Are there environmental or agricultural benefits in using forest residue biochar in boreal agricultural clay soil?

Are there environmental or agricultural benefits in using forest residue biochar in boreal agricultural clay soil? Helena Soinne a, ⁎, Riikka Keskinen b , Jaakko Heikkinen b ,JariHyväluoma c , Risto Uusitalo b , Krista Peltoniemi a , Sannakajsa Velmala a ,TainaPennanen a ,HannuFritze a , Janne Kaseva b , Markus Hannula d , Kimmo Rasa b a Natural Resources Institute Finland, Latokartanonkaari 9, FI-00790 Helsinki, Finland b Natural Resources Institute Finland, Tietotie 4, FI-31600 Jokioinen, Finland c HAMK University of Applied Sciences, Mustialantie 105, FI-31310 Mustiala, Finland d BioMediTech Faculty of Medicine and Health Technology, Tampere University, Tampere, Finland HIGHLIGHTS •Forest residue biochar significantly increased carbon stored in soil. •Biochar amendment was not effective in reducing leaching of nutrients. •Biochar did not change the clay soil water retention properties. •Soil bacterial or fungal communities were not changed. •A high biochar dose did not induce such positive impacts that would benefit farmers. GRAPHICAL ABSTRACT abstractarticle info Article history: Received 27 February 2020 Received in revised form 22 April 2020 Accepted 22 April 2020 Available online 28 April 2020 Editor: Fang Wang Keywords: Biochar Carbon sequestration Soil quality Soil productivity Nutrient leaching Microbial community Short-term agronomic and environmental benefits are fundamental factors in encouraging farmers to use biochar on a broad scale. The short-term impacts of forest residue biochar (BC) on the productivity and carbon (C) storage of arable boreal clay soil were studied in a field experiment. In addition, rain simulations and aggregate stability tests were carried out to investigate the potential of BC to reduce nutrient export to surface waters. A BC addition of 30 t ha −1 increased soil test phosphorus and decreased bulk density inthesurface soil but did not significantly change pH or water retention properties, and most importantly, did not increase the yield. There were no changes in the bacterial or fungal communities, or biomasses. Soil basal respiration was higher in BCamended plots in the spring, but no differences in respiration rates were detected in the fall two years after the application. Rain simulation experiments did not support the use of BC in reducing erosion or the export of nutrients from the field. Of the C added, on average 80% was discovered in the 0–45 cm soil layer one year after the application. Amendment of boreal clay soil with a high rate of BC characterized by a moderately alkaline pH, low surface functionalities, and a recalcitrant nature, did not induce such positive impacts that would unambiguously motivate farmers to invest in BC. BC use seems unviable from the farmer's perspective but could play a role in climate change mitigation, as it will likely serve as long-term C storage. © 2018 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). Science of the Total Environment 731 (2020) 138955 ⁎Corresponding author. E-mail address: helena.soinne@luke.fi(H. Soinne). https://doi.org/10.1016/j.scitotenv.2020.138955 0048-9697/© 2018 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). Contents lists available at ScienceDirect Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv 1. Introduction Various environmental and agronomic benefits are pursued by adding biochar (organic biomass carbonized under low or no oxygen conditions) to agricultural soils (Sohi et al., 2010;Atkinson et al., 2010). On a global scale, the option of mitigating climate change by sequestering atmospheric carbon dioxide (CO 2 ) to fairly stable biochar carbon (C) offers major potential (Brassard et al., 2016;Lehmann, 2007;Matovic, 2011;Smith, 2016;Woolf et al., 2010). Compared to uncharred biomass, biochar C mineralizes 10 to 100 times more slowly (Lehmann et al., 2009). Consequently, biochar addition can lead to increased soil C stocks for centuries (Wang et al., 2016). Other potential biochar amendment-induced environmental advantages include improved soil aggregate stability (Lu et al., 2014;Soinne et al., 2014)and increased nutrient retention capacity (Saarnio et al., 2018;Borchard et al., 2012;Laird et al., 2010), both of which decrease therisk of erosion and off-site nutrient transport (Le Bissonnais, 1996;Sollins et al., 1988). To attract farmers to invest in biochar, profits arising from the biochar treatment are required. The impacts of biochar amendments in soil have been extensively reviewed in the literature (Atkinson et al., 2010;Kavitha et al., 2018;Lehmann et al., 2011;Palansooriya et al., 2019;Sohi et al., 2010;Yu et al., 2019;Verheijen et al., 2009). Crop yield increase resulting from applied biochar can be achieved due to increased soil pH (Vaccari et al., 2011;Van Zwieten et al., 2010), improved nutrient