Excess fertility in residential‐scale urban agriculture soils in two western Oregon cities, USA

Abstract

Small‐scale, residential food gardens are arguably the most common form of urban agriculture (UA) in the world. Despite their ubiquity, we know relatively little about the characteristics of UA soils, in general, and of residential food gardens specifically. We thus sampled soils from 27 residential‐scale vegetable gardens in two western Oregon cities to describe the physical, chemical, and biological characteristics of residential‐scale UA soils. We distinguished growing sites by bed type: in‐ground beds (IGs) and raised beds (RBs). We assessed the proportion of soils that fell within published recommendations for vegetable production for various soil parameters. We found residential‐scale UA soils frequently exceeded recommended ranges for many fertility parameters. We also found differences in carbon/nitrogen ratio, active carbon, and sulfur, with RBs significantly higher than IGs. The excesses likely are due to routine overapplication of compost, soil amendments, and fertilizers by growers across their intensively managed urban spaces. Such overapplication and excess is likely to be exaggerated in RBs compared with IGs.

Abbreviations
• IG
• in-ground bed
• PMN
• potentially mineralizable nitrogen
• RB
• raised bed
• UA
• urban agriculture

INTRODUCTION

In an increasingly urbanized world, urban agriculture (UA) has been suggested as a means of building robust local food systems to meet future food needs (Akaeze & Nandwani, 2020; Burgin, 2018; Maheshwari et al., 2014; Tilman et al., 2002), foster sustainable and resilient cities (De Zeeuw et al., 2011; Maheshwari et al., 2014), promote public health and food justice (Morgan, 2015; Specht et al., 2014), grow an appreciation for agriculture within the general public (FAO, 2014; Morckel, 2015), and increase consumption of fresh fruit and vegetables (Evans et al., 2012). Approximately 30% of the world's population engages in UA, with about one-quarter producing food for sale (Badami & Ramankutty, 2015). In the United States, UA has recently attracted renewed interest, in large part, because of increased awareness of and desire for local foods (Mason & Knowd, 2010; McClintock et al., 2016; Wortman & Lovell, 2013) and policy initiatives (Taylor & Lovell, 2014). Despite its popularity, research into UA suffers from a general lack of funding for research in urban systems (Acuto et al., 2018). Additionally, "poor quality and weak study designs" have rendered most attempts to quantify UA contributions to food production unreliable (Warren et al., 2015). As a result, the unique needs of urban farmers are often unmet by university research efforts and extension services (Oberholtzer et al., 2014).

Unfettered by this paucity of relevant research, residential food gardens are perhaps the most common form of UA in many parts of the world (McClintock et al., 2016; Taylor & Lovell, 2014). The importance of residential food gardens within urban and suburban areas of the United States is expected to grow as interest in food gardening—particularly among millennials—is a major driver of yard and garden center sales (Garden Research, 2018). Cities across the country hold great agricultural potential (Grewal & Grewal, 2012) particularly if vacant land can be transformed into community gardens or other urban agricultural spaces. Residential food gardens represent 89% of all urban agricultural sites in Chicago, IL, covering 11.73 ha (29 acres) (Taylor & Lovell, 2014). In Portland, OR, a conservative estimate of more than 3,000 residential food gardens accounts for ∼8.1 ha (∼20 acres) (McClintock et al., 2013). In Madison, WI, over 45,000 residential food gardens cover more than 48.9 ha (121 acres) of land (Smith et al., 2013).

Research on residential food gardening includes efforts to map the spatial distribution of residential food gardens (McClintock et al., 2013; Smith et al., 2013; Taylor & Lovell, 2014) or to describe the social or economic benefits of home food gardens (Albaladejo-García et al., 2021; Dubová & Macháč, 2019; Gray et al., 2014; Langellotto, 2014; Schupp & Sharp, 2012). Broader investigations into residential gardens, not tied to food production, have documented the social and ecological drivers of the diversity of insects (Fetridge et al., 2008), birds (Lerman & Warren, 2011), or plants (Avolio et al., 2018).

Urban soils are largely ignored in the studies of growing food in urban areas (but see Gaston et al., 2019; Tresch et al., 2018a; Tresch et al., 2018b). Past research has documented the extremely heterogeneous nature of urban soils, known as the ‘urban soil mosaic’ (Pouyat et al., 2010). This mosaic results from the cumulative effects of anthropogenic factors in an urban area that have modified soils to the point that they should be considered separate from native, undisturbed soils in the region (Amundson et al., 2003). Most assessments of urban soils occur along roadsides and former industrial sites (Pouyat et al., 2010). These assessments typically focus on one of two categories: soil contaminants (Meuser, 2010) or soil compaction (Pavao-Zuckerman, 2008; Effland & Pouyat, 1997).

Given how little we know about urban soils, in general, it is important to point out that we know even less about urban agricultural soils. This has led to calls that UA soils research be prioritized as a focal area for soil scientists in the 21st century (Adewopo et al., 2014). A nationwide survey of urban farmers in the United States supports this conclusion: 85% of respondents identified ‘soil health’ as a key production challenge, and 77% identified ‘soil fertility’ as a priority for technical assistance and information (Oberholtzer et al., 2014). Despite their numerical dominance in UA (McClintock et al., 2013), few studies have assayed the current status of UA soils from residential-scale sites (but see Small et al., 2019; Tresch et al., 2018a, 2018b). Urban growers at both commercial and personal production levels need research directly addressing their needs to manage fertility and react to compromised or subprime soils.

Core Ideas

• Urban vegetable garden soils are overenriched in organic matter and some nutrients.
• Soil characteristics of in-ground garden beds are distinct from raised beds.
• Regularly importing soil or media to create gardens can contribute to excess fertility.

