The Environmental Risk Assessment (ERA) of pesticides in soil is based on a tiered approach that uses surrogate indicators of risk spanning from single species in the laboratory to soil organism communities in the field. This thesis aims to bridge the gap between laboratory approaches and field studies using a Terrestrial Model Ecosystem (TME). Chapter I provides a systematic approach that is applied in this thesis. The effects to soil communities and the corresponding detection limits depend largely on the intrinsic variability and the sensitivity of the systems. Since TME are meant to provide a high degree of realism comparable to field studies, the first chapter gives an overview on the characteristics of soil ecosystems and the ecology of the soil organisms. Chapter II gives an overview of the experimental approaches, the test compound lindane (gamma-hexachlorocyclohexane) and the assessment endpoints measured in both TME and field studies. The complex community-level data of different soil organism groups requires a variety of uni- and multivariate statistical methods. Chapter III involves the description and interpretation of effects of the persistent and toxic pesticide lindane on soil microarthropod communities that were detected in a one-year range-finding study in TME. The open, intact soil cores (diameter 300 mm, height 400 mm) included indigenous soil organisms of undisturbed grassland. Forty units were placed outdoors in an experimental facility. The key objective was at first to evaluate the dynamics and stability of microarthropod communities on grassland over a period that is relevant for assessing the intrinsic recovery potential of the TME communities following toxic stress. Sufficient numbers of organisms and replicates of the experimental units ensured that a statistical evaluation could be performed to estimate the sensitivity and the natural dynamics and the variability of the populations upon application of lindane applied at high rates of 7.5 and 75 kg active ingredient (a.i.)/ha. The results showed that TME soil cores maintained communities of soil organisms marked by typical diversity of improved grassland. Lindane applied at excessive rates caused clear dose-related and long-lasting effects on the communities of microarthropods. On the contrary, lumbricids, the total feeding activity and the growth of plant biomass were not affected by both treatments. Based on the results of the first effect study, a modified ‘dose-response study’ with the same compound lindane was designed (Chapter IV). Further organism groups were included, so that the effects on collembolans, oribatid mites, nematodes, soil fungi and plant biomass could be determined in forty-two TME. Lindane was applied in five concentrations between 0.032 mg a.i./kg dry soil and 3.2 mg a.i./kg dry weight soil, six-fold replicated each. Twelve TME served as untreated controls. Abundance and community structures of oribatids, collembolans, enchytraeids, nematodes and fungi were recorded. Oribatid mites’ community responded three months after treatment, although they were not significantly affected by the overall treatment regime. Collembolans in total and species-specific abundance as well as the community endpoints were adversely affected by moderate dosages of lindane. Effects were transient between three and five months after treatment with a recovery within one year. No significant effects have been detected for enchytraeids, nematodes and fungi. The study design and the obtained results allow for calculations of no observed effect concentrations below the highest treatment level for populations and for soil communities as defined entities, as well as effective concentrations as indicators of dose related responses. In Chapter V the ecology of the coring area is described by means of floristic and faunistic surveys. The distribution of collembolans is analysed by geostatistical methods and categorized as patchy or gradiental. These findings lead to conclusions on optimized soil coring strategies, which are proposed as small-scaled as possible to avoid excess variability. The temporal stability of TME is investigated under the propositions of the criteria ‘no tendency towards altered abundances and diversity structures’ and ‘similarity of communities compared to the original state’. The issue ‘for which kind of habitat is the particular TME representative’ is raised. It has been concluded that a high degree of similarity between field and TME samples remains manifest over time and the pattern persists the large seasonal variation of community structures. Prospective power analyses led to an estimation of the limits of effect detection. On average, the detectable differences of abundances of treatment groups compared to control level (MDD) was between five percent for nematodes and about fifty percent for enchytraeids, collembolans and oribatids, markedly varying between sampling dates.