Journal of Environmental Statistics http://www.jenvstat.org/rss Mon, 21 Oct 2019 03:30:10 GMT Mon, 21 Oct 2019 03:30:10 GMT Most recent publications from the Journal of Environmental Statistics Frailty Models for the Control Time of Wildland Fires in the Former Intensive Fire Management Zone of Ontario, Canada http://www.jenvstat.org/v09/i05/paper Vol. 9, Issue 5, Sep 2019

Abstract:

Using the control time of a forest or wildland fire, defined as the time from the start of suppression action to the time that it is declared under control, we extend the analysis from Morin et al. (2015) to investigate spatial trends in forest fire survival probability across Ontario’s Intensive Fire Management Zone for the period 1989 to 2004. The fire management compartments (FMCs) described in Woolford et al. (2009) form the spatial units of analysis. Spatial differences are explored in our study region by using proportional hazards shared frailty models which incorporate a random effect to modify the hazard for fires within each FMC. Estimates of this excess risk are used to visualize spatial patterns. We show that the frailty models achieve better fit, as compared to the models without frailty terms, and that the model assumptions are suitable for these data. Visualizing the estimated FMC-specific frailties suggest the following: lightning-caused fires in a region of northwestern Ontario have experienced shorter control times than comparable lightning fires that occur elsewhere; and, people-caused fires in that same region in northwestern Ontario as well as a region of southern Ontario may also have experienced shorter control times than comparable people-caused fires that have occurred elsewhere.

]]>
Thu, 26 Sep 2019 07:00:00 GMT /v09/i05 Douglas G. Woolford, Alisha Albert-Green, Amy A. Morin, David L. Martell
Statistical Surveillance Thresholds for Enhanced Situational Awareness of Spring Wildland Fire Activity in Alberta, Canada http://www.jenvstat.org/v09/i04/paper Vol. 9, Issue 4, Sep 2019

Abstract:

Wildland fire disasters across Canada and globally are increasing in frequency. Alberta’s spring wildfire season is a particularly challenging period. Situational awareness of the wildfire environment is critical for wildfire management agencies to be prepared when extreme events occur. We propose the use of simple initial attack (IA) and being held (BH) escape surveillance charts in near-real time with thresholds as tools for enhancing and tracking situational awareness. Since the discrete data sets we used are zero-inflated and over-dispersed we chose to model the exceedances over a threshold. We also used preceding December sea surface temperatures (SST) of the Pacific Ocean as an indicator of persistent spring wildfire activity. Our analysis indicates the tracking of IA and BH escapes and SST can provide additional decision support as part of an early warning system of spring wildfire risk.

]]>
Thu, 26 Sep 2019 07:00:00 GMT /v09/i04 Douglas G. Woolford, Cordy Tymstra, Mike D. Flannigan
Spatial Semi-parametric Spectral Density Estimation for Multivariate Extremes, with Application to Fire Threat http://www.jenvstat.org/v09/i03/paper Vol. 9, Issue 3, Sep 2019

Abstract:

We analyze the joint tail of two variables related to fire threat associated with Santa Ana Winds in Southern California. To do this, we apply a flexible model for the joint tail of asymptotically dependent multivariate distributions, when samples are taken at several locations across space. We use a spatial prior on the underlying multivariate extremal dependence structure, which enables us to borrow strength across space while still allowing for different joint tail distributions at different spatial locations, and to predict the joint tail of the distribution at un-observed locations. A simulation shows that this model is able to capture complex dependence structures well.

]]>
Thu, 26 Sep 2019 07:00:00 GMT /v09/i03 Benjamin A. Shaby, Mauricio Nascimento
Area Interaction Point Processes for Bivariate Point Patterns in a Bayesian Context http://www.jenvstat.org/v09/i02/paper Vol. 9, Issue 2, Sep 2019

Abstract:

In this paper we consider bivariate point patterns which may contain both attractive and inhibitive interactions. The two subpatterns may depend on each other with both intra- and interspecific interactions possible. We use area interaction point processes for quantifying both attractive and inhibitive interactions in contrast to pairwise interaction point processes, typically model regular point patterns. The ability to permit both attraction and repulsion is a valuable feature and allows for the modelling of different forms of interactions in a given community. The differentiation between intra- and interspecific interactions in one model accounts for the fact that the presence of a second species may “mask” or “magnify” existing intraspecific interactions. A Bayesian approach has been applied for estimating interaction parameters and for discriminating between eight competing research hypotheses. For the particular application to modelling the interactions of species in a highly biodiverse forest, this study reveals posterior support for an interspecific interaction of attraction between the two species considered and may serve to inform forest rehabilitation schemes relating to this forest. Overall, knowledge of the interactions of key species in any given forest would be invaluable to reforestation efforts if this forest is later ravaged by wildfires.

]]>
Thu, 26 Sep 2019 07:00:00 GMT /v09/i02 Peter Nightingale, Glenna F. Nightingale, Janine Bärbel Illian, Ruth King
Visualizing Multivariate Time Series of Aerial Fire Fighting Data http://www.jenvstat.org/v09/i01/paper Vol. 9, Issue 1, Sep 2019

Abstract:

Aerial wildland fire fighters have a unique challenge. They are able to fill their tank with water via a nearby body of water, drop this water on a fire, then return to repeat this process. For a given fire, the replications of the fills and drops are multivariate time series measured in space, and this data structure allows us to compare replications over space and time. We use control chart methodologies to determine which time series were unlike the others, then examine the data from both a univariate and a multivariate viewpoint to determine potential causes.

]]>
Thu, 26 Sep 2019 07:00:00 GMT /v09/i01 Douglas G. Woolford, Devan G. Becker, W. John Braun, Charmaine B. Dean
Modeling of Dengue Fever Death Counts Using Hidden Markov Model http://www.jenvstat.org/v08/i09/paper Vol. 8, Issue 9, Sep 2018

Abstract:

We explore the use of Poisson-hidden Markov model to describe an overdispersed data on monthly death counts due to Dengue fever. Independent Poisson mixture models of various components and stationary Poisson hidden Markov models of different states are fitted and the performance of each model is judged using model selection criteria. The sequence of hidden states are estimated based on the best fitted model. The method can be applied in identifying environmental factors affecting a stochastic process.

]]>
Sat, 15 Sep 2018 07:00:00 GMT /v08/i09 Joshni George, Seemon Thomas
On Identifying the Probability Distribution of Monthly Maximum Temperature of Two Coastal Stations in Bangladesh http://www.jenvstat.org/v08/i08/paper Vol. 8, Issue 8, Sep 2018

Abstract:

Rising temperature in the atmosphere causes sea level rise and affects low lying coastal areas and deltas of the world. The last decade of the twentieth century was globally the hottest since the beginning of worldwide temperature measurement during the nineteenth century. Many PDFs have been proposed in recent past, but in present study Weibull, Lognormal, Gamma, GEV, etc are used to describe the characteristics of maximum temperature. This paper attempts to determine the best fitted probability distribution of monthly maximum temperature. To identify the appropriate probability distribution of the observed data, this paper considers a data set on the monthly maximum temperature of two coastal stations (Cox’s Bazar and Patuakhali) over the period January, 1971 to November, 2015 and January, 1973 to November, 2015 respectively. To check the accuracy of the predicted data using theoretical probability distributions the goodness-of-fit criteria like KS, R², χ2, and RMSE were used in this paper. According to the goodness-of-fit criteria and from the graphical comparisons it can be said that Generalized Skew Logistic distribution (GSL) provides the best fit for the observed monthly maximum temperature data of Cox’s Bazar and Weibull (W) gives the best fit for Patuakhali among the probability distributions considered in this paper.