We thus sampled soils from 27 residential-scale UA sites (i.e., urban vegetable garden sites) in order to document their physical, chemical, and biological characteristics. We hypothesized that management of UA soils would differ by bed type: in-ground beds (IGs) or raised beds (RBs) (Figure 1). We made no a priori assumptions about the results of urban garden soils, on average. However, we hypothesized that RBs would have lower levels of heavy metals than IGs and that the chemical characteristics of IGs would fall within recommended agronomic ranges, more so than RBs. We made no a priori assumptions about UA soil characteristics because the soil could be compacted and contaminated or meticulously managed and enriched. We based the hypotheses of differences between beds on the observation that RBs are often filled with organic- and nutrient-rich materials that do not reflect the native soils of an area.

FIGURE 1. Examples of the two residential-scale urban agriculture bed types that were the focus of this study: raised beds (left) and in-ground beds (right)

MATERIALS AND METHODS Sample sites

We conducted our study in two cities of western Oregon: Corvallis and Portland. We recruited potential study sites from the pool of Oregon State University Extension Master Gardener volunteers in Benton, Multnomah, Marion, and Lane counties. We received 87 total responses to our invitation to participate. We screened out those sites that did not have a vegetable garden, where the respondent had managed their site for <1 yr, or when the respondent did not have authority to grant access to their site. Because of a low response rate in Marion (n = 3) and Lane counties (n = 4), we eliminated these potential study sites from our sampling scheme. Of the remaining sites in Benton and Multnomah counties, we eliminated sites that were located outside of Corvallis or Portland city limits, or that would have otherwise created inefficient sampling routes. Of the 34 remaining potential study sites, seven did not respond to attempts to schedule site visits. This left us with 27 total study sites: 11 in Corvallis and 16 in Portland (Figure 2). Six of these sites had both types of garden beds and so yielded additional samples resulting in 33 final samples: 21 RBs and 12 IGs.

FIGURE 2. Location of study sites in the cities of Corvallis, OR, and Portland, OR, USA

Between 23 and 26 Aug. 2017, we visited all sites and noted the UA bed type (RB or IG), measured the vegetable garden bed area (m²), and measured bed height (cm). For the purposes of this study, RBs were defined as those with an installed border that physically delineated the productive space from the ground surface and often elevated the soil within its borders, while IGs were those without protective borders. While others have classified beds with mounded media as a type of RB (e.g. Reeves et al., 2014; Edmunds, 2016), we chose to use the presence of physical borders as the defining aspect of an RB on the assumption that a hard border would affect management options like use of machinery or volume of imported media needed.

We also recorded the presence of all crops to the species level and further categorized them as annuals or perennials (Appendix 1). If there was a bare patch in the garden bed, we asked the gardener what had recently been harvested.

We did not collect personal or site-history data for this study.

Soil sampling

We collected soil samples on 30 Aug. 2017 (Corvallis, OR), and 31 Aug. 2017 (Portland, OR), using a 1-m long, 2-cm diam. auger to collect 30-cm cores. We chose this sample timing to be as late in the growing season as possible while still before autumn rainfall. In this way, we were hoping to capture season-end soil parameters rather than short-term peaks in nutrients following fertilization in the spring or the loss of nutrients to leaching from fall rains.

We aimed to sample soil toward the center of the beds but adjusted as needed to avoid irrigation components or plant materials. We took a soil sample every 2 m of bed length (Figure 3) removing any mulching materials prior to taking the sample. If the bed was <2 m long, we took a single sample as close to the middle as possible. For each bed at each site, we collected soil samples into a bucket where we manually mixed all soil cores and then scooped the collected soil into a 4.4 L (1 gallon) polyethylene bag that we labeled with the site number and bed type. To preserve sensitive biological and chemical data (Barker et al., 1969), we kept the samples in a cooler with ice until the end of the day when we moved them into cold storage at 3.3 °C. Soils were held in this cold room until they could be processed. Soil compaction was measured in the field on these same days. All other soil parameters were calculated from soil samples collected in the field and assayed in the lab.

FIGURE 3. Idealized soil sampling scheme in a sample garden bed. Soil core samples are represented by the red “×.”

Soil sample preparation

All samples were processed and analyzed at the Oregon State University Soil Health Laboratory (https://cropandsoil.oregonstate.edu/shl). In the lab, samples from the same site and same bed type were homogenized by passing the media through a 2-mm sieve. Especially clumped and dense samples were passed through an 8-mm sieve and left to air dry at ∼30 °C for 1 d before rejoining the intake process. We examined the material that did not pass through a sieve for identifiable rock and organic matter and discarded it then used a ceramic mortar and pestle to lightly crush the remaining material. We once again passed this material through the 2-mm sieve. After the final sieving, all matter >2 mm diam. was discarded. We placed each sample in a large tin then set in a drying rack to air dry at ∼30 °C. We placed dried samples in a labeled plastic bag and kept them at room temperature (21∼23 °C) for 2–7 d as they awaited processing.

Soil physical parameters

We used a Dickey–John soil compaction tester in the field to measure compaction within vegetable garden beds. We probed within a square meter of both ends of all beds. If the beds were longer than 4 m, another probe was inserted every 2 m. If the length of a bed was <2 m, we used a single probe. We recorded compaction to the nearest 25 psi in the top 40 cm of the soil or where the depth of the rating exceeded 300 psi (Duiker, 2002). We averaged readings to obtain a single value per bed then translated a site's penetrometer readings into a compaction category: little to none, slight, moderate, or severe (Duiker, 2002).

We found bulk density (D_b) by proxy for each soil sample using the best-fit revised empirical model proposed by De Vos et al. (2005). It uses organic matter percentage as determined by loss-on-ignition (LOI) testing: [Image Omitted. See PDF]

We used a rain simulator (Ogden et al., 1997) with minor adaptations (Moebius-Clune et al., 2016) to determine the percentage of wet-stable aggregates in a sample.

