Spatial Covariates Research Articles

Spatiotemporal modeling of ecological and sociological predictors of West Nile virus in Suffolk County, NY, mosquitoes.

Suffolk County, New York, is a locus for West Nile virus (WNV) infection in the American northeast that includes the majority of Long Island to the east of New York City. The county has a system of light and gravid traps used for mosquito collection and disease monitoring. In order to identify predictors of WNV incidence in mosquitoes and predict future occurrence of WNV, we have developed a spatiotemporal Bayesian model, beginning with over 40 ecological, meteorological, and built-environment covariates. A mixed-effects model including spatially and temporally correlated errors was fit to WNV surveillance data from 2008 to 2014 using the R package “R-INLA,” which allows for Bayesian modeling using the stochastic partial differential equation (SPDE) approach. The integrated nested Laplace approximation (INLA) SPDE allows for simultaneous fitting of a temporal parameter and a spatial covariance, while incorporating a variety of likelihood functions and running in R statistical software on a home computer. We found that land cover classified as open water and woody wetlands had a negative association with WNV incidence in mosquitoes, and the count of septic systems was associated with an increase in WNV. Mean temperature at two-week lag was associated with a strong positive impact, while mean precipitation at no lag and one-week lag was associated with positive and negative impacts on WNV, respectively. Incorporation of spatiotemporal factors resulted in a marked increase in model goodness-of-fit. The predictive power of the model was evaluated on 2015 surveillance results, where the best model achieved a sensitivity of 80.9% and a specificity of 77.0%. The spatial covariate was mapped across the county, identifying a gradient of WNV prevalence increasing from east to west. The Bayesian spatiotemporal model improves upon previous approaches, and we recommend the INLA SPDE methodology as an efficient way to develop robust models from surveillance data to develop and enhance monitoring and control programs. Our study confirms previously found associations between weather conditions and WNV and suggests that wetland cover has a mitigating effect on WNV infection in mosquitoes, while high septic system density is associated with an increase in WNV infection.

Layer-specific spatial prediction of As concentration in copper smelter vicinity considering the terrain exposure

Prevailing climatic conditions and local topography can be classified as the most influential environmental factors that affect the spatial dispersion of pollutants emanating from industrial sources. In this study, the combined effects of these factors were considered with respect to terrain exposure in order to explain the complex spatial trend of Arsenic (As) concentration that was atmospherically-deposited from one of the largest Copper Mining and Smelting Complexes in Europe, Bor in Serbia. Several exposure parameters were created and employed as spatial covariates within the so-called “Spline-Then-Krige” approach for producing maps of As concentration at three standard soil depth layers (0–5cm, 5–15cm and 15–30cm). The exposure parameters were created to quantify two different aspects of terrain exposure: Geometrical (Proximity) and Topographical exposure. Regression analysis confirmed the presence of a significant statistical association between the As data and all exposure parameters. The trend model showed good overall accuracy explaining 52% of the variance in As data for the surface soil layer, 49% for the middle layer and 35% for the deepest layer. Relative importance analysis revealed the importance of considering a more general model that includes interactions between exposure parameters. The kriging interpolation improved, to some extent, the regression accuracy for all three layers with R2 values ranging from 55% for the surface layer to the 36% for the deepest soil layer. The prediction maps show that As contamination levels are well above allowable Serbian agricultural concentration limits (As<25mg/kg) for approximately 78% of the mapping area, thereby indicating that long term smelting activity leaves significant consequences on soil even on deeper unexposed layers.

A hierarchical model for estimating the spatial distribution and abundance of animals detected by continuous-time recorders.

