Direct and indirect effects of urban gardening on aboveground and belowground diversity influencing soil multifunctionality

Urban gardens are popular green spaces that have the potential to provide essential ecosystem services, support human well-being, and at the same time foster biodiversity in cities. We investigated the impact of gardening activities on five soil functions and the relationship between plant (600 spp.) and soil fauna (earthworms: 18 spp., springtails: 39 spp.) in 85 urban gardens (170 sites) across the city of Zurich (Switzerland). Our results suggest that high plant diversity in gardens had a positive effect on soil fauna and soil multifunctionality, and that garden management intensity decreased plant diversity. Indices of biological activity in soil, such as organic and microbial carbon and bacterial abundance, showed a direct positive effect on soil multifunctionality. Soil moisture and disturbance, driven by watering and tilling, were the driving forces structuring plant and soil fauna communities. Plant indicator values proved useful to assess soil fauna community structure, even in anthropogenic plant assemblages. We conclude that to enhance soil functions, gardeners should increase plant diversity, and lower management intensity. Soil protective management practices, such as applying compost, mulch or avoiding soil tilling, should be included in urban green space planning to improve urban biodiversity and nature’s contribution to people.

. Urban gardens sampled in the city of Zurich. Allotment gardens are displayed in blue (N= 42) and domestic gardens in red (N= 43). Gardens were selected according to the garden type (domestic vs. allotment), the management intensity (extensive vs. intensive garden management), the degree of urbanisation (densely urbanised garden sites vs. peripheral areas). More information on the garden selection can be found in Tresch et al. 1 and Frey et al. 2 . This figure has been produced using the R package 'ggmap' 3 .

2/27
Domestic Allotment Figure S2. Example of allotment and domestic gardens in the city of Zurich. Within each garden two sampling plots (2 m x 2 m) with different garden land-use management were selected. Each of this sampling plots were later associated with one of the following garden land-use types: annual vegetable beds (vegetables; N= 47), perennial flowers and berries (flowers & berries; N= 52) or perennial lawn and meadows (grass; N= 71). This garden land-use types, rather than the two garden types, have been shown to contain the major differences in soil quality 1 and soil function decomposition 4 .  Figure S3. Pearson correlation matrix of selected soil characteristics based on the soil quality assessment of Tresch et al. 1 .
Only measurements with a very high goodness of fit statistic (p<0.001; see Table S1 1 ) for the NMDS ordination, characterising the differences in soil quality between the urban gardens of Zurich have been selected. Additionally, microbial information about gene copy numbers of Bacteria (16S) and Fungi (18S) from Tresch et al. 5 has been included in the biological soil characteristics. We dropped Boron because of the high correlation with Potassium (r=0.63) and soil basal respiration because of the correlation with C mineralisation (r=0.98). The overall variation inflation factor 6 Figure S4. PCA of soil characteristics. First four axes are needed according to the Kaiser-Guttman criteria 6 Figure S7. Alternative a priori SEM model investigating the causal relationships between urban gardening and soil multifunctionality. We expected that (1) different garden land-use types (vegetables, flowers & berries, grass) will have an effect on aboveground and belowground α and β-diversity. More specifically, we hypothesised that vegetables will have a negative effect on plant and soil fauna α and β-diversities and on soil multifunctionality compared to the other two garden land-use types. (2) Management intensity will negatively affect plant and soil fauna α and β-diversities and soil multifunctionality. (3) Higher plant α and β-diversity will increase soil fauna α and β-diversity and soil multifunctionality. (4) Soil fauna diversity aspects will positively influence soil multifunctionality. Soil characteristics, being affected by management and land-use types and urbanisation will have a direct effect on soil multifunctionality, depending on the measurements. (5) Urbanisation will have an effect on soil fauna and soil multifunctionality. Expected positive relationships are given in black and negative ones in red, grey arrows represent both positive and negative effects.  Figure S9. A priori SEM model with hypothesised direct and indirect effects of urban gardening on soil multifunctionality, including soil characteristics (cf. Figure 1). Expected positive relationships are given in black and negative ones in red, grey arrows represent both positive and negative effects. We expected that soil management will negatively affect plant and soil fauna diversity as well as soil multifunctionality (arrows 1 & 2). We hypothesised that higher plant diversity will have a positive effect on soil fauna and soil multifunctionality (arrows 3). We expected a positive effect of soil fauna diversity and biomass on soil multifunctionality (arrows 4). Urbanisation and soil characteristics (arrows 5 & 6) might have a positive or negative effect on soil fauna and soil multifunctionality.  Figure  2). Arrows represent unidirectional relationships among variables. Black arrows denote significantly (p<0.05) positive and red arrows significantly negative relationships (Table 4). Dashed grey arrows represent non significant relationships (p>0.05). The thickness of paths has been scaled based on the magnitude of the standardised regression coefficient. Conditional R 2 s, based on the variance of both the fixed and random effects, as well as marginal R 2 s, based on the fixed effect parts for each component models are given in the boxes of the response variables. Soil multifunctionality consists of five measurements related to important soil functions.  Table 4 for the complete SEM compositions. Residuals have to be independent and identically distributed, hence they should scatter around zero in the Tukey-Anscombe plots 11 . A few measurements do not fit well to the model as recognisable in the QQ-plots of the residuals, however the majority of the observations seem to fulfil the model assumptions well and since we did not assume a non-linear effect of the assessed variables with the response variables, we accepted the slight contradiction of model assumptions.

