Evaluating a Bayesian modelling approach (INLA-SPDE) for environmental mapping

Tyrosine hydroxylase plays crucial roles in larval cuticle formation and larval-pupal tanning in the rice stem borer, Chilo suppressalis

The striped stem borer, Chilo suppressalis (Walker), a notorious pest infesting rice, has evolved a high level of resistance to many commonly used insecticides. In this study, we investigate whether tyrosine hydroxylase (TH), which is required for larval development and cuticle tanning in many insects, could be a potential target for the control of C. suppressalis. We identified and characterized the full-length cDNA (CsTH) of C. suppressalis. The complete open reading frame of CsTH (MW690914) was 1683 bp in length, encoding a protein of 560 amino acids. Within the first to the sixth larval instars, CsTH was high in the first day just after molting, and lower in the ensuing days. From the wandering stage to the adult stage, levels of CSTH began to rise and reached a peak at the pupal stage. These patterns suggested a role for the gene in larval development and larval-pupal cuticle tanning. When we injected dsCsTH or 3-iodotyrosine (3-IT) as a TH inhibitor or fed a larva diet supplemented with 3-IT, there were significant impairments in larval development and larval-pupal cuticle tanning. Adult emergence was severely impaired, and most adults died. These results suggest that CsTH might play a critical role in larval development as well as larval-pupal tanning and immunity in C. suppressalis, and this gene could form a potential novel target for pest control.

Quantifying key vegetation parameters from Sentinel-3 and MODIS over the eastern Eurasian steppe with a Bayesian geostatistical model

Accurate estimation of grassland leaf area index (LAI), fractional vegetation cover (FVC), and aboveground biomass (AGB) is fundamental in grassland studies. The newly launched Ocean and Land Color Imager (OLCI) sensor onboard Sentinel-3 (S3) provides images with comparable spatial and spectral resolution with MODIS data. However, the use of S3 OLCI imageries for vegetation variable estimation is rarely evaluated. This study evaluated the potential of S3 OLCI and MODIS data for estimating grassland LAI, FVC, and AGB in the eastern Eurasian steppe. A Bayesian spatial model (Integrated Nested Laplace Approximation with Stochastic Partial Differential Equation, INLA-SPDE) was used to address spatial autocorrelation of in-situ observation data and to enhance our predictions. Our results showed that the models based on S3 OLCI data presented higher accuracy than models with MODIS data. The RMSEs decreased by 3.7-10.8 %, 3.7-7.5 %, and 1.6-14.2 % for LAI, FVC, and AGB predictions, respectively. Through combinations of multiple predictors, we confirmed the robustness of red edge bands for grassland variable estimation, the models employing red edge variables yielded 3.5 %, 3.2 %, and 0.4 % lower RMSEs than models with conventional visible and NIR bands for LAI, FVC, and AGB prediction, respectively. INLA-SPDE spatial model produced lower bias and higher prediction accuracy than random forest and random forests kriging method in most of the models; the INLA-SPDE predicted LAI and FVC maps also showed a better agreement with ground observations than MODIS and PROBA-V land products.

Synergetic use of DEM derivatives, Sentinel-1 and Sentinel-2 data for mapping soil properties of a sloped cropland based on a two-step ensemble learning method

