Fig 1.
Sampling locations in the UK and their relationship to average annual rainfall and average annual temperature, illustrating the temperature and soil moisture gradient used to help determine the farms sampled.
Sampling was carried out at SRUC, Craibstone Aberdeen (CBS), Torr Organic, Dumfries (TOR), Skeffling, Yorkshire (SKF), Loddington, Leicester (LOD), Manor Farm, Somerset (MAN), Elm Farm Berkshire (ELM) and Swanage, Dorset (DST). Figure created by Ron Corstanje, University of Cranfield and released under CC By 4.0 License.
Fig 2.
Abundances of earthworms by (a) species (total numbers). Total numbers = 13298 earthworms (endogeics = 77%, epigeics = 17%, anecics = 6%).* = continues to 6126 individuals; open bars indicates endogeic species, closed bars epigeics and checked bars anecics and (b) sampling site (mean abundance m-2 ± standard error, n = 15–20). F1 to F4 for each sampling site are fields 1 through 4 and for each farm are arranged in order from lowest management intensity on the left to highest intensity on the right. The intensity levels are relative and not necessarily comparable between farms.
Fig 3.
Panels show the sensitivity of the overall Bayesian belief network to main continuous variables used to predict values of total earthworm counts across all 7 farms.
Values of the continuous variables were split into 5 bins (where Bin 1 has the lowest values and 5 the highest) as defined for this dataset. For details of bin ranges see Table 1; the more intense the colour the more likely this value is for this number of earthworms. For example, at a depth of 5 cm, if temperature is in the range 12–16°C then earthworm abundance is most likely to be low (0–13 individuals), if the observed temperature is in the range 9.1–10.8°C, then the earthworm abundance will probably be in the range of 25–34 individuals, and at an observed temperature range 6.9–9.1°C then the abundance is most likely to be high (34–110 individuals), at a farm scale.
Table 1.
The ranges of soil properties associated with each of the 5 bins used in the BBN.
Fig 4.
Generalized linear model fits between measured soil parameters and numbers of earthworms at a species level across all sampling sites.
a) bulk density, b) nitrate-N, c) pH, d) soil moisture content (SM%). Details of the models including uncertainties are given in S7 Table. Only species where numbers vary with the particular soil parameter are shown. (Al-chl = Allolobophora chlorotica, A pros = Aporrectodea rosea, L-rub = Lumbricus rubellus, L-cas = Lumbricus castaneus, Ap-lim = Aporrectodea limicola, Ap-cal = Aporrectodea caliginosa, Ap-lon = Aporrectodea longa, M-mul = Murchieona muldall, L-fes = Lumbricus festivus, L-ter = Lumbricus terrestris.
Table 2.
Results from the Bayesian Belief Networks developed for each individual farm.
Fig 5.
Within field and farm distribution of earthworms for each of the four fields at three representative sites, a) CBS, b) DST and c) SKF. Contoured colours represent the interpolated (predicted) values of total earthworm counts, coloured dots represent the observed counts within each sampling point. Inserts show the variograms used to generate the interpolated surfaces, with the separation distances on the x-axis, and semivariance (earthworm counts, log for DST) on the y-axis.
Table 3.
Variogram parameters for the seven sites.