts for the appropriate water-based distances between sampling locations and is likely better at characterizing the uncertainty. However, more complex bodies of water may better showcase the capabilities and efficacy of the physical barrier INLA model.

Progression of a large syphilis outbreak in rural North Carolina through space and time: Application of a Bayesian Maximum Entropy graphical user interface

In 2001, the primary and secondary syphilis incidence rate in rural Columbus County, North Carolina was the highest in the nation. To understand the development of syphilis outbreaks in rural areas, we developed and used the Bayesian Maximum Entropy Graphical User Interface (BMEGUI) to map syphilis incidence rates from 1999-2004 in seven adjacent counties in North Carolina. Using BMEGUI, incidence rate maps were constructed for two aggregation scales (ZIP code and census tract) with two approaches (Poisson and simple kriging). The BME maps revealed the outbreak was initially localized in Robeson County and possibly connected to more urban endemic cases in adjacent Cumberland County. The outbreak spread to rural Columbus County in a leapfrog pattern with the subsequent development of a visible low incidence spatial corridor linking Roberson County with the rural areas of Columbus County. Though the data are from the early 2000s, they remain pertinent, as the combination of spatial data with the extensive sexual network analyses, particularly in rural areas gives thorough insights which have not been replicated in the past two decades. These observations support an important role for the connection of micropolitan areas with neighboring rural areas in the spread of syphilis. Public health interventions focusing on urban and micropolitan areas may effectively limit syphilis indirectly in nearby rural areas.

Microbial Find, Inform, and Test Model for Identifying Spatially Distributed Contamination Sources: Framework Foundation and Demonstration of Ruminant Bacteroides Abundance in River Sediments

Microbial pollution in rivers poses known ecological and health risks, yet causal and mechanistic linkages to sources remain difficult to establish. Host-associated microbial source tracking (MST) markers help to assess the microbial risks by linking hosts to contamination but do not identify the source locations. Land-use regression (LUR) models have been used to screen the source locations using spatial predictors but could be improved by characterizing transport (i.e., hauling, decay overland, and downstream). We introduce the microbial Find, Inform, and Test (FIT) framework, which expands previous LUR approaches and develops novel spatial predictor models to characterize the transported contributions. We applied FIT to characterize the sources of BoBac, a ruminant Bacteroides MST marker, quantified in riverbed sediment samples from Kewaunee County, Wisconsin. A 1 standard deviation increase in contributions from land-applied manure hauled from animal feeding operations (AFOs) was associated with a 77% (p-value <0.05) increase in the relative abundance of ruminant Bacteroides (BoBac-copies-per-16S-rRNA-copies) in the sediment. This is the first work finding an association between the upstream land-applied manure and the offsite bovine-associated fecal markers. These findings have implications for the sediment as a reservoir for microbial pollution associated with AFOs (e.g., pathogens and antibiotic-resistant bacteria). This framework and application advance statistical analysis in MST and water quality modeling more broadly.

Quantifying Uncertainty and Bridging the Scaling Gap in the Retrieval of Leaf Area Index by Coupling Sentinel-2 and UAV Observations

