Biochar
Biochar was prepared from pigeon pea (Cajanus cajan) stems in a Kon-Tiki kiln (Cornelissen et al. 2016). The temperature in the kiln was monitored 3–5 cm below the flame curtain using a Fluke 51 II Digital thermometer, equipped with an 80 cm external sensor probe (max temp 1372°C). The temperature in the kiln fluctuated between 600–750°C. More details about the measurements of temperature in Kon-Tiki kilns can be found in Cornelissen et al. (2016). The chemical characteristics of the biochar, analyzed according to Munera-Echeverri et al. (2018), include pH (10.4), Electrical Conductivity (1.4 mS cm− 1), total organic carbon (55.7%), Total N (0.69%), total H (1.1%), CEC (6.6 cmol(+) kg− 1), C:N ratio (81), and molar H:Corg ratio (0.24). Further details of the biochar and the methods used can be found in Table S1.
Experimental site and design
The experiment was conducted on a private farm in Mkushi, Zambia (S13°45′25.7″ E29°03′55.5″), in February 2018. The average annual precipitation and temperature are 1220 mm and 20.4°C, respectively. The soil is a sandy loam Acrisol (WRB 2015) containing 69% sand, 10% clay and 21% silt. Land use prior to the experiment was nine years of conservation agriculture, comprising planting basins, crop residue retention and crop rotation (maize-ground nuts). The basins were dug by hand using a hoe and each one had an area of ~ 0.07 m2 (40 cm length x 16.7 cm width x 20 cm depth) and a volume of ~ 13.4 L. The distance between basins was 80 cm and the distance between rows of basins was 90 cm (~ 13890 basins ha− 1) (Martinsen et al. 2019). The residues were placed in between rows of plants, therefore, the area dedicated to basins did not receive the crop residues. The addition of fertilizer was done inside the basins, but only in the seasons when maize was planted. By contrast, in the seasons with ground nuts no fertilizer was added, which agrees with the local practice of farmers. The effects of the first seven years of conservation agriculture on soil organic matter and soil properties can be found in (Martinsen et al. 2019).
In November 2017, the land was prepared under two different practices: Normal practice of conservation agriculture using basins (CA-NORM), and conservation agriculture with biochar addition inside basins (CA-BC). A total of 2.8 ton ha− 1 of dolomitic lime was applied in the basins of both practices (11 g kg− 1 soil inside basins). Each practice was randomly assigned to plots distributed in three blocks. Maize was planted on November 22, 2017. Each plot of the experimental blocks was about 20 m2, accommodating seven rows of plants. Each row had four planting basins with three maize plants per basin in CA-NORM and CA-BC. The three experimental blocks were surrounded by maize of the same variety planted in basins (Fig. S1).
In the CA-BC treatment, biochar was applied to the area occupied by basins only and mixed with soil to 20 cm depth. Fertilizer and seeds were place on top of the biochar-soil mixture prior to backfilling of ~ 5 cm soil on top. All CA-BC plots received 250 g pigeon pea biochar per basin, equivalent to 2.0% w w− 1 or 4-ton ha− 1. The CA-NORM and CA-BC plots received 17.1 ± 0.8 g of NPK (10% N, 20% P2O5, 10% K2O) per basin at planting, corresponding to 237 kg NPK fertilizer ha− 1. Four and eight weeks after planting (December 23, 2017 and January 20, 2018) urea was applied at a rate of 100 kg N ha− 1 each time to the two soil management practices.
15 N application
The 15N tracing experiment started 11 weeks after planting (on February 6, 2018) and the pulse of 15N was followed throughout 10 days. The experimental setup was a split-plot design, with each plot divided into 3 split-plots that corresponded to the two forms of 15N application and the unlabeled (water addition only) reference (Fig. S1). Two different 15N forms were tested: NH415NO3 and 15NH4NO3. In all six plots (two plots (CA-NORM and CA-BC) in each of three block), both 15N tracers (99.98 atom%) and the reference (H2O) addition were assigned to one row (Fig. S1), while one buffer row of maize was kept in between labeled rows to avoid cross-contamination (Fig. S1). In all CA-NORM and CA-BC plots, four basins were selected for 15N application (i.e. twelve plants in total) in each of the rows treated with NH415NO3, 15NH4NO3, and water, respectively. In each basin, the full area of 668 cm2 (40 cm x 16.7 cm) was treated. The total amount of 15N added per basin (CA-NORM and CA-BC) was 6 mg, which is equivalent to 0.1g 15N m− 2 (0.2 g N m− 2 added as NH4NO3). The 15N was added dissolved in distilled water (concentration of 24.0 mg 15N L− 1). Hand sprayers were used to add 250 ml of the labeled solutions (only water at the reference) evenly to the soil surface (Fig. S2). This volume was equivalent to about 4 mm per basin. Subsequently, 15 mm of clean water was added to all basins to wash the label into the soil. The reference received an equivalent total volume (19 mm water). Precipitation for the growing season 2017–2018 and during the 10 days 15N tracer experiment was monitored using a rain gage (Davis 6162 Wireless Vantage Pro2 Plus) and logged using 6510USB Davis WeatherLink Computer Interface.
