Search

Ensembles are a well established machine learning paradigm, leading to accurate and robust models, predominantly applied to predictive modeling tasks. Ensemble models comprise a finite set of diverse predictive models whose combined output is expected to yield an improved predictive performance as compared to an individual model. In this paper, we propose a new method for learning ensembles of process-based models of dynamic systems. The process-based modeling paradigm employs domain-specific knowledge to automatically learn models of dynamic systems from time-series observational data. Previous work has shown that ensembles based on sampling observational data (i.e., bagging and boosting), significantly improve predictive performance of process-based models. However, this improvement comes at the cost of a substantial increase of the computational time needed for learning. To address this problem, the paper proposes a method that aims at efficiently learning ensembles of process-based models, while maintaining their accurate long-term predictive performance. This is achieved by constructing ensembles with sampling domain-specific knowledge instead of sampling data. We apply the proposed method to and evaluate its performance on a set of problems of automated predictive modeling in three lake ecosystems using a library of process-based knowledge for modeling population dynamics. The experimental results identify the optimal design decisions regarding the learning algorithm. The results also show that the proposed ensembles yield significantly more accurate predictions of population dynamics as compared to individual process-based models. Finally, while their predictive performance is comparable to the one of ensembles obtained with the state-of-the-art methods of bagging and boosting, they are substantially more efficient. [ABSTRACT FROM AUTHOR]

In this paper, we introduce the R package gendist that computes the probability density function, the cumulative distribution function, the quantile function and generates random values for several generated probability distribution models including the mixture model, the composite model, the folded model, the skewed symmetric model and the arc tan model. These models are extensively used in the literature and the R functions provided here are flexible enough to accommodate various univariate distributions found in other R packages. We also show its applications in graphing, estimation, simulation and risk measurements. [ABSTRACT FROM AUTHOR]

Constrained ordination methods aims at finding an environmental gradient along which the species abundances are maximally separated. The species response functions, which describe the expected abundance as a function of the environmental score, are according to the ecological fundamental niche theory only meaningful if they are bell-shaped. Many classical model-based ordination methods, however, use quadratic regression models without imposing the bell-shape and thus allowing for meaningless U-shaped response functions. The analysis output (e.g. a biplot) may therefore be potentially misleading and the conclusions are prone to errors. In this paper we present a log-likelihood ratio criterion with a penalisation term to enforce more bell-shaped response shapes. We report the results of a simulation study and apply our method to metagenomics data from microbial ecology. [ABSTRACT FROM AUTHOR]

Freely-available satellite data streams and the ability to process these data on cloud-computing platforms such as Google Earth Engine have made frequent, large-scale landcover mapping at high resolution a real possibility. In this paper we apply these technologies, along with machine learning, to the mapping of peatlands–a landcover class that is critical for preserving biodiversity, helping to address climate change impacts, and providing ecosystem services, e.g., carbon storage–in the Boreal Forest Natural Region of Alberta, Canada. We outline a data-driven, scientific framework that: compiles large amounts of Earth observation data sets (radar, optical, and LiDAR); examines the extracted variables for suitability in peatland modelling; optimizes model parameterization; and finally, predicts peatland occurrence across a large boreal area (397, 958 km2) of Alberta at 10 m spatial resolution (equalling 3.9 billion pixels across Alberta). The resulting peatland occurrence model shows an accuracy of 87% and a kappa statistic of 0.57 when compared to our validation data set. Differentiating peatlands from mineral wetlands achieved an accuracy of 69% and kappa statistic of 0.37. This data-driven approach is applicable at large geopolitical scales (e.g., provincial, national) for wetland and landcover inventories that support long-term, responsible resource management. [ABSTRACT FROM AUTHOR]