To determine particle size and soil texture, we followed the pretreatment procedure outlined in Day (1965). We lost one IG sample because of unexpected overbubbling of the sample. We adjusted Gee and Bauder's (1986) procedure for hydrometer readings. Soil texture was determined by calculating the percentage of sand, clay, and silt particles in each sample and then using a soil texture calculator (USDA–NRCS, 2018). Percentage organic matter within a soil sample was determined by doubling percentage soil carbon (Pribyl, 2010).

Soil biological parameters

We measured active carbon following the protocol outlined in Moebius-Clune et al. (2016). We measured potentially mineralizable nitrogen (PMN) following an adaptation of the protocol outlined in Moebius-Clune et al. (2016), where samples were incubated in an aerobic environment for 1 mo. We added enough deionized water to bring the samples to approximate field capacity, as approximated from the bulk density proxy determined in Equation 1. The sum of the difference between ammonia (NH₄–N) and nitrate-N (NO₃–N) on Day 30 and Day 0, divided across 30 d, is recorded as PMN: [Image Omitted. See PDF]We found microbial respiration following the protocol outlined by Franzluebbers (1999) except that we used a Picarro CO₂ isotope analyzer to measure the concentration of CO₂ gas produced by samples 5 min after rewetting, again after 24 h of incubation at ∼22 °C, and once more after 72 h of incubation. One RB site was not analyzed, as we ran out of sampled media. The Picarro CO₂ isotope analyzer reports concentration of ¹²CO₂ and ¹³CO₂ for 2 min per sample as a rolling average. We recorded a reading from the final 10 s of each sample, then determined total respiration by summing the two isotope concentrations (¹²CO₂ + ¹³CO₂) across time.

Soil chemical parameters

We used methods outlined by Moebius-Clune et al. (2016) for soil pH with adjustments because of necessary dilution of the soil slurry beyond a water/soil ratio of 1:1. We followed an adapted protocol (Tiessen et al., 1981) to prepare samples for combustion analysis via Elementar Vario MACRO Cube to yield a measure of total ash, total organic matter, and percentages of carbon, nitrogen, and sulfur. We determined the C/N ratio using the total percentages found by this combustion analysis. We followed this same protocol to prepare samples for the Mehlich-III extraction process (Mehlich, 1984) to assess phosphorus, potassium, calcium, magnesium, manganese, sodium, iron, copper, aluminum, and zinc. We found boron content using a hot water bath to perform a calcium chloride extraction (Bingham, 1982) with an adjustment (Jones Jr., 2001) before final analysis of the filtrate with a PerkinElmer 2100 dual-viewing, inductively coupled plasma–optical emission spectrometer (Horneck et al., 1989).

Soil heavy metal content

We tested for the following heavy metals: arsenic, cadmium, cobalt, copper, chromium, lead, and nickel. We prepared samples for microwave digestion using an Anton Par Multiwave GO following Kingston and Jassie (1986) with an adjustment (Sah & Miller, 1992). We loaded samples into an Anton Par Multiwave GO before final analysis with the PerkinElmer 2100.

Determining recommended ranges for soil parameters

Soil parameter recommendations for residential gardeners (Collins et al., 2013) are often derived from or refer home growers to crop-specific vegetable production guides intended for commercial growers (e.g., Oregon State University, n.d.). Because residential gardeners often grow several crops with slightly or very different soil specifications within the same garden bed, we gleaned recommended ranges for soil parameters from general guides for agricultural soils (Horneck et al., 2011; Moebius-Clune et al., 2016). Where recommended ranges were available, we compared descriptive statistics for each soil sample parameter to its recommended range (Horneck et al., 2011; Moebius-Clune et al., 2016). Some parameters (e.g., bulk density proxy, active carbon, manganese, copper, microbial respiration) lacked either or both upper and lower limits in published recommendations for productive soils.

Statistical analyses

We used R (R Core Team, 2021) to analyze data from two approaches. First, we generated descriptive statistics for all soil parameters that we measured, and then we compared means and 95% confidence intervals against agronomic recommendations (see footnotes for Table 2). When an agronomic recommendation fell outside of the 95% confidence interval, we determined that sampled soils deviated from recommended levels.

Second, we used a series of unpaired t tests to compare soil parameters from RBs and IGs to identify parameters that differed by bed type. We assessed whether each parameter met test assumptions using Shapiro–Wilk's test to inspect normality and Levene's median test to examine residual variance. Several soil parameters (chromium, lead, bulk density proxy, boron, magnesium, potassium, sulfur, nitrogen, carbon, and organic matter) had nonnormal distributions and were subjected to Box-Cox transformation prior to statistical analysis. All parameters passed Levene's median test except for zinc, carbon, C/N ratio, and organic matter. We used Welch's t test, as it does not require equal variances. The P values were adjusted for multiple testing using the Benjamini–Hochberg correction. The P values below .05 after adjustment were considered significant.

RESULTS

Individual garden sites in Corvallis encompassed, on average, two-thirds more area than the Portland gardens (Table 1). There was no significant difference between the area of garden planted as RB vs. IG (F[1,31] = 0.56, p = .454). As expected, the height of RBs was significantly greater than IGs (F[1,31] = 14.525, p < .001). Vegetable garden beds were planted with a diverse array of 132 crops across all gardens (Appendix 1). Residential vegetable garden soils exceeded ranges recommended ranges for a broad array of soil parameters (Table 2).