Soil, gas and plant sampling and analyses
Soil samples were taken 1, 24, 72, 168 and 240 hours after 15N addition from 0–5 cm (n = 90) and 5–20 cm (n = 90). At each sampling time, soil samples from the labeled basins were taken and bulked for 15N analysis. Bulked soil samples consisted of six (0–5 cm) or four (5–20 cm) mixed cores taken with an 8 mm diameter auger. The samples were stored in a cooling box on ice and sub-samples were extracted on-site, within 4 hours with 1M KCl (Yu et al. 2017). The remaining soil was dried at 40°C for one week to determine the gravimetric moisture content (θg in g g− 1). The amount of dry soil in each bulked soil sample and the volume of the auger were used to estimate an average bulk density (BD in g cm− 3) for each split-plot at 0–5 cm and 5–20 cm depth. Volumetric moisture content (θv in cm3 cm− 3) was calculated as (θg x BD)/ρwater, where ρwater is the density of water (g cm− 3). Subsequently, dry soil samples were sieved (2 mm), milled in a mechanical mortar and packed into 8 x 5 mm tin capsules and shipped to the Stable Isotope Facility, of the University of California, Davis for 15N analysis. Total N and total C were determined using an Elementar Vario EL Cube elemental analyzer (Elemental Analysensysteme GmbH, Hanau, Germany) interfaced to an isotope ratio mass spectrometry (IRMS) to analyze 15N. The large combustion columns of the Elementar Vario EL Cube systems allows using big bulk samples. The weight of the sample was optimized according to the N content of the samples, which was about 73.454 ± 0.763 mg corresponding to about 36.97 ± 7.62 µg N. Other soil parameters were analyzed in dry soil samples sieved at 2mm. Exchangeable base cations (Ca2+, Mg2+, K+ and Na+) and extractable acidity (H+) were determined in ammonium acetate at pH 7 (Schollenberger 1945). Plant available phosphorus (P-AL) was determined by the ammonium lactate method described by Krogstad et al. (2008). Soil pH was determined in 0.01 M CaCl2 using a solid to solution ratio of 1:2.5.
The KCl extracts were prepared by adding 11 g of field moist soil and 40 ml of 1 M KCl to 50 ml centrifuge tubes, shaken horizontally at 200 strokes per minute for one hour and filtered using Whatman filters grade 589/3. The supernatants were frozen immediately and transported to the Norwegian University of Life Sciences (NMBU), where NO3− and NH4+ contents were analyzed using flow injection analysis (FIA star 5020, Tecator, Sweden). The 15N abundance in NO3− was determined following the denitrifier method of Zhu et al. (2018), which converts NO3− quantitatively to N2O before analyzing 15N by PreCon- GC-IRMS (Thermo Finnigan MAT, Germany).
Plant samples were taken after 24, 168 and 240 hours by collecting the aboveground biomass and digging out the entire root system of the three plants of one of the labeled basins under CA-NORM (n = 9) and CA-BC (n = 9). The roots were washed in the field, the plants cut at brace roots height and split into roots and aboveground biomass, and the fresh weight recorded. The plant samples were taken to the University of Zambia (UNZA), where they were oven-dried at 70°C and ground. The dry biomass was weighed, and the samples were transported to NMBU to be milled in a horizontal ball mill and weighed in tin capsules for 15N analysis at University of California, Davis using a PDZ Europa ANCA-GSL elemental analyzer interfaced to a PDZ Europa 20–20 IRMS. The sample weight was 4.96 ± 0.397 mg that corresponded to 66.89 ± 25.89 µg N. At the end of the growing season, maize yield, as well as total aboveground biomass were measured for each of the plots.
