Seasonal nitrogen and phosphorus leaching in urban agriculture: Dominance of non-growing season losses in a Southern Swedish case study

allot-ment


Introduction
In response to the negative externalities of current food supply systems, and questions about the sustainability of cities, urban agriculture (UA) has been a growing research topic over the last decade, especially because of its multifunctionality (Artmann and Sartison, 2018).The production of food within city boundaries has been praised for its contribution to social cohesion (Dobernig et al., 2016;Veen et al., 2016), food security (Eigenbrod and Gruda, 2015;Opitz et al., 2016), and biodiversity (Clucas et al., 2018;Orsini et al., 2014).UA not only provides healthy food and more green areas, but it also has a positive influence on physical and mental health and can enhance social capital among practitioners (Ackerman et al., 2014;Audate et al., 2021;Kabisch et al., 2017).
UA has also been identified as an opportunity for closing nutrient cycles within city's boundaries, although recycling may not always be efficient (Rufí-Salís et al., 2020).As hotspots of food consumption, cities import large amounts of nutrients such as phosphorus (P) and nitrogen (N) present in food (Lin et al., 2014).A large part of these nutrients end up in waste (food waste and human excreta) with limited recovery and reuse, in part because it can be logistically complex to send them back to rural areas.UA can play a role in recycling nutrients from urban organic waste through the use of compost and other locally produced recyclable fertilizer and soil amendment products (Metson and Bennett, 2015).There are technological options to concentrate nutrients in some waste streams (e.g., making struvite from human urine or wastewater to recover P (Wei et al., 2018)) but without further processing organic waste often remains heavy and bulky, making long distance transport unattractive.Its use in UA, when possible, could thus be a preferable choice for an efficient recycling of nutrients in organic waste.However, there are many types of UA and Goldstein et al. (2016) separated them into building-integrated-conditioned, building-integrated-non-conditioned, ground-based-conditioned and ground-based-non-conditioned.Even within one type UA practitioners have diverse gardening practices, including nutrient management.Previous work has shown that over-application of compost and other nutrient inputs in ground-based (in soil as opposed to in raised beds, on roofs or in buildings) UA leads to low nutrient use efficiencies, measured in experimental studies and as garden-gate balances, resulting in risk for eutrophication of water bodies (Hedlund et al., 2003;Khai et al., 2007;Small et al., 2019;Wielemaker et al., 2019).Even though the soil can function as a nutrient buffer, excessive amounts of N and P not taken up by crops can lead to leaching and surface runoff losses into surrounding waters.As UA expands worldwide, inefficient nutrient management might lead to eutrophication of urban waters, creating a serious environmental concern.
Only a few studies have measured nutrient losses to waters from UA and have done so in limited settings compared to the importance of accounting for drivers of loss already identified in rural settings (van de Vlasakker et al., 2022).Both in-situ and controlled experiments have demonstrated a high variability of nutrient leaching levels among garden plots; the type and rate of fertilizer application, soil type, and irrigation practices all influenced leaching (Abdalla, 2012;Kleinman et al., 2007;Safi et al., 2011;Small et al., 2018).These studies measured the losses occurring during the growing season, but it is well known from studies of rural agriculture that both N and P losses are highly seasonal, particularly in cold climates (Djodjic et al., 2000).In cold climate regions, the major losses usually occur outside the growing season (Bergström and Brink, 1986;Ulén et al., 2019), particularly if there are no plants (Delgado et al., 2007).There is a high risk that nutrients added in the fall as livestock manure and plant residues are lost with percolating snow melt and rain water, or with surface runoff, generated during the non-growing season (Fraters et al., 2007;Geel and Smit, 2006;Ulén et al., 2019).Better management to reduce such losses requires an improved understanding of interactions between agronomic practices, biogeochemical processes affecting nutrient turnover, and hydrological characteristics resulting in percolation and surface runoff generation (Liu et al., 2019).
Accounting for year-round losses associated with diverse real-world UA practices is particularly important in seasonal temperate environments.Although results from field trials and environmental monitoring from rural settings already indicate there is seasonal variation in losses, we cannot simply transpose results from one setting to another (Plach et al., 2019).UA takes place in different settings than rural agriculture and involves a very diverse set of practices regarding both soil preparation, crop choices, and fertilization.UA can also differ from rural agriculture in terms of its surrounding land use (affecting hydrological characteristics), knowledge of practitioners about nutrient management, and the goals of the production system (Deelstra and Girardet, 2000).Considering the high variability of practices (Goldstein et al., 2016;Small et al., 2019), and ongoing expansion of UA, collecting year-round data on nutrient management and losses in real-world settings, i.e. from non-experimental gardens, is key for promoting potential UA benefits while minimizing environmental risks.Such data would help to understand how important the non-growing season is compared to the growing season for nutrient losses and, consequentially, what practices can contribute to minimize them.
In this study, we investigated the importance of non-growing season vs. growing season for N and P leaching from UA gardens situated in the temperate climate of southern Sweden.We collected leachate from eight lysimeters installed in four different UA gardens between May 2020 and May 2021 to assess rates of nutrient leaching including total N (TN), total P (TP), nitrate-N (NO 3 --N) and phosphate-P (PO 4 3--P).We report on the variability in nutrient leaching over time and among nutrients in the context of real-world diverse management practices.