MotivationSeveral spatial capture-recapture (SCR) models have been developed to estimate animal abundance by analyzing the detections of individuals in a spatial array of traps. Most of these models do not use the actual dates and times of detection, even though this information is readily available when using continuous-time recorders, such as microphones or motion-activated cameras. Instead most SCR models either partition the period of trap operation into a set of subjectively chosen discrete intervals and ignore multiple detections of the same individual within each interval, or they simply use the frequency of detections during the period of trap operation and ignore the observed times of detection. Both practices make inefficient use of potentially important information in the data.Model and data analysisWe developed a hierarchical SCR model to estimate the spatial distribution and abundance of animals detected with continuous-time recorders. Our model includes two kinds of point processes: a spatial process to specify the distribution of latent activity centers of individuals within the region of sampling and a temporal process to specify temporal patterns in the detections of individuals. We illustrated this SCR model by analyzing spatial and temporal patterns evident in the camera-trap detections of tigers living in and around the Nagarahole Tiger Reserve in India. We also conducted a simulation study to examine the performance of our model when analyzing data sets of greater complexity than the tiger data.BenefitsOur approach provides three important benefits: First, it exploits all of the information in SCR data obtained using continuous-time recorders. Second, it is sufficiently versatile to allow the effects of both space use and behavior of animals to be specified as functions of covariates that vary over space and time. Third, it allows both the spatial distribution and abundance of individuals to be estimated, effectively providing a species distribution model, even in cases where spatial covariates of abundance are unknown or unavailable. We illustrated these benefits in the analysis of our data, which allowed us to quantify differences between nocturnal and diurnal activities of tigers and to estimate their spatial distribution and abundance across the study area. Our continuous-time SCR model allows an analyst to specify many of the ecological processes thought to be involved in the distribution, movement, and behavior of animals detected in a spatial trapping array of continuous-time recorders. We plan to extend this model to estimate the population dynamics of animals detected during multiple years of SCR surveys.

Examining the occupancy–density relationship for a low‐density carnivore

Summary The challenges associated with monitoring low‐density carnivores across large landscapes have limited the ability to implement and evaluate conservation and management strategies for such species. Non‐invasive sampling techniques and advanced statistical approaches have alleviated some of these challenges and can even allow for spatially explicit estimates of density, one of the most valuable wildlife monitoring tools. For some species, individual identification comes at no cost when unique attributes (e.g. pelage patterns) can be discerned with remote cameras, while other species require viable genetic material and expensive laboratory processing for individual assignment. Prohibitive costs may still force monitoring efforts to use species distribution or occupancy as a surrogate for density, which may not be appropriate under many conditions. Here, we used a large‐scale monitoring study of fisher Pekania pennanti to evaluate the effectiveness of occupancy as an approximation to density, particularly for informing harvest management decisions. We combined remote cameras with baited hair snares during 2013–2015 to sample across a 70 096‐km2 region of western New York, USA. We fit occupancy and Royle–Nichols models to species detection–non‐detection data collected by cameras, and spatial capture–recapture (SCR) models to individual encounter data obtained by genotyped hair samples. Variation in the state variables within 15‐km2 grid cells was modelled as a function of landscape attributes known to influence fisher distribution. We found a close relationship between grid cell estimates of fisher state variables from the models using detection–non‐detection data and those from the SCR model, likely due to informative spatial covariates across a large landscape extent and a grid cell resolution that worked well with the movement ecology of the species. Fisher occupancy and density were both positively associated with the proportion of coniferous‐mixed forest and negatively associated with road density. As a result, spatially explicit management recommendations for fisher were similar across models, though relative variation was dampened for the detection–non‐detection data. Synthesis and applications. Our work provides empirical evidence that models using detection–non‐detection data can make similar inferences regarding relative spatial variation of the focal population to models using more expensive individual encounters when the selected spatial grain approximates or is marginally smaller than home range size. When occupancy alone is chosen as a cost‐effective state variable for monitoring, simulation and sensitivity analyses should be used to understand how inferences from detection–non‐detection data will be affected by aspects of study design and species ecology.

Spatial clustering of curves with an application of satellite data

Water quality indicators are important to identify risks to the environment, society and human health. The European Community Water Framework Directive establishes guidelines for the classification of all water bodies across Europe and chemical and biological indicators were used to this scope. In particular, the Chlorophyll type A index (Chl-a) is a shared indicator of trophic status and monitoring activities may be useful to explain its spatial distribution and to discover local dangerous behaviours (for example the anoxic events). Differently by the classical approach based on an “average” values over a period, we propose a functional clustering model that takes into account temporal and spatial dependence of Chl-a concentrations in the Adriatic Sea for defining appropriate clusters of sites. We use satellite monthly data, during the period 2002–2012, and we model the spatial dependence among the sites by means of a Markov random field model. Compared to similar attempts in literature by Jiang and Serban (2012) our formulation includes spatial covariates. This inclusion allows for more flexibility to obtain more homogeneous and representative clusters of sites in the Adriatic Sea. The estimation of the model and the identification of the number of clusters are carried out using a pseudolikelihood function. A small simulation study complements the real data analysis.

