key: cord-0985086-vm8okvyy
authors: Salomon, M.J.; Watts-Williams, S.J.; McLaughlin, M.J.; Cavagnaro, T.R.
title: Urban soil health: a city-wide survey of chemical and biological properties of urban agriculture soils
date: 2020-07-15
journal: J Clean Prod
DOI: 10.1016/j.jclepro.2020.122900
sha: d9a7b64acba524c9982ee85b802a5a7e5da4097f
doc_id: 985086
cord_uid: vm8okvyy

The integration of urban green spaces into modern city planning is seen as a promising tool to offset the drawbacks of ever-expanding cities. Urban agriculture is a common method to implement such strategies and to increase urban sustainability with a special focus on food security. Due to their location, urban farms are highly influenced by past and present anthropogenic activities which can threaten both soil health and food safety. This study includes 12 urban agriculture sites in the metropolitan area of Adelaide, Australia. It is the first of its kind to focus on soil health in urban agriculture systems with a further emphasis on mycorrhizal fungi. Descriptive information about each site, the biodiversity of the selected plots and soil samples from different depths and locations were collected and analysed for chemical and biological parameters. Seven metals, total and plant-available (Colwell) phosphorus and available nitrogen were measured in soils. A glasshouse bioassay was also conducted to determine the abundance of beneficial arbuscular mycorrhizal fungi in the soils and the change of root colonization after inoculation with the mycorrhizal fungus Rhizophagus irregularis. Results showed a generally high biodiversity of plants that correlated with site activity (commercial or community garden) and which could potentially be used for urban biodiversity conservation. Metal concentrations in soils were below national guidelines levels for all samples, although sites with previous industrial history showed elevated levels when compared to sites without industrial history. The use of raised beds with introduced soils eliminated differences in previous land-use history, thereby providing a good option to support cleaner production. Gardening soils were considered highly fertile, with plant-available (Colwell) P concentrations exceeding recommended levels for most horticultural crops, while soils were adequately supplied with nitrogen. Most plant nutrients were derived from freely available urban waste streams and integrated via composting. Various urban waste streams could be used to counter-act imbalanced soil nutrients. Arbuscular mycorrhizal fungi were present in all sites, indicating that the practiced soil management is sustainable from a microbial perspective. Given their important role in supporting plant nutrition, and potential to reduce the need for external nutrient inputs, they provide an important focal point for achieving clean and sustainable urban food production. The results were incorporated into a framework for the management of urban soil health.

Mycorrhizal abundance for external nutrient inputs, they provide an important focal point for achieving clean and sustainable 28 urban food production. The results were incorporated into a framework for the management of urban 29 soil health. 30 31

The global population is expected to reach more than 9 billion by 2050, with most of this increase to 33 occur within urban areas (United Nations, 2019). In terms of land use, urban areas are projected to 34 grow up to 80 % by the year 2030, with most of this increase happening in developing countries 35 (Mahendra and Seto, 2019). As a consequence, around 2 % of the world's current arable land will be 36 lost due to urbanisation (Bren d'Amour et al., 2017). These developments lead to various social, 37 economic and environmental challenges that need to be addressed accordingly in the context of urban 38

planning. The integration of urban green spaces is seen as a promising strategy to offset many 39 drawbacks of ever-expanding cities and to increase urban sustainability. Urban green spaces can also 40 contribute to food security, which is of special importance for developing countries. This 41 implementation is called urban agriculture (Skar et al., 2019) . 42

Urban agriculture refers to food production systems inside city boundaries or densely populated areas. 43

As such, it makes significant contribution to social, economic and ecological quality (Miccoli et al., 44 2016) . It is a global phenomenon which is of special importance for food security in developing 45 countries. Estimates suggest that the scale of urban agriculture grows linearly with the urban growth 46 of countries in equatorial Africa (Lee-Smith, 2010) . Developing countries in Asia show a similarly 47 high participation of urban dwellers in agriculture, which is considered an important source of 48 livelihood (Zezza and Tasciotti, 2010) . Urban agriculture in more developed countries has a stronger 49 focus on social components rather than food production and is often associated as a leisure activity or 50 as a form of ecological activism. To date, the driving forces behind urban agriculture appear to be less 51 concerned with food security, and more so with social, cultural and ecological factors (Mok et al., 52 2014) . However, in the context of climate change, and other shocks to the food system (e.g. the recent 53

Covid-19 pandemic), there is renewed interest in urban agriculture as a means to secure a supply of 54 clean food in all regions of the world. Especially when regional transport of foods may be affected by 55 pandemic-induced controls on movement. In the pursuit of urban sustainability and sustainable food 56 production, there is further need to re-evaluate urban agriculture on a global scale (Skar et al., 2019) . 57

The selected sites were dominated by community gardens (n=10), but also included two commercial 108 production sites in an urban setting. The sites were surveyed in September-October (Austral Spring), 109 2017 and soil physicochemical properties were measured. The same soils were used in a glasshouse 110 bioassay experiment with the aim to assess their mycorrhizal potential. 111