Study area
The study took place in Linköping, Sweden, where there are several community/allotment and collective UA areas.For this paper we focus on a recently added park located southwest of Linköping's city center (58 • 39' N; 15 • 59' E).It was established in 2017 as part of a housing project and includes an urban gardening area with 79 individual gardens, each approximately 24 m 2 , which together form an allotment area.The park and urban gardening area are owned by the municipality.An association leased the area from the municipality and rents out the individual gardens to its members.Half of the members live in the surrounding neighborhood (Paradisodlarna, 2022).
Linköping has a humid continental climate according to the Köppen classification (Peel et al., 2007) with an average maximum temperature of 23 • C during summer and 0 • C during winter.The region is characterized by a relatively long non-growing season; in this paper we have defined this as the period from November 11th to May 11th based on the approximate start and end of the vegetation period (April/May--Oct/Nov) according to SMHI (2022) and the first date in our sampling campaign.Average annual precipitation in this part of Sweden usually reaches above 664 mm with the highest values noted in summer (SMHI, 2021).All meteorological data used in this study were obtained from a nearby Swedish national weather monitoring station (58 • 39' N; 15 • 52' E; altitude 94 m), located 4 km west of the study area (SMHI, 2021).

Individual gardeners and characteristics of plots
We recruited four gardeners by posting a message in their neighborhood Facebook group for urban gardening, by visiting the area in April 2020 and via referrals (snowball sampling), as part of a larger campaign to recruit UA practitioners all over the city to study nutrient cycling in UA.The aim was to get a wide variety of gardeners, noting that willingness to participate contributed to who was recruited.It was also necessary to sample all gardens within 1 day with the research team's resources, and as such there was an upper bound to the number of gardeners that could have participated.This study focuses on four gardens in the allotment area described in Section 2.1, where we collected water in the lysimeters year-round; the larger study with more gardens only examines growing-season leaching, but otherwise follows the same protocols as described below.
Each gardener had divided their garden into several cultivation plots for growing different crops.In each garden, we selected two plots for installation of two lysimeters.We informed the participants about the purpose of our study but did not interfere or give advice on how best to manage their plots.We were able to keep track of selected management practices through a standardized interview adapted from Metson and Bennett (2015) and Small et al. (2019) which we conducted in-person in the garden at the end of the growing season.In particular we collected information about N and P inputs, the garden cultivated area, and qualitative information on irrigation practices and crops harvested (Table 1, and Section 2.2.1).Interview results will be further explored in upcoming publications using a larger sample size; here we focus on the measured nutrient leaching.In addition to the interviews, we took weekly photos of each cultivation plot, showing that some of them were covered with nets or mulch for part of the year, and had informal conversations with the gardeners throughout the year.We also collected soil samples at the end of the growing season (see Section 2.2.2).

Fertilizer inputs
We asked each gardener which inputs they used in their garden and helped them determine how much of each was used by systematically looking at a list of possible inputs and physically examining the size of bags, buckets, and compost bins in the garden to minimize recall error (as well as purchase receipts when available, see Metson and Bennett, 2015).We then calculated N and P input for a given year based on the type and amount of each additive, and their respective nutrient content and bulk density.For garden products that were bought from shops, we used information regarding the nutrient content and density provided on the package of the product.For inputs that were from the garden or a non-standardized source, we used literature values (Table S1).The sum of nutrient inputs (kg ha − 1 y − 1 ) was calculated as: where i is the cultivated garden area in ha; Fert. is inorganic (mineral) fertilizer (kg yr − 1 ); Org. is organic fertilizer (e.g., fresh or composted manure, purchased plant or animal derived fertilizers such as bone meal, home-or community-made plant and food waste composts; kg yr − 1 ); Amend.stands for soil amendments containing nutrients (e.g., potting mix; kg yr − 1 ).We did not include other inputs such as irrigation water or seeds, as this was not within the scope of this study.We divided the total amount of nutrients by the cultivated garden area (m 2 ).The garden area was measured by our research team at the time of the interview as the total surface of cultivated garden area that year (i.e. the sum of all plots where inputs could have been spread, excluding any sheds or grassy area).This means we cannot account for differences in nutrient management practices among the plots within one garden and these annual input values remain a best estimate.Garden level fertilization practices did vary, but all used organic inputs exclusively (Table 1).In Garden 1, the main inputs were cow manure, chicken manure, plant-based compost, grass clippings, and bone meal.The gardener from Garden 2 only added a commercially available composted manure which contained around 10 % of cow manure, 10 % of chicken manure, and 80 % of dark peat.In Garden 3, mixed plant derived additives were the biggest inputs, although the gardener also used both chicken and horse manure.The majority of N and P came from grass clippings and a home-made liquid fertilizer (Bokashi).The gardener from Garden 4 only added plant derived fertilizers to the plot, including plant-based compost and grass clippings, along with some new garden soil.

