The weed seed bank of soils in a landscape segment in southern Bavaria ? I. Seed content, species composition and spatial variability

Abstract

The purpose of the present investigation was to examine the composition and the spatial variability of the arable soil seed bank in one large, coherent landscape segment, where the cultivation practice of neighboring farms and the gradients in soil properties determine the distribution of the weed seeds in soil. To this end, a 50×50 m grid of 338 reference points was fixed throughout a 143 ha area in southern Bavaria. The seed bank contained a median density of 4950 seeds per m2 able to germinate. These values showed remarkable variation throughout the investigated area. Thus, 27% of the sampling area contained less than 2500 seeds per m2, another 23% were below 5000 seeds per m2 and 19% ranged among 5000 and 10 000 germinating seeds per m2. Among the remaining sites, 22% showed seed densities between 10 000 and 20 000 and 9% more than 20 000 seeds per m2. To visualize the spatial distribution of the weed seed bank over the whole study area, the interpolation method kriging was used. For this purpose the spatial continuity must be described by means of a semivariogram analysis. The goodness of the interpolation results depends on the fit of the semivariogram model which shows the spatial dependency. Two important characteristic parameters of this model are the nugget and the range. The nugget expresses the sampling error and the local variability up to the 50 m sampling distance, which cannot be detected by the semivariogram. Therefore the interpolation cannot describe the seed bank variation on a scale below this threshold. In our investigations the nugget comprises about one third of the total variance. The estimation error increases towards the border of the investigated area. In general, this is due to missing sampling points at the field edges. The range observed in the present investigations was 340 m. Beyond this no more spatial variability exists and the measurements are independent of each other. The species occurring in more than five of the investigated samples were ordinated by Detrended Correspondence Analysis and classified by k-means cluster analysis of the ordination axes. Five weed communities were recognized from this analysis which distinctly differed in their relation to the soil type and to the preceding management.