Quantile regression of indoor air concentrations of volatile organic compounds (VOC).

There are many factors determining the concentration of volatile organic compounds (VOCs) in indoor air. On the basis of 601 population-based measurements we develop an explicit exposure model that includes factors, such as renovation, furniture, flat size, smoking, and education level of the occupants. As a novel method for the evaluation of concentrations of indoor air pollutants we use quantile regression, which has the advantages of robustness against non-Gaussian distributions (and outliers) and can adjust for unbalanced frequencies of observations. The applied bi- and multivariate quantile regressions provide (1) the VOC burden that is representative for the population of Leipzig, Germany, and (2) an inter-comparison of the effects of the studied factors and their levels. As a result, we find strong evidence for factors of general impact on most VOC components, such as the season, flooring, the type of the room, and the size of the apartment. Other impact factors are very specific to the VOC components. For example, wooden flooring (parquet) and new furniture increase the concentration of terpenes as well as the modifying factors high education and sampling in the child's room. Smokers ventilate their flats in an extent that in general reduces the VOC concentrations, except for benzene (contained in tobacco smoke), which is still higher in smoking than in non-smoking flats. Very often dampness is associated with an increased VOC burden in indoor air. An investigation of mixtures emphasises a high burden of co-occurring terpenes in very small and very large apartments.

A rigorous small area modelling-study for the Helicobacter pylori epidemiology.

This paper presents an investigation into spatial risk differences over small distances for the Helicobacter pylori infection in the city of Leipzig, Germany and two rural districts. A model, using Bayesian inference, was developed that adjusts the risk for individual-specific factors, and for spatial or individual over-dispersion, respectively. Additionally, the model takes into account conditional spatial autocorrelation. We found a significant positive association to the H. pylori infection risk for: "more than three children live in the household" (OR=2.4, p=0.001), "more persons live per sq.m than average" (OR=1.4, p=0.03), "home situated at main road" (OR=1.4, p=0.04) and "using well water" (OR=2.3, p=0.05). A protective effect was identified for "travelled to low prevalence region" (OR=0.4, p<0.0001) and "born in Germany" (OR=0.2, p<0.0001). Three administrative areas with significantly increased spatial risk were identified: one in the rural district and two in the city of Leipzig. The model explained 24.9% of the total deviance. Contrary to expectations, the largest part of deviance of the data was not explained by the identified significant risk factors, but by individual-specific heterogeneities. We conclude that further - so far not discussed - factors influence the risk and the spatial variation of the H.pylori infection. Furthermore, from the results we speculate about a possible impact of long-time air pollution and surface water.

Evaluation of human mobility models, for exposure to air pollutants.

The subject of the present study is to find the best way of making a simulation model for the exposure assessment of mobile individuals. For that purpose we consider and apply several very different approaches to model movement patterns in a more or less random way and compare them in terms of the exposure resulting for the individuals. The models combine random movement with agenda-driven movement. We do not aim to involve all details of real conditions into the models, but explain and review the general concepts and provide an inter-comparison of these concepts. Stationary and ergodic behaviour are explained and applied as well as performance criteria for a comparative evaluation of the investigated algorithms. In particular, the present study investigates the exposure to air contaminants of persons moving in heterogeneously polluted urban areas by help of movement simulations. For that purpose we applied four different movement algorithms: Lévy-modulated correlated random walk (LMCRW), Potential field controlled walk (PTW), Reference point mobility model (RPM), and RPM with a pre-defined daily agenda of targets (RPMA). We find that none of the studied algorithm results in purely random trajectories. PTW and RPMA prove to be suitable for human mobility modelling, because they provide conditions for very individual-specific trajectories and exposure. Suggesting these models we demonstrate the plausibility of their results for exposure to air-borne benzene and the combined exposure to benzene and nonane. It appears however that inter-individual variation in the individual-specific short-term exposure diminishes with runtime and when long-term exposure is considered.