Biochar production and characterisation

Biochar used in this experiment was produced from willow (Salix matsudana L.) chips. Air-dried feedstock (<12% moisture content) was pyrolysed at a maximum heating temperature of 350°C and residence time of 4 h. The characteristics of the biochar were: pH 7.8, organic C (C_org) 703 g kg⁻¹, the ratio of dichromate-oxidisable C out of C_org (C_ox/C_org) 51.4%, atomic H/C_org 0.63, liming equivalence 7.3% CaCO₃-eq (Garbuz et al. 2021). The biochar used in our study was classified as having a C storage class of 2, a liming class of 1, and a fertiliser class of zero (no nutrient value) (Camps-Arbestain et al. 2015). The choice of willow was based on the fact that it is readily available as it is used extensively in New Zealand in soil conservation and stream bank protection. Willow grows readily from a cutting and has an extensive rooting system.

Study sites

Two soils common under pastures in lower North Island of New Zealand were used in this study: (1) a Dystric Cambisol (IUSS Working Group WRB 2015), Brown soil in the New Zealand soil classification system (Hewitt 2010), from the experimental site of AgResearch Ballantrae Hill Country Research Station, Manawatu (40°18′35″S, 175°49′41″E); (2) a Sil-andic Andosol (IUSS Working Group WRB 2015), Allophanic soil (Hewitt 2010), from Hawera, Taranaki (39°36′28″S, 174°16′30″E). The Andosol used in this study is derived from volcanic ash. Rich in short-range order constituents, it offers high SOM protection, has a high anion retention, good physical properties, and resilience to treading pressure (Molloy 1998). The Cambisol is predominantly derived from loess materials. It has a low anion storage capacity and limited physical resilience to treading pressure. Two pastures grazed by dairy cows throughout the year on the Andosol were selected: one receiving dairy shed effluent (And-EF), and one not receiving effluent (And-NE). Both pastures receive 160 kg of N as fertiliser N ha⁻¹ year⁻¹, 300 kg of 20% potash superphosphate ha⁻¹ year⁻¹, and 1 kg selenium prill ha⁻¹ year⁻¹ (a standard fertiliser regime for NZ pastoral soils under dairy farming). The two selected pastures on the Cambisol were grazed by sheep throughout the year: one (Cam-LF) had received no superphosphate since 1980, and the other (Cam-HF) receives 375 kg superphosphate ha⁻¹ year⁻¹ since 1980 (Mackay et al. 2021).

Field-based mesocosm experiment

The field-based mesocosm experiment was conducted using large soil cores enclosed into PVC cylinders (150 mm Ø, 300 mm length). Four holes (5.1 cm Ø) were made in the wall of each cylinder to allow the free movement of soil organisms (fig. S1 in Garbuz et al. 2021). There were three treatments: (1) no amendments (negative control), (2) 1% of biochar application by weight (equivalent to approximately 10.9 Mg ha⁻¹), and (3) lime (positive control) applied at a rate corresponding to the liming equivalent of biochar (ca. 0.8 Mg ha⁻¹). Each treatment was replicated six times in each of the four pastures. During the southern hemisphere spring of 2017, the PVC cylinders were hammered into the ground and excavated with soil from each of the four pastures. At the laboratory, the turf layer (ca. 20 mm) was split off, and the top 150 mm of soil below the turf layer was removed from all cores. All earthworms from the topsoil were removed, counted, labelled with the core code, and cold-stored. For biochar and lime treatments, the soil was mixed with either biochar or lime, respectively, and put back into cylinders to the depth 20–170 mm; in control mesocosms the soil was also removed, and mixed without amendments. Earthworms and turf layer were placed back, and mesocosms installed in the field. Further details on design, preparation, and installation of mesocosms are provided in Garbuz et al. (2021). The experiment started in late October–November 2017, during the southern hemisphere spring. The sampling (18 cores from each pasture) occurred in November 2018, approximately 12 months after the start of the experiment. Climatological data for the two locations are provided in Supplementary Fig. S1.