Comparing the efficiency of digital and conventional soil mapping to predict soil types in a semi-arid region in Iran

The efficiency of different digital and conventional soil mapping approaches to produce categorical maps of soil types is determined by cost, sample size, accuracy and the selected taxonomic level. The efficiency of digital and conventional soil mapping approaches was examined in the semi-arid region of Borujen, central Iran. This research aimed to (i) compare two digital soil mapping approaches including Multinomial logistic regression and random forest, with the conventional soil mapping approach at four soil taxonomic levels (order, suborder, great group and subgroup levels), (ii) validate the predicted soil maps by the same validation data set to determine the best method for producing the soil maps, and (iii) select the best soil taxonomic level by different approaches at three sample sizes (100, 80, and 60 point observations), in two scenarios with and without a geomorphology map as a spatial covariate. In most predicted maps, using both digital soil mapping approaches, the best results were obtained using the combination of terrain attributes and the geomorphology map, although differences between the scenarios with and without the geomorphology map were not significant. Employing the geomorphology map increased map purity and the Kappa index, and led to a decrease in the ‘noisiness’ of soil maps. Multinomial logistic regression had better performance at higher taxonomic levels (order and suborder levels); however, random forest showed better performance at lower taxonomic levels (great group and subgroup levels). Multinomial logistic regression was less sensitive than random forest to a decrease in the number of training observations. The conventional soil mapping method produced a map with larger minimum polygon size because of traditional cartographic criteria used to make the geological map 1:100,000 (on which the conventional soil mapping map was largely based). Likewise, conventional soil mapping map had also a larger average polygon size that resulted in a lower level of detail. Multinomial logistic regression at the order level (map purity of 0.80), random forest at the suborder (map purity of 0.72) and great group level (map purity of 0.60), and conventional soil mapping at the subgroup level (map purity of 0.48) produced the most accurate maps in the study area. The multinomial logistic regression method was identified as the most effective approach based on a combined index of map purity, map information content, and map production cost. The combined index also showed that smaller sample size led to a preference for the order level, while a larger sample size led to a preference for the great group level.

Drivers of human-caused fire occurrence and its variation trend under climate change in the Great Xing'an Mountains, Northeast China

The Great Xing'an Mountains are an important boreal forest region in China with high frequency of fire occurrences. With climate change, this region may have a substantial change in fire frequency. Building the relationship between spatial pattern of human-caused fire occurrence and its influencing factors, and predicting the spatial patterns of human-caused fires under climate change scenarios are important for fire management and carbon balance in boreal forests. We employed a spatial point pattern model to explore the relationship between the spatial pattern of human-caused fire occurrence and its influencing factors based on a database of historical fire records (1967-2006) in the Great Xing'an Mountains. The fire occurrence time was used as dependent variable. Nine abiotic (annual temperature and precipitation, elevation, aspect, and slope), biotic (vegetation type), and human factors (distance to the nearest road, road density, and distance to the nearest settlement) were selected as explanatory variables. We substituted the climate scenario data (RCP 2.6 and RCP 8.5) for the current climate data to predict the future spatial patterns of human-caused fire occurrence in 2050. Our results showed that the point pattern progress (PPP) model was an effective tool to predict the future relationship between fire occurrence and its spatial covariates. The climatic variables might significantly affect human-caused fire occurrence, while vegetation type, elevation and human variables were important predictors of human-caused fire occurrence. The human-caused fire occurrence probability was expected to increase in the south of the area, and the north and the area along the main roads would also become areas with high human-caused fire occurrence. The human-caused fire occurrence would increase by 72.2% under the RCP 2.6 scenario and by 166.7% under the RCP 8.5 scenario in 2050. Under climate change scenarios, the spatial patterns of human-caused fires were mainly influenced by the climate and human factors.