Understanding the spatial variability of soil organic matter (SOM), soil total nitrogen (STN), soil total phosphorus (STP), and soil total potassium (STK) is important to support site-specific agronomic management, food production, and climate change adaptation. High-resolution remote sensing imageries have emerged as an innovative solution to investigate the spatial variation in agricultural soils with machine learning (ML) algorithms. However, the predictive power of the individual and combined effects of Sentinel-1 (S1) synthetic aperture radar (SAR) and Sentinel-2 (S2) multispectral images for mapping soil properties, especially STN, STP, and STK, have rarely been investigated. Moreover, single ML model may achieve unstable performance for predicting multiple soil properties due to strong spatial heterogeneity. This study explored the combine use of S1, S2, and DEM derivatives to map SOM, STN, STP, and STK content of a sloped cropland of northeastern China. A two-step method with a weighted sum of four ML models was proposed to improve the accuracy and robustness in predicting multiple soil properties. Our results showed that single ML model has various performance in predicting the four soil properties. The optimal ML models could explain approximately 56 %, 53 %, 56 % and 37 % of the variability of SOM, STN, STP, and STK, respectively. Using the weights estimated through a 10-fold cross-validation procedure, the two-step ensemble learning model was retrained and showed more robust performance than the four ML models, in which the prediction accuracy was improved by 2.38 %, 1.40 %, 3.52 %, and 3.29 % for SOM, STN, STP, and STK, respectively. Our results also showed that the optical S2 derived features, especially the two S2 short-wave infrared bands, enhanced vegetation index, and soil adjusted vegetation index, were more important for soil property prediction than S1 data and DEM derivatives. Compared with individual sensor, a combination of S1 and S2 data yielded more accurate predictions of STN and STP but not for SOM and STK. The results of this study highlight the potential of high-resolution S1 and S2 data and the two-step method for soil property prediction at farmland scale.

Bayesian spatiotemporal modelling analysis of hemorrhagic fever with renal syndrome outbreaks in China using R-INLA

Hemorrhagic fever with renal syndrome (HFRS) is a category B infectious disease caused by Hantavirus infection, which can cause acute kidney injury and has a high mortality rate. At present, China is the country most severely afflicted by HFRS in the world, and it is critical to carry out efficient HFRS prevention and management in a scientific and accurate manner. The study used data on the incidence of HFRS in mainland China from 2015 to 2018, built a Bayesian hierarchical spatiotemporal distribution model, and applied the Integrated Nested Laplace Approximation algorithm to analyse the factors influencing the development of HFRS, the spatial and temporal distribution characteristics, and the threshold exceedance locations. The results revealed that the woodland and grassland area (RR = 1.357, 95% CI: 1.005-1.791), economic level (RR = 1.299, 95% CI: 1.007-1.649), and traffic level (RR = 2.442, 95% CI: 1.825-3.199) were all significantly and positively associated with the development of HFRS, with traffic level having the strongest promoting effect. The seasonal cycle was obvious in time, with peaks in May-June and October-December each year, most notably in November. Spatially, there was a south-heavy north-light trend, with a high risk of incidence largely in places rich in mountain and forest vegetation, of which Guizhou, Guangxi, Guangdong, and Jiangxi provinces continuing to have a high incidence in recent years, and the evolution of the epidemic in Hubei and Hunan was becoming more serious. When the early warning threshold was set at 0.2, the detection impact was best, and Guizhou, Guangxi, Guangdong, Jiangxi, Hainan, and Tianjin were positioned near the critical point of the exceedance threshold with the highest risk of incidence. It is recommended that the relevant managers call for active vaccination of outdoor workers, such as those working in agriculture and construction sites, implement rat prevention and extermination before winter arrives, and warn high-risk and medium-high-risk areas to conduct early outbreak surveillance. Move the prevention and control gates forward based on the exceedance threshold for doing preventive and control detection and epidemic research and judgement work.

Nanoparticle facilitated stacked-dsRNA improves suppression of the Lepidoperan pest Chilo suppresallis

In recent years, gene knockdown technology using double-stranded RNA (dsRNA) has been widely used as an environment-friendly pest control strategy, but its instability and limited cellular uptake have limited its overall effect. Studies have shown that the efficiency of single dsRNA can be improved by using various nanomaterials. However, the effect of stacked-dsRNA wrapped by nanomaterial on pests remains unclear. In the present study, both CYP15C1 and C-factor genes were cloned from the midgut of C. suppressalis, and the transcript of C-factor is most highly expressed in heads. Feeding a dsCYP15C1 or dsC-factor - nanomaterial mixture can downregulate the gene expression and significantly increase larval mortality. More importantly, feeding the stacked-dsRNA wrapped by nanomaterial can significantly increase the mortality of C. suppressalis, compared with feeding dsCYP15C1 or dsC-factor - nanomaterial mixture alone. These results showed that CYP15C1 and C-factor could be potential targets for an effective management of C. suppressalis, and we developed a nanoparticle-facilitated stacked-dsRNA strategy in the control of C. suppresallis. Our research provides a theoretical basis for gene function analysis and field pest control, and will promote the application of RNAi technology in the stacked style of pest control.

