Introduction

Nitrogen (N) is a key driver of plant productivity and N2 fixed by nodulated legumes is estimated to contribute 33–46 Tg N annually to global agricultural systems (Herridge et al. 2008). Declining levels of organic matter in soils results in a reduced capacity to supply N for crop growth (Dalal and Mayer 1986) and, as a result, crop, fodder and shrub legumes continue to play a critical role in maintaining or improving soil N status (Peoples et al. 1995). In addition to fixing N, legumes contribute to soil carbon (C) via release of root exudates and tissue residues, act as a disease- and pest-breaks in cereal-based rotations, and can enhance the phosphorus (P) nutrition of subsequent crops through undefined mechanisms (Karpenstein-Machan and Stuelpnagel 2000; Nuruzzaman et al. 2005; Rose et al. 2010).

While nodulated legumes offer a promising option as a rotation crop in farming systems, N2 fixation by rhizobia-legume symbioses is often limited by soil constraints such as acidity, which affects more than 1500 Mha worldwide (Graham and Vance 2000). In addition, the binding of key nutrients including P and molybdenum (Mo) in acid soils (Smith et al. 1997) limits plant uptake of these minerals with further consequences for legume productivity. The deleterious effects of P deficiency in particular to reduce biological N2 fixation and uptake of soil mineral N by legumes has been well established (e.g., Giller and Cadisch 1995; Graham and Vance 2000).

While liming can overcome the constraints of acid soils and increase legume growth, the accessibility of lime may be a limitation to its use in low-input farming systems. In contrast, biochar (pyrolysed organic matter) has been shown to increase N2 fixation in acid soils (Rondon et al. 2007; Guerna et al. 2015) and may be a viable option for low-input farming systems by exploiting locally-available feedstocks (Cowie et al. 2012). Rondon et al. (2007) attributed the enhanced N2 fixation in common bean (Phaseolus vulgaris L.) with biochar application in their controlled-environment study to increased boron (B) and Mo availability. The authors also suggested that higher potassium (K), calcium (Ca), and P availability were factors, although higher pH, reduced mineral N and aluminium (Al) saturation may have contributed also, but to a lesser extent. However, given that the acid soil in the study of Rondon et al. (2007) was limed prior to sowing, and that the biochar contained significant quantities of P, Mo and B, it is difficult to attribute the changes in legume growth and N2 fixation to any long term influence of the biochar on soil properties as opposed to a simple fertilisation effect over the 75-day-study. Indeed, the authors conclude that longer-term field studies were required to better understand the impacts of biochar of N2 fixation. A more recent glasshouse study by Guerna et al. (2015) demonstrated that biological N2 fixation increases following biochar amendment, resulting from increased plant P uptake which was associated with a 360 % increase in mycorrhizal colonisation. In contrast, Mia et al. (2014) suggested that improved biological N2 fixation in red clover (Trifolium pratense L.) following biochar amendment resulted from improved K nutrition in a moderately acid soil. The authors suggest this benefit could be short term only due to the relatively high availability of K from the biochars used.

The aim of the present study was therefore to investigate the impact of biochar on legume growth and N2 fixation in an acidic soil in the field over several seasons and to elucidate any causal mechanisms for enhanced legume performance. In addition to overcoming nutrient deficiencies in acid soils, a number of other possibilities may also explain any enhanced N2 fixation due to biochar application. Further, the presence of free ammonium (NH4 +) or nitrate (NO3 −) in soils can inhibit N2 fixation (Herridge and Betts 1988). Biochar addition can reduce mineral N concentrations through a range of mechanisms (Cayuela et al. 2014; Clough et al. 2013) and this reduction in free soil NO3 − and NH4 + may enhance the percentage of N2 fixed by legumes grown with biochar amendments.

