Figure 3: Change in soil biodiversity in response to
environmental stressors associated with (a) climate change, (b) land-use
intensification, (c) pollution, and (d) nutrient enrichment. In panel
(b), the estimate for macro- and meso-fauna is shown in black, and
micro-fauna in grey as body size remained as an additive term in the
model. Hedges’ g was used as the effect size in all models. Negative
effect sizes indicate that the environmental stressor causes a reduction
in biodiversity, and a positive effect size indicates an increase in
biodiversity. Error bars indicate 95% confidence intervals. Effect
sizes where error bars do not cross the dashed vertical zero line, are
significantly different from zero. The values of n indicate the
number of cases of each environmental stressor in the model. Grey
shading is for enhancing readability only.
For the environmental stressors model investigating different stressors
related to land-use intensification, the effect of body size remained in
the model, although as an additive effect only. Overall, macro-fauna was
more negatively impacted by each of the land-use intensification
stressors, while micro-fauna were impacted less (estimate = 0.53; ±95%
CIs = 0.22, 0.84; p-value < 0.001). Meso-fauna was not
significantly different from macro-fauna, and thus was not presented
separately (Figure 3b).
For both macro-/meso- and micro-fauna, the change from an organic system
to an inorganic system had the biggest negative impact on biodiversity
(estimate = -1.12; ±95% CIs = -1.48, -0.78; p-value <0.0001,
for macro-fauna), with the increase in tillage (i.e., comparing reduced
tillage practices to conventional tillage) having the second biggest
negative impact (estimate = -0.91; ±95% CIs = 1.17, -0.64; p-value
< 0.0001, for macro-fauna). The impact of increased fire
(intensity or frequency), harvesting, and grazing also had significant
negative impacts on biodiversity, although to a lesser extent (and not a
significant effect when adjusting environmental stressor estimates for
micro-fauna).
Within the model focussed on pollution stressors, the main and
interactive effect of body size was removed from the model. The effect
of pollutant type was significant, with both pesticides (estimate =
-0.41; ±95% CIs = -0.78, -0.04; p-value = 0.03) and metals (estimate =
-1.03; ±95% CIs = -1.35, -0.70; p-value < 0.0001) having a
significantly negative effect on soil fauna communities (Figure 3c).
Further inspection of the raw effect sizes, when accounting for
different sources of metals and pesticides (using FAO 2018 categories; )
show that there was variation of the effect sizes within each category
(Figure S4). Notably, effect sizes of metals from mining/smelting
demonstrate the greatest variation, often being more negative.
For the nutrient enrichment model, as with most of the models, the body
size variable was removed as both an interactive and main effect,
leaving just the impact of different nutrient enrichment stressors
(Figure 3d). Of the 8 different stressors, five did not have any
significant impact on biodiversity (synthetic fertilizers, Ca-liming +
Wood ash, compost, sludge and multiple fertilizer types) but all trended
towards a positive impact, except Ca-liming + Wood ash. The impacts of
manure + slurry, other organic fertilizers, and residue + mulch, were
all similar, and all significantly positive (estimate = 0.76; ±95% CIs
= 0.47, 1.05; p-value < 0.001, estimate = 0.57; ±95% CIs =
0.20, 0.95; p-value = 0.003, estimate = 0.67; ±95% CIs = 0.37; 0.97,
p-value < 0.001, respectively).
For the invasive species environmental stressors model, as with the
pollution model, both the terms for the body size and the different
types of invasive species were removed from the model completely.
However, in line with the main model, the overall intercept of the
models was not significantly different from zero (estimate = -0.15;
±95% CIs = -0.55, 0.25; p-value = 0.47).