Soil physico-chemical and biological properties

Soil samples for chemical analysis and microbial biomass were collected with a corer (30 mm Ø) from five depths: 0–20 mm (the turf), 20–95 mm, 95–170 mm, 170–200 mm, and 200–300 mm, and air-dried. The following variables were measured: soil bulk density (BD), pH, total C, total N (TN), nitrate-N (NO₃⁻–N) and ammonium-N (NH₄⁺–N), Olsen P. Inorganic C was negligible (<0.05%) in the lime-treated soil after 12 months of incubation, and thus total soil C was all organic (C_org). Mesofauna abundance (Collembola, Oribatida and Gamasina) was sampled by taking 50 mm × 50 mm × 50 mm cores from the topsoil (20–95 mm) in each mesocosm cylinder. Fungal and bacterial biomass were measured in mixed topsoil (20–170 mm) using the substrate-induced respiration (SIR) method with selective inhibition (Nakamoto and Wakahara 2004). Fungal (C_f) and bacterial (C_b) biomass C was calculated according to Anderson and Domsch (1978). The sum of C_f and C_b was considered as the microbial biomass. Earthworms from each mesocosm cylinder (full depth) were hand sorted, identified to species when possible, and counted. Plant roots from each cylinder were collected by hand, washed over a 3-mm sieve, oven-dried (40°C) and weighed. Soil properties are summarised in Garbuz et al. (2021) and we relate them to enzyme activities in this study.

Soil enzymes analysis

Soil samples for enzyme analysis were collected with a corer (30 mm Ø) from the same five depths as soil chemistry and microbial biomass samples. All soils were sieved (<2 mm) and air-dried prior to analysis. Alkaline and acid phosphatases, nitrate-reductase, urease, cellulase, peroxidase and dehydrogenase activity were measured in each of the five depths in all three treatments of the two contrasting soils. Details on soil enzyme analysis are provided in Garbuz et al. (2020) and in the supplementary information.

Statistical analysis

Statistical analysis was carried out using SAS 9.4. Normality of data sets was evaluated by the Shapiro–Wilk test. A multicollinearity analysis was done to check simple correlations and variance inflation factors for variables. The data were normalised using z-score prior to analysis. Analysis of variance (ANOVA) with contrast statements and Tukey HSD test were used to investigate the effect of factors: pasture (And-NE, And-EF, Cam-LF, Cam-HF), treatment (control, biochar, and lime) and pasture × treatment interaction on enzyme activities. When the interaction term was not significant, main effects were reported; if the interaction effect was significant, the four pastures were considered separately. Finally, for biochar-treated and control mesocosms we constructed a hypothetical model of causal relationships underlying the observed patterns in soil enzyme activity, biota, and nutrients, and used path analysis (proc CALIS in SAS 9.4) to calculate coefficients associated with each path in the model. Due to sample size, the hypothetical model of causal relationships underlying the observed patterns in soil C-enzymes activity was limited to cellulase and dehydrogenase.

Enzymes of the C cycle

Cellulase and dehydrogenase activities were higher in the biochar-treated soil than in the control (all P < 0.001), while the effect on peroxidase was site-dependent (Tables 2 and 3). In all pastures treated with lime, the activity of peroxidase was increased over that of the control and biochar-treated soil (all P < 0.001). Lime had no effect on cellulase activity, while its effect on dehydrogenase was site-dependent (Tables 2 and 3).

Path analysis (biochar-treated and negative control mesocosms only) identified significant pathways for cellulase and dehydrogenase (shown as C-enzymes) that offer a plausible representation of causal relationships (Fig. 1). In both soil types, the activity of cellulase (and C-enzymes for the Cambisol) was closely correlated with microbial biomass, with biochar addition acting as a significant driver for this variable, through changes in soil pH and the SOM pool (Fig. 1). The pathway involving roots in cellulase and dehydrogenase activity (C-enzymes) was important only in Cambisol pastures. In both soil types, the main source of C used by microbes appears to be root-derived, as opposed to the source of N, which appears to be derived from the soil pool, as described below. Pathways involving earthworms and mesofauna were not significant for cellulase and dehydrogenase, suggesting that these enzymes’ activity is largely microbial-driven. On the other hand, pathways for peroxidase indicated the singular importance of pH for activity of this enzyme, with other factors not being significant (data not shown).

Fig. 1. 

Pathways for cellulase and dehydrogenase: (a) cellulase only in the Andosol pastures, (b) cellulase + dehydrogenase (C-enzymes) in the Cambisol pastures. Pathways for dehydrogenase in Andosol were not significant. Double-sided (red) arrows are correlations; dashed (blue) arrows represent pathways which were not quantified in the model. Both models significant, P < 0.001.

