Modelling and mapping spatio-temporal trends of heavy metal accumulation in moss and natural surface soil monitored 1990-2010 throughout Norway by multivariate generalized linear models and geostatistics

- Comprehensive analysis of correlation between heavy metal deposition and accumulation.
- Generalized linear models (GLM) can reveal a better fit than respective linear models.
- Integrated use of GLM and geostatistics yield a high spatial resolution.
- Atmospheric deposition, land use, elevation and precipitation are relevant factors.
- From 1995 to 2010 lead concentration in both moss and surface soil decreases.

Objective. This study explores the statistical relations between the accumulation of heavy metals in moss and natural surface soil and potential influencing factors such as atmospheric deposition by use of multivariate regression-kriging and generalized linear models. Based on data collected in 1995, 2000, 2005 and 2010 throughout Norway the statistical correlation of a set of potential predictors (elevation, precipitation, density of different land uses, population density, physical properties of soil) with concentrations of cadmium (Cd), mercury and lead in moss and natural surface soil (response variables), respectively, were evaluated. Spatio-temporal trends were estimated by applying generalized linear models and geostatistics on spatial data covering Norway. The resulting maps were used to investigate to what extent the HM concentrations in moss and natural surface soil are correlated. Results. From a set of ten potential predictor variables the modelled atmospheric deposition showed the highest correlation with heavy metals concentrations in moss and natural surface soil. Density of various land uses in a 5 km radius reveal significant correlations with lead and cadmium concentration in moss and mercury concentration in natural surface soil. Elevation also appeared as a relevant factor for accumulation of lead and mercury in moss and cadmium in natural surface soil respectively. Precipitation was found to be a significant factor for cadmium in moss and mercury in natural surface soil. The integrated use of multivariate generalized linear models and kriging interpolation enabled creating heavy metals maps at a high level of spatial resolution. The spatial patterns of cadmium and lead concentrations in moss and natural surface soil in 1995 and 2005 are similar. The heavy metals concentrations in moss and natural surface soil are correlated significantly with high coefficients for lead, medium for cadmium and moderate for mercury. From 1995 up to 2010 the modelled moss and natural surface soil estimates indicate a decrease of lead concentration in both moss and natural surface soil. In the case of the moss data the decrease of accumulation is more pronounced. By contrast, the modelled cadmium and mercury concentrations do not exhibit any significant temporal trend. Conclusions. In Europe, there is hardly any nation-wide investigation of statistical correlations between the accumulation of heavy metals in moss and natural surface soil and potential influencing factors such as atmospheric deposition. This study could show that assessments of heavy metal concentrations in natural surface soil could complement biomonitoring with moss but should not replace it since the heavy metal concentrations in mosses reliably traces the spatial pattern of respective atmospheric deposition. Generalized linear models extend established methods for estimating spatial patterns and temporal trends of HM concentration in moss and natural surface soil.