Spatio-temporal modeling of traffic accidents incidence on urban road networks based on an explicit network triangulation

Traffic deaths and injuries are one of the major global public health concerns. The present study considers accident records in an urban environment to explore and analyze spatial and temporal in the incidence of road traffic accidents. We propose a spatio-temporal model to provide predictions of the number of traffic collisions on any given road segment, to further generate a risk map of the entire road network. A Bayesian methodology using Integrated nested Laplace approximations with stochastic partial differential equations (SPDE) has been applied in the modeling process. As a novelty, we have introduced SPDE network triangulation to estimate the spatial autocorrelation restricted to the linear network. The resulting risk maps provide information to identify safe routes between source and destination points, and can be useful for accident prevention and multi-disciplinary road safety measures.

Unveiling spatial variability in herbicide soil sorption using Bayesian digital mapping

Regional mapping herbicide sorption to soil is essential for risk assessment. However, conducting analytical quantification of adsorption coefficient (K_d ) in large-scale studies is too costly; therefore, a research question arises on goodness of K_d spatial prediction from sampling. The application of a spatial Bayesian regression (BR) is a newer technique in agricultural and natural resources sciences that allows converting spatially discrete samples into maps covering continuous spatial domains. The objective of this work was to unveil herbicide sorption to soil at a landscape scale by developing a predictive BR model. We integrated a large set of ancillary soil and climate covariables from sites with K_d measurements into a spatial mixed model including site random effects. The models were fitted using glyphosate and atrazine K_d s, determined in 80 and 120 sites, respectively, from central Argentina. For model assessment, measurements of global and point-wise prediction errors were obtained by cross-validation; residual variability was estimated by bootstrap to compare BR with regression kriging. Results showed that the BR spatial predictions outperformed regression kriging. The glyphosate K_d model (root mean square prediction error, 13% of the mean) included aluminum oxides, pH, and clay content, whereas the atrazine K_d model strongly depended on soil organic carbon and clay and on climatic variables related to water availability (root mean square prediction error, 27%). Spatial modeling of a complex edaphic process as herbicide sorption to soils enhanced environmental interpretations. An efficient approach for spatial mapping provides a modern perspective on the study of herbicide sorption to soil.

Bayesian spatial modelling of terrestrial radiation in Switzerland

The geographic variation of terrestrial radiation can be exploited in epidemiological studies of the health effects of protracted low-dose exposure. Various methods have been applied to derive maps of this variation. We aimed to construct a map of terrestrial radiation for Switzerland. We used airborne γ-spectrometry measurements to model the ambient dose rates from terrestrial radiation through a Bayesian mixed-effects model and conducted inference using Integrated Nested Laplace Approximation (INLA). We predicted higher levels of ambient dose rates in the alpine regions and Ticino compared with the western and northern parts of Switzerland. We provide a map that can be used for exposure assessment in epidemiological studies and as a baseline map for assessing potential contamination.