As part of a broad program to assess the potential of biochar to enhance legume productivity in the field, we investigated the impact of well-characterised and contrasting biochars, in comparison to a commercially available compost, lime and nil amendment controls, on the growth, grain yield and N2 fixation of nodulated faba bean (Vicia faba L.). The field studies were undertaken on an acidic ferralsol, a soil type that occupies 750 Mha of tropical and sub-tropical agriculture worldwide (FAO 1998). These acidic soils often present constraints in supply of P to crops due to the complexation of phosphate with Al and iron (Fe) oxides, clays and organic matter (Moody 1994). To elucidate causal mechanisms we resolved key changes in soil parameters and plant nutrient uptake as a result of amendment addition to establish correlations between these parameters and legume growth and N2 fixation. By growing a maize crop in the first season after amendment addition (and between subsequent faba bean crops) to nullify any immediate nutritional benefits (particularly P, Mo and B) from the addition of the amendments, we were able to focus on the longer-term impacts that biochar properties have on soil parameters and legume growth. Resolving the causal factors affecting N2 fixation and legume growth on the acidic ferralsol following biochar, compost and lime amendments should enable better prediction of the effects of these amendments across a broader range of environments.

Materials and methods

Organic amendments

Two biochars, one from poultry shed waste and the other from papermill residue, were produced by Pacific Pyrolysis Ltd., Somersby NSW Australia, using a continuous 300 kg h−1 pilot slow-pyrolysis unit. Poultry shed waste was obtained from a commercial poultry producer and papermill residue was collected from an Australian paper mill previously described in Van Zwieten et al. (2010). Both biomass feedstock’s were pyrolysed at a highest treatment temperature (HTT) of 550 °C, at a heating rate of 5–10 °C min−1 and with a maximum residence time at HTT of 45 min. The compost was a commercial product, certified to Australian Standards AS 4454 (Composts, soil conditioners and mulches).

Total carbon (TC) concentration of the compost was 21 %, compared with 38 and 42 % for paper mill (PM) and poultry litter (PL) biochars, respectively (Table 1). Total organic carbons (TOCs) were less than TCs for the compost and the two biochars, with the PL biochar showing the largest reduction. Total N concentrations were 2.5 and 2.6 % for the two biochars, compared with 1.2 % for the compost. The mineral N concentrations were low for all three organic amendments, with almost 100 % of the mineral N present as ammonium. Total P, citrate-soluble P and total K were highest in the PL biochar compared to PM biochar and compost.

Table 1 Chemistry of the compost and biochars used in the field study
Full size table

The PL biochar and PM biochar had pH values of 8.9 and 6.8 respectively, with liming values of 8.8 and 18 % that of agricultural lime (CaCO3) (Table 1). Compost had a liming value of 6.3 %. The PL biochar had 44 cmol NH4OAc extractable K, and 13 cmol NH4OAc extractable Na. The PM biochar had lower extractable cations, and was dominated by Ca. Similarly, extractable Ca dominated in the compost. The PL biochar contained relatively high levels of the heavy metals copper (Cu), chromium (Cr), zinc (Zn) and arsenic (As), compared with the PM biochar and compost, while the compost contained relatively high levels of lead (Pb).

Study site and experimental design

The 3 years cereal-legume rotation experiment had amendments applied in December 2007 and was conducted on a Rhodic Ferralsol (described in detail in Macdonald et al. 2014) at the Wollongbar Primary Industries Institute, NSW, Australia (28.49.34″S, 153.23.5″E; elevation 140 m). In summary, the soil is fine-textured and iron-rich with kaolinite, gibbsite and goethite mineralogy. It has a bulk density of 1 g cm−3 and a CEC of 22 meq(+) 100 g−1. The phosphorus buffering index of the soil was high at 483 (Slavich et al. 2013). The climate is sub-tropical with a summer-dominant rainfall. Three months before the commencement of the trial, the site was prepared by mowing existing pasture and removing most of the above ground biomass. Glyphosate was then sprayed at the recommended rate and, after 2 weeks, the site was rotary hoed to 100 mm depth of soil. The site was rotary hoed again immediately prior to application of amendments to control weeds. The field site was surveyed and plots 5 × 4 m marked out in a grid with three replicate blocks of ten treatments.