TABLE 1 The heights and areas of garden beds in both cities

	Bed area			Bed height
	Total	Range	Median	Range	Median
	m2			cm
Total (n = 33)	831	1–101	18	0–81	13
Corvallis (n = 14)	446	35–101	35	0–36	6
Portland (n = 19)	385	1–76	12	0–81	15
In-ground beds (n = 12)	324	1–55	27	0–28	0
Raised beds (n = 21)	507	2–76	15	3–81	20

TABLE 2 The mean and standard error for each tested parameter across all beds (n = 33), raised beds (RB; n = 21), and in-ground beds (IG; n = 11). Recommended ranges are cited from general agronomic guidelines

Parameter	Recommended range	Sample mean	RB mean	IG mean
Soil physical parameters
Bulk density proxy (g ml−1)	<1.6a	1.2 (0.02)	1.1 (0.04)	1.3 (0.02)
Wet aggregate stability (g 100 g−1 dry soil)	>35b	55.4 (2)	57.8 (3)	50.6 (2)
Penetrometer (psi)	<300b	151 (11)	134 (11)	186 (9)
Organic matter (% of soil by mass)	3–6c	13 (1)	15 (1)	10 (0.7)
Soil biological parameters
Active carbon (mg kg−1)	>400b	859 (41)	950 (35)*	669 (37)
Potentially mineralizable nitrogen (NO3–N + NH4–N μ [g soil]−1 day−1)	None found	0.6 (0.03)	0.6 (0.03)	0.5 (0.03)
Nitrate–nitrogen (NO3–N) ppm	N/Ad	0.53 (0.02)	0.57 (0.03)	0.52 (0.04)
CO2 respiration (μg CO₂ [g dry soil]−1 day−1)	500–1,500b	319 (11)	361 (3)	248 (2)
Soil chemical parameters
pH	6–8.2e	6.4 (0.05)	6.4 (0.05)	6.5 (0.03)
C/N	24:1f	14.1:1 (0.5)	15:1 (0.5)*	12.6:1 (0.2)
Carbon (%)	>1.25g	6.5 (0.6)	7.3 (0.6)	4.8 (0.3)
Nitrogen (%)	0.1–0.15e	0.5 (0.04)	0.5 (0.04)	0.4 (0.02)
Sulfur (%)	N/Ae	0.07 (0.003)	0.08 (0.003)*	0.06 (0.003)
Phosphorus (mg kg−1)	27–135e	273 (18)	283 (18)	262 (17)
Potassium (mg kg−1)	171–912e	659 (144)	772 (182)	448 (40)
Calcium (mg kg−1)	N/Ae	4196 (235)	4524 (181)	3510 (196)
Magnesium (mg kg−1)	60–300e	609 (48)	624 (46)	545 (49)
Manganese (mg kg−1)	1–5e	36 (2)	37 (2)	35 (2)
Plant-available copper (mg kg−1)	>0.64	13 (1)	12 (1)	15 (1)
Zinc (mg kg−1)	>1.5e	36 (3)	39 (3)	31 (2)
Boron (mg kg−1)	0.5–2e	0.7 (0.1)	0.8 (0.1)	0.5 (0.04)
Soil heavy metals (mg kg−1)
Arsenic	None found	4 (0.5)	4 (0.4)	5 (0.7)
Lead	None found	44 (13)	52 (15)	30 (6)
Cobalt	None found	14 (0.7)	14 (0.7)	15 (0.4)
Copper	None found	50 (3)	51 (3)	47 (2)
Chromium	None found	24 (1)	23 (2)	25 (1)
Cadmium	None found	0.2 (0.03)	0.2 (0.03)	0.2 (0.03)
Nickel	None found	18 (1)	18 (1)	18 (0.6)

Note. Original recommendation given for Bray-P or Ammonium Acetate-K extraction, converted to Mehlich-3 using Culman et al. (2019). Italicized text indicate means that fell outside the recommended range for that parameter. Significant differences (≤.05 probability level) are marked by an asterisk (*) in the column with the greater outcome. Values in parentheses represent standard error.

a

Recommended ranges were derived from Daddow and Warrington (1983).

b

Recommended ranges were derived from Moebius-Clune et al. (2016).

c

Recommended ranges were derived from Fenton et al. (2008).

d

Recommended ranges were derived from Sullivan et al. (2019).

e

Recommended ranges were derived from Horneck et al. (2011).

f

Recommended ranges were derived from Soil Survey Staff (2011).

g

Recommended ranges were derived from Pribyl (2010).

Soil physical parameters

The soils in this study had an abundance of organic matter (12.9 ± 1.1%). The RBs (14.8 ± 1.18%) were not significantly more enriched in organic matter than IGs (9.6 ± 0.679%; t_23.36 = −2.63, p = .065). Soils in this study were well below the recommended threshold of <300 psi (150.9 ± 11.02 psi). Of 33 beds (21 RBs and 12 IGs), 30 showed no evidence of compaction. We found no significant differences between bed types for wet aggregate stability or bulk density proxy. Soil texture was similar among study sites and between bed types, although Corvallis soils had slightly more clay (Figure 4). Most soil samples (59%) were classified as loam. The rest of the sites were classified as sandy loam (24%), clay loam (12%), silt loam (3%), and sandy clay loam (3%).

FIGURE 4. Soil texture of garden sites with Portland, OR, in blue, Corvallis, OR, in red, in-ground beds as squares, and raised beds as circles. Modified from Soil Survey Staff (1951)

Soil biological parameters

Active carbon (858.8 ± 41.36 mg kg⁻¹) was significantly greater in RBs (950.1 ± 35.39 mg kg⁻¹) than for IGs (699.1 ± 37.06 mg kg⁻¹) (t_22.15 = −3.31, p = .046), doubling the recommended minimum. The PMN is a combination of four parameters—initial and final concentrations of both ammonia and nitrate-N. Ultimately, PMN did not differ significantly between bed types (t_22.33 = −0.58, p = .608). Nitrate-N was also not significantly different between bed-types (F_32,1 = 0.258, p = .615).

Soil chemical parameters

Soil pH (overall, 6.4 ± 0.05; RBs, 6.4 ± 0.05; IGs, 6.5 ± 0.03) was well within suggested range for annual vegetables.