PestGrass
PestFlower PestVeg How often do you use pesticides, fungicides or herbicides to protect your lawn?
How often do you use pesticides, fungicides or herbicides (without slug pellets) to protect your flowers?
How often do you use pesticides, fungicides or herbicides (without slug pellets) to protect your vegetables? Never (1) Never (1) Never (1) Less than once per year (2) Less than once per year (2) Less than once per year (2)  How often do you use fertilisers for your flowers? Never (1) Never (1) Never (1) Every 4 to 5 years (2) Every 2 to 3 years (2) Every 2 to 3 years (2) Every 2 to 3 years (3) Once a year (3) Once a year (3) Once a year (4) 2 to 3 times per year (4) 2 to 3 times per year (4) More than once a year (5) More than three times per year (5) More than three times per year (5)

Weeds
PestTrees Leaves How often do you remove most of the weeds in your garden?
How often do you use insecticides, fungicides or herbicides to protect your trees and shrubs?
How often do you remove most of the leaves in your garden?
Do you follow the principle of mixed cultivation (planting different varieties of vegetables and/or flowers in the same cultivation plot)?
Do you leave islands of flowers when you mow your lawn?

19/27
WaterGrass WaterVeg WaterFlower How often do you water your lawn?
How often do you water your vegetable beds?
How often do you water your flower beds?
Never (1) Never (1) Never (1) When dry (2) When dry (2) When dry (2) once a week (3) once a week (3) once a week (3) twice a week (4) twice a week (4) twice a week (4) More than twice a week (5) More than twice a week (5) More than twice a week (5) CareGrass DiggingForbs DiggingCrops How often do you scarify your lawn (including reseeding) How often do you till your soil in the flower beds?

DrySticks
FstCutGrass Mulch Do you leave withered flowers and sticks during the winter in your garden?
When is the first time point of cutting your lawn?
Do you use organic material (mulch) to cover your vegetable beds? Never (5) April (5) Never ( (1) Always (1) ForkForbs ForkCrops CutTrees How often do you loosen your soil with a fork without turning it around (or milling)?
How often do you loosen your soil with a fork without turning it around (or milling)?
How often do you cut most of your forbs and trees? More than once per year (5) More than once per year (5) More than once per year (5) Once per year (4) Once per year (4) Once a year (4) Every 2 years or less (3) Every 2 years or less (3) Every 2 years (3) Every 3 years or less (2) Every 3 years or less (2) Every 3 to 5 years(2) Never (1) Never (1) Less than every 5 years (1)    Table S6. Strength of indirect and total pathway estimates of the final SEM ( Figure 2, Table 3), calculated by multiplying the standardised coefficients along the path to the response variable and adding the direct pathways 33 . Note that only significant direct and indirect pathways (P<0.05, SEM    Table S8. Estimated LMEM coefficients of soil multifunctionality (A) and its single components: belowground decomposition of green tea bags (B), aboveground decomposition of leaf litter (C), C mineralisation (D), N mineralisation (E) and water holding capacity (F), as a function of garden land-use types (cf. effect plots Figure S9). Garden ID was set as random effect in all models. Given are the mean, the 2.5% and the 97.5% quantiles of the Bayesian posterior distribution. Bold numbers indicate significant fixed effects, with credible intervals not crossing zero 11 Table S10. Soil fauna phylogenetic diversity assessed as phylogenetic species variability (PSV).