Enzymes of the N and P cycle

In both Cambisol pastures urease was higher in the biochar-treated soil than in the control or lime. In the Andosol, the soil receiving effluents (And-EF) had higher urease (all P < 0.005) than the same soil without effluent (And-NE), but no effect of biochar was detected. The activity of nitrate reductase was highest in soil treated with lime in both Andosol pastures, but also higher in biochar-treated soil than in control (Tables 2 and 3). Higher nitrate-reductase activity was correlated with higher earthworm abundances in sites with lower fertility (And-NE and Cam-LF) (both P < 0.001); data on earthworm abundance is provided in Garbuz et al. (2021). Path analysis suggests that the main drivers for nitrate-reductase activity in both the Andosol and Cambisol pastures were available N (NO₃⁻–N) and soil pH (Fig. 2). Pathways involving soil fauna (earthworms and mesofauna) were not significant for nitrate reductase in either soil type. Pathways involving plant roots and microbial biomass in nitrate-reductase activity were important only in Cambisol pastures. Alkaline phosphatase activity was higher (and acid phosphatase lower) in the biochar- and lime-treated soils compared to the control (both P < 0.005).

Fig. 2. 

Pathways for nitrate reductase in the (a) Andosol and (b) Cambisol pastures. Pathways for fauna (earthworms and mesofauna) were not significant. Double-sided (red) arrows are correlations; dashed (blue) arrows represent pathways which were not quantified in the model. Both models significant, P < 0.001.

Enzymes of the C cycle

The positive effect of biochar on dehydrogenase activity was attributed to an increase in microbial biomass C in biochar-treated soils (data for microbial biomass provided by Garbuz et al. 2021). Root C inputs to the rhizosphere, in response to higher root biomass, have been shown to stimulate soil enzyme activity and increase microbial biomass (Brzostek et al. 2013). Dehydrogenase is an important component of microbial metabolic functions (Casida 1977) and is often used as an indicator of soil microbial response to land use practice changes (Watts et al. 2010; Järvan et al. 2014). Biochar application can have a positive effect on dehydrogenase activity and increased C mineralisation in the soil as it has been observed in the short term (positive priming) (Ouyang et al. 2014; Lehmann et al. 2021) which can be related, to some extent, to the presence of labile C supplied in biochar as well as its labile C to labile N ratio if residual N is present (Wang et al. 2012). The high C_ox/C_org ratio and the relatively high atomic H/C_org ratio suggests the presence of a considerable fraction of labile C in our biochar. This probably favoured microbial growth and positive priming of native SOM as inferred from a mass balance calculation (Garbuz et al. 2021). However, over one year period, biochar-treated soils in our study showed a 1.2 to 4.0 Mg C ha⁻¹ gain in root C, especially high in Cambisol pastures, which may generate a negative priming over time (Garbuz et al. 2021). Biochar can promote long-term C storage through stabilisation of rhizo-deposits and organic ligands, in general, on biochar surface, as it can act as a new reactive surface (Lehmann et al. 2021).

The increased cellulase activity in biochar-treated soils was correlated with increased root biomass, suggesting that the increased supply of root necromass provided additional substrate for cellulase activity (Sajjad et al. 2002; Sinsabaugh 2010), yet the path analysis suggests that, in the Andosols, this only occurs indirectly through the impact of the enhanced root growth on microbial biomass. The amount of cellulose in our biochar is negligible given that the biochar was produced at 350°C and cellulose is fully carbonised at temperatures above 240°C (Demirbaş 2001). The path analysis indicated both similarities and differences in the drivers of C-enzyme activities in the Andosol and Cambisol pastures. The similarities reflect that, as expected, the most important driver is soil biological activity, while the differences are likely to be a manifestation of different types of SOM, soil aggregation and microbial community structure between the two soils. For example, the fact that in the Cambisol, but not in the Andosol, roots have a direct impact on both measured enzymes involved in the C cycle, could be related to the existence of a more plant-derived OM in the Cambisol, and a more microbial-derived SOM in the Andosol (Herath et al. 2015) due to differences in clay mineralogy and prevalence of microaggregates in the latter (Angst et al. 2021). The increase in root biomass reported by Garbuz et al. (2021) would have caused a direct increase in microbial-derived SOM in both soil types, but this increase would have been diluted in the Andosol where this SOM fraction is already more abundant as suggested by Wang et al. (2015b).

Peroxidase activity generally increases with soil pH (Sinsabaugh 2010). With an increase in pH, the bonds of organic molecules (ligands) with mineral surfaces become weaker, and desorbed SOM becomes more easily degraded and oxidised by peroxidases (Sinsabaugh 2010; Tian and Shi 2014).

Enzymes of the N cycle

Nitrate reductase activity is affected by factors such as nitrate concentration and soil pH, with an optimum at pH 7 (Abdelmagid and Tabatabai 1987). This is consistent with the enzyme activity trends observed in this study: (1) nitrate-reductase activity was highest in the Andosol that received effluents (And-EF) and in the Cambisol with high fertility pasture containing an active legume component (Cam-HF); (2) both lime and biochar increased nitrate-reductase activity, with lime having a more pronounced effect. Andosols are abundant in micropores and microaggregates which can remain saturated with water for long periods under udic moisture conditions (Buurman et al. 2007), as experienced during the study, thus creating favourable anaerobic conditions for nitrate reduction. This, and high substrate availability, could explain the higher nitrate-reductase activity in this soil (Abdelmagid and Tabatabai 1987).