The experiment was designed to investigate the mechanisms from which single application of poultry litter (PL) biochar or papermill (PM) biochar (application rate 10 t ha−1 dry wt) impacted growth, grain production and biological N2 fixation in winter-grown faba bean crops. The three other treatments were compost (25 t ha−1 wet wt, equivalent to 13 t ha−1 dry wt), agricultural lime (3 t ha−1) and nil amendment. The five soil treatments were combined with the presence or absence of fertiliser (urea 400 kg ha−1, single superphosphate 300 kg ha−1 and muriate of potash 140 kg ha−1, equivalent to 184 kg N ha−1, 26 kg P ha−1 and 70 kg K ha−1), applied prior to sowing sweet corn (Zea mays L.) across all plots in the summers before, between and following the two faba bean crops. Amendments were applied in December 2007 and the faba bean crop was sown approximately 6 months after application and again in the following season, i.e., 18 months after application. There was no fertiliser applied to the faba bean crops. The site was rotary hoed to incorporate the soil amendments into the 0–100 mm soil layer. The nil amendment plots were also rotary hoed. All fertilisers were subsequently surface applied.

The 10 treatments, i.e., five soil treatments × 2 fertiliser levels, were allocated to plots in a randomised complete block layout with three replicates. Soil bulk densities across the various treatments in the 0–100 mm profile ranged between 1.00 and 1.02 and were not statistically different at 3 months after the application of amendments (data not shown).

Each plot was mechanically sown with six rows of corn in November 2007, 2008 and 2009 and harvested in February 2008, 2009 and 2010. Above ground biomass was removed shortly after harvest (data not presented). Faba bean (cv Cairo) was sown in May in 2008 and 2009 at rates of 100 kg seed ha−1. Prior to sowing, the faba bean seed was inoculated at the recommended rate with commercial Group F inoculant containing Rhizobium leguminosarum bv. viciae strain WSM1455 (Herridge et al. 2014).

Plant and grain sampling and assessment of N2 fixation

Faba bean biological N2 fixation was determined using the natural 15 N abundance technique (Unkovich et al. 2008). Plants were sampled at maximum shoot biomass, i.e., late pod-fill (Schwenke et al. 1998). For each plot, a 1 m row of faba bean was harvested to ground level together with an adjacent 1-m row of planted oats (Avena sativa) which was used as the non N2-fixing reference. All plant materials were dried in a dehydrator at 700C for 48 h, ground using a Culatti Micro Hammer-mill fitted with a 1 mm screen, then sub-sampled for analysis. The acid extractable elements in biomass were determined according to USEPA 6010 using a Varian 720-EC ICP-OES, Inductively Coupled Plasma, Optical Emission Spectrometer (ICP-OES). Grain yield was determined by harvesting four rows per plot (equivalent to the width of one pass) using a Wintersteiger small-plot harvester. Measurements were converted to a per-hectare basis.

Subsamples of dried and ground material from the late pod-fill sampling were analysed for %N and δ15N. The subsamples (ca. 5 mg) were precisely weighed using a 5-place balance into tin capsules and submitted to the University of California Stable Isotope Facility for analysis by a PDZ Europa ANCA-GSL elemental analyser interfaced to a PDZ Europa 20–20 isotope ratio mass spectrometer (Sercon Ltd., Cheshire, UK). The percentage of legume N derived from biological N2 fixation (%Ndfa) was estimated using standard equations (Unkovich et al. 2008) and a B value of −0.39‰ (Schwenke et al. 1998). Total N2 fixed by faba bean was calculated by multiplying %Ndfa by shoot N (kg ha−1) by a constant of 1.4 to account for below-ground N (Herridge et al. 2008).

Soil sampling and analysis

On six occasions during the study, soil samples were taken to monitor effects of soil amendments and fertiliser on soil properties. Three soil cores (50 mm diameter) were taken from each plot (0–100 mm), air-dried, composited and passed through a 1 mm sieve prior to analysis. Soil and biochar chemical analyses were done in a National Association of Testing Authorities Australia (NATA) facility accredited to ISO17025.