A comparative study on disease risk model in exploratory spatial analysis

The present work mainly focuses on the issue of risk model in spatial data analysis. Through the analysis of morbidity data of influenza A (H1N1) across China's administrative regions from 2009 to 2012, a comparative study was carried out among Poisson model, Poisson-Gamma model, log-normal model, EB estimator of moment and Bayesian hierarchical model. By using R programming language, the feasibility of the above analysis methods was verified and the variability of the estimated values generated by each model was calculated, the Bayesian model for spatial disease analysis was improved, and estimator considering uncorrelated spatial model, correlated spatial model and covariate factors was proved to be the best by comparing DIC values of the models. By using the Markov chain for simulative iteration, iterative convergence was illustrated by graphs of iteration track, autocorrelation function, kernel density and quantile estimation. The research on spatial variability of disease morbidity is helpful in detecting epidemic areas and forewarning the pathophoresis of prospective epidemic disease.

A comparative study on disease risk model in exploratory spatial analysis

The present work mainly focuses on the issue of risk model in spatial data analysis. Through the analysis of morbidity data of influenza A (H1N1) across China's administrative regions from 2009 to 2012, a comparative study was carried out among Poisson model, Poisson-Gamma model, log-normal model, EB estimator of moment and Bayesian hierarchical model. By using R programming language, the feasibility of the above analysis methods was verified and the variability of the estimated values generated by each model was calculated, the Bayesian model for spatial disease analysis was improved, and estimator considering uncorrelated spatial model, correlated spatial model and covariate factors was proved to be the best by comparing DIC values of the models. By using the Markov chain for simulative iteration, iterative convergence was illustrated by graphs of iteration track, autocorrelation function, kernel density and quantile estimation. The research on spatial variability of disease morbidity is helpful in detecting epidemic areas and forewarning the pathophoresis of prospective epidemic disease.

Single-index model for inhomogeneous spatial point processes

Abstract: We introduce a single index model for the intensity of an inhomogeneous spatial point process, relating the intensity function to an unknown function ρ of a linear combination of measurements of a p-dimensional spatial covariate process. Such a model extends and generalizes a commonly used model where ρ is known. We derive an estimating procedure for ρ and the coefficient parameters β and show consistency and asymptotic normality of estimates of β under some regularity assumptions. We present results of some simulation studies showing the effectiveness of the procedure. Finally, we apply the procedure to a dataset of fast food restaurant locations in New York City.

Anthropogenic impacts in protected areas: assessing the efficiency of conservation efforts using Mediterranean ant communities.

In countries with high levels of urbanization, protected areas are often subject to human disturbance. In addition to dealing with fragmentation, land managers also have to confront the loss of characteristic ecosystems due to biotic homogenization, which is the increasing similarity of species assemblages among geographically separate regions. Using ants as a model system, we explored whether anthropogenic factors negatively affect biodiversity of protected areas of a regional network. We first analysed the effect of fragmentation and human activity on ant biodiversity within protected areas. Secondly, we tested whether homogenization could occur among protected areas. We sampled 79 plots in the most common habitats of 32 protected areas in southern Spain and calculated ant community richness and diversity indices, endemic richness, and Bray–Curtis similarity indices (between pairs of plots). We related these indices with patch fragmentation and human disturbance variables, taking into account environmental, spatial and landscape covariates. We used ANOSIM to test for differences between similarity indices, specifically among levels of anthropogenic disturbance. Species richness was positively correlated with the distance from the border of the protected areas and the number of endemic species was negatively correlated with the degree of fragmentation. Ant communities were similar within each protected area but differed across regions. Human disturbance was not correlated with community similarity among sampling points. Our approach suggests how the ability of European protected areas to sustain biodiversity is limited because they remain susceptible to anthropogenic impacts. Although ant communities maintained their biological distinctiveness, we reveal how fragmentation within protected areas is important for community richness and endemism maintenance.

Patterns of red tree vole distribution and habitat suitability: implications for surveys and conservation planning

AbstractEnvironmental regulations often require wildlife surveys prior to habitat disturbance to avoid impacts or as the basis for planning mitigation, yet project‐level surveys may not provide the insights needed to guide long‐term management. Management of the red tree vole (Arborimus longicaudus) has largely been based on such surveys. As an alternative approach, we evaluated distribution patterns using frequency of red tree vole occurrence and habitat suitability models to guide conservation planning. We developed a suite of models based on subsets of covariates from two previously developed models and evaluated the extent to which spatial covariates improved the models. We used presence–absence data that were collected from 364 randomly selected 1‐ha Current Vegetation Survey and Forest Inventory and Analysis plots to develop models and describe occurrence patterns. The best models included a spatial covariate, maximum tree diameter, distance from suitable habitat, forest age class, and the interaction between maximum tree diameter and forest age class. We compared performance of the previously published models, our best model, and an ensemble model that used predictions from all three models. Under the ensemble model, correct classification rates were relatively high and considerably improved, suggesting that the application of all three models provided greater accuracy than any individual model. We argue that habitat models, coupled with spatial patterns of the frequency of occurrence, can provide useful tools for addressing species management and may provide more insight than project‐level surveys. The use of habitat suitability models can therefore be closely tied to red tree vole management decisions and conservation strategies, as well as reducing survey costs that otherwise often make projects infeasible.