The intelligent water drop algorithm is a swarm-based metaheuristic algorithm, inspired by the characteristics of water drops in the river and the environmental changes resulting from the action of the flowing river. Since its appearance as an alternative stochastic optimization method, the algorithm has found applications in solving a wide range of combinatorial and functional optimization problems. This paper presents an improved intelligent water drop algorithm for solving multi-depot vehicle routing problems. A simulated annealing algorithm was introduced into the proposed algorithm as a local search metaheuristic to prevent the intelligent water drop algorithm from getting trapped into local minima and also improve its solution quality. In addition, some of the potential problematic issues associated with using simulated annealing that include high computational runtime and exponential calculation of the probability of acceptance criteria, are investigated. The exponential calculation of the probability of acceptance criteria for the simulated annealing based techniques is computationally expensive. Therefore, in order to maximize the performance of the intelligent water drop algorithm using simulated annealing, a better way of calculating the probability of acceptance criteria is considered. The performance of the proposed hybrid algorithm is evaluated by using 33 standard test problems, with the results obtained compared with the solutions offered by four well-known techniques from the subject literature. Experimental results and statistical tests show that the new method possesses outstanding performance in terms of solution quality and runtime consumed. In addition, the proposed algorithm is suitable for solving large-scale problems. [ABSTRACT FROM AUTHOR]

Fluctuations in environmental parameters are increasingly being recognized as essential features of any habitat. The quantification of whether environmental fluctuations are prevalently predictable or unpredictable is remarkably relevant to understanding the evolutionary responses of organisms. However, when characterizing the relevant features of natural habitats, ecologists typically face two problems: (1) gathering long-term data and (2) handling the hard-won data. This paper takes advantage of the free access to long-term recordings of remote sensing data (27 years, Landsat TM/ETM+) to assess a set of environmental models for estimating environmental predictability. The case study included 20 Mediterranean saline ponds and lakes, and the focal variable was the water-surface area. This study first aimed to produce a method for accurately estimating the water-surface area from satellite images. Saline ponds can develop salt-crusted areas that make it difficult to distinguish between soil and water. This challenge was addressed using a novel pipeline that combines band ratio water indices and the short near-infrared band as a salt filter. The study then extracted the predictable and unpredictable components of variation in the water-surface area. Two different approaches, each showing variations in the parameters, were used to obtain the stochastic variation around a regular pattern with the objective of dissecting the effect of assumptions on predictability estimations. The first approach, which is based on Colwell’s predictability metrics, transforms the focal variable into a nominal one. The resulting discrete categories define the relevant variations in the water-surface area. In the second approach, we introduced General Additive Model (GAM) fitting as a new metric for quantifying predictability. Both approaches produced a wide range of predictability for the studied ponds. Some model assumptions–which are considered very different a priori–had minor effects, whereas others produced predictability estimations that showed some degree of divergence. We hypothesize that these diverging estimations of predictability reflect the effect of fluctuations on different types of organisms. The fluctuation analysis described in this manuscript is applicable to a wide variety of systems, including both aquatic and non-aquatic systems, and will be valuable for quantifying and characterizing predictability, which is essential within the expected global increase in the unpredictability of environmental fluctuations. We advocate that a priori information for organisms of interest should be used to select the most suitable metrics for estimating predictability, and we provide some guidelines for this approach. [ABSTRACT FROM AUTHOR]

Distribution function estimation of the random variable of river flow is an important problem in hydrology. This issue is directly related to quantile estimation, and consequently to return level prediction. The estimation process can be complemented with the construction of confidence intervals (CIs) to perform a probabilistic assessment of the different variables and/or estimated functions. In this work, several methods for constructing CIs using bootstrap techniques, and parametric and nonparametric procedures in the estimation process are studied and compared. In the case that the target is the joint estimation of a vector of values, some new corrections to obtain joint coverage probabilities closer to the corresponding nominal values are also presented. A comprehensive simulation study compares the different approaches, and the application of the different procedures to real data sets from four rivers in the United States and one in Spain complete the paper. [ABSTRACT FROM AUTHOR]