use efficiency (Chan et al., 2008), and enhanced soil hydraulic properties (Abel et al., 2013;Laird et al., 2010). However, although the mean effect of biochar amendments on crop yields seems positive (Jeffery et al., 2011), the effect on yield in several experiments has been insignificant or even negative (Borchard et al., 2012, 2014; Jeffery et al., 2011;Tammeorg et al., 2014a;Tammeorg et al., 2014b Nelissen et al., 2015). In a recent global-scale meta-analysis by Jeffery et al. (2017), biochar was shown to stimulate yields in the tropics, probably through liming and fertilization effects, but not in the usually more fertile soils of temperate latitudes. In general, the highest positive agronomic responses can be expected from acid, coarse, and nutrient-poor soils (Liu et al., 2013;Jeffery et al., 2017). Negative impacts of biochar application may occur due to the reduced availability of nutrients e.g. by over-liming or N immobilization (Jeffery et al., 2017;Spokas et al., 2012). In soils with elevated P concentrations, biochar-induced increase in P availability may accelerate the risk of P loss (Saarnio et al., 2018). Furthermore, biochar application has potential of encasing soil microbial activity and diversity by providing favorable microbial habitats and substrates for their metabolic activities (Palansooriya et al., 2019), but on the other hand, biochar may contain toxic compounds that have negative impacts on soil microbes (Gomez et al., 2014). Therefore, both the direct positive and negative effects of biochar on soil biota and subsequent impacts on soil processes need to be considered (Chintala et al., 2014;Lehmann et al., 2011). Besides climatic andsoil conditions, the characteristics of the biochar itself govern potential responses in soil (Novak et al., 2009). The process conditions of the thermal treatment influence the biochar properties (e.g. C content, stability, internal porosity, pH, and surface area and surface activity) to a great extent, but the type of feedstock, namely its C and ash content and composition, also have a major effect on the char quality (Ronsse et al., 2013;Sun et al., 2014;Windeatt et al., 2014;Lee et al., 2013;Zhao et al., 2013). An extensive forestry sector in the boreal climate zone results in a large amount of potential biochar raw material in the form of forest residue (e.g. small trees, branches, and stumps). After clear cutting and thinning, wood material unsuitable for industrial use is either left on site or harvested as bioenergy. The annual gross potential of forest residues in Finland and Sweden is N100 million solid m 3 , while the realistically achievable amount is around 40 million solid m 3 (Routa et al., 2013). From the perspective of C sequestration, forest residue decomposes relatively quickly when left on site. Depending on the diameter of branches, after 100 years of decomposition only 2–16% of the initial branch biomass remains (Repo et al., 2011). When incinerated as bioenergy, the C release into the atmosphere is immediate. Thus, the C sequestration potential of forest residue biochar is remarkable. However, the properties and performance of forest residue-based biochar in the agricultural context need to be verified to motivate farmers to adopt such actions. In this study, a field experiment was established to assess the impacts of forest residue biochar (BC) on a boreal low-productive clay soil. We aimed to find out whether short-term gains in productivity or nonyield-related factors can be achieved by BC to justify recommending the treatment and prompt the interest of farmers and relevant authorities in wider BC use. The effects of BC addition on plant growth, soil structure, fertility (pH and available nutrient concentrations), moisture characteristics, fungal and bacterial biomass and communities, gross microbiological activity, and soil C storage were examined over two consecutive growing seasons after the BC application. In addition, the potential of BC to reduce nutrient export to surface water was investigated under simulated rainfall. 2. Material and methods 2.1. Forest residue biochar BC used in theexperiment was produced bya continuously operated pilot-scale mobile pyrolysis unit (Raussin metalli Ky, Sippola, Finland) in the fall of 2016. The raw material used was chipped forest residue from thinnings and/or clear cuttings containing stems, branches, and small wood, but not stumps. The maximum temperature the material washeated to was about 450°C. The produced BC wasmixed and milled using a large-scale hammer mill, and thereafter packed into 5 large bags of equal weight. After homogenization, the BC dry matter content was 59%. BC was analyzed for pH (H 2 O) and electrical conductivity (EC) in 1:5 solid-to-water paste (Table 1). Oxygen (O) content was calculated as the difference between the initial dry mass and sum of ash (815 °C), C, nitrogen (N), and hydrogen (H) contents