Soil characteristics
In October 2020, 10-15 cores were taken at each plot (randomly) in a 0.25 m 2 circle around the tube that allowed us to sample water in the lysimeters down to 0.2 m using a stainless-steel gouge auger (Swedish Board of Agriculture, 2010).In total 1 kg of fresh soil was collected from each plot, and sent to the commercial laboratory Agrilab AB, Uppsala, where the samples were mixed and dried (Termaks, 40 • C) for 72 h.Then the samples were prepared for further analysis following grinding and sieving with a 2 mm sieve according to standard methods (ISO, 2006).Texture and volatile solids content (loss on ignition) was determined according to the Swedish standard method (SIS, 2020).Plant available (extracted with ammonium lactate, AL) P (P-AL) and cation content (Ca-AL, K-AL, Mg-AL, Al-AL and Fe-AL) were determined also using the ammonium lactate method according to the Swedish standard method (SIS, 1993).Generally, soil samples were extracted with 0.01 molar ammonium lactate and 0.4 molar acetic acid followed by detection of metals with flameless atomic absorption spectrometry (SIS, 1986) and P with inductively coupled plasma atomic emission spectroscopy (ICP-AES) according to Boumans (1979).pH was measured in an extract of 1:5 (volume based) soil:deionized water.Carbon and N content of the soils were determined through dry combustion according to standard methods (ISO, 1995(ISO, , 1998, respectively) , respectively) using an elemental analyzer for macrosamples (TruMac ® CN, Leco corp, S:t Joseph, MI, USA) at the Swedish University of Agricultural Sciences laboratory.Soil characteristics for all the plots were very similar (Table 2), probably because the studied urban agriculture area is relatively new and has been used for less than three growing seasons.

Field set-up
Water leachate was collected from eight simplified lysimeters according to the method developed in Small et al. (2018).Briefly, each lysimeter consists of a 1 L wide-mouth HDPE bottle attached to a polypropylene funnel (both purchased from VWR Collection, ⌀101 mm or ⌀147 mm) with stem (⌀ 30 mm, also from VWR Collection) closed with inert rockwool to prevent soil intrusion into the leachate collection bottle.Lysimeters were equipped with Tygon tubing (Saint-Gobain brand, inner ⌀3 mm, 0.9 m long), which reached from bottom of the bottle to the soil surface and closed with a 3-way valve (Becton Dickinson Connecta).Each plot contained one lysimeter, which was buried in the center (top of the funnel at 0.3 m below the soil surface) to minimize interference with crop roots.

Field sampling
Water from all lysimeters was collected weekly for 52 weeks (May 2020-May 2021).On each occasion we collected the water with a 60 mL syringe via the 3-way valve on the tubing and recorded the total volume of leachate.If the volume was sufficient we collected four 25 mL samples in four separate 50 mL HDPE bottles for TN, TP, PO 4 3--P and NO 3 --N analyses.Samples for PO 4 3--P and NO 3 --N were pre-filtered in situ (polyethersulfone syringe filters, pore size of 0.45 µm), whereas leachate samples for TN and TP analyses were kept unfiltered.If there was not enough water, unfiltered samples were prioritized and if the volume was between 2 and 5 mL, only one sample for analysis of TP and TN was taken.We also prioritized PO 4 3--P before NO 3 --N if too little water was collected in any given week.Collection bottles for P species were acid washed (2 N HCL) prior to sampling.If a sufficient volume of water could be collected (> 150 mL), temperature, conductivity, and pH of the leachate were also measured in the field on non-acidified samples (Hanna Instruments, HI-991301).

Table 1
Management practices among gardeners in the UA study area in Linköping, Sweden in 2020.Plant-derived inputs refer both to composted (or fermented) plant material and the addition of non-composted clippings (e.g., grass and hay).For all gardens, the described water management reflects the frequency of watering during (warm) summer days based on interview data (not measured).In the winter season, there was no irrigation.

Leachate analysis
Samples were transported from the field to the laboratory in a dark cooler and stored in a refrigerator or freezer until further analysis.TN samples were kept at 4 • C pending analysis by TOC/TN analyzer (Schimadzu Kyoto, Japan, detection limit 4 μg L − 1 ), which was performed the following day.Samples containing more than 5 mg TN L − 1 , were diluted with deionized water prior to analysis.A standard (5 mg TN L − 1 ) was used to correct results.Samples for TP, PO 4 3--P and NO 3 --N were stored frozen (− 18 • C) until further analysis.NO 3 --N concentrations were analyzed with ion chromatography (30 Compact IC Flex, Metrohm with the column Metrosep A supp 5 150/4.0,calibrated for 0.1 mg L -1 ).The leachate concentrations of TP and PO 4 3--P were detected spectrophotometrically (Nanocolor UV/VIS, Macherey-Nagel, Germany), after digestion with peroxodisulphate for TP, with the ascorbic acid-molybdate method following the Swedish standard procedures for TP (method SS 028127, detection limit 2 μg L − 1 ) and PO 4 3--P (method SS 028126, detection limit 2 μg L − 1 ).If the absorbance value of a measured sample was higher than the absorbance of the highest standard concentration, the original sample was diluted and the whole analytical procedure repeated.
To calculate leaching for all nutrient species the following equation was applied: where L loss (kg ha − 1 week − 1 ) is leaching loss of a given nutrient (TP, TN, PO 4 3--P, NO 3 --N), C (mg L − 1 ) is the concentration of a given nutrient, V is the volume of the collected leachate (L), dT (days) is the time difference between consecutive sampling occasions when leachate was collected (1 week), and A (m 2 ) is the funnel area (0.16 or 0.23) of each lysimeter.
Cumulative nutrient leaching (kg ha − 1 y − 1 ) was calculated for growing and non-growing season separately as a sum of the total nutrient loss for each season.Cumulative annual leaching was calculated as a sum of the growing and non-growing season values for each plot.Maximum potential water input with precipitation was calculated as a weekly cumulative volume of precipitation (based on data from the weather station, SMHI, 2021), which fell on the soil surface above the funnel area of the lysimeter during the week prior to leachate collection.To calculate Leachate/Precipitation ratio for each week the following equation was applied: where LP ratio is Leachate/Precipitation ratio, V is the volume of the collected leachate (L) and Vp is total volume of precipitation (L), which fell on the funnel area of the lysimeter during the week prior to leachate collection.A ratio above 1 during the growing season indicates a substantial contribution from irrigation water whereas during the nongrowing season this suggests contribution from snowmelt, soil oversaturation, or discrepancies between weather station data and actual precipitation in the respective gardens.
To test the hypothesis that one season (growing or non-growing) dominates leaching, we ran a Wilcoxon signed-rank test, which allowed us to account for pairing (i.e., each lysimeter has a value for growing and non-growing season).