Carbon content (6.5 ± 0.56%) showed a wide range (2.5–15.2%), all well above the recommended minimum. The C/N ratio of the sites (14.1 ± 0.49) showed RBs (15.1 ± 0.54) as significantly greater than IGs (12.6 ± 0.19) (t_30.27 = −3.02, p = .049), all with greater nitrogen than suggested.

Sulfur (0.07 ± 0.003%) was significantly elevated in RBs (0.08 ± 0.003%) compared with IGs (0.06 ± 0.003%) (t_29.99 = −3.30, p = .046). Calcium (4,363 ± 235 mg kg⁻¹) averaged double the upper limit of the recommended range (1,000–2,000 mg kg⁻¹). On average, the following chemical parameters fell within agronomic recommendations: pH, carbon, nitrogen, potassium, manganese, plant-available copper, and zinc. Most soil elements were not significantly different between bed types including nitrogen, phosphorus, potassium, magnesium, manganese, copper, zinc, and boron.

Heavy metal content

Heavy metals (e.g., arsenic, cadmium, cobalt, copper, chromium, lead, and nickel) fell well below recommended maximums for all gardens. In addition, we found no significant differences in heavy metal content between bed types. One IG had an arsenic level that was concerning (17 mg kg⁻¹)—four times the sample average (4 ± 9 mg kg⁻¹)—and above the recommended threshold of 16 mg kg⁻¹.

Across all sites, lead levels were below 400 mg kg⁻¹, the level at which restrictions on gardening activities are recommended (Brewer et al., 2016). However, eight sites had lead levels >50 mg kg⁻¹, the level at which gardeners are cautioned to limit dust exposure (Brewer et al., 2016). These soils came from a raised bed in Corvallis (197 mg kg⁻¹) and four RBs (63–346 mg kg⁻¹) and three IGs in Portland (68–83 mg kg⁻¹).

DISCUSSION

This study presents a comprehensive reporting of the physical, biological, and chemical characteristics of residential-scale UA soils. The physical properties of a soil strongly influence several key soil processes (e.g., retention and transport of water and nutrients) and are associated with the agronomic potential of a soil (McCarty et al., 2016). The biological components of soil influence soil evolution (Jenny, 1941) and soil productivity (Raynaud et al., 2021; Schnitzer et al., 2011). We did not directly measure soil microbes but instead measured proxies of soil microbial activity. For example, PMN represents how much organic nitrogen is being consumed and excreted by microbes. Carbon dioxide respiration is a direct measure of carbon respired by microbes.

In addition, we examined the differences between garden soils in RBs vs IGs. We did not collect personal data about the site management, so any of the growers could have applied any fertilizer before we arrived for sampling. However, we timed our sampling to be as late as possible in efforts to avoid fertilization, as we assumed growers would be unlikely to fertilize right before the fall and winter rains. Our observations of the sites showed no obvious signs of recent fertilization nor did any grower make any such mention to us.

Excesses

The excess nutrients that we documented likely come from routine compost applications and nutrient buildup over time (Ugarte & Taylor, 2020; Small et al., 2019; Ninkov et al., 2018; Seiter & Horwath, 2004). The story of urban soils is one where anthropogenic factors (Joimel et al., 2020; Taguchi et al., 2020) are the primary drivers of soil enrichment (Asabere et al., 2018). Simply by being urban, these soils are under elevated management scrutiny (Dobson et al., 2021; Bretzel et al., 2018) to such a degree that it is “a common phenomenon in disturbed urban soils” to see soil carbon increase over time (Midgley et al., 2021). However, the response of garden soils to organic matter additions also demonstrates the ability of small-site managers to redirect the trajectory of their soils. For example, an analysis of pesticide labels for home use vs. agricultural use of imidacloprid (a systemic insecticide) has shown that label directions could result in home-scale application rates that are up to 120 times higher than the maximum label rate approved for agricultural crops (Hopwood et al., 2012). However, these excessive inputs do not translate to increased harvests (Shrestha et al., 2020; Small et al, 2019). Instead, excesses, such as phosphorus (Small et al., 2019), seem to concentrate in the native soil just below the managed soil if they are not washed away with the groundwater. The excessive calcium, phosphorus, and potassium that we documented from garden soils in this study are unlikely to harm crops nor is this excess likely to encourage rapid growth (Browne, 1942). Instead, these overenriched soils that exceed agricultural recommendations are not only economically wasteful but may also promote nutrient loss from the production beds (Riaz et al., 2018; Dewaelheyns et al., 2013). This matter can be further affected by the amount of residue and the species of the previous crop (Goss et al., 1991).

Walsh et at. (2005) examined ‘urban stream syndrome,’ a model describing the consistently observed ecological degradation of streams that drain from urban landscapes. Gittleman et al. (2017) estimate that 0.23 and 4.06 cm of water runs off per 6.54 cm² of New York City community garden during a typical (3.81-cm rain event) and heavy storm (12.7-cm rain event), respectively. However, we do not yet know the extent to which excess nutrients might be lost from urban gardens following a rain or irrigation event. But this combination of overenriched soils in a landscape surrounded by impervious surfaces suggests that urban gardeners might be contributing to urban runoff and waterway contamination (Taguchi et al., 2020).

The importance of management

We chose to investigate the differences between RBs and IGs because we assumed that the choice between bed type would have a significant effect on the management of those garden soils. For example, the border of an RB might discourage tillage or otherwise limit equipment access. Borders are also likely to reduce physical erosion or other displacement of input and fertility. Additionally, an RB typically requires the import of media to fill it and so dramatically change the content and nature of those soils compared with IGs.

We found stable soil parameters beneath the general excess of soil nutrients such as bulk density proxy, pH, and texture. These factors are important, as they affect nearly all aspects of chemical and biological interactions within the soil matrix (Mengel & Kirkby, 2001). With such foundational parameters showing such little variance, we suggest that results from this study could be comparable with other urban garden soils. Differences between sites or bed types are less likely to be related to the underlying, native soils but are instead likely are due to management practices (Lin et al., 2018; Tresch et al., 2018a).