Total C and N were measured by Dumas combustion using an Elementar vario MAX CN analyser with combustion chamber set at 900 °C and oxygen flow rate of 125 ml min−1. The pH was measured in 0.01 M CaCl2 (1:5) according to method 4B2 (Rayment and Higginson 1992). Solubilised cations were assessed using 1 M NH4OAc described in method 15E1 (Rayment and Higginson 1992) and analysed with a Varian 720-EC ICP-OES, Inductively Coupled Plasma, Optical Emission Spectrometer (ICP-OES). Liming values of amendments were measured as carbonate equivalent using method 19A1 (Rayment and Higginson 1992). The acid extractable elements and metals were determined with the ICP-OES according to USEPA 6010. Plant available Mo in soil was measured using method 12 B 1 (Rayment and Lyon 2010). The method is based on a 0.01 M EDTA extraction buffered to pH 8.6 with NH4HCO3. Biochars and compost were tested for total P, water-soluble P, citrate-soluble and -insoluble P according to AOAC Official Methods 977.01 and 963.03 (AOAC International 2000).

Plant analysis

Plant material (0.25 g of dried and ground plant material) was digested in a teflon vessel by microwave using H2O2/HNO3 according to Wu et al. (1997). Diluted extracts were analysed using a Varian 720-EC ICP-OES, Inductively Coupled Plasma, Optical Emission Spectrometer (ICP-OES).

Statistical analysis

Linear mixed models were used to estimate and compare the various attributes measured for each amendment and season. The model treated each attribute as a response to fixed effects defined by a five-level factor (soil amendment), a two-level factor (fertiliser) and a factor for the interaction between those terms. Where seasonal data were available the fixed effects were extended to include seasons and the interaction between seasons and the other factors. Each model included random effects to estimate variation due to the replicate blocks and, when seasonal data was available, a random effect to include variation due to individual plots.

An analysis of variance was derived from each model in order to conduct null hypothesis significance tests for each factor. The average of each attribute at all combinations of soil amendment, fertiliser and season was predicted together with an estimate of standard error (SE). Formal statistical comparison of the averages was enabled by calculation of least significant difference (LSD) at the 5 % level of probability. Variance heterogeneity due to seasons was observed with some attributes clearly showing different levels of variation within seasons. When that was the case, an extra variance component was included in the model in order to inflate the corresponding standard errors.

To investigate the interaction between the N2 fixation observations and selected soil and plant yield attributes from the 2008 crop, a principal components analysis was conducted and presented as a biplot. Nine key attributes were identified as soil factors: pH, EDTA available Mo, total C, available Al, and plant factors being grain yield, N fixed by the crop, total P, B and Mo uptake by the crop. The 10 (soil amendments and fertiliser) by 9 (attributes) matrix of averages was created, centered and scaled so that each attribute had mean zero and variance of one over the amendments. Principal component scores and loadings were calculated from this matrix and presented as a biplot with attributes represented by arrows and amendments represented by points. A brief guide to interpreting the biplot is that attributes with arrows in similar direction tend to be positively correlated. Attributes with arrows of opposite direction tend to be negatively correlated. Attributes with arrows at right angles tend to have zero correlation. Similarity of amendments with respect to all attributes is indicated by proximity of their points and the degree of association between amendments and attributes can be indicated by projecting a perpendicular line from the treatment to the attribute vector. All data analyses and statistical graphs were constructed using the statistical package R (R development core team 2012).

Results

Corn yields during three seasons

In 2008 there was no significant effect of amendments on fresh cob yields in unfertilised plots, with yields ranging 17.1–20.2 t ha−1 (Table 2). Likewise within fertilised plots, amendments had no impact on yield. However, in comparing across fertilised and unfertilised treatments, fertiliser with lime and biochar amendments yielded significantly (p ≤ 0.05) higher than the non-fertilised control. All biomass (cob and stover) was removed from plots, resulting in the removal of 193–265 kg N ha−1 in unfertilised plots and > 300 kg N ha−1 in fertilised plots (Supplementary Table 1). The removal of other key nutrients at harvest in 2008 included 19–31 kg P ha−1, 70–136 kg K ha−1, 52–82 kg Ca ha−1, 4–7 kg Mg ha−1 and 8–13 kg S ha−1 (Supplementary Table 1).