Leaf area index (LAI) estimates can inform decision-making in crop management. The European Space Agency’s Sentinel-2 satellite, with observations in the red-edge spectral region, can monitor crops globally at sub-field spatial resolutions (10–20 m). However, satellite LAI estimates require calibration with ground measurements. Calibration is challenged by spatial heterogeneity and scale mismatches between field and satellite measurements. Unmanned Aerial Vehicles (UAVs), generating high-resolution (cm-scale) LAI estimates, provide intermediary observations that we use here to characterise uncertainty and reduce spatial scaling discrepancies between Sentinel-2 observations and field surveys. We use a novel UAV multispectral sensor that matches Sentinel-2 spectral bands, flown in conjunction with LAI ground measurements. UAV and field surveys were conducted on multiple dates—coinciding with different wheat growth stages—that corresponded to Sentinel-2 overpasses. We compared chlorophyll red-edge index (CIred-edge) maps, derived from the Sentinel-2 and UAV platforms. We used Gaussian processes regression machine learning to calibrate a UAV model for LAI, based on ground data. Using the UAV LAI, we evaluated a two-stage calibration approach for generating robust LAI estimates from Sentinel-2. The agreement between Sentinel-2 and UAV CIred-edge values increased with growth stage—R2 ranged from 0.32 (stem elongation) to 0.75 (milk development). The CIred-edge variance between the two platforms was more comparable later in the growing season due to a more homogeneous and closed wheat canopy. The single-stage Sentinel-2 LAI calibration (i.e., direct calibration from ground measurements) performed poorly (mean R2 = 0.29, mean NRMSE = 17%) when compared to the two-stage calibration using the UAV data (mean R2 = 0.88, mean NRMSE = 8%). The two-stage approach reduced both errors and biases by >50%. By upscaling ground measurements and providing more representative model training samples, UAV observations provide an effective and viable means of enhancing Sentinel-2 wheat LAI retrievals. We anticipate that our UAV calibration approach to resolving spatial heterogeneity would enhance the retrieval accuracy of LAI and additional biophysical variables for other arable crop types and a broader range of vegetation cover types.

Evaluating the Efficacy and Optimal Deployment of Thermal Infrared and True-Colour Imaging When Using Drones for Monitoring Kangaroos

Advances in drone technology have given rise to much interest in the use of drone-mounted thermal imagery in wildlife monitoring. This research tested the feasibility of monitoring large mammals in an urban environment and investigated the influence of drone flight parameters and environmental conditions on their successful detection using thermal infrared (TIR) and true-colour (RGB) imagery. We conducted 18 drone flights at different altitudes on the Sunshine Coast, Queensland, Australia. Eastern grey kangaroos (Macropus giganteus) were detected from TIR (n=39) and RGB orthomosaics (n=33) using manual image interpretation. Factors that predicted the detection of kangaroos from drone images were identified using unbiased recursive partitioning. Drone-mounted imagery achieved an overall 73.2% detection success rate using TIR imagery and 67.2% using RGB imagery when compared to on-ground counts of kangaroos. We showed that the successful detection of kangaroos using TIR images was influenced by vegetation type, whereas detection using RGB images was influenced by vegetation type, time of day that the drone was deployed, and weather conditions. Kangaroo detection was highest in grasslands, and kangaroos were not successfully detected in shrublands. Drone-mounted TIR and RGB imagery are effective at detecting large mammals in urban and peri-urban environments.

Measuring Regional Atmospheric CO2 Concentrations in the Lower Troposphere with a Non-Dispersive Infrared Analyzer Mounted on a UAV, Ogata Village, Akita, Japan

We have developed a simple measuring system prototype that uses an unmanned aerial vehicle (UAV) and a non-dispersive infrared (NDIR) analyzer to detect regional carbon dioxide (CO2) concentrations and obtain vertical CO2 distributions. Here, we report CO2 measurement results for the lower troposphere above Ogata Village, Akita Prefecture, Japan (about 40° N, 140° E, approximately −1 m amsl), obtained with this UAV system. The actual flight observations were conducted at 500, 400, 300, 200, 100, and 10 m above the ground, at least once a month during the daytime from February 2018 to February 2019. The raw CO2 values from the NDIR were calibrated by two different CO2 standard gases and high-purity nitrogen (N2) gas (as a CO2 zero gas; 0 ppm). During the observation period, the maximum CO2 concentration was measured in February 2019 and the minimum in August 2018. In all seasons, CO2 concentrations became higher as the flight altitude was increased. The monthly pattern of observed CO2 changes is similar to that generally observed in the Northern Hemisphere as well as to surface CO2 changes simulated by an atmospheric transport model of the Japan Meteorological Agency. It is highly probable that these changes reflect the vegetation distribution around the study area.