Some researchers have suggested that urban soils are expected to have low levels of soil microbial activity (Lorenz, 2015), high levels of soil compaction (Lorenz, 2015; Pavao-Zuckerman, 2008), and high levels of contamination (Pickett et al., 2011). However, like Ninkov et al. (2018), we found UA soils in this study to be of high fertility. This emphasizes the importance of site location and demonstrates dramatic differences between urban soils, for example, brownfields vs. gardens. All UA soils sampled had evidence of active microbial communities (e.g., PMN, active carbon, and CO₂ respiration), which could benefit crop production via cycling nutrients out of organic matter and into plant-accessible forms of nitrogen, phosphorus, or other elements but may also promote leaching of mobile compounds like nitrates. Nearly all soils sampled had low to no compaction. All soil samples were below recommended thresholds for heavy metals. This follows previous research (Pouyat et al., 2010; Amundson et al., 2003) that suggests that anthropogenic factors supersede traditional soil formation factors. In essence, no matter the initial conditions, if it is in an urban location, a soil can be significantly altered in a very short time span.

Raised beds are often recommended to urban gardeners, specifically (Edmunds, 2016; Finster et al., 2004), and to home gardeners, in general (Bell et al., 2014; Langellotto-Rhodaback et al., 2011), as a means of working around soil issues and concentrating soil management efforts to a small, defined space. Despite anecdotal claims to the benefits of RBs avoiding poor soil texture (Langellotto-Rhodaback, 2014), we found little evidence to suggest RBs are superior to IGs in this regard. However, compared to IGs, RBs were significantly elevated in active carbon, C/N, and sulfur, and RBs showed an interesting, nonsignificant elevation in organic matter compared to IGs.

Raised beds are recommended as a way for gardeners to grow food in areas where metals in the soil are a concern (Brewer et al., 2016). However, the three highest lead levels were found within RBs. Soil samples from other sites near these lead concentrations (located 0.6–1.3 km away) did not exhibit elevated levels of lead, which suggests that environmental deposition is not the cause of the elevated lead levels. We do not know the source of the metals in these three soil samples, but it is entirely possible it is from lead paint from the house itself. There is some research showing how legacy contaminants from pesticides increase heavy metal content in garden soils (Ninkov et al., 2018). The highest arsenic level in our study came from a site that was located on the border of what had been an apple orchard.

However, given the wide variety of compost products on the market, it is possible that the growing media used to fill the raised beds contained heavy metals. We advise growers to be cautious with their compost purchases. If the compost is not labeled with nutritional claims, it is exempt from analysis and contamination limits (Association of American Plant Food Control Officials, 2018). If the compost does contain nutrient information, it must be registered and appropriately labeled. However, the disclosure of heavy metal content is often online and not on the bagged product (Oregon Department of Agriculture, 2018). Even small amounts of heavy metals can accumulate across numerous additions to a site, which might result in levels above safe thresholds for those enriched soils.

Compost and organic matter

Conventional agriculture and agronomic recommendations operate from an assumption of deficit. Fenton et al. (2008) recommend farmers maintain their soils 3–6% organic matter by mass, and some farms struggle to meet this minimum. Samples in this study averaged double this highest typical recommendation at 13% organic matter by mass, including four samples at >20%. Relatively little is known about the consequences of excessive organic matter in agricultural soils. This may be because such excesses are not common in commercial agriculture where loss of soil organic carbon is often a concern (see Lal, 2007; Magdoff & Weil, 2004). Researchers have discovered accelerated nitrogen mineralization in organic soils (Broadbent, 1986), but these are Histosols—bog-type soils that generally experience extended anaerobic conditions. However, urban gardeners may apply compost in excess in order to meet fertility needs (Alvarex-Campos & Evanylo, 2019; Lorenz, 2015), resulting in elevated phosphorus and potassium in their soils and subsequent loss to local watersheds (Small et al., 2019; Goss et al., 1991).

Growers with high levels of organic matter should ensure the lab they use for soil sample analysis is prepared for these samples. Many processes and analyses were complicated by extended timelines or modifications from standard protocols. For example, typical soil screening might take 20 min (A. Villaseñor, personal communication, 2018); samples in this study averaged 2 h to process and screen soil particles from >2 mm organic particulate. Additionally, preparing samples for textural analysis typically requires 1–3 d of oxidation treatment with hydrogen peroxide (A. Villaseñor, personal communication, 2018); most of the samples in this study showed evidence of active oxidation even after five wk of treatment.

Compost brings variable supplies of nutrients. While some research suggests organic matter can provide an adequate supply of micronutrients (Chen & Stevenson, 1986), we found a notable lack of boron in our samples. In a study characterized by excess, boron stands unique in that only one site exceeded recommendations, while more than half fell below the minimum concentration. This could be due to the variability both of boron content and soil-boron testing for western Oregon soils (Parker & Gardner, 1982).

The overreliance on organic matter to fertilize home garden soils, combined with the ease at which soil nutrients can be applied in small-plot gardens, likely contributed to the documented excess of several garden nutrients. Additionally, the lack of upper limits for many soil parameters may lead growers to take a ‘more is better’ approach (Moebius-Clune et al., 2016). We propose this to be one of the driving forces behind overenrichment of urban agricultural soils. We suggest an alternative approach to managing urban soils for producing crops—rather than expecting urban soils to be deficient, test them and respond with targeted applications and save money and environmental damage by reducing fertilizer applications (Figure 5).