Precipitation and collected leachate volumes
The sampling year was characterized by lower precipitation than average (664 mm) annual values recorded for Linköping area (SMHI, 2021).During our sampling period, the total precipitation (rain and snow) was 530 mm with 329 mm in the growing and 201 mm in the non-growing season.Despite 39 % lower precipitation during the non-growing season, the mean of the leachate volumes collected in the non-growing season was significantly higher than that in the growing season (p < 0.05; Fig. S1).In most plots, the lysimeters collected from 53 % to 95 % of the annual leachate volume during the non-growing season (Fig. 1 bottom panel).On multiple occasions throughout the year, the collected weekly leachate volume exceeded the maximum potential water input with precipitation (Fig. S2).
Overall, the highest leachate volumes were noted during the nongrowing season, especially during and after the snow thawing period (Fig. 2).However, in some lysimeters (plots 1, 2 and 5) the volumes were already high in the beginning of the non-growing season.The largest annual leachate volumes (12.6, 16.1, 12.5 L) were collected from plots that were uncovered for most of the time (plots 1, 2, 8 as per Table 1, Fig. 1, and Fig. 2).Plot 3, which was covered with an anti-insect net throughout the whole year (Table 1), was a clear outlier, with very little water in total (2.8 L) collected and most of it (74 %) during the summer when the plants were irrigated (Fig. 1).

TN and TP leachate loss
Like with water volumes, significantly more TN leachate was, on average, collected in the non-growing season (p < 0.05, Fig. S1), but plot 3 was an exception from this pattern (Fig. 1).For TP, the mean loss was also higher during the non-growing season, though not statistically significant (Fig S1, p = 0.08).The proportion lost during the nongrowing season was related to the proportion of leachate water collected during that season, whereas no such relation could be discerned for TN (Fig. 1).Both the concentrations and the quantitative leaching of TN and TP were characterized by high variability among gardens and plots (Fig. 3 Annual cumulative TN leachate ranged from 39 kg ha − 1 y − 1 to 191 kg ha − 1 y − 1 across all plots and was the highest in plot 5 and the lowest in plot 3 (Fig. 1).Leaching rates for TN were 4-15 times higher during the non-growing season than the growing season in all the plots with an exception for the covered plot (plot 3), where leaching losses were 1.5 times lower (Fig. 1).Initially, there was an increase in TN leaching rates in the beginning of the growing season (middle of May through June) for two plots covered with net (Fig. 3).Throughout the growing season leaching rates were low and an increase occurred only in December, which continued during the non-growing season until snow fall occurred.After the snow thawing period, there was a considerable raise in TN leachate loss.
Cumulative yearly TP leachate ranged from 0.94 to 2.43 kg ha − 1 y − 1 across plots and the highest quantities were leached from plots without cover (1, 2, 4, 8).Similarly to for TN, the non-growing season was characterized by 3-5 times greater TP leaching compared to the growing season for most of the plots, with 20-30 times higher for plots 6 and 4 (Fig. 1).In contrast, the leaching in the covered plot (3) were higher during the growing season than in the non-growing season, and in plot 8 almost half of the leaching occurred in the growing season.These TP loss proportions corresponded well to the proportion of water collected in the respective seasons for these two plots.
Throughout the year, TP leaching exhibited a similar temporal pattern as for TN, except that no early loss of TP was observed in May/ June (except for plots 1 and 2, Fig. 4).The leachate losses were low during the growing season and higher in the non-growing season with a considerable increase during the snow melt period for most plots.However, plots 3 and 8 deviated from that pattern; in plot 3 large TP leaching was observed in June and in plot 8 almost half the annual TP leaching loss occurred from June to August (Fig. 4).

Fig. 2. Cumulative volume of leachate collected in lysim-
eters in allotment garden plots in Linköping (colored lines with plot numbers at the end of each line, and a unique symbol for each plot, see Table 1 for plot descriptions) and the cumulative precipitation over the lysimeter area (grey line and circles).Colored backgrounds indicate growing (green to the left) and non-growing (grey to the right) seasons, and stars and white dotted lines indicate weeks with snowfall.Plot 1 is marked with squares, plot 2 with diamonds, plot 3 with upside-down triangles, plot 4 with right-side-up triangles, plot 5 with x, plot 6 with +, plot 7 with dots and x, and plot 8 with circles.The colors of plots are unique but grouped by garden.For example, plot 1 and 2 are in the same garden so they are both a shade of blue.
Although the lowest leaching losses were observed in plot 3 that received the lowest input, there was no clear positive relation between N or P inputs, assessed from interview information (Table 1), and measured nutrient leaching in the other plots.In fact, the highest total TN leaching was observed in plots 5 and 6, where the estimated N input was low compared to other gardens.