Table 2 The effect of amendments and fertiliser treatments on fresh weight of corn cobs (t ha−1) over three seasons. Average standard error of the means was 2.40 (2008), 1.49 (2009) and 2.04 (2010)
Full size table

Corn cob yields in 2009 were unaffected by addition of amendments applied in December 2007 in the unfertilised treatments, but lime, PM biochar and PL biochar treatments resulted in significantly higher yields than the control treatment in the presence of fertiliser (Table 2). Further, the PM biochar treatment had significantly higher cob yields than the compost treatment. In 2010 there was no effect of any amendment on cob yields regardless of fertiliser treatment.

Yields of faba bean and nutrient uptake

Significant differences between nil amendment and the four soil amendments, either with or without applied fertiliser, were observed for biomass and grain yields, shoot fixed and total N contents and shoot P content in 2008, while there were no differences between any of the five treatments for shoot B content. Control treatments showed significantly lower shoot Mo content than all amendments except the PL biochar treatment in the absence of fertiliser in 2008, yet in the plus fertiliser treatment shoot Mo content was significantly higher in the control treatment than all other treatments (p ≤ 0.05; Table 3). In the 2008 season in the absence of fertiliser, compost and PM biochar amendments resulted in significantly higher biomass yields than the PL biochar amendment, and significantly higher grain yields than the lime treatment (p ≤ 0.05). In terms of nutrient uptake in 2008, PL biochar resulted in significantly lower shoot N contents compared to PM biochar and compost treatments in the absence of fertiliser, but no differences in shoot P or B content were observed among amendments. Interestingly, shoot Mo content was significantly higher in the lime treatment compared to other amendments in the presence or absence of fertiliser. In the fertilised plots in 2008, all amended plots had significantly higher shoot P content and biomass and grain yields than control plots, but there was no difference among amendments for any yield or nutrient uptake parameters (Table 3).

Table 3 Effect of biochar, compost and lime on faba bean yield and shoot nutrient content
Full size table

The wet weather following the sowing of the second faba bean crop in May 2009 (350 mm in May and 150 mm in June) led to severe infection of all plots with chocolate spot (Botrytis fabae). This was then compounded by severe moisture stress when < 50 mm rain fell from July to September (Supplementary Fig. 1), leading to extremely poor growth in all plots (grain yields in the nil fertiliser treatment of 0.29 t ha−1 in 2009 compare to 1.78 t ha−1 in 2008; Table 3). The only significant treatment effects observed in 2009 were significantly higher shoot biomass yields in all amended plots compared to control plots in the fertilised treatment, and significantly higher shoot biomass yields in PM biochar plots compared to control plots in the absence of fertiliser.

Faba bean N2 fixation

The δ15N values were similar for the nil and plus fertiliser treatments, but varied substantially between the two seasons particularly for the non N2-fixing reference, oats (Table 4). The δ15N values for faba bean ranged −0.12–0.98‰ in 2008 and 1.24–2.92‰ in 2009, compared with δ15N ranges for the non N2-fixing oats of 3.22–4.59‰ (2008) and 6.54–12.21‰ in 2009.

Table 4 Effects of soil amendments on δ15N (‰) of above-ground faba bean and non N2-fixing reference (oats) and estimated %Ndfa of faba bean, all growing in an acidic ferralsol during 2008 and 2009
Full size table

The proportion of faba bean shoot N derived from N2 fixation (%Ndfa) was high for all treatments with average estimates across all treatments of 77 % in 2008 and 72 % in 2009 (Table 4). Overall, the lime and PM biochar amendments resulted in the highest %Ndfa values with the compost amendment the lowest, although the only significant difference (p < 0.05) was the higher %Ndfa in PM biochar in the fertilised treatment in 2009 (89 %) compared to the compost treatment (62 %). The amount of N2 fixed in the aboveground biomass (fixed shoot N content) was significantly lower in controls than for all amended plots with or without fertiliser in 2008 (76 kg ha−1 for unfertilised vs 84 kg ha−1 for fertilised treatment; Table 3). In the absence of fertiliser, fixed shoot N was significantly lower in the PL biochar plots than all other amended plots, while in the presence of fertiliser the compost plots had significantly lower fixed shoot N than all plots with other amendments (p ≤ 0.05; Table 3). Data from 2009 had high variability and no significant differences were detected.