Effective pretreatment of spectral reflectance is vital to model accuracy in soil parameter estimation. However, the classic integer derivative has some disadvantages, including spectral information loss and the introduction of high-frequency noise. In this paper, the fractional order derivative algorithm was applied to the pretreatment and partial least squares regression (PLSR) was used to assess the clay content of desert soils. Overall, 103 soil samples were collected from the Ebinur Lake basin in the Xinjiang Uighur Autonomous Region of China, and used as data sets for calibration and validation. Following laboratory measurements of spectral reflectance and clay content, the raw spectral reflectance and absorbance data were treated using the fractional derivative order from the 0.0 to the 2.0 order (order interval: 0.2). The ratio of performance to deviation (RPD), determinant coefficients of calibration (), root mean square errors of calibration (RMSEC), determinant coefficients of prediction (), and root mean square errors of prediction (RMSEP) were applied to assess the performance of predicting models. The results showed that models built on the fractional derivative order performed better than when using the classic integer derivative. Comparison of the predictive effects of 22 models for estimating clay content, calibrated by PLSR, showed that those models based on the fractional derivative 1.8 order of spectral reflectance ( = 0.907, RMSEC = 0.425%, = 0.916, RMSEP = 0.364%, and RPD = 2.484 ≥ 2.000) and absorbance ( = 0.888, RMSEC = 0.446%, = 0.918, RMSEP = 0.383% and RPD = 2.511 ≥ 2.000) were most effective. Furthermore, they performed well in quantitative estimations of the clay content of soils in the study area. [ABSTRACT FROM AUTHOR]

In many scientific studies, the underlying data-generating process is unknown and multiple statistical models are considered to describe it. For example, in a factorial experiment we might consider models involving just main effects, as well as those that include interactions. Model-averaging is a commonly-used statistical technique to allow for model uncertainty in parameter estimation. In the frequentist setting, the model-averaged estimate of a parameter is a weighted mean of the estimates from the individual models, with the weights typically being based on an information criterion, cross-validation, or bootstrapping. One approach to building a model-averaged confidence interval is to use a Wald interval, based on the model-averaged estimate and its standard error. This has been the default method in many application areas, particularly those in the life sciences. The MA-Wald interval, however, assumes that the studentized model-averaged estimate has a normal distribution, which can be far from true in practice due to the random, data-driven model weights. Recently, the model-averaged tail area Wald interval (MATA-Wald) has been proposed as an alternative to the MA-Wald interval, which only assumes that the studentized estimate from each model has a N(0, 1) or t-distribution, when that model is true. This alternative to the MA-Wald interval has been shown to have better coverage in simulation studies. However, when we have a response variable that is skewed, even these relaxed assumptions may not be valid, and use of these intervals might therefore result in poor coverage. We propose a new interval (MATA-SBoot) which uses a parametric bootstrap approach to estimate the distribution of the studentized estimate for each model, when that model is true. This method only requires that the studentized estimate from each model is approximately pivotal, an assumption that will often be true in practice, even for skewed data. We illustrate use of this new interval in the analysis of a three-factor marine global change experiment in which the response variable is assumed to have a lognormal distribution. We also perform a simulation study, based on the example, to compare the lower and upper error rates of this interval with those for existing methods. The results suggest that the MATA-SBoot interval can provide better error rates than existing intervals when we have skewed data, particularly for the upper error rate when the sample size is small. [ABSTRACT FROM AUTHOR]

Coastlines are the boundary between the ocean and land and are an important line feature in spatial data. Coastlines must be adapted via map generalization when they are stored in multi-scale spatial databases. This article presents a new method of coastline generalization that considers the buffer consistency from the original coastline to the generalized coastlines. This method uses the geographical distance field to identify the feature points and bends that influence the buffer consistency from the original coastline to the generalized coastline, and the process is also used to simplify the bends and maintain the concave and convex characteristics in the generalization. This method is compared to the bend-simplification method, and the results indicate that the proposed method can preserve the shape characteristics of coastlines. Moreover, the seaward buffer boundaries from the original coastline to the generalized coastline are consistent when the distance is greater than the tolerance of the coastline generalization. [ABSTRACT FROM AUTHOR]