Source apportionment and spatial distribution of potentially toxic elements in soils: A new exploration on receptor and geostatistical models

Potentially toxic element (PTE) pollution is considered as the main soil environmental problem in the world. Source apportionment and spatial pattern of soil PTEs are essential for soil management. US-EPA positive matrix factorization (EPAPMF) and sequential Gaussian simulation (SGS) are general modeling tools for source apportionment and spatial distribution, respectively. Factor analysis with nonnegative constraints (FA-NNC) and stochastic partial derivative equations (SPDE) provided potential tools for this issue. We compared the performance of FA-NNC with PMF and the performance of SPDE with SGS, based on a dataset containing 9 PTEs in 285 topsoil samples. Three factors were determined by the two receptor models, and the source contributions were similar, suggesting that FA-NNC can validly identify quantitative sources of soil PTEs. The average source contributions were calculated based on the PMF and FA-NNC. Natural sources dominated the contents of As, Co, Cr, Cu, Ni, and Zn and affected 56.0%, 38.7%, and 84.8% of the Cd, Hg, and Pb concentrations, respectively. A total of 59.8% of Hg and 12.0% of Pb were associated with atmospheric deposition from coal combustion, industrial and traffic emissions, respectively. Agricultural and industrial activities contributed 37.2% of Cd concentration. SPDE proved to be an effective geostatistical technique to simulate the spatial patterns of soil PTEs with higher prediction accuracy than SGS. Co, Cr, Cu, and Ni had similar spatial patterns with hotspots randomly distributed across the study area. The common hotspots of As, Cd, Hg, Pb, and Zn in central parts inherited their high geochemical background in mudstone, while intensive human inputs in these areas also contributed to the accumulation of Cd, Hg, and Pb.

Using a Bayesian modelling approach (INLA-SPDE) to predict the occurrence of the Spinetail Devil Ray (Mobular mobular)

To protect the most vulnerable marine species it is essential to have an understanding of their spatiotemporal distributions. In recent decades, Bayesian statistics have been successfully used to quantify uncertainty surrounding identified areas of interest for bycatch species. However, conventional simulation-based approaches are often computationally intensive. To address this issue, in this study, an alternative Bayesian approach (Integrated Nested Laplace Approximation with Stochastic Partial Differential Equation, INLA-SPDE) is used to predict the occurrence of Mobula mobular species in the eastern Pacific Ocean (EPO). Specifically, a Generalized Additive Model is implemented to analyze data from the Inter-American Tropical Tuna Commission’s (IATTC) tropical tuna purse-seine fishery observer bycatch database (2005–2015). The INLA-SPDE approach had the potential to predict both the areas of importance in the EPO, that are already known for this species, and the more marginal hotspots, such as the Gulf of California and the Equatorial area which are not identified using other habitat models. Some drawbacks were identified with the INLA-SPDE database, including the difficulties of dealing with categorical variables and triangulating effectively to analyze spatial data. Despite these challenges, we conclude that INLA approach method is an useful complementary and/or alternative approach to traditional ones when modeling bycatch data to inform accurately management decisions.

Application of spatial analysis to investigate contribution of VOCs to photochemical ozone creation

This study was concerned with the temporal analysis of benzene, toluene, ethylbenzene, xylenes (BTEXs), and ozone in Rochester, New York, between 2012 and 2018. Additionally, the influence of ozone precursors (e.g., BTEXs and NO₂) and meteorological variables (e.g., relative humidity (RH), temperature along with wind speed) on ozone dispersion was investigated in the eastern half of the USA using the integrated nested Laplace approximation and stochastic partial differential equation (INLA-SPDE). The benzene variability at seasonal scale was characterized by higher values during the cold seasons. On the contrary, the long-term temporal trend of ozone depicted a repetitive cyclic behavior while an episode, with values exceeding 5 μg/m³, was detected associated with benzene in 2015. The spatial analysis by INLA-SPDE indicated that 1,3,5-trimethylbenzene and benzene were the key ozone precursors influencing ozone formation. It was demonstrated that increase of temperature had a considerable impact on ozone build-up whereas the increment of RH leads to decrease in ambient values of ozone. The amounts of root mean squared error (RMSE), mean absolute error (MAE), and bias for the validation data (e.g., 32 samples) were 0.005, 0.004, and 0.0008, exhibiting a reasonable out-of-sample forecasting by the INLA-SPDE model. The distribution map of ozone highlighted a hot spot in the state of Florida.