20/27
Soil fauna phylogenetic diversity, which is usually well correlated with functional diversity in most biodiversity ecosystem functioning studies 34 , was assessed as phylogenetic species variability (PSV), representing the mean of the phylogenetic correlations among species, in this case springtail and earthworms, in a community 35 . Phylogenetic trees ('rotl' package 36 ) were constructed based on the open tree of life project 10 with branch lengths ('ape' package 9 ) to calculate PSV 35 .

24/27
Table S11. Measurement details of soil functions and properties used to calculate soil multifunctionality.
We used five measurements to calculate soil multifunctionality with the averaging approach 37 . 1.) The soil function litter decomposition aboveground was measured with litter bags (18 cm x 18 cm; see Finerty et al. 38 ) with a mesh size of 4 mm on the top and on the bottom a mesh size of 1 mm, in order to prevent smaller pre-decomposed fragments from being lost during the recollection phase. We placed one litter bag on top of the soil layer in each urban garden plot (N=170) for six months (December 2015-May2016), during which most of the leaf litter accumulating in gardens will be decomposed by soil organisms. We only used litter bags with 4 mm mesh size for this calculation and not the ones with 1 mm mesh size on the top of the litter bags, in order to include also macrofauna decomposers in the proxy for aboveground decomposition. Additional leaf litter traits (e.g. C to N ratio or leaf tensile strength) can be found in Tresch et al. 4 Table A.1. The litter material (Zea mays L.) has been oven dried at 40°C and separated manually into leaf and stem parts before weighing. The starting weight in each litter bag was 2±0.01 g leaf and 2±0.01 g stem material (central leaf vein). Furthermore, only leaf litter has been used, since the mean mass loss (79.6±2.2 %) has been significantly higher compared to the more recalcitrant stems (37.9±20.8 %) 4 .

2.)
Litter decomposition belowground of mainly soil microfauna 39 was measured by the mass loss of green tea bags in accordance to the tea bag index method by Keuskamp et al. 39 . Per garden plot, four replicated tea bags for each tea type (green and rooibos tea) were buried at a depth of 8 cm for 90 days (mid-October until mid-January 2016). The mass loss, expressed as percentage change before and after decomposition was calculated after drying at 60°C and subsequent incineration of the tea bags without the nylon net 5 , in order to subtract small soil particles (< 0.25 mm, the size of the tea bag mesh) which possibly entered the tea bags during the phase of decomposition 1 . Only green tea decomposition has been used for the calculation of soil multifunctionality, because of higher mean decomposition rates 39  3.) Soil nutrient supply has been assessed by the measurements of N min and C min . N min was measured in an extract with 0.01 M CaCl 2 (1:4 w/v) following Krauss et al. 40 .
4.) C min rates were calculated as cumulative values after 4 weeks by incubating 30 g moist soil (40-50 % water holding capacity) at 20°C. CO 2 flux calculations were based on the increase of CO 2 concentration in the head-space over 6 hours, measured once per week for the 4 week time period with a gas chromatograph (7890A, Agilent Technologies, USA) as described in Tresch et al. 1 Table S12. The linearity of the enrichment was tested according to Krause et al. 41 .

5.)
The capacity of the soil to store water was measured by the soil water holding capacity (WHC). We measured WHC with a cylinder method, where field moist soil is saturated with water on a sand bath following Schinner et al. 42 . Table S12. Urban garden land-use types by garden types. Table A) displays all sampled urban garden plots. Table B) illustrates total number of observations without NA's used for the SEM and other statistical analyses. The discrepancy in observations is due to many reasons, such as missing litter bags on some sites or missing values in laboratory analyses for the measurement of the soil quality indices.