]]>
Sat, 15 Sep 2018 07:00:00 GMT /v08/i08 Md. Moyazzem Hossain, Gazi Mahmud Alam, Faruq Abdulla
Prediction Limits for the Mean of a Sample from a Lognormal Distribution: Uncensored and Censored Cases http://www.jenvstat.org/v08/i07/paper Vol. 8, Issue 7, Sep 2018

Abstract:

For some regulatory purposes, it is desired to compare average on-site pollution concentrations in a narrowly defined geographic area with a large collection of background measurements. An approach to this problem is to treat this as a statistical prediction for the mean of a future sample based on a background sample. In this article, assuming lognormality, a fiducial approach is described for constructing prediction limits for the mean of a sample when the background sample is uncensored or censored. The fiducial prediction limits are evaluated with respect to coverage probabilities, and are compared with those based on another approximate method. Monte Carlo simulation studies for the uncensored case indicate that the fiducial methods are accurate and practically exact even for small samples, and they are very satisfactory for the censored case. Algorithms for computation of confidence limits are provided. The methods are illustrated using two real data sets.

]]>
Sat, 15 Sep 2018 07:00:00 GMT /v08/i07 K. Krishnamoorthy, Md. Sazib Hasan
A simulation comparison of estimators of spatial covariance parameters and associated bootstrap percentiles http://www.jenvstat.org/v08/i06/paper Vol. 8, Issue 6, Sep 2018

Abstract:

A simulation study is implemented to study estimators of the covariance structure of a stationary Gaussian spatial process and a spatial process with t-distributed margins. The estimators compared are Gaussian restricted maximum likelihood (REML) and curve-fitting by ordinary least squares and by the nonparametric Shapiro-Botha approach. Processes with Matérn covariance functions are considered and the parameters estimated are the nugget, partial sill and practical range. Both parametric and nonparametric bootstrap distributions of the estimators are computed and compared to the true marginal distributions of the estimators.

Gaussian REML is the estimator of choice for both Gaussian and t-distributed data and all choices of the Matérn covariance structure. However, accurate estimation of the Matérn shape parameter is critical to achieving a good fit while this does not affect the Shapiro-Botha estimator. The parametric bootstrap performed well for all estimators although it tended to be biased downward. It was slightly better than the nonparametric bootstrap for Gaussian data, equivalent to it for t-distributed data and worse overall for the Shapiro-Botha estimates.

A numerical example, obtained from environmental monitoring, is included to illustrate the application of the methods and the bootstrap.

]]>
Sat, 15 Sep 2018 07:00:00 GMT /v08/i06 Gabrielle Kelly, Raquel Menezes
Statistical Models for Evaluating Water Pollution: The Case of Asejire and Eleyele Reservoirs in Nigeria http://www.jenvstat.org/v08/i05/paper Vol. 8, Issue 5, Sep 2018

Abstract:

Water pollution is a major environmental problem due to rapid population growth that over exploit and pollute the water resources. In this work the physico-chemical study of Asejire and Eleyele reservoirs are carried out to examine the water pollution levels. Eleyele and Asejire reservoirs are the two major sources of pipe-borne water in Ibadan with a population of about four million people. Water samples were collected from both sites from January 2003-December 2007 and analysed for 13 physico-chemical parameters. The data were subjected to Principal Component Analysis (PCA) to define the parameters responsible for the main variability in water pollution. The PCA produces 5 significant main components explaining 66.6% and 69.8% variance in Asejire and Eleyele reservoir, respectively. Generalized Linear Model (GLM) is applied to study the variability in turbidity level which shows that four parameters in each reservoir are important to explain the turbidity variation. Also many parameters in Asejire lie within the SON and WHO permissible limits while in Eleyele reservoir many parameters lie out. This therefore is an indication that water in Eleyele reservoir is more polluted than in Asejire reservoir.

]]>
Sat, 15 Sep 2018 07:00:00 GMT /v08/i05 Privatus Christopher, K. O. Obisesan
Hints of latent drivers investigating university student performance http://www.jenvstat.org/v08/i04/paper Vol. 8, Issue 4, Sep 2018

Abstract:

Job market, nowadays, asks for higher and higher skills and competences. Therefore, also the measurement and assessment of the university students performance are crucial issues for policy makers. Although the scientific literature provides several papers investigating the main determinants of university student performance, often results are very different, and they seem to hold just in very specific contexts. This paper aims to contribute to the international literature, focusing on the role of student specific characteristics, supporting the idea that unobservable variables (such as motivation, aptitudes or abilities) should be more investigated.