Leaching of NO 3 --N and PO 4 3--P
Dissolved ion forms (NO 3 --N and PO 4 3--P) constituted the main proportion of TN and TP, respectively (Table S2).NO 3 --N accounted for 73-92 % of TN on average and PO 4 3--P comprised 62-85 % of TP.However, occasionally, these dissolved forms represented a very small fraction of TN and TP (1-2 %).The average NO 3 --N leachate loss ranged from the lowest 4 kg ha − 1 week − 1 in plot 3 to 16 kg ha − 1 week − 1 in plot 6.In contrast, the highest mean leaching of PO 4 3--P was noted in plot 3 and the lowest in plots 6 and 8, with values of 0.17, 0.05, and 0.05 kg ha − 1 week − 1 , respectively.Thus, plot 3 was an outlier in terms of the share of dissolved forms of N and P. Because the volume of leachate collected was occasionally low, NO 3 --N and PO 4 3--P could not always be analyzed in addition to TN and TP.Still, there was no difference in the mean proportion of dissolved N or P in water from plots with few samples than those with more samples analyzed.

High and variable total nutrient leaching
In this study, leachate TN losses below the 0.3 m root zone in smallscale UA plots were exceptionally high (39-191 kg ha − 1 ) when compared to average N leaching from arable land to drainage pipes, which was estimated at an average 22 kg N ha − 1 y − 1 for Sweden (Johnsson and Hoffmann, 1998).However, TN leachate losses in rural agriculture are known to be variable, where N fertilization levels and precipitation are two of the most important (among several) factors that drive such variability.Bergström and Brink (1986) measured N leaching Fig. 3. Cumulative TN leachate collected in lysimeters in allotment garden plots in Linköping and cumulative precipitation (line and grey circles) over the lysimeter area.The top panel shows results for plots 1-4, and the bottom panel plots 5-8 for visual clarity.Colored backgrounds indicate growing (green to the left) and non-growing (grey to the right) seasons, and stars and white dotted lines indicate weeks with snowfall.Plot 1 is marked with squares, plot 2 with diamonds, plot 3 with upside-down triangles, plot 4 with right-side-up triangles, plot 5 with x, plot 6 with +, plot 7 with dots and x, and plot 8 with circles.The colors of plots are unique but grouped by garden.For example, plot 1 and 2 are in the same garden so they are both a shade of blue.
to tile drains in experimental fields with cereal crops and found that at high N fertilization levels (200 kg ha − 1 ) the leaching was as high as 90 kg ha − 1 in a wet year, but around 35 kg ha − 1 in a year with average precipitation.Similarly, Stålnacke et al. (2014) analyzed TN losses over time from 35 small agricultural catchments around the Baltic Sea, where the mean annual loss varied from 7 to 102 kg ha − 1 , and found that ~54 % of the variability among catchments could be explained by differences in water discharge.
Our results demonstrate that water volumes and nutrient (N and P) leachate losses from individually managed small-scale UA varied both among gardens and between plots managed by the same gardener (Figs.2-4).In contrast to the studies cited above, the observed differences among plots in TN leachate were not clearly related to differences in leachate water volume (Fig. 1).A multitude of factors, including fertilizer management, soil and crop cover, irrigation (discussed in Section 4.2.2) and soil management may all be contributing to the observed variability.
UA is often prone to suboptimal nutrient use due to overfertilization, resulting in low N and P use efficiency (Harada et al., 2018) and high garden-gate and soil P surpluses (Metson and Bennett, 2015;van de Vlasakker et al., 2022;Wielemaker et al., 2019).Three of the four gardens in this study received high N and P inputs, which for some plots were several tenfolds' higher than commonly used in rural agriculture (Table 1).Commercial open-field vegetable production systems do receive more fertilizers per area than other crops.Rates > 300 kg ha − 1 yr − 1 are not unusual among North American vegetable growers according to Congreves and Van Eerd (2015), who evaluated the impact of different management practices on minimizing N loss.The authors reviewed studies of cole crops (e.g., broccoli and kale) and reported average N use efficiencies of less than 50 % for all types, suggesting that a substantial amount of the N added would be prone to losses.This can partly be attributed to relatively shallow root systems of vegetable crops, resulting in limited uptake capacity (De Pascale et al., 2018) and to intensive fertilization and irrigation to maximize yields (Cameira and Mota, 2017).In their review of diffuse N losses and available mitigation strategies for horticultural systems, Cameira and Mota (2017) cite studies with reported leaching losses as high as 200-300 kg ha − 1 yr − 1 for various cole crops, i.e. about ten times higher than the mean losses from Swedish farmland mentioned above.
Another possible reason why the observed leaching from our gardens were much higher than measured leaching from agricultural fields could be that we collected the leachate water at a shallower depth (0.3 m vs. ca 1 m).Our sampling depth would not account for some of the subsoil transformations that may decrease nutrient concentrations in leachate Fig. 4. Cumulative TP leachate collected in lysimeters in allotment garden plots in Linköping and cumulative precipitation (grey line and circles) over the lysimeter area.The top panel shows results for plots 1-4, and the bottom panel plots 5-8 for visual clarity.Colored backgrounds indicate growing (green to the left) and non-growing (grey to the right) seasons, and stars and white dotted lines indicate weeks with snowfall.Plot 1 is marked with squares, plot 2 with diamonds, plot 3 with upside-down triangles, plot 4 with right-side-up triangles, plot 5 with x, plot 6 with +, plot 7 with dots and x, and plot 8 with circles.The colors of plots are unique but grouped by garden.For example, plot 1 and 2 are in the same garden so they are both a shade of blue.
water.Changes in soil water nutrient content may also depend on crop rooting depths and their N-uptake capacity (Thorup-Kristensen and Sørensen, 1999).Crops in our study belong to moderately deep rooted vegetables (Zemek et al., 2020a) so some removal of N below our measurement depth should also occur.However, at high level of N supply (as in our study), crop uptake capacity becomes more important than the root depth.Thus, with such high N input (Table 1), large leaching of N can be expected also at 1 m depth, especially considering that deeper rooting crops (as in our study) possess high potential for N leaching compared to shallow rooting plants (Zemek et al., 2020b and references therein).
Similar to N, the leaching of P in our gardens were 1.7-4.5 times higher than average P leaching from conventional farming in Sweden (Djodjic et al., 2004(Djodjic et al., , 2002)), but not as variable among plots as the TN leaching.As for N, this is likely related to high doses of P added as fertilizers (Table 1) but also to generally high levels of plant available P in the soils.The plant available P in the garden soils (Table 2) were in the highest P-class (V) according to the Swedish classification system (Swedish Board of Agriculture, 2021).For those levels, no additional P is needed for vegetable crops such as peas, and only 15 kg P ha − 1 for P demanding vegetable like potatoes (Swedish Board of Agriculture, 2021), which shows that the application rates of organic fertilizers in this study resulted in large over-application of P.
Our results indicate that there is a potential risk for high N and P losses from small-scale UA.Further studies are necessary to determine the impact such losses may have on ground and surface-water in urban areas.It is for example well known that the properties of the subsoil are important for the ultimate fate of nutrients that leach beyond the root zone, e.g.dissolved P leaching from the topsoil may be adsorbed if the subsoil has a high sorption capacity (Andersson et al., 2013;Lehmann and Schroth, 2003).Similar to findings in a recent literature review (van de Vlasakker et al., 2022), our results suggest that a substantial nutrient surplus (inputs > crop harvest) is common in urban gardens.Potentially they may become hotspots for N and P losses (Small et al., 2019), but this will depend on factors such as their location and the watershed geohydrological conditions, as has been shown from many studies of hotspot areas in agricultural watersheds (Kovacs et al., 2012;Schoumans and Chardon, 2003).Hence, more in-depth investigations of urban watersheds with a substantial proportion of small-scale gardening are motivated to identify under which conditions urban agriculture areas may significantly impact water quality.