All sites were within a 15 km radius of the City of Adelaide (see Supp. Figure land use (see below). Using publicly available data, a total of 17 urban agriculture sites were 116 identified as potential survey sites. Selection criteria were a minimum size of 200 m² and evidence of 117 active food production. Of the 17 sites identified, representatives of 12 sites agreed to being included 118 in this study. For confidentially, the precise locations and names of some sites are not identified here. 119 2.2. Survey: site characterisation and sampling. 120

Prior to visiting sites, further information was gathered using publicly available web sites as well as 121 current and historical satellite imagery. This contextual information includes local land use context, 122 garden size and number of garden beds. Information on historical land use was supplemented and/or 123 confirmed during site visits. Upon arrival at each site, the number of beds was recorded and if 124 production took place in raised beds or not. At each site, gardening beds with evidence of active 125 farming were identified and four or five representative beds randomly selected for more detailed 126 investigation and sampling (see Supp. Figure 2B) . 127

The dimensions of the beds sampled at each site were measured, and the identity and abundance of 128 plants species being grown at the time was recorded. The source of the soil (i.e., indigenous or 129 imported potting soil) in the production areas was recorded, and where possible, information on the 130 nature of amendments (e.g., manure, compost, etc.) was recorded. Although no sites were formally 131 certified as organic, all sites followed basic principles and ethos of organic farming. These principles 132 mainly included the use of organic pesticides over synthetic ones and abstinence of any mineral 133

Soil was collected from each bed by taking five soil cores from the 0-10 cm soil layer using a 10 cm 135 diameter auger. Those five cores were then combined at the bed level to produce one composite 136 sample per bed. At two of the sites, cropping was in rows rather than beds, thus soil samples were 137 taken from an area of 1.5 x 2.5 m, which was equivalent to the typical bed size at the other sites. 138

In an effort to characterise underlying soil conditions at each site, soil samples were also taken from 139 across the site in the non-cultivated area (e.g. in the space between the beds), later referred to as the 140 'underlying soil'. These samples were taken from four separate locations randomly distributed across 141 the site (i.e. n = 4). Samples were taken from the underlying soil for the 0 -10 cm and 10 -30 cm soil 142 layers using a 5 cm diameter auger; at some sites it was not possible to sample to a depth of 30 cm 143 due to high soil strength. All soils collected were stored in air-tight plastic bags and placed in a travel 144 refrigerator at 4 °C until their return to the laboratory, where they were processed immediately. 145

Upon return to the laboratory, soil samples were carefully mixed and any coarse woody (or other) 147 debris removed using a 2 cm sieve. The sieved soil was then divided into subsamples for analysis as 148

follows. The first sub-sample was used for determination of soil gravimetric moisture content after 149 drying at 105 °C for 48 hours. The second sub-sample was used for colorimetric determination of 150 mineral N (ammonium and nitrate) on 2 M KCl soil extracts as described in Cavagnaro et al. (2006) . 151

The third sub-sample was air-dried at 40 °C for at least 48 hours and used for further physicochemical 152 analysis: soil pH and EC (1:5 water extract) was measured using a TPS WP-81 pH, TDS, 153 determined on soil digests in aqua regia and perchloric acid, followed by analysis for the individual 159 elements: Arsenic (As), cadmium (Cd), copper (Cu), manganese (Mn), nickel (Ni), phosphor (P), lead 160 (Pb) and zinc (Zn), by inductively coupled plasma optical emission spectroscopy (ICP-OES, 161

PerkinElmer Avio 200). The reference soil ACU-4 was used as certified reference material with 162 recovery rates between 89 % and 106 %. The instrument detection limits (on a soil basis) were 0.028 163 mg kg -1 for As, 0.012 mg kg -1 for Cd, 0.1 mg kg -1 for Cu and Mn, 0.028 mg kg -1 for Ni and 0.1 mg kg -164 1 for Pb and Zn. 165

Mycorrhizal fungi are often cited as a key indicator of soil health and as having a role to play in clean 167 and sustainable production systems. In order to investigate the potential for indigenous and introduced 168 (Rhizophagus irregularis, see below) AMF to colonise the roots of plant grown in the soils collected 169 from the sites, a glasshouse bioassay experiment was undertaken. Due to the limited amount of soil 170 from some sites following physicochemical analysis, it was not possible to conduct the supplemented 171 inoculation (i.e. R. irregularis) treatment on every collected sample; however, 80 % of the soils could 172 be inoculated, with n=50 in the test of indigenous AMF inoculum potential, and n=40 in the test of 173 impacts on soil after supplemental inoculation with R. irregularis. 174