determined by Vario MAX CHN Elementar-Analysator. Other elements listed in Table 1 were analyzed after aqua regia extraction with ICP-OES. Exchangeable cations were extracted with neutral ammonium acetate (1 M CH 3 COONH 4, pH 7.0) and the concentrations of Ca, Mg, K, and Na were analyzed from this solution by an ICP-OES (Perkin Elmer Optima 8300). The cation exchange capacity (CEC) was determined as adsorbed NH 4 -N exchanged out by 2M KCl solution. Surface acidic functionalities were Table 1 Chemical characteristics of the forest residue biochar (BC). Variable (unit) Value Variable (unit) Value Variable (unit) Value Variable (unit) Value pH 8.23 K (g kg −1 ) 3.7 Cu (mg kg −1 ) 10 CEC (cmol kg −1 )12 EC (μScm −1 ) 174 Fe (g kg −1 ) 3.0 Ni (mg kg −1 ) 2.7 Ca exc (g kg −1 ) 4.2 Ash, 815 °C (%) 6.3 Mg (g kg −1 ) 1.2 Cr (mg kg −1 ) 2.5 K exc (g kg −1 ) 2.10 C (%) 80 P (g kg −1 ) 0.99 Co (mg kg −1 ) 0.96 Mg exc (g kg −1 ) 0.20 H (%) 3.1 Al (g kg −1 ) 0.62 Cd (mg kg −1 ) 0.89 Na exc (mg kg −1 )20 N (%) 0.8 Mn (mg kg −1 ) 310 As (mg kg −1 ) 0.32 Carboxyls (meq g −1 ) 0.03 O (%) 9.8 Zn (mg kg −1 ) 247 Pb (mg kg −1 )b3.0 Lactones (meq g −1 ) 0.09 Ca (g kg −1 ) 9.6 S (mg kg −1 ) 143 Mo (mg kg −1 )b2.0 Phenols (meq g −1 ) 0.16 2H. Soinne et al. / Science of the Total Environment 731 (2020) 138955 determined by Boehm titration (Boehm et al., 1964) according to Kim et al. (2012) 2.2. X-ray microtomography The internal structure of fourrandomly selected BC particles wasimaged with x-ray computed microtomography. The sample size was c. 1 mm. Imaging of the samples was performed with a Zeiss Xradia MicroXCT-400 (Zeiss, Pleasanton, CA, USA) device. In total, 1600 full 360° projections were acquired for each sample. The exposure time for each projection was 3 s, and the source voltage and current were 40 kV and 250 μA respectively. A 20× objective was used with binning 2, resulting in a pixel size of 1.132 μm. No filters were used in the imaging. The image stacks were reconstructed with the filtered back projection algorithm using Zeiss XMReconstructor software. For quantitative analysis of the micrometer-scale pore structure, the original grayscale images were segmented into a binary representation of pore and solid phases. Image processing was performed in several steps. Grayscale images were denoised by variance-weighted (VaWe) mean filtering (Gonzalez and Woods, 2001). The denoised images were segmented into pores and solids, using the segmentation method developed by Sauvola and Pietikäinen (2000). The segmented binary images were further filtered with a majority filter, and finally, all isolated solid objects with a volume b1000 voxels were removed from the images. The binary image thus obtained was used in the image analysis. Image analysis was used to determine the porosity and pore-size distribution of the imaged BC samples. The pore size distributions of the samples were determined using an approach based on mathematical morphology (Horgan, 1998) by successive application of morphological opening with an increasing structuring element diameter (e.g. Hilpert et al., 2003). The pore size defined by morphological opening is closely related to the stationary distribution of water in pore space. This method is therefore suitable for the present purpose, because it relates to the pore sizes used in the interpretation of soil moisture characteristic curves. 2.3. Field study and soil properties The BC field experiment was conducted at the Natural Resources Institute Finland (Luke) in Jokioinen, southwestern Finland on a clay soil field. Clay soils of the area have typically been classified into (Vertic Luvic) Stagnosols (Lilja et al., 2017). Like the majority of the fields in Finland, the experimental field was artificially drained. The previous yield data and practical experience suggested relatively poor growing conditions for the chosen experimental area. In the top 20 cm layer, clay, silt, andsandcontent was 64%, 21%, and 15% respectively. C content (dry combustion, Leco TruMac CN analyzer) was on average 5.1% in the 0–10 cm soil layer and 4.6% in the 10–20 cmsoil layer. Soil test phosphorus (STP) and other nutrients extracted with acid ammonium acetate (AAAc, pH 4.65; Vuorinen and Mäkitie, 1955) were at least good according to the Finnish guidelines for the status of a soil test. The study area comprised 10 plots of 60 m 2 (6 × 10 m) arranged in five blocks, for which the BC and control treatments were randomized (randomized block design). The BC application was done in the autumn 2016 at the rate of 30 Mg dry matter ha −1 . Spreading of BC was done manually to ensure an even application over the plots. Immediately after the spreading of BC(October 2016), all plots were cultivated to approximately 10–12 cm depth, using the Kongskilde cultivator. In 2017 and 2018, the field was sown for oats and was fertilized (N 90 kg ha −1 , P 10 kg ha −1 ) according to the