Morphometric differences were investigated among five fish species of subfamily Barbinae from the Ganga river system through traditional morphometrics and the truss network system. Species taken into account were Puntius chola (Hamilton 1822), Puntius sophore (Hamilton 1822), Pethia ticto (Hamilton 1822), Pethia conchonius (Hamilton 1822) and Systomus sarana (Hamilton 1822). Although, taxonomists carefully examine external body features to discriminate these species, there is still a risk of misidentification during a visual assessment. In the present study, the traditional morphological analysis included 22 morphometric measurements and 10 meristic counts. Truss network system of 14 landmarks was interconnected to yield 91 distance variables. The principal component analysis (PCA), discriminant function analysis (DFA) and cluster analysis (CA) were employed in order to determine morphometric variations. In traditional analysis, 29 characters out of 32 were found significant (p<0.05). Eight principal components were extracted through PCA explaining 85.30% of the total variance in samples, DFA correctly classified 100.0% of original grouped cases and 100.0% of cross-validated grouped cases. Truss analysis showed that all the 90 characters were significant (p<0.05). PCA extracted four principal components explaining 96.45% of the total variance. DFA correctly classified 96.1% of original grouped cases and 92.1% of cross-validated grouped cases. The results acquired from the traditional as well as truss analyses indicate significant morphometric heterogeneity. However, variations are not the same for the two different methods (traditional and truss) employed for the analyses. Shape differences among species were evident from relative warps (RW) supporting truss network analysis. Geometric morphometric methods (GMM), but limited use of Procrustes methods revealed even very small dissimilarity between groups. In spite of determining the morphometric differentiation among species, the present study also provides a useful insight on the application and complementary role of truss analysis with traditional morphometric analysis in the correct classification of the selected species. [ABSTRACT FROM AUTHOR]

Though malaria control initiatives have markedly reduced malaria prevalence in recent decades, global eradication is far from actuality. Recent studies show that environmental and social heterogeneities in low-transmission settings have an increased weight in shaping malaria micro-epidemiology. New integrated and more localized control strategies should be developed and tested. Here we present a set of agent-based models designed to study the influence of local scale human movements on local scale malaria transmission in a typical Amazon environment, where malaria is transmission is low and strongly connected with seasonal riverine flooding. The agent-based simulations show that the overall malaria incidence is essentially not influenced by local scale human movements. In contrast, the locations of malaria high risk spatial hotspots heavily depend on human movements because simulated malaria hotspots are mainly centered on farms, were laborers work during the day. The agent-based models are then used to test the effectiveness of two different malaria control strategies both designed to reduce local scale malaria incidence by targeting hotspots. The first control scenario consists in treat against mosquito bites people that, during the simulation, enter at least once inside hotspots revealed considering the actual sites where human individuals were infected. The second scenario involves the treatment of people entering in hotspots calculated assuming that the infection sites of every infected individual is located in the household where the individual lives. Simulations show that both considered scenarios perform better in controlling malaria than a randomized treatment, although targeting household hotspots shows slightly better performance. [ABSTRACT FROM AUTHOR]

Betweenness has been used in a number of marine studies to identify portions of sea that sustain the connectivity of whole marine networks. Herein we highlight the need of methodological exactness in the calculation of betweenness when graph theory is applied to marine connectivity studies based on transfer probabilities. We show the inconsistency in calculating betweeness directly from transfer probabilities and propose a new metric for the node-to-node distance that solves it. Our argumentation is illustrated by both simple theoretical examples and the analysis of a literature data set. [ABSTRACT FROM AUTHOR]

A warming climate increases thermal inputs to lakes with potential implications for water quality and aquatic ecosystems. In a previous study, we used a dynamic water column temperature and mixing simulation model to simulate chronic (7-day average) maximum temperatures under a range of potential future climate projections at selected sites representative of different U.S. regions. Here, to extend results to lakes where dynamic models have not been developed, we apply a novel machine learning approach that uses Gaussian Process regression to describe the model response surface as a function of simplified lake characteristics (depth, surface area, water clarity) and climate forcing (winter and summer air temperatures and potential evapotranspiration). We use this approach to extrapolate predictions from the simulation model to the statistical sample of U.S. lakes in the National Lakes Assessment (NLA) database. Results provide a national-scale scoping assessment of the potential thermal risk to lake water quality and ecosystems across the U.S. We suggest a small fraction of lakes will experience less risk of summer thermal stress events due to changes in stratification and mixing dynamics, but most will experience increases. The percentage of lakes in the NLA with simulated 7-day average maximum water temperatures in excess of 30°C is projected to increase from less than 2% to approximately 22% by the end of the 21st century, which could significantly reduce the number of lakes that can support cold water fisheries. Site-specific analysis of the full range of factors that influence thermal profiles in individual lakes is needed to develop appropriate adaptation strategies. [ABSTRACT FROM AUTHOR]