]]>
Sat, 15 Sep 2018 07:00:00 GMT /v08/i04 Giada Adelfio, Giovanni Boscaino
Likelihood-based detection of cluster centers for Neyman–Scott point processes http://www.jenvstat.org/v08/i03/paper Vol. 8, Issue 3, Sep 2018

Abstract:

This study deals with the problem of estimating the unobservable cluster centers for a special type of Neyman"Scott point processes, in which the cluster sizes (numbers of members in each cluster) are distributed according to the Poisson distribution. The key point of the solution is the conversion among different forms of conditional intensities, λ(t | ·)dt = P{N[t,t + dt) = 1 | ·} = E[N[t,t + dt) | ·], where · represents a σ-algebra generated by some information from the process N. Some recursive formulae associated with the filtering gain (information gain represented by the ratio of the likelihood of the point process when we know more information to the likelihood when we know less) are derived. These recursive equations can be solved numerically by using Monte Carlo integration. The proposed method is illustrated by two simulation experiments, a purely temporal and a multi-type spatiotemporal case.

]]>
Sat, 15 Sep 2018 07:00:00 GMT /v08/i03 Jiancang Zhuang
Rater Classification by Means of Set-theoretic Methods Applied to Forestry Data http://www.jenvstat.org/v08/i02/paper Vol. 8, Issue 2, Sep 2018

Abstract:

We consider a situation where r raters select subsets from a set of n items by marking them by ‘0’ or ‘1’, as in classification problems, approval voting and in general subset voting. The number r of raters is small in comparison to the number n of items. We intend to classify the raters, to understand their behavior and to go beyond the possibilities of classical statistical methods such as Fleiss’ kappa, cluster analysis or latent class analysis. We use a non-parametric set-theoretic approach, which is natural for the given dichotomous setting. We recommend the determination of a set-theoretic mean, the Vorob’ev expectation, to play a role similar to the classical mean of a sample. In particular, we use distances of the raters’ subsets from the mean as characteristics of the individual raters. Furthermore, we introduce a new measure of conformity of a given rater with all others, characterizing the extent to which the rater deviates from the whole group of raters. We demonstrate the use of these methods in a case study, where the raters are forest managers and the items are trees in a forest thinning experiment. Our aim is to contribute to an understanding of the psychological processes involved, when forest managers mark trees for forest operations.

]]>
Sat, 15 Sep 2018 07:00:00 GMT /v08/i02 Andreas Wünsche, Dietrich Stoyan, Arne Pommerening
Introduction to Special Issue on Novel or Unusual Ideas in Environmental Statistics http://www.jenvstat.org/v08/i01/paper Vol. 8, Issue 1, Sep 2018

Abstract:

Innovative ideas are often unfairly rejected by journals during the publication process, in favor of more standard, mainstream articles. Brief comments on this problem, which motivated this special issue, are given.

]]>
Sat, 15 Sep 2018 07:00:00 GMT /v08/i01 Rick Paik Schoenberg
A Bayesian Model-Based Approach for Determining Multivariate Tolerable Regions http://www.jenvstat.org/v07/i06/paper Vol. 7, Issue 6, Nov 2015

Abstract:

A crucial concern of toxicologists is to determine an acceptable exposure level(s) to a hazardous substance(s). Often lab experiments produce data featuring multiple hazards and multiple outcome measures. The current practice evaluates each hazard and outcome combination separately, which leads to multiple statistical tests that suffer from inflated Type I error rates. This paper introduces a Bayesian model-based approach for analyzing data of similar nature. This approach is dimension-preserving in that it permits simultaneous quantification of an acceptable exposure level among multiple hazards. Furthermore, we introduce the concept of significance probabilities to assess the importance of the outcomes in determining an acceptable exposure level. The proposed methodology is motivated and illustrated through analyzing the dataset from a rodent study of pesticides on neurotoxicity conducted by Moser et al. (2005).

]]>
Wed, 18 Nov 2015 08:00:00 GMT /v07/i06 Epiphanie Nyirabahizi, Edward L. Boone, J. Paul Brooks, Xi Chen