Soil chemistry

Analysis of the soils just prior to sowing of the first, i.e., 2008, faba bean crop ca. 7 months following application of soil amendment revealed differences among treatments in EDTA extractable Mo, pH and Bray P (Table 5). Amendment with PM biochar resulted in significantly higher available Mo compared to either control or compost in plots that did not receive fertiliser. In fertilised plots, the limed treatment had significantly lower Mo than either the PM or PL biochar treatments, but these biochar treatments were not different to the control.

Table 5 Soil chemistry before sowing faba bean into an acidic ferralsol in April 2008
Full size table

In both fertilised and unfertilised control plots, the pH was low (pH = 4.3) at the time of sowing the 2008 faba bean crop. All soil amendments significantly increased soil pH in the absence of fertiliser, but only the lime, compost and PM biochar increased pH in the presence of fertiliser. A significant acidification trend was observed across the 4 crop cycle, except in the unamended controls, which were generally significantly lower than amendments (Fig. 1). The PL biochar had the least influence on soil pH of the four amendments.

Fig. 1
figure 1

Average soil pH under each amendment for two crop rotations. * indicates date of amendment, and the triangle indicates sowing of the first corn crop. C (corn) and B (bean) indicate standing crop when samples were taken. Vertical bar span the least significant difference at 5 % critical value for comparing amendments at each observation date

Full size image

Poultry litter biochar had significantly higher Bray P in both fertilised and unfertilised plots, compared to controls or other amendments (Table 5). There were no statistical differences between the other amendments. In the absence of fertiliser, there was a decline in available P in the PL biochar plots, but levels still remained significantly higher than for the other treatments during the sampling period (Fig. 2). In the presence of applied fertilisers, there was no significant decline in available P during the sampling period. Poultry litter biochar again had the greatest influence on available P in soil. Fertilised plots tended to have higher Bray P than plots that did not receive fertiliser.

Fig. 2
figure 2

Average soil Bray P under each amendment for two crop rotations. * indicates date of amendment, and the triangle indicates sowing of the first corn crop. C (corn) and B (bean) indicate standing crop when samples were taken. Vertical bar span the least significant difference at 5 % critical value for comparing amendments at each observation date

Full size image

The control and lime amendments had higher nitrate levels in the fertilised plots compared to the unfertilised plots, but there were no differences between soil amendments. There were no treatment effects on soil NH4 + concentrations (Table 5).

Bi-plot showing association of yields with various parameters

Biological N2 fixation was very closely related to plant Mo uptake (Fig. 3). This was influenced by liming and fertiliser. Yield of faba bean, however, was most closely related to plant P and B uptake, as well as soil pH and EDTA-extractable Mo concentration. Amendment with PM biochar had the strongest influence over these parameters although lime with fertiliser application, PL biochar and compost also had some influence. High exchangeable Al was linked closely to nil amendment controls which had significantly lower pH values compared to amendments (Table 5).

Fig. 3
figure 3

Bi-plot of the first two principal components calculated from the average 2008 season faba bean yield (Y), N2 fixed by the crop (Nfix), P, B and Mo shoot uptake and associated soil parameters; exchangeable aluminum (Al), soil carbon (C), EDTA molybdenum (Mo) and pH. Amendments are indicated by name in combination with presence (+) or absence (-) of fertiliser. Arrows in similar direction tend to be positively correlated. Attributes with arrows of opposite direction tend to be negatively correlated. Attributes with arrows at right angles tend to have zero correlation. Similarity of amendments with respect to all attributes is indicated by proximity of their points and the degree of association between amendments and attributes can be indicated by projecting a perpendicular line from the treatment to the attribute vector