Human activities introduce a variety of chemicals to the Laurentian Great Lakes including pesticides, pharmaceuticals, flame retardants, plasticizers, and solvents (collectively referred to as contaminants of emerging concern or CECs) potentially threatening the vitality of these valuable ecosystems. We conducted a basin-wide study to identify the presence of CECs and other chemicals of interest in 12 U.S. tributaries to the Laurentian Great Lakes during 2013 and 2014. A total of 292 surface-water and 80 sediment samples were collected and analyzed for approximately 200 chemicals. A total of 32 and 28 chemicals were detected in at least 30% of water and sediment samples, respectively. Concentrations ranged from 0.0284 (indole) to 72.2 (cholesterol) μg/L in water and 1.75 (diphenhydramine) to 20,800 μg/kg (fluoranthene) in sediment. Cluster analyses revealed chemicals that frequently co-occurred such as pharmaceuticals and flame retardants at sites receiving similar inputs such as wastewater treatment plant effluent. Comparison of environmental concentrations to water and sediment-quality benchmarks revealed that polycyclic aromatic hydrocarbon concentrations often exceeded benchmarks in both water and sediment. Additionally, bis(2-ethylhexyl) phthalate and dichlorvos concentrations exceeded water-quality benchmarks in several rivers. Results from this study can be used to understand organism exposure, prioritize river basins for future management efforts, and guide detailed assessments of factors influencing transport and fate of CECs in the Great Lakes Basin. [ABSTRACT FROM AUTHOR]

The Annual Crop Planning (ACP) problem was a recently introduced problem in the literature. This study further expounds on this problem by presenting a new mathematical formulation, which is based on market economic factors. To determine solutions, a new local search metaheuristic algorithm is investigated which is called the enhanced Best Performance Algorithm (eBPA). eBPA’s results are compared against two well-known local search metaheuristic algorithms; these include Tabu Search and Simulated Annealing. The results show the potential of the eBPA for continuous optimization problems. [ABSTRACT FROM AUTHOR]

Controlling a complex network towards a desired state is of great importance in many applications. Existing works present an approximate algorithm to find the input nodes used to control partial nodes of the network. However, the input nodes obtained by this algorithm depend on the node matching order and cannot achieve optimum results. Here we present a novel algorithm to find the input nodes for target control based on preferential matching. The algorithm elaborately arranges the matching order of the nodes to reduce the size of the input node set. The results on both synthetic and real networks indicate that the proposed algorithm outperforms the previous algorithm. [ABSTRACT FROM AUTHOR]

Increasingly, user generated content (UGC) in social media postings and their associated metadata such as time and location stamps are being used to provide useful operational information during natural hazard events such as hurricanes, storms and floods. The main advantage of these new sources of data are twofold. First, in a purely additive sense, they can provide much denser geographical coverage of the hazard as compared to traditional sensor networks. Second, they provide what physical sensors are not able to do: By documenting personal observations and experiences, they directly record the impact of a hazard on the human environment. For this reason interpretation of the content (e.g., hashtags, images, text, emojis, etc) and metadata (e.g., keywords, tags, geolocation) have been a focus of much research into social media analytics. However, as choices of semantic tags in the current methods are usually reduced to the exact name or type of the event (e.g., hashtags ‘#Sandy’ or ‘#flooding’), the main limitation of such approaches remains their mere nowcasting capacity. In this study we make use of polysemous tags of images posted during several recent flood events and demonstrate how such volunteered geographic data can be used to provide early warning of an event before its outbreak. [ABSTRACT FROM AUTHOR]