Finnish guidelines, taking into account the results of the national soil test procedure. In 2017, the thermal growing season (daily mean temperatures consistently above 5 °C) started on May 1 and ended on October 18. The field was sown with oats on May 19 and harvested on September 28. The temperature sum of 2017 was 1200 °C and below the long-term average of 1300 °C (1981–2010, statistics from the Finnish Meteorological Institute), whereas the amount of rain duringthe growingseason (c. 400 mm) exceeded the long-term average (350 mm). In 2018, the growing season started already on April 14. After a very dry spring, oats were sown on May 31 and harvesting was done on September 12. The temperature sum of 2018 was 1700 °C and above the longterm average, whereas the amount of rain during the growing season (c. 300 mm) was below average. The mean daily temperature during 2017 and 2018, and the rain sum of both years, is presented in the Supplementary material (Supplementary material 1). Grain yields were calculated using the mass and moisture content of grains harvested from a 20 m 2 area in each plot. The element contents of seeds were analyzed after HNO 3 digestion (ICP-OES, Perkin Elmer Optima 8300) and N content was analyzed by the Kjeldahl method (Foss Kjeltec TM 8400). The schedule of soil and plant sampling and subsequent analyses are summarized in Table 2. 2.4. Soil analyses Before sowing in 2018, soil samples for an analysis of bulk density and total C (TC) and nitrogen (TN) were taken from each plot in four different layers (0–10, 10–20, 20–30, and 30–45 cm) with a 45.2 mm diameter auger. These samples of known volume were dried at 40 °C and weighed for the bulk density measurement. Thereafter, the samples were ground and analyzed for C and N (Leco TruMac CN). Calculation of soil C stock in the soil profile was based on the measured bulk density and C content. In the fall of 2018 after the harvest, composite soil samples from each plot were collected from 0–10 and 10–20 cm layers for analysis of pH and EC (1:5 H 2 O). In addition, these samples were analyzed for soil test phosphorus (STP) and Ca, Mg, K, Na, and S in accordance with the Finnish national soil test procedure (Acid ammonium acetate, AAAc, extraction, Vuorinen and Mäkitie, 1955). Mehlich3 (Meh3) extraction (Mehlich, 1984) was carried out to analyze P (colorimetric method), Al, and Fe (ICP-OES). 2.5. Soil moisture characteristics In the fall of 2018 after the harvest, undisturbed soil samples were taken in moist soil conditions. First, a 2.5 cm layer of surface soil was removed. A 200 cm 3 steel cylinder (diameter 7 cm) was then gently pressed to a depth of 2.5–7.5 cm (i.e. within the mixing depth of BC). Table 2 Samples taken from the forest residue biochar (BC) field experiment and analyzes performed during the study period 2017–2018. Sampling time Measure, sampling purpose Fall 2016, after harvest BC application Spring 2017, before sowing Monoliths for rainfall simulation Fall 2017, at the time of harvest Yield mass and nutrient contents in the seeds Spring 2018, before sowing Monoliths for rainfall simulation Bulk density and TC, TN down to 45 cm Samples for the aggregate stability test (0–5 cm) Samples for basal respiration and DNA extraction (0–10 cm) Fall 2018, before harvest Samples for basal respiration and DNA extraction (0–10 cm) Samples for microbial biomass and PLFA (0–10 cm) Fall 2018, at the time of harvest Yield mass and nutrient contents of the seeds Fall 2018, after harvest Samples for acid ammonium acetate extraction (AAAc) (0–10 and 10–20 cm) Samples for pH, EC (0–10 and 10–20 cm) Samples for Mehlich3-P,\\Al, and\\Fe (0–10 and 10–20 cm) Soil cores for pF-curve (2.5 cm–7.5 cm) Steady-state infiltration rates 3H. Soinne et al. / Science of the Total Environment 731 (2020) 138955 The cylinders were excavated from the soil, closed with plastic lids, and stored at +4 °C in closed plastic bags to avoid drying. Five replicate samples were taken from each of the 10 plots. Samples were saturated with water for two weeks, and a pressure chamber approach was thereafter used to determine thesoil moisture characteristic curve in suction pressure steps of 3.2, 5.0, 10.0, 15.8, 25.1, 39.8, 100, and 316 kPa. A vapor pressure equilibrium approach with saturated ammonium oxalate and sodium chloride solution was used to determine moisture content at 1500 kPa (permanent wilting point) and 39,000 kPa. Shrinkage of the samples during the drying process was measured both vertically and horizontally in each of the suction pressure steps. Thevertical shrinkage was measured at five locations per sample using a Vernier caliper, and the horizontal shrinkage from four locations using a feeler gauge. The effect of BC application on soil porosity was studied using the soil water characteristic data. The suction pressures were converted to approximate pore sizes using the Young-Laplace equation, and the water content of the samples were interpolated at 5 μm pore intervals using linear interpolation. Finally, the effect of BC on porosity was calculated as the difference between the pore size distribution of samples with BC and control. The calculation procedure is described in detail in Rasa et al. (2018). 