The culture of R. irregularis (WFVAM10) has been used in previous studies and was found to result 175 in good mycorrhizal root colonization (Watts-Williams and Cavagnaro, 2012). The culture is 176 regularly propagated in a closed pot culture system with Tagetes patula nana as a host plant. On 177 average, 7 spores g -1 inoculum were present, as well as a variable number of infected root pieces. This 178 source of mycorrhiza inoculum has previously been found to provide high levels of AM colonisation 179 under a range of conditions. 180

The glasshouse bioassay was performed as follows: tomato (Lycopersicon esculentum cv. 76R) seeds 181 were surface-sterilized and pre-germinated on double autoclaved sand mixture, before being 182 transplanted into the final substrate after the development of the first true leaf. The final substrate 183 consisted of 150 g of the collected garden bed soils mixed with 150 g of double autoclaved fine sand. 184 R. irregularis inoculum was added (10 % w/w) for the supplemented treatment while keeping the 185 same final weight. Plants were grown in an environmentally controlled greenhouse from November to 186

December, 2017 (Austral Spring-Summer) and randomized weekly. Plants were watered daily using 187 reverse osmosis (RO) water, and no other nutrients were added. 188

Plants were destructively harvested 36 days after transplanting, and roots and shoots were separated 189 before being dried at 65 °C. At harvest, a subsample of the fresh roots was taken and stored in 50 % 190 ethanol for 24 hours. Mycorrhizal colonization was quantified using the gridline intersect method 191 after staining with ink and vinegar (Vierheilig et al., 1998) . Shoots were ground to a fine powder 192 before being analysed for the elements calcium (Ca), copper (Cu), iron (Fe), potassium (K), sulphur 193 (S), magnesium (Mg), manganese (Mn), phosphor (P) and zinc (Zn) by ICP-AES (as described 194

above). To obtain information about the presence of indigenous mycorrhizal spores in the collected 195 soil samples, a subsample of the collected soils (n = 27) was processed according to (Merryweather 196 and Moyersoen, 1997) as follows: depending on the available soil, between 10 -30 g dry soil was 197 weighed as biological triplicates and wet-sieved on 27 µm and 450 µm sieves for spore extraction. 198

The extract was then centrifuged in a 50 % sugar solution for further cleaning. The supernatant was 199 separated and washed three times with RO water. Spores were then placed onto a 45 mm glass dish 200 with four circular walls in between (nematode counting dish) and counted using a dissecting 201 microscope (Olympus SZ-PT) between 80 -100 x magnification. 202

Survey: the data was not normally distributed and was therefore analysed using the non-parametric 204

Kruskal-Wallis one-way analysis of variance with Bonferroni correction. In order to identify 205 differences between the variables 'location' (garden beds, underlying soil 0-10 and 10-30 cm) or 206 'previous industrial history' (yes/no) (see below), site means (e.g. averaged across beds) were used as 207

replicates. However, when comparing sites, individual samples were used as replicates. Where 208 significant differences were identified, post hoc tests were performed using Fisher's Least Significant 209

Difference. In order to explore the relationship between different variables (e.g. total P and plant-210 available (Colwell) P), simple linear regression modelling was undertaken. 211

Bioassay: data was not normally distributed and Kruskal-Wallis one-way analysis of variance with 212

Bonferroni correction was used in order to reveal differences between groups. Where significant 213 differences were identified, post hoc tests were performed using Fisher's Least Significant Difference. 214

Individual samples were used as replicates and analysed with the grouping factor Inoculation (none/R. 215 irregularis). Again, simple linear regression modelling was used to explore relationship between 216 different variables (e.g., shoot P concentration and soil P concentration). 217

All data was analysed with the software R in the version 3.5.0, using the package 'agricolae' 1.2 218 (CRAN, 2018) for non-parametric Kruskal-Wallis analysis with Fisher's Least Significant Difference 219 as post hoc test. Principal component analysis was performed using the function 'prcomp' and 'lm' 220 was used for the coefficient of determination R 2 . 221

The sites included in this study (Table 1) , were on average approximately 0.1 ha in size, but ranged 224 from 210 to 15,000 m 2 . Whereas at nine of the sites production was predominantly conducted in 225 raised beds using introduced soil or potting mix, at two of the sites it was in beds formed from the 226 natural soil and supplemented with self-made or externally sourced compost. The remaining site grew 227 crops in the natural soil without any organic amendments. Across all sites, an average of 35% of the 228 available area was dedicated to production (as garden beds, chickens, beekeeping and fruit trees), and 229 the remainder was used for pathways, storage facilities (e.g. sheds), and other non-production oriented 230

activities. There was an average of 29 beds at each of the 10 community garden sites, which was 231 similar to the average number of gardeners (23) at each of these sites. At the two commercial sites, 232 production was set up in rows rather than beds. While the community gardens provided a mix of 233 activities ranging from food production to social inclusion and educational activities, the two 234 commercial enterprises focused solely on food production. Compost was produced and used at all but 235 two of the sites (one commercial and one community garden). Further nutrients were imported, 236 typically in the form of commercially available municipal green-waste compost and/or animal 237 (predominantly horse) manure. Most sites were located between residential allotments and often in 238 close proximity to park lands or other nature reserves (see Table 1 ). 239

