Indirect gradient analysis at different spatial scales of prorated and non-prorated earthworm abundance and biomass data in temperate agro-ecosystems

We investigated how earthworm communities of agricultural systems vary in abundance, biomass and species composition at different spatial scales. In four farms representative of the main agro-ecological regions of Belgium 14 parcels were sampled using a combined method (hand sorting after formalin extraction). Parcels in both grassland and arable land were studied. In each parcel two or three sample plots (1 m2) were randomly selected; these were further split up in two, resulting in four or six 0.5 m2 subplots per parcel. Principal component analysis was applied to earthworm numbers and biomass. Innovative was that we compared the results from data matrices with raw data with prorated matrices, where unidentified juveniles were added to species following certain rules. To test the degree of variability between the different spatial levels, average distances between subplots, plots and parcels based on ordination scores were compared. A general linear model with a nested structure for subplots, plots, and parcels was built to find absolute differences between farms. Earthworm numbers ranged from 10 to 463 individuals m–2, biomass varied between 3 and 186 g m–2 and species numbers ranged from one to seven per parcel. Ordinations separated earthworm communities along a geographical—mainly soil—gradient on the first axis and distinguished between land uses along the second axis. Both in numbers of individuals and in biomass differences in earthworm populations increased from subplot to parcel level. As expected, differences within a pair of subplots were small. Plots within the same parcel tended to be highly similar, but in some cases large dissimilarities were encountered, reflecting a clustered population structure of earthworm communities at the field scale, and, possibly, abiotic gradients. The largest differences were observed at the parcel level. The latter may be attributed to differences in land use. Prorating had a narrowing effect on the data through which differences became less pronounced.