Seasonal variability in nutrient leaching
Previous studies of UA have focused on N and P leaching during the growing season.Our results for the growing season (0.04-1.3 kg ha − 1 for P and 9.2-42 kg ha − 1 for N) are in the lower range of previously reported leaching losses of P (Shrestha et al., 2020;Small et al., 2018) and N (Small et al., 2018) in temperate regions.Importantly, our results clearly show that the dominant part of annual N and P leaching losses occurs during the non-growing season, i.e. between November and April (Fig. 3 and Fig. 4).Up to 95 % of water, 88 % of N and 97 % of P leachate losses were measured during that season (Fig. 2), which is in line with what is known for conventional farming at northern latitudes (Plach et al., 2019;Ulén et al., 2019).

Impact of plot cover and irrigation during the growing season
Although leaching was highest during the non-growing season, leaching losses were also observed during the growing season and both irrigation and occasional rainfall could be driving losses in those weeks.We occasionally collected high water volumes when little or even no rainfall was noted (Fig. S2).Although we did not measure water input from irrigation (though we inquired about general practices), excessive irrigation could be the reason that collected leachate volumes occasionally exceeded the maximum possible water input via precipitation during the growing season in some plots (Fig. S2).It was particularly visible for covered plot 3 and for plot 8.In those plots, P leaching in July and September, respectively, coincided with water percolation volumes, but exceeded maximum potential water input with precipitation (Fig. 2, Fig. 4, Fig. S2).We also observed higher water percolation and higher TP leaching losses during the growing season from plot 2 than from the other plots.Non-optimal irrigation increases N and P leaching (Barton and Colmer, 2006;Li et al., 2018) and could be an important factor driving those occasional high nutrient losses during the growing season.In contrast, well balanced irrigation may influence nutrient losses more than fertilization (Li et al., 2018), especially to counteract a reduced plant nutrient uptake during drought periods.This suggests that optimal irrigation is an important step to properly manage nutrients (Sigua et al., 2017), also in small UA settings.
Ground and plant covering might also have contributed to spatial variability in leaching losses in our study.Although most plots followed a similar seasonal pattern, one plot was a clear outlier.Plot 3 experienced overall low leaching but the growing season contributed to 74 % of the water, 62 % of the N and 78 % of the P leaching losses (Fig. 1).This plot was located in the same area, its soil characteristics were similar to other gardens and it also received similar nutrient input levels as plot 4.However, plot 3 was considerably different regarding individual management, namely the usage of fine netting (picture SI), which covered the entire plot down to the ground throughout the year.This type of garden fabric is usually used in agriculture to prevent insect predation while allowing sun penetration (Giannoulis et al., 2021).Additional benefits of net usage include moisture retention and elimination the potential damage caused by higher rainfall (Marino, 2021).The use of such a cover type seems to have blocked most of the rain and snow, which in turn considerably reduced water leachate volumes.We can infer that the nutrient and water leaching during the growing season in this plot were driven by irrigation.Another garden (plot 5 and 6) was also covered with a down to the ground net, yet, with a bigger mesh size, which likely retained moisture and blocked some precipitation (picture SI).However, the net was removed at the beginning of the non-growing season, which seems to have resulted in elevated water losses compared to time period when the net was present (Fig. 2).Thus, our findings seem to indicate there may be potential for leachate mitigation with the use of a fine net over crops in small-scale UA.
We cannot rule out that precipitation as rain caused occasional nutrient leaching, but our data do not indicate that direct rainfall was a main driving force for leaching during the growing season.This is not surprising as intense rainfall events rather cause surface runoff than leaching (Yao et al., 2021).However, in conventional rural farming, precipitation dynamics can sometimes significantly affect nitrate-N leaching (Hess et al., 2020) and may play a crucial role during the summer month losses (Jabloun et al., 2015).Although, in our study, the rainfall-nutrient leaching relationship was not straightforward, partially due to lack of quantitative information about irrigation, we emphasize that, precipitation dynamics still need to be accounted for to better manage nutrient losses from UA.It may be especially important under projected changes in rainfall patterns and intensity in northern Europe related to climate change, which together with warmer temperatures, can increase nutrient leaching (Patil et al., 2010).Our results support the importance of assessing annual leaching, and identifying factors that affect the leaching during the non-growing season.