FIGURE 5. A flowchart to guide urban agriculture growers and gardeners to establish and maintain productive soils. The first step is securing access to land then take a soil sample from this site and send it to a lab for analysis. Use the soil analysis report to decide whether the site's soil is ready to grow crops or still needs changes to its foundational parameters like texture or pH. Otherwise, managers need only apply deficient nutrients as fertilizers, and they can maintain a productive soil without generating fertility excesses and potentially damaging their surrounding ecosystems

We found no difference in nitrate-N between bed types and suggest three possible explanations, while the first is highly unlikely given the broad diversity and differences in crops we documented across the gardens (Appendix 1): (a) perhaps every garden produced approximately the same harvest from its soil and so pulled a relatively even, average amount of nitrate from the soil; (b) maybe the differences in organic matter are quickly neutralized with the region's tremendous rainfall of 101–203 cm (40–80 inches) per year (Schaefer et al., 2008); or (c) irrigation might provide a similar neutralizing effect as precipitation.

Distinct needs of urban growers

The field of urban soils is relatively young (Kaye et al., 2006), which often leaves urban growers advised and trained to manage their soils with the same research that informs commercial agriculture rather than recommendations focused on their unique circumstances. For example, cover crop calculators for farms and for gardens differ only the in the units: acres for farms vs. square feet for gardens (OSU Small Farms Program, 2018). Soil sampling instructions assume that gardeners will sample over a large area (Fery & Murphy, 2013) rather than in the discrete, small beds that were typical for gardeners in this study. Home gardeners might not understand that commercial farms rely on cover crops and rotation in addition to compost importation to maintain soil organic matter (Magdoff & Weil, 2004; Hodges, 1991).

Urban food production has the potential to positively contribute to the sustainability and resilience of local food systems and to transform urban spaces (Saldivar-Tanaka, 2004; Moore et al., 2008). In fact, a case study of Cleveland, OH, suggests that most postindustrial cities’ food needs could be grown within ∼160 km (100 miles) of urban areas in North America (Grewal & Grewal, 2012).

Limitations

This study was designed to avoid data that comes from involving human subjects. We thus focused our study on the gardens and did not ask questions of the gardeners so that we would not have to go through the Institutional Review Board process. As such, a limitation of this study was that we did not collect information like garden age, source of compost (or other fertility import), or irrigation source. Although this information would have provided useful context for our results, we do not have this data.

CONCLUSION

We conducted an observational study of soil health in residential vegetable gardens in Corvallis and Portland, OR. The relatively small sample size (27 gardens and 33 discrete bed types from two cities) of this observational study limits our ability to make broad generalizations. We found that gardens are generally overenriched in several soil parameters including organic matter, phosphorus, calcium, magnesium, and potassium. These results are particularly notable, as sampled soils came from the gardens of certified Oregon State University Extension Master Gardeners, all of whom have received training in soil health and fertility.

Several interesting patterns emerged from the data, which could be the focus of future research. For example, future studies could (a) develop an alternative set of laboratory tests for soils with elevated levels of organic matter, (b) examine the leaching potential of urban garden soils, (c) develop data-driven models for sustainable management of urban garden fertility, and (d) investigate how to modify regional Master Gardener education materials to better enable gardeners to sustainably manage their soils. Nearly any investigation will advance our knowledge in the field and will help gardeners and other urban growers make more informed decisions for their land.

This research highlights a burgeoning need to know if there is an optimal media to fill raised beds, and if so, what are the characteristics or composition of this media? Raised beds provide excellent ways to overcome many traditional limitations to plant growth but their construction often leaves management struggling for years with collapsing or diminishing beds. The common solution of constant compost application can itself become an issue of overenrichment. What is the best option for urban growers facing limited land access?

ACKNOWLEDGMENTS 

We thank the gardeners and urban farmers who allowed us to visit their sites and sample their soils. We thank Tammy Winfield for creating Figure 2. This work was supported in part by a donation from Y. Sherry Sheng and Spike Wadsworth. 

AUTHOR CONTRIBUTIONS

Mykl Nelson: Conceptualization; Data curation; Investigation; Methodology; Project administration; Visualization; Writing – original draft; Writing – review & editing. Gwynne Mhuireach: Data curation; Formal analysis; Writing – review & editing. Gail A. Langellotto: Conceptualization; Funding acquisition; Methodology; Project administration; Supervision.

CONFLICT OF INTEREST

The authors declare no conflicts of interest.