Mapping peat in Scotland with remote sensing and site characteristics

Mapping the spatial distribution of organic soil is important for decision making in land management. Organic soil types have a strong effect on carbon storage, water availability and quality, biodiversity and many other ecosystem services. A large proportion of Scotland's soil is organic, and is classed as peats or organomineral. Existing soil maps for Scotland include a 1:25 000 map of eastern and southern parts of the country that identifies the major soil subgroup at each location, or a 1:250 000 map covering the whole country that gives proportions of the major soil group within mapping units. A detailed map of the organic soils of Scotland that disaggregates these mapping units is therefore required. Existing data were used from several Scottish soil surveys and remote sensing data (Landsat 8) to develop a model for predicting whether a site has mineral, organomineral (mineral with peaty topsoil) or organic (peat) soil. The model was applied across Scotland at a grid resolution of 100 m. Predictive accuracies of 68.1% (mineral), 61.3% (organomineral) and 81.8% (organic) were obtained, with an overall prediction accuracy of 69.8%. Training the model with two classes only, non‐organic (including mineral and organomineral) and organic, gave accuracies of 83.9 and 93.6%, respectively, and an overall accuracy of 86.4%. A map of prediction confidence was also produced. The results show how the approach can be used to disaggregate the existing map units and predict mineral and organic soils at locations within a map unit. The map is being used to prioritize monitoring efforts in catchments across Scotland.Highlights Use of remote sensing and spatial covariates to map peat in Scotland. Resulting peat maps will be useful for decision making on land management and carbon storage in soil. Peat is mapped accurately in most areas, although uncertainty is greater in complex landscapes. Combination of remote sensing and spatial covariates improves accuracy of mapping.

Regionalization of monthly rainfall erosivity patterns in Switzerland

Abstract. One major controlling factor of water erosion is rainfall erosivity, which is quantified as the product of total storm energy and a maximum 30 min intensity (I30). Rainfall erosivity is often expressed as R-factor in soil erosion risk models like the Universal Soil Loss Equation (USLE) and its revised version (RUSLE). As rainfall erosivity is closely correlated with rainfall amount and intensity, the rainfall erosivity of Switzerland can be expected to have a regional characteristic and seasonal dynamic throughout the year. This intra-annual variability was mapped by a monthly modeling approach to assess simultaneously spatial and monthly patterns of rainfall erosivity. So far only national seasonal means and regional annual means exist for Switzerland. We used a network of 87 precipitation gauging stations with a 10 min temporal resolution to calculate long-term monthly mean R-factors. Stepwise generalized linear regression (GLM) and leave-one-out cross-validation (LOOCV) were used to select spatial covariates which explain the spatial and temporal patterns of the R-factor for each month across Switzerland. The monthly R-factor is mapped by summarizing the predicted R-factor of the regression equation and the corresponding residues of the regression, which are interpolated by ordinary kriging (regression–kriging). As spatial covariates, a variety of precipitation indicator data has been included such as snow depths, a combination product of hourly precipitation measurements and radar observations (CombiPrecip), daily Alpine precipitation (EURO4M-APGD), and monthly precipitation sums (RhiresM). Topographic parameters (elevation, slope) were also significant explanatory variables for single months. The comparison of the 12 monthly rainfall erosivity maps showed a distinct seasonality with the highest rainfall erosivity in summer (June, July, and August) influenced by intense rainfall events. Winter months have the lowest rainfall erosivity. A proportion of 62 % of the total annual rainfall erosivity is identified within four months only (June–September). The highest erosion risk can be expected in July, where not only rainfall erosivity but also erosivity density is high. In addition to the intra-annual temporal regime, a spatial variability of this seasonality was detectable between different regions of Switzerland. The assessment of the dynamic behavior of the R-factor is valuable for the identification of susceptible seasons and regions.