Understanding the Stochastic Partial Differential Equation Approach to Smoothing

Correlation and smoothness are terms used to describe a wide variety of random quantities. In time, space, and many other domains, they both imply the same idea: quantities that occur closer together are more similar than those further apart. Two popular statistical models that represent this idea are basis-penalty smoothers (Wood in Texts in statistical science, CRC Press, Boca Raton, 2017) and stochastic partial differential equations (SPDEs) (Lindgren et al. in J R Stat Soc Series B (Stat Methodol) 73(4):423–498, 2011). In this paper, we discuss how the SPDE can be interpreted as a smoothing penalty and can be fitted using the package , allowing practitioners with existing knowledge of smoothing penalties to better understand the implementation and theory behind the SPDE approach.

Supplementary materials accompanying this paper appear online.

Knockdown of the aminopeptidase N genes decreases susceptibility of Chilo suppressalis larvae to Cry1Ab/Cry1Ac and Cry1Ca

Bacillus thuringiensis (Bt) insecticide is currently the most widely used bioinsecticide. Bt expressing cry genes are some of the most successful foreign-genome-inserting genes used in transgenic insect-resistant crop development. Cry toxins are resistant to lepidopteran pests, such as Chilo suppressalis, a major insect pest of rice worldwide. Since Cry toxins exert their activity by binding to specific receptors in the midgut of target insects, identification of functional Cry toxin receptors in the midgut of C. suppressalis larvae is crucial to evaluate potential resistance mechanisms and develop effective strategies for inhibiting insect resistance. In this study, we isolated two aminopeptidase N genes (APN6 and APN8) from C. suppressalis and determined that they were expressed in the foregut. APN6 was highly expressed at the fourth instar, and APN8 was highly expressed in adult and pupa. Knockdown of CsAPN6 and CsAPN8 by RNA interference resulted in significantly decreased susceptibility of larvae to Bt rice varieties TT51 (expressing cry1Ac/cry1Ab fusion genes) and T1C-19 (expressing cry1Ca), but not T2A-1 (expressing cry2Aa). These findings suggest that both APN6 and APN8 are involved in the toxicity of Cry1Ac/Cry1Ab and Cry1Ca toxins.

Flexible modelling of spatial variation in agricultural field trials with the R package INLA

KEY MESSAGE

Established spatial models improve the analysis of agricultural field trials with or without genomic data and can be fitted with the open-source R package INLA. The objective of this paper was to fit different established spatial models for analysing agricultural field trials using the open-source R package INLA. Spatial variation is common in field trials, and accounting for it increases the accuracy of estimated genetic effects. However, this is still hindered by the lack of available software implementations. We compare some established spatial models and show possibilities for flexible modelling with respect to field trial design and joint modelling over multiple years and locations. We use a Bayesian framework and for statistical inference the integrated nested Laplace approximations (INLA) implemented in the R package INLA. The spatial models we use are the well-known independent row and column effects, separable first-order autoregressive ([Formula: see text]) models and a Gaussian random field (Matérn) model that is approximated via the stochastic partial differential equation approach. The Matérn model can accommodate flexible field trial designs and yields interpretable parameters. We test the models in a simulation study imitating a wheat breeding programme with different levels of spatial variation, with and without genome-wide markers and with combining data over two locations, modelling spatial and genetic effects jointly. The results show comparable predictive performance for both the [Formula: see text] and the Matérn models. We also present an example of fitting the models to a real wheat breeding data and simulated tree breeding data with the Nelder wheel design to show the flexibility of the Matérn model and the R package INLA.