All sites together had a total plant species richness of 73 species in the production areas surveyed, and 240 at the individual site level, ranging from one to 21 species (Table 1) . On the bed level, species 241 richness ranged from one to twelve species. The most abundant crops were varieties of onions, 242 lettuce, cabbage, broad beans and carrots, all of which are typical winter crops grown in South 243

Australia. Plant richness and biodiversity (Shannon-Index) varied greatly between the sites and in 244 some cases beds only contained one plant species. The Shannon-Index was used as a biodiversity 245 index which accounts for both species abundance and evenness (Tuomisto, 2010) . However, most 246 garden beds showed a high plant species richness with different crops grown in close proximity. This 247 likely reflects the fact that beds typically service the needs of an individual grower. 248

In an attempt to identify potential contamination of these urban soils (referred to as garden beds) and 250 in the underlying soil (sampled from between the beds, using soil layers 0-10 and 10-30 cm, referred 251 to as 'underlying soil 10' and 'underlying soil 30'), soil elemental concentrations were compared to 252 National Environmental Protection Measure Health Investigation "A" Guideline Levels (NEPM-HIL) 253 as stated by NEPM (1999) ( Table 2) . Across all sites, the concentrations of As, Cu, Cd, Mn, Ni, Pb 254

and Zn were well below the NEPM-HIL A guideline levels, indicating that minimal risks to human 255 health are posed by the soil either in the beds or the underlying soils (see Table 2 ). Importantly, for As 256 and Cd, concentrations were below detection limits (0.028 mg kg -1 ) in the majority of samples and 257 were therefore omitted from statistical analysis. 258

One of the motivations for undertaking production in raised beds was a perceived risk that there may 259 be contamination in soil at the site(s), as a legacy of previous land use (e.g. industrial or unknown) at 260 the site. To explore this concern, the results were compared for concentrations of Cu, Mn, Ni, Pb and 261

Zn between sample locations (garden beds, underlying soil 0-10 and 10-30 cm) using the sites as 262 replicates (Figure 1 ). Whereas this analysis revealed significantly higher concentrations of Ni in the 263 underlying 10-30 cm soil layer than in the garden beds (p = 0.04), there were no significant 264 differences between the sampling locations for Cu, Mn, Pb and Zn. Variability within sites was high 265 with a number of outliers identified (see Supp. Figure 3) . 266

Sites were further classified on the basis of their prior land use; industrial (n = 3) or non-industrial (n 267 = 9) (Figure 1 ). When comparing metals in the garden beds at sites with industrial versus non-268 industrial land use histories, there were no significant differences detected. However, for the 269 underlying soil layers (0-10 and 10-30 cm, respectively), there were significant differences for Cu, Ni, 270

Pb and Zn, with the industrial sites having higher concentrations than the non-industrial ones (see 271 Figure 1 and Supp. Figure 3) . 272

Concentrations of plant-available (Colwell) P in the garden beds ranged from 36 to 1,265 mg kg -1 soil 274 and showed high variability within and between sites. Most of the garden beds contained relatively 275 high concentrations of plant-available (Colwell) P (median = 442.5 mg kg -1 soil), exceeding the 276 critical concentration of plant-available (Colwell) P for most horticultural crops (e.g. lettuce = 115 mg 277 kg -1 soil, Hartemink (2000)) (see Figure 2A ). Only sites 4 and 10 differed significantly from all other 278 sites, with these beds having significantly lower concentrations of plant-available (Colwell) P than at 279 all other sites. Concentrations of total P in the soil were also measured and were significantly higher 280 in the garden beds than in the underlying soil layers (0-10 and 10-30 cm) (Supp. Figure 4) . A 281 regression analysis between concentrations of total P and plant-available (Colwell) P in the garden 282 beds resulted in a positive, albeit moderate, correlation (R 2 = 0.43). 283

Total nitrogen (N) in the soil collected from the garden beds was generally high (median = 0.7%). 284

Mineral N in the garden beds was comprised from an approximate equimolar ratio of ammonium and 285 nitrate (median = 6.2 mg kg -1 and 6.1 mg kg -1 soil, respectively), and did not differ significantly 286 between sites (Table 3 and Figure 2B ). However, variability within sites was high; for example, at site 287 5 mineral N ranged from 7.6 to 26.5 mg kg -1 soil. Total N in the underlying soil (0-10 cm) was lower 288 than in the beds (median = 0.4%). Mineral N in the underlying soil was dominated by ammonium 289 rather than nitrate (median = 4.3 mg NH 4 -N kg -1 soil and 0.6 mg NO 3 -N kg -1 soil). 