Stochastic transmission dynamic models are especially useful for studying the early emergence of novel pathogens given the importance of chance events when the number of infectious individuals is small. However, methods for parameter estimation and prediction for these types of stochastic models remain limited. In this manuscript, we describe a calibration and prediction framework for stochastic compartmental transmission models of epidemics. The proposed method, Multiple Shooting for Stochastic systems (MSS), applies a linear noise approximation to describe the size of the fluctuations, and uses each new surveillance observation to update the belief about the true epidemic state. Using simulated outbreaks of a novel viral pathogen, we evaluate the accuracy of MSS for real-time parameter estimation and prediction during epidemics. We assume that weekly counts for the number of new diagnosed cases are available and serve as an imperfect proxy of incidence. We show that MSS produces accurate estimates of key epidemic parameters (i.e. mean duration of infectiousness, R0, and Reff) and can provide an accurate estimate of the unobserved number of infectious individuals during the course of an epidemic. MSS also allows for accurate prediction of the number and timing of future hospitalizations and the overall attack rate. We compare the performance of MSS to three state-of-the-art benchmark methods: 1) a likelihood approximation with an assumption of independent Poisson observations; 2) a particle filtering method; and 3) an ensemble Kalman filter method. We find that MSS significantly outperforms each of these three benchmark methods in the majority of epidemic scenarios tested. In summary, MSS is a promising method that may improve on current approaches for calibration and prediction using stochastic models of epidemics. [ABSTRACT FROM AUTHOR]

Background: In the Point-Centred Quarter Method (PCQM), the mean distance of the first nearest plants in each quadrant of a number of random sample points is converted to plant density. It is a quick method for plant density estimation. In recent publications the estimator equations of simple PCQM (PCQM1) and higher order ones (PCQM2 and PCQM3, which uses the distance of the second and third nearest plants, respectively) show discrepancy. This study attempts to review PCQM estimators in order to find the most accurate equation form. We tested the accuracy of different PCQM equations using Monte Carlo Simulations in simulated (having ‘random’, ‘aggregated’ and ‘regular’ spatial patterns) plant populations and empirical ones. Principal Findings: PCQM requires at least 50 sample points to ensure a desired level of accuracy. PCQM with a corrected estimator is more accurate than with a previously published estimator. The published PCQM versions (PCQM1, PCQM2 and PCQM3) show significant differences in accuracy of density estimation, i.e. the higher order PCQM provides higher accuracy. However, the corrected PCQM versions show no significant differences among them as tested in various spatial patterns except in plant assemblages with a strong repulsion (plant competition). If N is number of sample points and R is distance, the corrected estimator of PCQM1 is 4(4N − 1)/(π ∑ R2) but not 12N/(π ∑ R2), of PCQM2 is 4(8N − 1)/(π ∑ R2) but not 28N/(π ∑ R2) and of PCQM3 is 4(12N − 1)/(π ∑ R2) but not 44N/(π ∑ R2) as published. Significance: If the spatial pattern of a plant association is random, PCQM1 with a corrected equation estimator and over 50 sample points would be sufficient to provide accurate density estimation. PCQM using just the nearest tree in each quadrant is therefore sufficient, which facilitates sampling of trees, particularly in areas with just a few hundred trees per hectare. PCQM3 provides the best density estimations for all types of plant assemblages including the repulsion process. Since in practice, the spatial pattern of a plant association remains unknown before starting a vegetation survey, for field applications the use of PCQM3 along with the corrected estimator is recommended. However, for sparse plant populations, where the use of PCQM3 may pose practical limitations, the PCQM2 or PCQM1 would be applied. During application of PCQM in the field, care should be taken to summarize the distance data based on ‘the inverse summation of squared distances’ but not ‘the summation of inverse squared distances’ as erroneously published. [ABSTRACT FROM AUTHOR]

Signal detection in functional magnetic resonance imaging (fMRI) inherently involves the problem of testing a large number of hypotheses. A popular strategy to address this multiplicity is the control of the false discovery rate (FDR). In this work we consider the case where prior knowledge is available to partition the set of all hypotheses into disjoint subsets or families, e. g., by a-priori knowledge on the functionality of certain regions of interest. If the proportion of true null hypotheses differs between families, this structural information can be used to increase statistical power. We propose a two-stage multiple test procedure which first excludes those families from the analysis for which there is no strong evidence for containing true alternatives. We show control of the family-wise error rate at this first stage of testing. Then, at the second stage, we proceed to test the hypotheses within each non-excluded family and obtain asymptotic control of the FDR within each family at this second stage. Our main mathematical result is that this two-stage strategy implies asymptotic control of the FDR with respect to all hypotheses. In simulations we demonstrate the increased power of this new procedure in comparison with established procedures in situations with highly unbalanced families. Finally, we apply the proposed method to simulated and to real fMRI data. [ABSTRACT FROM AUTHOR]