Comparison of methods for addressing the point-to-area data transformation to make data suitable for environmental, health and socio-economic studies

Soil lead (Pb) provides an important exposure pathway to the human body through soil ingestion and dust inhalation and is closely associated with human health as well as social behaviour. The challenge of transforming different spatial supports arises when linking point data (Pb concentration) to areal data (health status or social behaviour). A detailed review of methodologies for integrating point and areal data has been carried out. Among a number of methodologies, eight methods: (1) average, (2) median, (3) centroids inverse distance weighted (IDW), (4) average block IDW, (5) median block IDW, (6) centroids ordinary kriging (OK), (7) average block OK and (8) median block OK, have been compared using Pb data set in the Greater London Authority (GLA) area. The results indicated that the method of median block IDW was recommended for further investigation of the relationship between Pb concentration and socio-economic factors in the ward-level of the GLA area. The reasons were (i) spatial interpolations were useful for predicting unobserved values when simple average and median could not work in the locations where there were no samples collected in some areal units; (ii) the median value was more suitable than the average value for a skewed data set; (iii) the block method reduced estimation error and provided more representative values of areal units than the centroid method; (iv) IDW reserved more spatial variation than OK, containing more local maxima (hotspot) and local minima. Despite that it is still hard to decide the optimal method, this study has highlighted the point-to-area transformation issue and provided valuable examples to compare the different methods.

Identifying heavy metal pollution hot spots in soil-rice systems: A case study in South of Yangtze River Delta, China

The soil-rice system in China is subjected to increasing concentrations of heavy metals (HMs) which derived from various sources. It is very critical to investigate the concentrations, spatial characteristics and hot spots of HMs content in the soil-rice system. This study presents work completed on 915 soil-rice sample pairs collected from South of Yangtze River Delta, China. These samples were evaluated for HM concentrations. Ordinary Kriging and the Getis-Ord index were used to explore spatial distributions and pollution hot spots. Averaged HMs content in soil is shown to be Zn > Cr > Pb > Cu > Ni > As > Hg > Cd, and concentrations in rice arrange as Zn > Cu > Cr > Ni > As > Cd > Pb > Hg. Compared with Chinese maximum permissible limits, mean content of all HMs in farmland soil are at safe levels and averaged content of all HMs in rice were also at safe levels except As and Ni. Ni was most polluted HM in soil Most of and showed relatively high content in farmland soil in southeastern part. As and Ni are the most polluted in rice, with highest content distributed in the northwestern and southern area, respectively. The majority of HMs pollution hot spots in soil clustered in the central area. Pollution hot spots of Ni and As in rice are mainly concentrated in the central part and southeastern part, correspondingly. Our results found a weak link between content and spatial pattern of pollution status of HMs in soil and rice. The results are anticipated to contribute to more efficient and accurate control of HMs pollution in soil-rice system, and assist decision-makers achieve a balance between cost and regulation of HM pollution.

Heavy Metal Pollution Delineation Based on Uncertainty in a Coastal Industrial City in the Yangtze River Delta, China

Assessing heavy metal pollution and delineating pollution are the bases for evaluating pollution and determining a cost-effective remediation plan. Most existing studies are based on the spatial distribution of pollutants but ignore related uncertainty. In this study, eight heavy-metal concentrations (Cr, Pb, Cd, Hg, Zn, Cu, Ni, and Zn) were collected at 1040 sampling sites in a coastal industrial city in the Yangtze River Delta, China. The single pollution index (PI) and Nemerow integrated pollution index (NIPI) were calculated for every surface sample (0–20 cm) to assess the degree of heavy metal pollution. Ordinary kriging (OK) was used to map the spatial distribution of heavy metals content and NIPI. Then, we delineated composite heavy metal contamination based on the uncertainty produced by indicator kriging (IK). The results showed that mean values of all PIs and NIPIs were at safe levels. Heavy metals were most accumulated in the central portion of the study area. Based on IK, the spatial probability of composite heavy metal pollution was computed. The probability of composite contamination in the central core urban area was highest. A probability of 0.6 was found as the optimum probability threshold to delineate polluted areas from unpolluted areas for integrative heavy metal contamination. Results of pollution delineation based on uncertainty showed the proportion of false negative error areas was 6.34%, while the proportion of false positive error areas was 0.86%. The accuracy of the classification was 92.80%. This indicated the method we developed is a valuable tool for delineating heavy metal pollution.