Of the 90 plants included in the bioassay, 13 died within the first 14 days after transplantation, with 297 symptoms of tomato stem rot evident on those seedlings. One seedling was omitted from further 298 analysis due to a mutated growth phenotype. Of the 76 remaining plants, 27 were inoculated with the 299 AMF R. irregularis. 300

Plants growing in the indigenous soil without the R. irregularis treatment showed a mycorrhizal root 301 colonization between 3 and 56 %. Inoculation with R. irregularis increased average colonization 302 significantly from 26 to 31 % ( Figure 3B ). Altogether, 17 samples had increased colonization, three 303 samples had a neutral response and seven were negatively affected. This change in root colonization 304 was highly variable between samples collected from beds within a given site. For example, two 305 separate beds within site 4 showed the greatest increase (4/B1) and decrease (4/B3) in mycorrhizal 306 colonization with inoculation with R. irregularis ( Figure 3A) . 307

Correlation between plant-available (Colwell) P and mycorrhizal root colonization was low (R 2 = 308 0.08), and some samples with high concentrations of plant-available (Colwell) P showed a strong 309 increase in mycorrhizal root colonization with inoculation (e.g., samples 2/B4 or 6/B4). The 310 abundance of AMF spores in the tested subsample ranges from 3 to 44 spores g -1 dry soil with a mean 311 of 11 spores ( Figure 3A) . 312

Shoot biomass varied greatly between samples, similar to the measured variability of soil mineral N 313 and plant-available (Colwell) P. However, shoot biomass was significantly lower in the R. irregularis 314 inoculated plants (mean = 0.8 mg kg -1 ), than in the non-inoculated control (mean = 1.0 mg kg -1 ) 315 ( Figure 3C ). 316

Shoot P concentrations were significantly higher in the R. irregularis treatment (mean = 4.0 mg kg -1 ) 317 than in the non-inoculated treatment (mean = 3.3 mg kg -1 ). Conversely, concentrations of Fe were 318 lower in the R. irregularis treatment (mean = 0.05 mg kg -1 ) than in the non-inoculated control 319 treatment (mean = 0.07 mg kg -1 ). There were no significant differences for Zn (Supp. Figure 6) . 320

Regression analysis between concentrations of P, Mn and Zn in the plant tissue and soil resulted in R 2 321 < 0.01 for P and Zn and R 2 = 0.55 for Mn. 322

The PCA showed that shoot biomass was most closely correlated to soil total N, total P, Colwell P, 323 and total C (see Supp. Figure 5B ). Strong negative correlations were found between shoot biomass 324 and mycorrhizal root colonization and, to a lesser degree, soil nitrate. 325

The sites included in this urban agriculture study ranged in size, number of participants, and their 327 focus (commercial and community gardens). The nature of most sites was relatively uniform with 328 plants being grown in raised beds with relatively high plant biodiversity compared to conventional 329 agriculture systems. While concentrations of potentially toxic metals in soils were well below 330 guideline levels, they were higher on sites with a history of industrial land use. Whereas systems had 331 relative low levels of mineral N and adequate levels of total N, plant-available (Colwell) P was very 332 high. Collected soils were abundant in AMF spores and a greenhouse bioassay showed high 333 mycorrhizal root colonization, even in soils with high P concentration. Following, these results are 334 discussed in the context of soil health and safety as well as their significance towards sustainable and 335 clean food production. 336

There were two broad types of sites identified in this study: community gardens and commercial sites. 338

Both types differed in their configuration, farming methods and plant biodiversity. All community 339 gardens showed a strong multifunctional character by combining mainly social and ecological 340

functions. As such, they allocated more space to non-production areas and wheelchair accessible 341 pathways to allow social gatherings for the community. Food production in most community gardens 342 took place in raised beds, while both commercial sites were growing plants in the natural soil. The 343 decision to use raised beds and imported soil was in many cases due to perceived concerns around 344 potential soil contamination and was in some cases mandated by local government. In general, plant 345 biodiversity in the community garden was higher than in the commercial sites and included many 346 ornamental plants and perennials such as Rosmarinus officinalis or Physalis peruviana. The higher 347 diversity of crops grown in the community gardens is likely due to using the garden as a kitchen 348 garden, whereas the commercial sites put an emphasis on producing saleable amounts of product. 349

