CHAID Models on boundary conditions of metal accumulation in mosses collected in Germany 1990, 1995 and 2000
ABSTRACT The European heavy metals in mosses surveys allow mapping the metal accumulation in mosses indicating atmospheric deposition. Yet, there is still great uncertainty on how local and regional phenomena influence the atmospheric metal bioaccumulation. Therefore, the presented study aims at ranking factors that affect the spatial patterns of the metal concentrations in the mosses. Applying chi-square automatic interaction detection (CHAID) to the German moss measurements and related sampling site-specific descriptions taken from the surveys in 1990, 1995 and 2000 and supplementary land cover data, the spatial variation in metal concentrations in mosses were proved to depend mostly on different moss species, canopy drip and distance to the sea. Most of these findings could be corroborated by classification tree analyses on the same data as presented in another study. The results of both the studies should be verified by applying the same methodology using additional emission and deposition data and monitoring information from other countries participating in the UNECE moss surveys.
- SourceAvailable from: Svatava Kubešová
- [Show abstract] [Hide abstract]
ABSTRACT: The European Heavy Metals in Mosses Surveys (UNECE-ICP Vegetation) is a programme performed every 5 years since 1990 in at least 21 European countries. The moss surveys aim at uncovering the spatiotemporal patterns of metal and nitrogen bioaccumulation in mosses. In France, the moss survey was conducted for the third time in 2006. Five hundred thirty-six monitoring sites were sampled across the whole French territory. The aim of the presented study is to give an integrative picture of the metal bioaccumulation for the entire French territory without geographical gaps. Furthermore, confounding factors of the metal bioaccumulation in mosses should be investigated. Element loads of arsenic (As), cadmium (Cd), chromium (Cr), copper (Cu), iron (Fe), mercury (Hg), nickel (Ni), lead (Pb), antimony (Sb), vanadium (V) and zinc (Zn) measured in the French campaign 2006 were aggregated to a multi-metal index (MMI). This index was first introduced in the German moss surveys and represents the mean rank of each monitoring site or estimated raster cell regarding all elements referred to. Hence, the spatial variability of the metal bioaccumulation in France could be assessed as a whole. A comparison of the MMI map with the spatial patterns of the Cu loads in mosses was then drawn, as Cu originates to a large extent from urban sources. Applying CHAID, the MMI and the Cu loads in the mosses were further investigated with regard to confounding factors. The said results were discussed on the basis of recent scientific publications. The MMI surface map shows high values in strongly industrialized and urbanized regions as well as at sites of high altitude, lying, for example in the Massif Central and the French Alps. Accordingly, the CHAID decision tree consequently shows the altitude to be the statistically most significant influencing factor of the MMI followed by the sampled moss species. As for the MMI map, the surface map for Cu mirrors urban agglomerations, as high values can be found in the areas of Greater Paris, Lyon and Marseille. The CHAID tree for Cu revealed the sampled moss species and the ratio of urban land uses within 5 km of the sampling sites to be the main influencing factors. The aggregation of metal bioaccumulation data was adopted for the French monitoring campaign. The influence of altitude, moss species-specific accumulation rates and urban emissions on the bioaccumulation is confirmed by international scientific publications. Nevertheless, the confounding factors in France differ from those derived from the German data, where the MMI was mainly associated to canopy drip effects and the growth patterns of the sampled mosses. The Cu and the MMI maps give a comprehensive overview of the metal bioaccumulation in France without geographical gaps. Hence, this approach allows summarising the spatial patterns of eleven element loads in mosses by use of geostatistics and percentile statistics. The presented metal integrating approach should be applied on data from past French moss surveys and on those to come. Additionally, the decision tree analyses should be carried out to examine possibly changing boundary conditions of the metal accumulation in mosses over time.Environmental Science and Pollution Research 05/2009; 16(5):499-507. DOI:10.1007/s11356-009-0146-0 · 2.76 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: The aim of this study was, for the first time ever, to thoroughly identify the factors influencing Cd, Hg and Pb concentrations in mosses sampled within the framework of the European Heavy Metals in Mosses Surveys 1990–2005. These investigations can be seen as a follow up of a previous study where only the moss data recorded in the survey 2005 was included in the analysis (Schröder et al. 2010). The analyses of this investigation give a complete overview on the statistical association of Cd, Hg and Pb concentrations in mosses and sampling site-specific and regional characteristics, encompassing data from 4661 (1990), 7301 (1995), 6764 (2000) and 5600 (2005) sampling sites across Europe. From the many metals monitored in the European moss surveys, Cd, Hg and Pb were used as examples, since only for these three metals deposition measurements are being recorded in the framework of the European Monitoring and Evaluation Programme (EMEP). As exemplary case studies revealed that other factors besides atmospheric deposition of metals influence the element concentrations in mosses, the moss datasets of the above mentioned surveys were analysed by means of bivariate statistics and decision tree analysis in order to identify factors influencing metal bioaccumulation. In the analyses we used the metadata recorded during the sampling as well as additional geodata on, e.g., depositions, emissions and land use. Bivariate Spearman correlation analyses showed the highest correlations between Cd and Pb concentrations in mosses and EMEP modelled total deposition data (0.62 ≤ rs ≤ 0.73). For Hg the correlations with all the tested factors were considerably lower (e.g. total deposition r s ≤ 0.24). Decision tree analyses by means of Classification and Regression Trees (CART) identified the total deposition as the statistically most significant factor for the Cd and Pb concentrations in the mosses in all four monitoring campaigns. For Hg, the most significant factor in 1990 as identified by CART was the distance to the nearest Hg source recorded in the European Pollutant Emission Register, in 1995 and 2000 it was the analytical method, and in 2005 it was the sampled moss species. The strong correlations between the Cd and Pb concentrations in the mosses and the total deposition can be used to calculate deposition maps with a regression kriging approach on the basis of surface maps on the element concentrations in the mosses. KeywordsBiomonitoring-Evaluation procedures-Statistics-Spatiotemporal analysesJournal of Atmospheric Chemistry 06/2009; 63(2):109-124. DOI:10.1007/s10874-010-9160-3 · 1.63 Impact Factor