The rainfall erosivity factor in the Czech Republic and its uncertainty

Abstract. In the present paper, the rainfall erosivity factor (R factor) for the area of the Czech Republic is assessed. Based on 10 min data for 96 stations and corresponding R factor estimates, a number of spatial interpolation methods are applied and cross-validated. These methods include inverse distance weighting, standard, ordinary, and regression kriging with parameters estimated by the method of moments and restricted maximum likelihood, and a generalized least-squares (GLS) model. For the regression-based methods, various statistics of monthly precipitation as well as geographical indices are considered as covariates. In addition to the uncertainty originating from spatial interpolation, the uncertainty due to estimation of the rainfall kinetic energy (needed for calculation of the R factor) as well as the effect of record length and spatial coverage are also addressed. Finally, the contribution of each source of uncertainty is quantified. The average R factor for the area of the Czech Republic is 640 MJ ha−1 mm h−1, with values for the individual stations ranging between 320 and 1520 MJ ha−1 mm h−1. Among various spatial interpolation methods, the GLS model relating the R factor to the altitude, longitude, mean precipitation, and mean fraction of precipitation above the 95th percentile of monthly precipitation performed best. Application of the GLS model also reduced the uncertainty due to the record length, which is substantial when the R factor is estimated for individual sites. Our results revealed that reasonable estimates of the R factor can be obtained even from relatively short records (15–20 years), provided sufficient spatial coverage and covariates are available.

Modeling fire size of wildfires in Castellon (Spain), using spatiotemporal marked point processes

The extent of fires, their periodicity and their impact on terrestrial communities have been a major concern in the last century. Wildfires play an important role in shaping landscapes and as a source of CO2 and particulate matter, contributing to the green house effect and to global warming. Modeling the spatial variability of wildfire extent is therefore an important subject in order to understand and to predict future trends on their effect in landscape changes and global warming. The most common approaches have been through point pattern analysis or by Markov random fields. Those methods have made possible to build risk maps, but for many forest managers knowing the fire size besides the location of the fire is very useful. In this work we use spatial marked point processes to model the fire size of the forest fires observed in Castellón, Spain, during the years 2001–2006. Our modeling approach incorporates spatial covariates as they are useful to model spatial variability and to gain insight about factors related to the presence of forest fires. Such information may be of great utility to predict the spreading of ongoing fires and also to prevent wildfire outburst by controlling risk factors. We describe and take advantage of the Bayesian methodology including Integrated Nested Laplace Approximation (INLA) and Stochastic Partial Differential Equation (SPDE) in the modeling process. We present the results of different models fitted to the data and discuss its usefulness to fire managers and planners.

Models of Eucalypt phenology predict bat population flux.

Fruit bats (Pteropodidae) have received increased attention after the recent emergence of notable viral pathogens of bat origin. Their vagility hinders data collection on abundance and distribution, which constrains modeling efforts and our understanding of bat ecology, viral dynamics, and spillover. We addressed this knowledge gap with models and data on the occurrence and abundance of nectarivorous fruit bat populations at 3 day roosts in southeast Queensland. We used environmental drivers of nectar production as predictors and explored relationships between bat abundance and virus spillover. Specifically, we developed several novel modeling tools motivated by complexities of fruit bat foraging ecology, including: (1) a dataset of spatial variables comprising Eucalypt‐focused vegetation indices, cumulative precipitation, and temperature anomaly; (2) an algorithm that associated bat population response with spatial covariates in a spatially and temporally relevant way given our current understanding of bat foraging behavior; and (3) a thorough statistical learning approach to finding optimal covariate combinations. We identified covariates that classify fruit bat occupancy at each of our three study roosts with 86–93% accuracy. Negative binomial models explained 43–53% of the variation in observed abundance across roosts. Our models suggest that spatiotemporal heterogeneity in Eucalypt‐based food resources could drive at least 50% of bat population behavior at the landscape scale. We found that 13 spillover events were observed within the foraging range of our study roosts, and they occurred during times when models predicted low population abundance. Our results suggest that, in southeast Queensland, spillover may not be driven by large aggregations of fruit bats attracted by nectar‐based resources, but rather by behavior of smaller resident subpopulations. Our models and data integrated remote sensing and statistical learning to make inferences on bat ecology and disease dynamics. This work provides a foundation for further studies on landscape‐scale population movement and spatiotemporal disease dynamics.