Those results suggest that especially the community gardens present a big potential for urban 350 biodiversity conversation and provide important ecosystem functions (Goddard et al., 2010) . The 351 sustainable character of the commercial sites lies mainly within their focus on food production, 352 combined with their proximity to the consumers and short transportation routes. Although not part of 353 this study, it is likely that food produce of both commercial sites is associated with less greenhouse 354 gas emissions than conventionally produced food (Lee et al., 2015) . Both the community gardens and 355 the commercial sites made efficient use of valuable urban space in a densely populated area. Their 356 actual configuration is a reflection of their surroundings and the needs of the local residents and they 357 all followed a strong multifunctional character (Lovell, 2010) . This multifunctionality allows all sites 358 to mitigate various challenges that arise from expanding cities (Mahendra and Seto, 2019) . various mechanisms such as atmospheric deposition, runoff from metal surfaces, bonfires, burial of 365 metal-containing waste, pesticides, or fertilizers (Alloway, 2004) . All samples in this study were 366 below the NEPM HIL-A guidelines for the tested metals, however, sites with industrial historical land 367 use had significant higher concentrations of Cu, Ni, Pb and Zn in the underlying soil layer than sites 368 with non-industrial history. In contrast, there was no significant difference in concentrations of metals 369 in soils from gardening beds when sites with and without industrial land use histories were compared. 370

The use of raised beds with introduced soils appears to have been an effective way to safely (from a 371 metal perspective) undertake food production in sites with industrial histories. Although it is unlikely 372 for developed countries to undertake any form of food production in areas with known soil 373 contamination, raised beds represent one option to help ensure a safe and secure food supply system, 374 in countries facing food shortages (Kessler, 2013) . 375

Concentrations of Zn in site 11 (mean = 258 mg kg -1 ) were well above the typical levels of about 57 -376 100 mg kg -1 in organically managed soils (Noulas et al., 2018) . This finding might not only be caused 377 by its industrial history, but also the use of Zn-based pesticides or the application of municipal 378 composts (Heiger-Bernays et al., 2009 ). While speculative, this highlights the need to consider 379 potential introduction of heavy metals, and indeed other contaminants, with external inputs. These 380 levels of Zn are of interest from an agricultural perspective but are still within the critical guideline 381 levels by a factor of 28. Although the re-use of urban waste products comes with certain reservations, 382 it did not negatively affect the sites included in this study (from a metal perspective). On the contrary, 383

it is likely that the use of organic amendments from urban waste streams saved a substantial amount 384 of energy due to the omission of mineral fertilizer (Favoino and Hogg, 2008) , however, that was not a 385 focus in this study. 386

With the exception of two sites, plant-available (Colwell) P in the soil collected from the garden beds 388 was very high, and well in excess of required levels for horticultural production (Hartemink, 2000) . 389

High levels of plant available P in these soils is likely a reflection of easily accessible nutrient sources 390 that are high in P, such as horse manure (Airaksinen et al., 2001) , coupled with the highly immobile 391 nature of P in the soil (Hartemink, 2000) . Similar results were found in various urban agriculture nitrate. However, most plant N is derived (following mineralization) from organic forms in the soil 396 which is also represented in the total N analysis. Concentrations of total N in the soil from the garden 397 beds ranged from 0.03 % to 1.3 % with a median of 0.7 %. When comparing those values against the 398 critical concentrations for wheat (0.1 %) (Hartemink, 2000) , most garden bed soils can be considered 399 adequately supplied with N. This divergence between high amounts of total N and low amount of 400 mineral N might be caused by the highly dynamic cycling of N in soils which is affected by many 401 environmental factors (Hartemink, 2000) . All things considered, nutrient management in urban 402 agriculture systems is characterised by an over-supply of urban waste products which leads to excess 403 or imbalanced soil nutrient concentrations. Such imbalances between nutrient inputs and outputs 404 should be closely monitored to avoid build-up in the soil. Excess nutrients may pose a risk due to run-405 off or can interfere with the uptake of other plant nutrients (Fageria, 2001 ). However, the use of 406 mainly organic urban waste products also resulted in high total N concentrations which is a significant 407 parameter for good soil health (Hartemink, 2000) . One solution to counteract excess or imbalanced 408 nutrients in the context of urban agriculture is to either reduce nutrient inputs or to use a blend of 409 different organic materials with different nutrient profiles. For example, after communicating the 410 issue of high P concentrations to participants of the study, one community garden incorporated spent 411 coffee ground as nutrient source which has a broad N:P ratio of about 30:1 (Liu and Price, 2011). 412

Other common composting materials with high N:P ratios are straw (N:P = 8:1) or wood chips (N:P = 413 7:1) (Wurff et al., 2016) . The results of the PCA revealed that all sites which used raised beds with 414 introduced soils shared a close relationship. This indicates that most soils and composts originate from 415 a similar source, probably due to its easy accessibility. However, most developed cities provide a 416 variety of freely available organic materials with different nutrient profiles. In order to use this 417 resource in a sustainable way, it is necessary for gardeners to familiarize themselves with the 418 principles of balanced nutrient management. 419