Fluxes of N2O were measured 1, 24, 48, 72, 120, 192, 216 and 240 hours after 15N addition (n = 192). A closed static chamber of 170 cm2 (14.7 cm diameter) and 2.95 L headspace was gently pressed inside the planting basins under CA-NORM and CA-BC (n = 144). Gas samples were collected using a 20 ml syringe coupled to a 3-way valve; gas samples were transferred to pre-evacuated 10 ml glass vials crimp-sealed with a butyl septum (Chromacol). Samples were taken 1, 15, and 30 minutes after chamber deployment. Temperature inside the chambers was recorded at the beginning and the end of chamber deployment. The glass vials were shipped to Norway and analyzed for N2O by automated gas chromatography (GC Model 7890A, Agilent, USA). The N2O fluxes were estimated by linear regression of N2O concentration change over time and calculated as µg N2O-N m− 2 h− 1.
Calculations
Analysis of 15N in KCl-extractable NO3−, bulk soil and plant
The atom% 15N of KCl-extractable NO3− was calculated according to Stevens and Laughlin (1994)), using the mass to charge ratios (m/z) 45 and 46 of the N2O assuming a non-random distribution to account for double substituted 15N2O produced in the denitrifier method:
Atom% 15NO3− = 100 (45R + 246R -17R − 218R) / (2 + 245R+246R)
where 45R is the ratio of the ion currents (I) at m/z 45 and 44 (45I/44I); 46R = 46I/44I; 17R (17O/16O) = 3.8861 x 10− 4; 18R (18O/17O) = 2.0947 x 10− 3. Oxygen isotopes were assumed to be at natural abundance.
Atom% 15N excess values of NO3− (atom% 15NNO3−) and bulk soil (atom% 15Nsoil) were calculated by subtracting the atom% 15N of the non-labeled reference treatments
15 N mass balance
The mass of 15N recovered (g m− 2) in each N pool was calculated as:
Mass 15N = (Xsample * Ncontent * Mass)
where Xsample is the 15N fraction in the sample calculated as suggested by Providoli et al. (2005):
Xsample = (Fsample – Freference) / (Ftracer – Freference)
Here, Fsample is the fractional abundance of 15N in the samples (15N/(15N + 14N)). Freference is the fractional abundance of 15N in the reference treatments. Ftracer is the fractional abundance of applied tracer (0.9998, i.e. 99.98 atom%). Ncontent is the concentration of N in plant material, bulk soil and KCl-extractable NO3− (g g− 1), respectively. Mass is the total plant biomass and soil per unit area of basin. The 15N recovered (%) was expressed relative to the amount of 15N applied (0.1g 15N m− 2 basin or planting row). The recovery of 15N in the soil residual N was defined as the recovery of 15N in the bulk soil minus the 15N recovery in the KCl-NO3− pool. Note that soil residual 15N comprises the 15N in the NH4+ pool and the organic N pool.
Data analysis
Statistical analyses were performed using the package lme4 (Bates et al. 2015) of R software (R-Core-Team 2020). The effects of soil management on soil properties, biomass, grain yield and effect of soil management and N form on 15N recovery were tested by using linear mixed effect models with block as a random factor. Soil parameters at 0 to 5 cm and 5 to 20 cm were analyzed separately. Further linear mixed effect models were used to test differences between the two soil management practices and the change over time of volumetric soil moisture, ammonium and nitrate in soil, and N2O fluxes inside basins, and recovery of 15N. Soil management practice nested in block was used as random factor. Model checking was based on plotting (histograms and QQ plots) and visual inspection of residuals, fitted values and predicted random effects to assess normality and potential outliers. Also, Shapiro-Wilk test of the residuals was performed (α = 0.05). One outlier of volumetric soil moisture was removed without comprising the results of the post hoc test. KCl-NO3−, KCl-NH4+, the recovery of 15N in soil NO3−, and N2O fluxes were log transformed to fulfill model assumptions. The temporal autocorrelation between repeated measurements was assumed constant between the different treatment combinations. The parsimonious models were chosen using Akaike information criterion (AIC) and R2 values. Differences between treatments were assessed by least-squares means using the function difflsmeans of the package lmerTest (Kuznetsova et al. 2017). Spearman correlation matrix was set up to find significant correlations between variables. Bidirectional stepwise regression analysis was performed to link natural logarithm of average N2O fluxes in each plot at each sampling date with the variables that were significantly correlated with N2O production.