2.6. Tension infiltrometry The steady-state infiltration rates of water into the soil were determined in October 2018 with tension infiltrometers manufactured by Soil Measurement Systems LLC (Huntington Beach, CA, USA). The used devices consist of a porous disk (diameter 20 cm) which is separated from the water reservoir and bubble towers. At each measurement site, the vegetation was first cut to soil level, and the soil surface was then gently leveled and smoothed. A thin layer of moist fine sand was used to ensure good contact between the infiltration disk and the soil surface. Measurements were performed at supply pressure heads of −6, −3, and −1 cm in an ascending sequence. The measurements were continued until a steady-state infiltration was reached. The hydraulic conductivity at each supply pressure was calculated using the method described by Ankeny et al. (1991), which is based on Wooding's theory of unconfined three-dimensional infiltration (Wooding, 1968). 2.7. Rainfall simulation and aggregate stability test Undisturbed soil monoliths (height 40 cm, diameter 30 cm) for rainfall simulation (Uusitalo et al., 2012) were collected by a tractor-driven auger (see Persson and Bergström, 1991) from each plot on May 3, 2017 and May 22, 2018. The samples were transported to the laboratory and stored at about +6 °C in the dark for b3weeks. Before rainfall simulation tests, the bottom of the samples was prepared for natural plains of cleavage, and the voids were filled with coarse quartz sand and finally secured in sample holders. The samples were slowly saturated with water one day from the bottom to the level of soil surface. The water level was maintained over two days and then allowed to drain overnight. A stationary drop-former was used in the rainfall simulation tests (Uusitalo and Aura, 2005). The rainfall intensity was adjusted to 5 mm h −1 (deionized water with kinetic energy about 80 J m −2 h −1 ), which is typical intensity for rain in South-West Finland. The samples were subjected to 5 h of rainfall in two subsequent days. Water samples were collected during these rainfall periods and in addition, the percolation water drained overnightbetween days 1 and2 was collected. Consequently,four water samples per each monolith were collected: (i) water draining overnight after the initial saturation, (ii) percolation water obtained during the first rainfall event, (iii) water draining overnight after the first day of rainfall simulations, and (iv) percolation water obtained during the second rainfall event. Each of these water samples were analyzed separately, but the results are presented as their mean values. Turbidity of each water sample was measured immediately (2100AN IS Turbidimeter, Hach Company) followed by EC and pH analyses. A part of the water sample was filtered through 0.2 μm Nuclepore (Whatman) membrane and both filtered and unfiltered fractions were frozen and stored at −18 °C until further analyses. The filtered subsamples were analyzed for dissolved molybdate-reactive phosphorus (DRP), NO 3 -N and NH 4 -N. Following autoclave digestion with peroxodisulphate and sulphuric acid, the unfiltered samples were analyzed for total P and N concentrations. The P and N species were determined using a continuous flow injection analyzer (LaChat autoanalyzer). Before analyses for dissolved organic carbon (DOC) with a Shimadzu TOC analyzer, water samples were passed through Whatman GF/C glass filters. In the spring 2018, soil surface samples (0–5 cm) were taken for aggregate stability tests from each plot at 3 different locations. The samples were kept at sampling moisture and dry sieved to retain an aggregate fraction of 2–5 mm. The sieved aggregates were air-dried and 4 g (dry matter) from each sample was subjected to wet sieving (Heikkinen et al., 2019) with an Eijkelkamp 08.13 apparatus equipped with 0.25-mm mesh size sieves. The aggregates placed on the sieves were pre-moistened for 15 min in cans containing 100 ml of deionized water. The pre-moistening was followed by a 3 min up-anddown run with the wet sieving apparatus. Water-stable aggregates remaining on the sieves were oven-dried (110 °C) and weighed. The proportion of water-stable aggregates (WSA%) was calculated by dividing the weight of the material on the sieves by the total weight of the initial material. The suspension containing soil material that passed through the sieves was transferred to a 100-ml container and left to settle for 21 h. Samples for turbidity measurements were pipetted from the surface of the container, and turbidity (in nephelometric turbidity units, NTU) was determined using a HACH 2100AN IS turbidimeter (Hach Company). The turbidity values were divided by the dry mass of the aggregates taken for the analysis. 