Mycorrhizal fungi were present in all soils collected in this survey. On average, 11 AMF spores g -1 421 dry soil were present in the samples that were used in the bioassay. Such spore abundance is similar to 422 organic agriculture soils where up to 14 AMF spores g -1 soil were found (Oehl et al., 2004) . The true 423 mycorrhiza potential of the soil samples is probably still higher, as root pieces or extraradical hyphae 424 in the soils act as another inoculum source but were not measured in this study. The mycorrhizal 425 potential is also reflected by the high percentage root colonization of plants without R. irregularis 426 inoculation. The inoculation with R. irregularis suggested that most soils have higher mycorrhizal 427 potential and can support higher root colonization. In that way, the addition of R. irregularis further 428 bolstered the mycorrhizal root colonization for most samples which might be explained by the fast-429 growing nature of this AMF species (Malbreil et al., 2014) . Interestingly, site 4 showed a high 430 variability in its response to inoculation with R. irregularis as those samples showed either a positive, 431 neutral or negative response. This response to inoculation cannot be explained within the 432 methodology of this study and might be linked to other microbial processes that impact mycorrhizal 433 growth (Miransari, 2011) . Such a spatial variability of soil microorganisms has been reported 434 previously by Štursová et al. (2016) . To this date it is not possible to compare the AMF spore 435 numbers of this study with other urban agriculture sites, as no such data are available. 436

Given the ample supply of plant nutrients at most sites, it is surprising to find such an abundance of 437 AMF in the soil. Most scientific literature even described an inhibition of mycorrhizal development at 438 high levels of soil P. The results of this study might suggest that nutrient uptake is not the major 439 driver behind mycorrhizal symbiosis in urban agriculture soils, or, is at least redundant from a nutrient Healthy soils are the foundation of urban green spaces, regardless of whether those spaces are 468 intended for leisure or food production. As such, protecting urban soils from anthropogenic influences 469

and improving soils wherever possible should be a priority in every urban planning framework. The 470

following preliminary framework outlines the main steps involved in managing urban soil health 471 based on the results of this study and with an emphasis on urban agriculture and urban green spaces 472 (see Supp. Fig. 7) . 473

The basis of this framework is to minimize the impacts of anthropogenic activities on urban soils, for 474 example through environmental policies (De Kimpe and Morel, 2000). Future urban development is 475 then classified as "hazardous" or "safe" depending on the expected effect on the surrounding soil. 476

Hazardous activities are such that are likely to result in adverse soil properties that can only be fixed 477 at high cost (e.g. organic soil pollutants or potentially toxic metals). Activities that only have limited 478 effects on soil health or effects that can be overcome in the context of urban green spaces are 479 considered "safe". Most community gardens in this study were operating on "safe" zones, where soil 480 compaction due to previous urban development was an issue that could be overcome by using raised 481

The soil health of the urban surrounding is mapped according to the two categories (De Kimpe and 483

Morel, 2000). Where soil contamination is of no concern, urban green spaces are encouraged, for Year Plan identifies specific goals which are consistent with the goals of this proposed framework, 496 such as increasing the liveability of Adelaide by planting 20 million trees by 2020 and transforming 497

Adelaide into a "green liveable city". However, the importance of healthy urban soil is only briefly 498 mentioned in the Plan and is not bolstered by any specific strategies to achieve this goal. This 499 framework could be implemented as an addition to the Plan to solve this deficit and ultimately 500 improve the overall soil quality of Adelaide for future generations. 501

The urban agriculture sites in this study provided multiple benefits towards the local community 503 which included social services, eco-biodiversity, food production and recycling of urban waste 504 streams. All sites showed strong multi-functional characteristics that allowed for efficient space use in 505 a densely populated area. All soil samples were within the national guidelines for concentrations of 506 potentially toxic metals, although higher concentrations were observed in industrially affected soils 507 than in non-industrial soil. The use of raised beds and introduced soil was a successful method to 508 offset those differences caused by the previous industrial legacy. Soils that were used for plant 509 production had an adequate supply of N and very high levels of plant-available P, which mostly 510 stemmed from freely accessible urban waste streams rather than mineral fertilization. One case 511 example showed that organic amendments can be sourced from different urban waste streams with 512 different nutritional values to avoid such imbalanced nutrient concentrations. Although most soils had 513 imbalanced concentrations of plant nutrients, they were managed sustainably from a microbial 514 perspective and contained a high abundance of mycorrhizal propagules. This naturally developed 515 mycorrhizal assemblage is likely to provide important ecosystem functions in the context of urban 516 agriculture. The findings of this study were incorporated into a preliminary framework for the 517 management of urban soil health. This framework aims to facilitate the planning and implementation 518 of urban green spaces by mapping the soil health of urban areas. 519 520 Acknowledgements 521 MJS acknowledges support from the University of Adelaide and the provided Adelaide Scholarship 522

International. SJWW acknowledges support from the University of Adelaide Ramsay Fellowship. We 523 thank Ms. Bogumila Tomczak and Mr. Colin Rivers for technical assistance. Further, we would like 524 to express our gratitude towards the gardeners and representatives of all sites involved in this study. 525

The following sites agreed to being acknowledged by name: 

Quality of different bedding materials 538 and their influence on the compostability of horse manure