1 Appendix

Full list of crops recorded from all 26 garden study sites

Common name	Latin binomial	No. of sites
Annual crops
Tomato	Solanum lycopersicum L.	25
Basil	Ocimum basilicum L.	22
Bean	Fabaceae spp.	20
Kale	Brassica oleracea L.	20
Squash, summer	Cucurbita pepo L.	19
Pepper, hot	Capsicum spp.	18
Lettuce	Lactuca sativa L.	17
Eggplant	Solanum melongena L.	15
Cucumber	Cucumis sativus L.	14
Pepper, sweet	Capsicum annuum L.	14
Onion	Allium cepa L.	13
Potato	Solanum tuberosum L.	13
Squash, winter	Cucurbita spp.	13
Carrot	Daucus carota L. subsp. sativus (Hoffm.) Schübl. & G. Martens	12
Chard	Beta vulgaris L. subsp. vulgaris	12
Pea	Pisum sativum L.	12
Cabbage	Brassica oleracea L.	11
Garlic	Allium sativum L.	11
Parsley	Petroselunum crispum (Mill.) Fuss	11
Sunflower	Helianthus annuus L.	11
Beet	Beta vulgaris L. subsp. vulgaris	10
Broccoli	Brassica oleracea L. var. italica Plenck	9
Nasturtium	Tropaeolum majus L.	8
Leek	Allium porrum L.	7
Purple-coneflower	Echinacea purpurea (L.) Moench	6
Arugula	Eruca vesicaria (L.) Cav. subsp. sativa (Mill.) Thell.	5
Borage	Borago officinalis L.	5
Brussels sprouts	Brassica oleracea var. gemmifera DC.	5
Cilantro	Coriandrum sativum L.	5
Fennel	Foeniculum vulgare Mill.	5
Parsnip	Pastinaca sativa L. subsp. sativa	5
Cauliflower	Brassica oleracea L. var. botrytis L.	4
Shallot	Allium cepa L.	4
Bok choy	Brassica rapa L. subsp. chinensis (L.) Hanelt	3
Celery	Apium graveolens L.	3
Corn	Zea mays L.	3
Dill	Anethum graveolens L.	3
Ground cherry	Physalis longifolia Nutt.	3
Horseradish	Armoracia rusticana G. Gaertn. et al.	3
Kohlrabi	Brassica oleracea L. var. gongylodes L.	3
Radish	Raphanus sativus L. var. sativus	3
Shiso	Perilla frutescens (L.) Britton var. crispa (Thunb.) W. Deane	3
Spinach	Spinacia oleracea L.	3
Tomatillo	Physalis philadelphica Lam.	3
Cabbage, napa	Brassica rapa L. subsp. pekinensis (Lour.) Hanelt	2
Chicory	Cichorium intybus L.	2
Collard	Brassica oleracea L. var. viridis L.	2
Mustard	Brassica juncea (L.) Czern. (various)	2
Oca	Oxalis tuberosa (Molina)	2
Okra	Abelmoschus esculentus (L.) Moench	2
Pumpkin	Cucurbita spp.	2
Purslane	Portulaca oleracea L.	2
Sunchoke	Helianthus tuberosus L.	2
Sweet potato	Ipomoea batatas (L.) Lam. var. batatas	2
Turnip	Brassica rapa L. subsp. rapa	2
Yacón	Smallanthus sonchifolius (Poepp.) H. Rob.	2
Amaranth	Amaranthus spp.	1
Barley	Hordeum vulgare L.	1
Burnet	Sanguisorba minor (Scop.)	1
Celeriac	Apium graveolens L. var. rapaceum (Mill.) DC.	1
Chinese kale	Brassica oleracea L. var. alboglabra (L. H. Bailey) Musil	1
Kalettes	Kale × brussel sprout hybrid	1
Melon	Cucurbitaceae spp.	1
Papalo (yerba porosa)	Porophyllum ruderale (Jacq.) Cass.	1
Perilla	Perilla frutescens (L.) Britton	1
Romanesco	Brassica oleracea var. botrytis ‘Romanesco’	1
Scallion	Allium fistulosum L.	1
Sorrel	Rumex acetosa L.	1
Stevia	Stevia rebaudiana (Bertoni) Bertoni	1
Milk thistle	Silybum marianum (L.) Gaertn.	1
Vietnamese coriander	Persicaria odorata (Lour.) Soják	1
Watercress	Nasturtium officinale W. T. Aiton	1
Perennial crops
Raspberry	Rubus idaeus L.	16
Strawberry	Fragaria ×ananassa Duchesne ex Rozier	16
Blueberry	Vaccinium angustifolium Aiton	15
Apple	Malus domestica (Suckow) Borkh.	14
Fig	Ficus carica L.	14
Blackberry	Rubus ursinus Cham. & Schltdl.	13
Rhubarb	Rheum ×rhabarbarum L.	13
Sage	Salvia officinalis L.	12
Asparagus	Asparagus officinalis L.	11
Artichoke	Cynara cardunculus L.	10
Chive	Allium schoenoprasum L.	10
Lavender	Lavandula angustifolia Mill. subsp. angustifolia	9
Rosemary	Rosmarinus officinalis L.	9
Grapevine	Vitis vinifera L.	8
Pear	Pyrus communis L.	8
Currant	Ribes spp.	7
Oregano	Origanum vulgare L. subsp. vulgare	7
Thyme	Thymus vulgaris L.	7
Mint	Mentha spicata L.	6
Plum	Prunus domestica L. subsp. domestica	6
Hop	Humulus lupulus L.	5
Persimmon	Diospyros spp.	5
Cherry	Prunus avium (L.) L.	4
Peach	Prunus persica (L.) Batsch	4
Marjoram	Origanum majorana L.	3
Tarragon	Artemisia dracunculus L.	3
Aronia	Aronia spp.	2
Gooseberry	Ribes hirtellum Michx.	2
Hardy kiwi	Actinidia arguta (Siebold & Zucc.) Planch. ex Miq.	2
Lemon	Citrus ×limon (L.) Osbeck	2
Lemon verbena	Aloysia citrodora Paláu	2
Tayberry	Rubus ×loganobaccus L. H. Bailey	2
Asian pear	Pyrus pyrifolia (Burm. f.) Nakai var. culta (Makino) Nakai	1
Chamomile	Chamaemelum nobile (L.) All.	1
Choke cherry	Prunus virginiana L.	1
Elderberry	Sambucus spp.	1
Fiddlehead	Unknown	1
Pineapple guava	Acca sellowiana (O. Berg) Burret	1
Goji	Lycium barbarum L.	1
Gumi	Elaeagnus multiflora Thunb.	1
Hazelnut	Corylus americana Marshall	1
Honeyberry	Melicoccus bijugatus Jacq.	1
Jostaberry	Ribes ×nidigrolaria Rud. Bauer & A. Bauer	1
Lemongrass	Cymbopogon citratus (DC.) Stapf	1
Lime	Citrus ×aurantiifolia (Christm.) Swingle	1
Mulberry	Morus spp.	1
Papaya	Carica papaya L.	1
Passion fruit	Passiflora edulis Sims	1
Pawpaw	Asimina triloba (L.) Dunal	1
Quince	Cyodonia oblonga Mill.	1
Savory	Satureja spp.	1
Seaberry (sallowthorn)	Hippophae rhamnoides L.	1
Serviceberry	Amelanchier canadensis (L.) Medik.	1
Snowberry	Symphoricarpos albus (L.) S. F. Blake	1
Thimbleberry	Rubus nutkanus Moc. ex Ser.	1
Walnut	Juglans spp.	1