2.8. Basal respiration, DNA extraction, bioinformatics, qPCR, and amplicon sequencing In 2018, samples were collected for microbiological analyses. Samples from the surface soil (0–10 cm) of each plot (composite samples comprised of ten subsamples) were collected before sowing in the spring and after the harvest in the fall. After the sampling and prior to the analyses, samples were stored cold (+4 °C). Prior to respiration measurements, the soil samples were kept at 14 °C for two days. The basal respiration rate was determined from fresh soil (20 ml), as the amount of CO 2 evolved during24h incubation and wasmeasured as described by Pietikäinen and Fritze (1995). The DNA of all control and BC-amended plots (n = 5) in both the spring and fall samples was extracted with a NucleoSpin soil kit (Macherey Nagel, Germany) in line with the manufacturer's protocol. For amplicon sequencing, the replicate samples of the DNA extractions from the fall sampling were combined to form one control and one BC-amended sample. These DNA samples were delivered for sequencing at Tartu University's Institute of Genomics. For bacteria, targeting the 16S V4 region of the 16S SSU rRNA gene was amplified in a twostep PCR, using the 16S rRNA 515F and 806R primers (Caporaso et al., 2011, 2012), and for fungi targeting, using an internal transcribed spacer 2 (ITS2) region with ITS4 (White et al., 1990) and gITS7 (Ihrmark et al., 2012) primers with an 8 bp dual index for 24 cycles. The final PCR fragments were run as paired-end 2 × 300 bp with MiSeq platform (Illumina), using MiSeq v3 kit producing about 20–25 M reads per flow cell. Quantitative PCR for the fungal ITS region and bacterial and archaeal 16S rRNA gene was conducted as described in Peltoniemi et al. (2015) for all samples separately. Sequence assembly, quality filtering, removal of artifacts, primerdimers and primers from raw 16S rRNA and ITS2 sequence reads, along with clustering and taxonomical annotations, were conducted 4H. Soinne et al. / Science of the Total Environment 731 (2020) 138955 with a PipeCraft 1.0 pipeline (Anslan et al., 2017). PipeCraft utilizes several implemented tools of e.g. mothur v1.36.1 (Schloss et al., 2009), vsearch v1.11.1, and CD-HIT v4.6 (Fu et al., 2012), which were used in pre-processing, assembling, chimera filtering, and clustering steps. Raw sequence reads were processed according to the manual, with slight modifications for demultiplexed sequence data. Briefly, assembly of paired-end reads and initial quality filtering was conducted with vsearch (v1.11.1; github.com/torognes/vsearch;Rognes et al., 2016)according to the following parameters: minimum overlap 15; max differences 99; minimum length 150 bp; e_max 1; max ambiguous 0 and trunc qual 20 for fungi and 10 for bacteria. On average, 37% of the raw reads were filtered out after the assembly. Chimera filtering was performed for the reoriented reads using vsearch de novo filtering with parameters: annotation 0.97 and abskew 2; and for ITS both referencebased filtering was used with Unite ITS2 ref. v7.1 as a database. Primers and primer artifacts were also filtered out from sequences at this step. In addition, fungal ITS2 region was extracted from reads with ITSx (Bengtsson-Palme et al., 2013). In the next step, sequence reads were clustered and OTU table created with CD-hit with parameters: threshold 0.97; and min size 2. In the last step, bacterial OTUs were taxonomically annotated by searching for representative sequences with BLAST using the reference 16S rRNA (SILVA_123_SSURef_Nr99_tax_silva. fasta) obtained from SILVA (Quast et al., 2012;Yilmaz et al., 2014). For the fungi, the ITS2 database (sh_genral_release_dynamic_01.12.2018. fasta) from UNITE (Nilsson et al., 2018) was used. After the first quality filtering steps, raw sequence data for bacteria consisted of 40,013 reads clustering in 1967 OTUs. For fungi, 47,366 reads clustered in 691 OTUs. Second quality filtering was done based on the results of the alignments: We filtered out bacterial and fungal OTUs that had an e-value higher than e-25, query coverage b90%, and identity b90% with the database match. OTUs with an affiliation other than bacteria or fungi, as well as singleton OTUs and reads with a relative proportion below 0.0001%, were removed from the data. Furthermore, fungal OTUs referring to exactly the same species hypothesis (Kõljalg et al., 2013)were consolidated. Bacterial OTUs were not consolidated. This resulted in 36,195 reads that were affiliated to 889 bacterial OTUs, and 41,259 fungal reads that were affiliated to 296 OTUs. The raw sequence data was deposited in the sequence read archive (SRA) of the NCBI/EMBL database, BioProject PRJNA599338, with the accession numbers SAMN13745490 and SAMN13745497. 