Contamination of soils in domestic gardens and allotments: A brief overview

Urban agriculture and food security, nutrition and health

Urban agriculture in Bragança, Northeast Portugal: 544 assessing the nutrient dynamic in the soil and plants, and their contamination with trace metals

Future urban land expansion and implications for global croplands

Arbuscular 550 mycorrhizas, microbial communities, nutrient availability, and soil aggregates in organic tomato 551 production

agricolae: Statistical Procedures for Agricultural Research

Urban soil management: a growing concern

On-farm production of inoculum of indigenous 556 arbuscular mycorrhizal fungi and assessment of diluents of compost for inoculum production

Nutrient interactions in crop plants

Critical evaluation of municipal solid waste composting and potential 561 compost markets

The potential role of compost in reducing greenhouse gases

568 Government of South Australia, 2010. The 30-Year Plan for Greater Adelaide

Soil Analysis -An interpretation manual

Characterization and low-cost remediation of soils contaminated 574 by timbers in community gardens

576 Assessment of soil health in urban agriculture: Soil enzymes and microbial properties

Functioning of mycorrhizal associations along the 579 mutualism-parasitism continuum

Urban gardening: managing the risks of contaminated soil

Reducing greenhouse gas emissions with urban 584 agriculture: A Life Cycle Assessment perspective

Cities feeding people: an update on urban agriculture in equatorial Africa

Greenhouse gas emission reduction effect in the 589 transportation sector by urban agriculture in Seoul

The future of urban agriculture and biodiversity-ecosystem 592 services: Challenges and next steps

Evaluation of three composting systems for the management of spent 595 coffee grounds

Multifunctional urban agriculture for sustainable land use planning in the United 598

Upward and outward growth: Managing urban expansion for more 600 equitable cities in the global south

Genomics of Arbuscular Mycorrhizal Fungi: Out 602 of the Shadows

Small-scale urban agriculture results in high yields but 605 requires judicious management of inputs to achieve sustainability

Working with mycorrhizas in forestry and agriculture

Feeding the cities through urban agriculture the community 610 esteem value

Integrating community 613 gardens into public parks: An innovative approach for providing ecosystem services in urban areas

Interactions between arbuscular mycorrhizal fungi and soil bacteria

Lead (Pb) and other 619 metals in New York City community garden soils: Factors influencing contaminant distributions

Strawberry 622 fields forever? Urban agriculture in developed countries: a review. Agronomy for Sustainable 623 Development

National environment protection (assessment of site contamination) measure

Zinc in soils, water and food crops

Impact of 628 long-term conventional and organic farming on the diversity of arbuscular mycorrhizal fungi

Why farmers should 632 manage the arbuscular mycorrhizal symbiosis

How healthy is urban 635 horticulture in high traffic areas? Trace metal concentrations in vegetable crops from plantings 636 within inner city neighbourhoods in

Mycorrhizae: An Overview

Urban agriculture as a keystone contribution towards securing sustainable and 644 healthy development for cities in the future

Small-scale spatial heterogeneity of 647 ecosystem properties, microbial community composition and microbial activities in a temperate 648 mountain forest soil

The effect of site-and landscape-scale factors on lead contamination of 650 leafy vegetables grown in urban gardens

A consistent terminology for quantifying species diversity? Yes, it does exist

Delivering 657 affordable housing through the planning system in urban renewal contexts: converging government 658 roles in Queensland

Evolutionary ecology of mycorrhizal functional diversity in 661 agricultural systems

Ink and Vinegar, a Simple Staining Technique 664 for Arbuscular-Mycorrhizal Fungi

Arbuscular mycorrhizas modify tomato responses to soil 666 zinc and phosphorus addition

Fertile cities: Nutrient management 669 practices in urban agriculture

Harvest to harvest: Recovering nutrients with New 672 Sanitation systems for reuse in Urban Agriculture. Resources, Conservation and Recycling 128

Handbook for composting and 675 compost use in organic horticulture

Review on Remediation Technologies of Soil Contaminated by 677 Heavy Metals

Urban agriculture, poverty, and food security: Empirical evidence from a 680 sample of developing countries

Change of mycorrhizal root colonization after addition of R. irregularis in percentage, number of spores present per gram of dried indigenous soil (numbers) and corresponding plant-available Colwell P (dots)

 Site  1  2  3  4  5  6  7  8  9  10  11  12  Size (m 2 )  680  210  880  15,000  2100  710  700  1000  2400  600  300  1100  Year established  2011  2010  2005  1907  1992  2003  2014  2010  1994  2016  2012 

• Urban agriculture is a promising solution for food supply but not well researched • We investigated 12 urban sites regarding soil contamination and fertility • The formation of mycorrhizas in soils was quantified in a greenhouse bioassay • Metal contamination was identified, but not to levels posing concerns • Nutrients imbalances were identified, in particular overfertilization with P

☒ The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.☐The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:Matthias Salomon, on behalf of all authors