Substrate availability (NO₃⁻–N) in biochar-treated topsoil in our experiment was higher or the same as in control and lime treatments (Garbuz et al. 2021) but, compared to lime, biochar mesocosms had a lower nitrate-reductase activity. In our study, biochar-treated soil had significantly lower bulk density and higher root biomass (Garbuz et al. 2021); we speculate that this would be associated with increased soil aeration, thereby limiting the activity of nitrate reductase (Abdelmagid and Tabatabai 1987; Joseph et al. 2015). Moreover, it is well known that biochar has an effect on redox-regulated N transformations (Chacón et al. 2017). In fact, when produced at low temperature, such as the one used in this study, biochar can act as an electron shuttle (Chacón et al. 2017; Dai et al. 2021), favouring the full reduction of NO₃⁻ to N₂ (Obia et al. 2015). Yet the impact of biochar electrochemical properties on N-reductase activity is hard to discern with the data available.

There is evidence that biochar and lime affect different groups of soil bacteria responsible for denitrification (Bai et al. 2015; Jha et al. 2016; Harter et al. 2017; Weldon et al. 2019). The positive correlation between earthworm abundance and the activity of nitrate reductase in two of the pastures supports the idea of a link between earthworms and denitrifying bacteria, suggested by some authors (Burtelow et al. 1998; Depkat-Jakob et al. 2010) through increased nitrate input by earthworms. Garbuz et al. (2020), working with the same soils under glasshouse conditions, also showed synergetic interactions between lime, earthworms, and increased nitrate-reductase activity.

The high urease activity in the Andosol might reflect the higher urine input from the lactating dairy cows, up to 55 L urine cow⁻¹ day⁻¹ (Betteridge et al. 1986), compared to 3 L urine sheep⁻¹ day⁻¹ (Ledgard et al. 2008) from sheep grazing in Cambisol. Urease activity has been reported to be strongly correlated with soil bacterial biomass (Amini Kiasari et al. 2018). In our experiment bacterial biomass increased with the addition of biochar (Garbuz et al. 2021), and was correlated with urease activity in all pastures except And-NE. The latter may be explained by the fact that the bacteria–urease correlation relies on a specific group of ureolytic bacteria, but not on the whole bacterial community (Lloyd and Sheaffe 1973). In other studies, urease activity in response to biochar addition has also been unpredictable; for example, rice husk biochar had both negative and positive effects on urease activity in two different acid soils – Ultisol with a pH 5.8, C_org 16 g kg⁻¹ and available N 1.6 g kg⁻¹ and Alfisol with a pH 4.4, C_org 3.7 g kg⁻¹, and available N 0.0156 g kg⁻¹ (Huang et al. 2017; Oladele 2019).

Enzymes of the P cycle

As expected, the addition of alkaline material (lime or biochar) drove an increase in the alkaline phosphotase to acid phosphotase (AlP/AcP) ratio, reflecting the sensitivity of phosphatases to soil pH (Acosta-Martínez and Tabatabai 2000), with activity of acid phosphatase decreasing and alkaline phosphatase increasing. Olsen P values in the soil amended with biochar were higher than in the control in both Andosols and in the Cambisol with high fertility (Garbuz et al. 2021), so the higher alkaline phosphatase activity in the biochar-treated soils compared with the control reflects increased substrate availability. However, this was not the case in the Cambisol with low fertility, indicating that despite an increase in the AlP/AcP ratio, this does not always translate into an increase in plant-available P. In addition to causing an increase in available P through desorption due to increased pH, biochar amendment may also influence P availability because of the increase in SOM. An enrichment in organic ligands would result in chelation of Al³⁺ and Fe³⁺ that would otherwise precipitate P (Gao and DeLuca 2016; Gao et al. 2019).

The higher phosphatase activity in the Andosol than Cambisol soil can be explained by the fact that the short-range order inorganic constituents (e.g. allophane) abundant in Andosols have the capacity to immobilise phosphatase (Chatterjee et al. 2014; Jordanova 2017) and protect this enzyme from adverse conditions (Shindo et al. 2002). Phosphatase activities are positively correlated to SOM, which enhances the stability of these enzymes (Bonmati et al. 1991). As root biomass was enhanced in the biochar treatment (Garbuz et al. 2021), we can hypothesise that root-derived SOM increased alkaline phosphatase activity.