2.9. Microbial community composition by PLFA analyses Biomass and phospholipid fatty acid PLFA analyses were performed for samples taken in the fall after the harvest. Prior totheanalyses,samples were stored cold (+4 °C). The phospholipid extraction and analysis of PLFAs were carried out as described by Frostegård et al. (1993). Briefly, 2.5 g of fresh soil was extracted with a chloroform:methanol:citrate buffer mixture (1:2:0.8) and lipids were thereafter separated into neutral lipids, glycolipids, and phospholipids on a silicic acid column. The phospholipids were subjected to mild alkaline methanolysis, and fatty acid methyl esters were detected with a gas chromatograph (GC) using a flame ionization detector and 50-m HP-5 capillary column (see GC-run configuration in Pennanen et al., 1999). In total, 43 different PLFAs were identified from each sample, and they were expressed as a mole percentage (mol % = area % of a single PLFA from the area sum of all identified PLFAs). All the identified PLFAs were used to calculate the total biomass indicator PLFA tot . The following PLFAs were considered to be predominantly of bacterial origin: i15:0; a15:0; 15:0; i16:0; 16:1ω9; 16:1ω7t; i17:0; a17:0; 17:0; cy17:0;18:1ω7;andcy19:0.Theywerechosenasanindexof bacterialbiomass PLFA bact (Frostegård and Bååth, 1996). The amount of 18:2ω6was used as an indicator of fungal biomass, PLFA fung , because it is suggested to be mainly of fungal origin in soil (Federle, 1986) and is known to correlate well with the amount of ergosterol (Frostegård and Bååth, 1996). 2.10. Statistical analyses TheBC treatment induced differences in soil chemical properties and aggregate stability were tested with a one-way analysis of variance for randomized block design using BC treatment as a fixed effect. Turbidity values indicating colloid detachment during the wet sieving did not meet the requirements of analysis of variance (Shapiro-Wilk test for normality) and were analyzed with a nonparametric Wilcoxon ranksum test. In the analyses of yields and nutrient content in the seeds measured in 2017 and 2018, as well as in microbiological properties that were measured twice during the growing season, a two-way analysis of variance was used with BC treatment and the sampling event as fixed effects. Pairwise comparisons between the control and BC treatment in each year were performed using Tukey's test. For the nutrient contents in the seeds that were not normally distributed, a nonparametric Wilcoxon rank-sum test was used. The area% of each PLFA was used to analyze the variation in PLFA composition using principal components analysis (PCA) followed by Analysis of Variance (ANOVA) to test differences in PC scores and PLFA biomass values. Statistical analyses of near-saturated hydraulic conductivity (K unsat ) and soil water holding capacity were based on a repeated measures design with two treatments (BC and control), three supply pressure heads (−6, −3, and, −1 cm) or eleven matric potentials respectively, and five replicates in blocks. Generalized linear mixed models (GLMM) with lognormal (with an identity link) and beta (with a logit link) distributions were used for K unsat and soil water holding capacity respectively, with a treatment and suction pressure head or suction pressure point respectively, and their interaction as fixed effects. The effects of block and treatment within a block were used as random effects. Correlation between supply pressure heads or suction pressure points within treatment in a block was considered in using an unstructured and homogeneous autoregressive covariance structures respectively (Gbur et al., 2012). In the rain simulation experiment, means of pH, nutrient concentration, and turbidity (NTU) in BC-amended and control soils in 2017 and 2018 were compared by GLMMs. Treatment and year, and their interaction, were used as fixedeffects,andtheblockandtheinteraction of block and year as random effects. Correlation between years was taken into account using a heterogeneous or homogeneous compound symmetry covariance structure. The assumptions of gamma (with log link) and lognormal (with identity link) distributions were used for TP and turbidity respectively. Other dependent variables were assumed to be normally distributed. All models were fitted by using the residual pseudo likelihood (TP and soil water holding capacity) or restricted maximum likelihood (others) estimation method. The degrees of freedom were calculated using the Kenward-Roger method (Kenward and Roger, 2009). The normality of residuals was checked using boxplot, and residuals were plotted against the fitted values. Westfall's (1997) method was used for pairwise comparisons of means, with a significance level of α= 0.05. The analyses were performed using the GLIMMIX procedure of the SAS Enterprise Guide 7.15 (SAS Institute Inc., Cary, NC, USA). 