Full size image

Discussion

A significant and well-defined constraint to productivity of legumes and efficacy of biological N2 fixation is soil acidity and the related problems of Al- and Mn-toxicity, and deficiency of Mo, Ca, and P (Graham and Vance 2000). Acidity limits both survival and persistence of nodule bacteria in soil, and the process of nodulation itself. The rhodic ferralsol used in this study, while having good physical properties, had a low pH, low available Mo and high exchangeable Al, all of which are well defined constraints to successful legume production. It is possible that some of these constraints (e.g., Ca and P deficiency) may have been overcome in the 2008 faba bean crop as a result of a flush of nutrients following the conversion of a pasture system to a cropping system in early 2008. However, we contend that this is unlikely given that 19–31 kg P ha−1 and 52–82 kg Ca ha−1were removed from plots following the first (2008) corn crop (Supplementary Table 1). The fact that faba bean grain yields were higher in the control plots that received NPK fertiliser than control plots without fertiliser (2.20 vs 1.78 t ha−1; Table 3) further suggests that nutrient limitations existed. The 2009 faba bean crop suffered from moisture stress and infection by chocolate spot. We attribute this poor growth of faba bean (biomass reduced by over 100 % compared to the previous season) to moisture and disease stress rather than any specific depletion of nutrients or pH effect. Indeed, the pH of the soil in the 2009 faba bean crop in the compost and biochar treatments were not significantly different to the pH of the 2008 faba bean crop, when these treatments resulted in the highest yields in the unfertilised plots (Table 3). This assertion is further supported by the fact that maize yields in the following season (2010) did not show any drastic yield reductions compared to previous maize crops (Table 2). As such, the discussion focusses on the growth and biological N2 fixation of the 2008 faba bean crop.

Fixation of N2 in 2008 (shoot fixed N content; Table 3) was most strongly associated with the Mo status of shoot tissue (Fig. 3), supporting the results of earlier studies on acid soils that have found strong correlations between legume Mo uptake and N2 fixation (Johansen et al. 2007; Rondon et al. 2007). However, rectifying Mo deficiency in the earlier studies was achieved by the addition of Mo fertiliser (Johansen et al. 2007) or a biochar that contained significant Mo that was apparently available to plants over the 75 days growth period (Rondon et al. 2007). In the field trials conducted in the present study, Mo supplied in the amendments could not explain the enhanced Mo nutrition of the legumes and corresponding increases in %Ndfa and therefore shoot fixed N content. The compost (at 13 t ha−1, dry weight equivalent) had the highest concentration of Mo (Table 1) and supplied substantially more Mo to soils than the lime or biochar amendments. Despite this, the compost treatment resulted in lower %Ndfa (Table 4). Given that the liming treatment resulted in a significantly greater shoot Mo content than all other amendments regardless of fertiliser application (except the nil amendment in the presence of fertiliser, Table 3) and that lime had a strong association with Mo uptake and N2 fixation (Fig. 3), enhanced soil Mo availability via increased pH appears the most plausible explanation. Molybdenum availability in acidic soils is typically limited due to sorption to Fe- and Al-oxides, with maximum sorption in the range pH 4–5 (Smith et al. 1997). The increase in pH (CaCl2) in the rhodic ferralsol from 4.3 in the unamended soil to 5.0 or greater in the lime and PM biochar treatments (Table 5) presumably enhanced the bioavailability of native Mo in the soil.