Managing distance and covariate information with point-based clustering.

BackgroundGeographic perspectives of disease and the human condition often involve point-based observations and questions of clustering or dispersion within a spatial context. These problems involve a finite set of point observations and are constrained by a larger, but finite, set of locations where the observations could occur. Developing a rigorous method for pattern analysis in this context requires handling spatial covariates, a method for constrained finite spatial clustering, and addressing bias in geographic distance measures. An approach, based on Ripley’s K and applied to the problem of clustering with deliberate self-harm (DSH), is presented.MethodsPoint-based Monte-Carlo simulation of Ripley’s K, accounting for socio-economic deprivation and sources of distance measurement bias, was developed to estimate clustering of DSH at a range of spatial scales. A rotated Minkowski L1 distance metric allowed variation in physical distance and clustering to be assessed. Self-harm data was derived from an audit of 2 years’ emergency hospital presentations (n = 136) in a New Zealand town (population ~50,000). Study area was defined by residential (housing) land parcels representing a finite set of possible point addresses.ResultsArea-based deprivation was spatially correlated. Accounting for deprivation and distance bias showed evidence for clustering of DSH for spatial scales up to 500 m with a one-sided 95 % CI, suggesting that social contagion may be present for this urban cohort.ConclusionsMany problems involve finite locations in geographic space that require estimates of distance-based clustering at many scales. A Monte-Carlo approach to Ripley’s K, incorporating covariates and models for distance bias, are crucial when assessing health-related clustering. The case study showed that social network structure defined at the neighbourhood level may account for aspects of neighbourhood clustering of DSH. Accounting for covariate measures that exhibit spatial clustering, such as deprivation, are crucial when assessing point-based clustering.

Functional-coefficient spatial autoregressive models with nonparametric spatial weights

We apply local linear regression and sieve estimation technique to estimate functional coefficients and an unknown spatial weighting function, respectively, via a nonparametric GMM estimation method, where we allow both exogenous and endogenous spatial covariates. A consistency result is derived to support the method. Moreover, a two-step estimator is constructed for the functional coefficients, and under certain conditions, we show that this estimator can be oracle efficient in the sense that its limiting distribution is the same regardless of whether or not the spatial weights are known. Both simulated and real data examples are used to illustrate our theory.

Meta-analysis of field scale spatial variability of grassland soil CO2 efflux: Interaction of biotic and abiotic drivers

In this study eight temperate grassland sites were monitored for soil CO2 efflux (Rs) and the spatial covariate soil water content (SWC) and soil temperature (Ts) at fine scale in over 77 measurement campaigns. The goals of this multisite study were to explore the correlations between environmental gradients and spatial patterns of Rs, SWC and Ts, which are not site-specific and to quantify the relevance of biotic and abiotic controls over spatial patterns along increasing vegetation structural complexity. These patterns in water-limited ecosystems in East-Central Europe are likely to be influenced by summer droughts caused by the changing climate.A consistent experimental setup was applied at the study sites including 75 sampling locations along 15m circular transects. Spatial data processing was mainly based on variography. Two proxy variables were introduced to relate the site characteristics in terms of soils, water status and vegetation. Normalised SWC (SWCn) reconciled site-specific soil water regimes while normalised day of year integrated temperature and vegetation phenology.A principal component analysis revealed that the progressing closure of vegetation in combination with large Rs and SWCn values, as well as low Ts and Rs variability support the detectability of spatial patterns found in both the abiotic and biotic variables. Our results showed that apart from SWC the pattern of soil temperature also had an effect on spatial structures. We detected that when the spatially structured variability of Ts was low, a strong negative correlation existed between SWCn and the spatial autocorrelation length of Rs with r=0.66 (p<0.001). However, for high spatially structured variability of Ts, occurring presumably at low Ts in spring and autumn, the correlation did not exist and it was difficult to quantify the spatial autocorrelation of Rs. Our results are indicative of a potential shift from homogeneity and dominance of biotic processes to an increased heterogeneity and abiotic regulation in drought prone ecosystems under conditions of decreasing soil moisture.