Importance of seasonal organic-N mineralization
The amount of mineral N that is subject to leaching after the crop has been harvested is the sum of soil mineral N remaining after harvest and non-growing season mineralization of plant residues left on and in the soil.A review of commercial horticulture in Switzerland showed that the crop N demand varies with a factor two-three among vegetables, with a soil mineral N (SMN) target level of about 100-150 kg N ha − 1 yr − 1 for low demand crops like garlic, beans, and squash, up to 250-300 kg N ha − 1 yr − 1 for N demanding crops like kale or cabbage (Zemek et al., 2020a).SMN represents the mineral N level required to reach maximum yield based on fertilization trials.In our study, three out of four gardeners applied amounts of composted and fresh plant residues that was more than enough to meet SMN (Table 1).
The fertilizers were mostly applied as a one-time application, which is known to cause the highest potential nutrient leaching (Nachmansohn, 2016), and thus this may have been a contributing factor to the large leaching observed in our study.This could also explain the observed increase in TN leachate early in the season in most of the plots (Fig. 3).When the soil temperature increases, microbial mineralization rates also increase and mineral-N and P from the organic fertilizers are released into the soil (Buffam et al., 2016).Consequently, ammonium-N is rapidly nitrified to highly mobile nitrate-N, which in combination with limited uptake rates of young plants in spring, can result in leachate losses.
As the plants grew, crop uptake matched the soil continuous supply of inorganic-N resulting in low leaching during the rest of the growing season.However, at the very high levels of TN supplied in all but two plots, it is possible that some nitrate-N remained in the soil after harvest.Thorup-Kristensen and Sørensen (1999) reported 10-100 kg NO 3 --N ha − 1 yr − 1 in the upper meter of soil at fertilization levels of 300 kg N ha − 1 yr − 1 depending on the rooting depth of three different vegetables.
In our case, we measured the leachate at a depth of 0.3 m, which means that the rooting depth probably was of less importance for the results, but the high doses of organic-N could indeed have resulted in nitrate-N remaining in the soil at the end of the growing season.Additionally, crop residues supply more organic-N in the end of the growing season.Zemek et al. (2020a) showed that the most N demanding crops (cabbage species) also result in the largest amount of organic-N left as crop residues after the harvest, increasing the risk for nitrate-N leaching in the non-growing season.
In our study, the soil in most of the plots was left without plant cover, and thus nutrient uptake, after the harvest.The lowered evapotranspiration in combination with precipitation led to gradual saturation of the soil, resulting in water percolation starting early December, about 2 months after harvesting.Plot 3 was an exception from this pattern, as the cover material (mesh-net) functioned as a "tent", preventing precipitation from reaching the soil.In the other plots, continued mineralization of organic fertilizers and plant residues followed by nitrification, added to the soil pool of mineral-N prone to wash-out with percolating water.The fact that nitrate-N accounted for on average 73-99 % of the TN lost with percolating water confirms this (Table S2).Interestingly, the TN leachate from plot 4 was still high (100 kg N ha − 1 yr -1), even if with fertilizer, it received the lowest input among all uncovered plots (90 kg N ha − 1 yr − 1 ).The beans grown in that plot have a symbiotic relationship with nitrogen-fixing Rhizobium bacteria, resulting in additional nitrogen inputs; thus this crop belongs to the group of vegetables that leaves a high amount of N in the soil after harvest (Zemek et al., 2020a).This, in combination with possible organic N remaining in the soil from previous years, may explain the relatively high TN leachate also from this plot.