Despite this contention, it is acknowledged that soil N status may have influenced the %Ndfa among treatments to some degree because of the different types and amounts of N applied in the various amendments. While is it well established that high soil mineral N can inhibit leguminous biological N2 fixation (Herridge and Betts 1988), uncertainties exist with the influence of organic N application via plant biomass, compost, or indeed biochar. The cultivation of a maize crop prior to the sowing of the faba bean crop removed much of the readily available N (193–350 kg N/ha exported in removed corn biomass, Supplementary Table 1). This was reflected in the concentrations of NH4 + and NO3 − in the soil prior to sowing the 2008 faba bean crop, where no significant differences were detected amongst amendments (Table 5). However, N mineralisation rates in compost vary widely depending on the source and properties of the compost (Hartz et al. 2000). For example, Pu et al. (2012) recently found negligible N mineralisation from compost added to a sub-tropical ferrosol, while other reports vary from 10 % (Wen et al. 1995) to 50 % (Buchanan and Gliessman 1991) mineralisation over a cropping season. Assuming this range of mineralisation, 16–78 kg N/ha may have been available to the faba bean crop in the 2008 growing season, and this may have contributed to the lower %Ndfa in the compost treatment. In contrast, while the biochars used in our study contained 2.5 and 2.6 % total N for PM biochar and PL biochar respectively, it is unlikely that any major quantity of this N would mineralise in the short term. Using similar biochars to our current study (paper sludge and poultry litter), Singh et al. (2012) reported minimum turnover times for the biochar carbon of 100 years, and as temperature of production increased from 400 to 550 °C, residence time increased significantly. This further suggests that pyrogenic N within biochar may remain relatively unavailable and thus not play an important role in supplying mineral N to the soil and reducing biological N2 fixation of a legume growing in that soil.

Interestingly, while Mo uptake was the major driver for N2 fixation, a number of other factors contributed to legume grain yield production in the acid soil. Grain yield was strongly associated with pH, P uptake and to some extent, B uptake (Fig. 3). Macdonald et al. (2014), using a similar soil type, demonstrated that biochars have the ability to increase soil pH resulting in relatively large gains in biomass attributed to improved P nutrition and alleviated Al toxicity. It should be noted here that soil characteristics are important in effecting this outcome, as Macdonald et al. (2014) also showed nil effects on a neutral vertisol as well as negative impacts on plant growth due to increased electrical conductivity in an acidic arenosol. The compost added an estimated 819 kg ha−1 lime equivalents, PL biochar 880 kg ha−1 lime equivalents and PM biochar 1.8 t ha−1 lime equivalents. In-field effects on pH, however, were similar between the limed plots (equivalent to 3 t ha−1 agricultural lime) and PM biochar (Fig. 2), suggesting the longer-term liming effect of biochar amendments may be underestimated using the laboratory analysis described in Rayment and Lyon (2010). In previous work (Van Zwieten et al. 2010), PM biochar was shown to have a range of acid neutralising sites based on oxides, hydroxides and carbonate, thus our addition of 3 t ha−1 lime into soil was designed to match the anticipated in-field pH response to PM biochar. This was achieved with closely matching soil pH values across the project timeline (Fig. 1). While the PL biochar had the highest P input from all amendments (Table 1), resulting in sustained increases in soil Bray P (Fig. 2), shoot biomass yields were significantly less than the compost and lime treatments in 2008 in the unfertilised plots (Table 3). This may be due to the relatively limited influence PL biochar had on soil pH (Fig. 1). This suggests that both alleviatied Al toxicity and improved P nutrition were required for higher yields. Similarly, Guerna et al. (2015) showed that growth and %Ndfa of common bean (Phaseolus vulgaris L.) were closely correlated to plant P uptake in a degraded and slightly acid humic acrisol. The authors surmised that the increased P uptake, however, resulted mostly from improvements in mycorrhizal colonization, with a lesser direct contribution from improved soil pH and nutrient status.

Conclusions

Overcoming the major nutritional constraints to legume production in acid soils over the mid- to longer- term relies on increasing Mo uptake into shoots via amelioration of soil pH. The improvement in crop yields rely on addressing additional constraints such as P availability, especially in high P fixing soil, and overcoming Al toxicity. The study here has demonstrated that biochars can provide an important liming effect in addition to their immediate fertiliser value. Given that the acid-neutralising capacity of biochars depends on both the feedstock and pyrolysis temperature, it may be worth exploiting this to produce biochars with a high acid-neutralising capacity in low-input farming systems constrained by acid soils. The causal mechanisms for enhanced N2 fixation and crop productivity identified in the present study differed from those identified in short term glasshouse studies. This highlights the need for further field studies over multiple seasons to understand the longer-term impacts of biochar and other amendments on addressing constraints to biological N2 fixation and legume productivity.