Linking landscape patterns to sources of water contamination: Implications for tracking fecal contaminants with geospatial and Bayesian approaches

Microbial source tracking (MST) techniques have been designed to identify the host source of fecal contamination in water. However, current MST techniques cannot provide geographic origins of particular sources because they do not provide any spatial information beyond the points of observation. In this study, the associations between landscape patterns and the major sources of microbial contamination were examined and the application of geospatial techniques (e.g., remote sensing and geographic information systems) and Bayesian modeling was explored to track microbial sources over the landscape. The land cover information of three watersheds (the lower Dungeness Watershed, the Middle Rio Grande Watershed, and the Arroyo Burro Watershed) in the United States was obtained either by classifying high resolution satellite images or directly using land cover datasets (e.g., National Land Cover Dataset, 2006 and 2011). Then, the relationship between land use/land cover (LULC) and microbial sources from these three geographically disparate watersheds were analyzed using Bayesian hierarchical models. The results showed the predictive positive associations between human sources of fecal contamination and developed area, between dog sources and grassland, and between bird sources and water, but negative associations between human sources and forest and water areas. Furthermore, the diversity of microbial sources had positive associations with landscape fragmentation and diversity indices. This study demonstrates associations between landscape patterns and major microbial sources and offers new insight in tracking the dominant sources of fecal contamination in water using geospatial and Bayesian techniques.

Evaluation of Ground Surface Models Derived from Unmanned Aerial Systems with Digital Aerial Photogrammetry in a Disturbed Conifer Forest

Detailed vertical forest structure information can be remotely sensed by combining technologies of unmanned aerial systems (UAS) and digital aerial photogrammetry (DAP). A key limitation in the application of DAP methods, however, is the inability to produce accurate digital elevation models (DEM) in areas of dense vegetation. This study investigates the terrain modeling potential of UAS-DAP methods within a temperate conifer forest in British Columbia, Canada. UAS-acquired images were photogrammetrically processed to produce high-resolution DAP point clouds. To evaluate the terrain modeling ability of DAP, first, a sensitivity analysis was conducted to estimate optimal parameters of three ground-point classification algorithms designed for airborne laser scanning (ALS). Algorithms tested include progressive triangulated irregular network (TIN) densification (PTD), hierarchical robust interpolation (HRI) and simple progressive morphological filtering (SMRF). Points were classified as ground from the ALS and served as ground-truth data to which UAS-DAP derived DEMs were compared. The proportion of area with root mean square error (RMSE) <1.5 m were 56.5%, 51.6% and 52.3% for the PTD, HRI and SMRF methods respectively. To assess the influence of terrain slope and canopy cover, error values of DAP-DEMs produced using optimal parameters were compared to stratified classes of canopy cover and slope generated from ALS point clouds. Results indicate that canopy cover was approximately three times more influential on RMSE than terrain slope.

Geostatistical Prediction of Microbial Water Quality Throughout a Stream Network Using Meteorology, Land Cover, and Spatiotemporal Autocorrelation

Predictive modeling is promising as an inexpensive tool to assess water quality. We developed geostatistical predictive models of microbial water quality that empirically modeled spatiotemporal autocorrelation in measured fecal coliform (FC) bacteria concentrations to improve prediction. We compared five geostatistical models featuring different autocorrelation structures, fit to 676 observations from 19 locations in North Carolina's Jordan Lake watershed using meteorological and land cover predictor variables. Though stream distance metrics (with and without flow-weighting) failed to improve prediction over the Euclidean distance metric, incorporating temporal autocorrelation substantially improved prediction over the space-only models. We predicted FC throughout the stream network daily for one year, designating locations "impaired", "unimpaired", or "unassessed" if the probability of exceeding the state standard was ≥90%, ≤10%, or >10% but <90%, respectively. We could assign impairment status to more of the stream network on days any FC were measured, suggesting frequent sample-based monitoring remains necessary, though implementing spatiotemporal predictive models may reduce the number of concurrent sampling locations required to adequately assess water quality. Together, these results suggest that prioritizing sampling at different times and conditions using geographically sparse monitoring networks is adequate to build robust and informative geostatistical models of water quality impairment.