Soil heterogeneity and P leaching variations
Variation in water and nutrient fluxes are often related to soil heterogeneity (Manzoni et al., 2008;Raynaud and Leadley, 2004), including soil type and P content (Bergström et al., 2015).All plots in our study had relatively similar soil characteristics (Table 2), making the macro-level soil type and nutrient content unlikely drivers of the spatial heterogeneity in TP leaching losses that we identified.Still, infiltration rates at the plot-scale are affected by the soil structure and physical disruption of aggregates, e.g. when gardeners prepare the soil for seeding, may lead to reduced infiltration rates and lower water holding capacity (Basche and DeLonge, 2019;Logsdon et al., 1993).Additionally, plant root growth and macrofauna (e.g.earthworms) activity affect the structure, for example by creating macropores that lead to preferential flow paths through the soil profile (Bouma, 1991).This would increase the risk for elevated P losses, particularly in clay soils as those of the gardens in our study (Bronswijk et al., 1995;Liu et al., 2012;Shipitalo et al., 2000).Both these factors could have contributed to the considerable differences in leachate volumes collected from closely located lysimeters in plots 5 and 6 (Fig. 1), where 50 % more water and TP was collected from plot 5 over the year.Similarly, we collected twice as much water from plot 8 than from plot 7 over the year although both plots were managed by the same gardener.In contrast to for plot 7, almost half of both the water and the TP leaching from plot 8 was collected during the growing season, pointing towards the importance of water flow in driving the TP leaching.This was in sharp contrast to the TN leaching from the same plot which mainly were collected during the non-growing season.Information from the respective gardeners suggest plots received similar soil preparation and fertilizer amounts, suggesting that small-scale differences in soil properties may be an important factor to explain variations in observed TP leaching in UA.
Considering the above, mitigation strategies should include both crop management (e.g.residue management after harvest and/or the use of cover crops and crop rotations) and fertilizer and irrigation water management approaches.According to De Pascale et al. ( 2018), the most desirable would be application of state-of-the-art practices like plant biostimulants, sensors and other forms of monitoring soil and plants, to improve the use of N and irrigation water.However, it may be challenging to disseminate such knowledge to small scale UA growers; the costs of some of these techniques may also limit their future application among hobby and small-scale gardeners.General guidelines to avoid over irrigation and fertilization based on grouped soil tests for an allotment/community garden area may be more suitable.

Importance of snowmelt for annual TN and TP leaching
The non-growing season is relatively long in northern latitudes and includes greater snowfall and snowmelt, which facilitates nutrient leaching in agricultural catchments (Costa et al., 2020).Indeed, we also observed elevated nutrient leaching during the snowmelt period.Initially, we noted some nutrient leaching in the beginning of the snow cover period, which is evidence that the soil was not frozen and leaching losses could still occur (Fig. 3, Fig. 4).Snow cover may be an important regulator of soil processes; it insulates and prevents the soil from freezing and also provides moisture, which in turn affects biological processes and thus nutrient cycling (Jones, 1999).Eventually, due to low air temperatures (− 20 C), the soil froze and remained frozen for about 2 months, during which period the leachate losses of TN and TP were low.At the on-set of snowmelt TN and TP leachate rapidly increased.
Nutrient leachate during snowmelt is often a major cause for hydrological nutrient exports in both agricultural and natural watersheds (Elliott, 2013;Liu et al., 2014).N mineralization rapidly raises in previously frozen soils and increases the pool of labile organic-N, which means that snowmelt can contribute to high losses (Djodjic et al., 2002).Consequently, prolonged infiltration caused by slowly thawing snow and precipitation can lead to high soil saturation, which elevates the risk for high nutrient losses also after snow melt period (Liu et al., 2013).In our study the period from the on-set of snowmelt until early May accounted for 32-75 % and 20-83 % of the annual TN and TP leaching, respectively.
This confirms the extraordinary importance of the snowmelt period for nutrient leaching, not only in conventional farming (Plach et al., 2019;Ulén et al., 2019), but also in UA in northern latitudes.
In summary, our data emphasize that temporal dynamics of nutrient leaching during and after snowmelt should be accounted for in UA gardens in a northern climate.

Conclusions
This study demonstrates the importance of non-growing season nutrient leaching in accounting for N and P leaching from UA gardens in a temperate region.Even if most active nutrient management happens in the spring and summer, a majority of leaching in such a temperate climate occurred in the winter and early spring, especially following snowmelt.Our findings also illustrate that these leaching losses are relatively high and that they are characterized by large amounts of within and among garden variability.Careful monitoring must thus occur year-round and account for high spatial heterogeneity to avoid under-estimating leaching losses.Subsequently it would likely be necessary to create garden management guidelines that account for how nutrients can be lost outside of the growing season for areas where leaching from UA may impact local or downstream water quality if it is hydrologically connected and represents a large enough area of a watershed.Our highly variable results indicate that there is potential to substantially reduce nutrient leaching from UA by adapted gardening practices, e.g.soil covering and timely application of fertilizers; however, a better understanding of cause-effect relationships is needed before such guidelines can be formulated.

CRediT authorship contribution statement
All the authors have seen and approved the final version of the manuscript being submitted.It is original authors' work.It has not received prior publication and isn't under consideration for publication elsewhere.

Fig. 1 .
Fig. 1.Total annual leachate loss of water in liters (bottom panel), TN (middle panel) and TP (top panel) as kilogram per hectare per year (kg ha − 1 y − 1 ) from each study area plot (see Table 1 for plot descriptions).Data are arranged according to total volume of leachate collected.Percentage values above each bar indicate how much of the TN and TP leachate loss, and water volume, occurred during the non-growing season.

Table 2
Average values of soil characteristics of eight UA gardening plots in Linköping, Sweden, in October 2020.AL stands for ammonium lactate, the standard method used for extraction in Sweden (for more detailed description see Methods text), P is phosphorus, K is potassium, Mg is magnesium, Ca is calcium, Al is aluminum, and Fe is iron.