3. Results 3.1. Yields and harvest quality There were no differences in yield masses between the control and BC-treated plots. However, yields were significantly lower in the second and much dryer year (Supplementary material 2), but even then, oat growth did not benefit from the BC treatment. The contents of major elements (N, K, P, S, Mg, Ca) in oat seeds were not affected by the BC amendment (Supplementary material 2). For elements in low contents (mg kg −1 and ug kg −1 ), only Ni, Zn, and Mo showed differences between treatments. In the first harvest after the BC addition in 2017, Mo and Zn were significantly higher in seeds 5H. Soinne et al. / Science of the Total Environment 731 (2020) 138955 grown on BC-treated soil. In the second harvest in 2018, the Ni content in oat seeds was significantly lower in BC-treated plots (Supplementary material 2). 3.2. Soil pH, EC, and available nutrients Two years after the BC treatment, soil pH or EC did not differ between control and BC-treated plots (Table 3). In BC-treated plots, STP (acid ammonium acetate extractable P) was higher in the 0–10 cm soil layer than in the control plots, but no differences were detected in the 10–20 cm soil layer. A similar trend was seen in the soil P Meh3 for the surface soil, but the difference between the treatments was not significant (p= 0.0814). Of major cations, only soil test Ca was higher in BC-treated plots. 3.3. Bulk density, carbon, and nitrogen in soil profile Bulk densities measured in the spring of 2018 were lower at the 0–10 cm soil surface layer of BC-treated plots (p=0.0113),butnodifferences were detected in the 10–20 cm layer (Fig. 1). BC addition increased C content in the 0–10 cm layer (p= 0.0051), but below this, no clear differences were detected. The BC treatment showed no effect on soil N content in the two different layers. When measured to afixed soil depth (45 cm), the stock of C in BC-treated plots (163 Mg ha −1 ) was higher (p= 0.0167) than in control plots (144 Mg ha −1 ), but the stock of N did not differ between the treatments, being 11.4 and 11.1 Mg ha −1 in BC and control plots respectively. 3.4. Soil moisture characteristics and infiltration Compared to the control, the BC amendment tended to increase soil water holding capacity up to the suction pressure of 25 kPa (Fig. 2). The lower the suction pressure, the greater the difference tended to be. The differences in soil water holding capacity between the control and samples with BC were not, however, statistically significant at pb0.05 level (p=0.07–0.88), but at the pb0.1 level a significant difference was found at near-saturated conditions (0.25 kPa). The application of BC especially increased the share of pores smaller than a 10-μm diameter (Fig. 2). Another peak was found in pore size range of around 25–30 μm. The near-saturated hydraulic conductivity was not affected by a BC addition at any supply pressure head (−1, −3, and, −6cm p= 0.3745, p= 0.2694, and p= 0.1212 respectively, Fig. 3). 3.5. Porosity of forest residue biochar, X-ray microtomography Visual inspection of 3D images showed, that three of the BC samples had a pore structure typical of deciduous trees, while the structure of one sample was clearly different and was probably from a coniferous tree (Supplementary material 3). Porosity of the samples varied such that the clearly different sample had the lowest porosity (0.37), while for the three other samples, porosities ranged between 0.45 and 0.49. The pore size distributions of the samples are shown in Fig. 4. Three samples had a bimodal pore size distribution, and one sample had unimodal distribution. In the samples with bimodal distribution, the pore diameter of the first maxima (tracheids) was in the range 5–10 μm. The size of the pores leading to the second maxima (vessels) differed, being c. 30–35 μminsampleS1(cf.Fig. 4), Table 3 Soil pH and electrical conductivity (EC, μScm −1 ), Mehlich3-P, -Al, and -Fe (Meh3, mg kg −1 soil) and AAAc(pH4.65) extractable P and cations (Ca, Mg and K) (mg kg −1 soil) at two depths (0–10 cm and 10–20 cm) measured two years after the BC treatment. CI = 95% confidence interval. Statistically significant differences were tested separately for both layers. 0–10 cm 10–20 cm Control CI BC CI pControl CI BC CI p pH 5.7 [5.6, 5.8] 5.8 [5.7, 5.9] 0.337 5.7 [5.7, 5.8] 5.8 [5.7, 5.8] 0.683 EC 99 [89, 109] 97 [87, 107] 0.798 96 [88, 104] 96 [88, 105] 0.964 Al Meh3 1719 [1690, 1747] 1720 [1692, 1749] 0.908 1631 [1595, 1666] 1638 [1602, 1673] 0.726 Fe Meh3 509 [481, 537] 525 [497, 553] 0.318 479 [472, 486] 479 [472, 486] 0.917 P Meh3 57 [46, 68] 70 [58, 81] 0.092 40 [37, 43] 37 [34, 40] 0.173 P AAAc 11.9 [10.8, 12.9] 13.5 [12.5, 14.6] 0.039 10.0 [9.2, 10.8] 9.6 [8.8, 10.4] 0.409 Ca AAAc 2491 [2427, 2555] 2608 [2544, 2672] 0.023 2551 [2482, 2620] 2597 [2529, 2666] 0.257 Mg AAAc 706 [677, 736] 711 [681, 740] 0.784 744 [707, 782] 739 [702, 776] 0.783 K AAAc 787 [737, 836] 849 [800, 899] 0.069 738 [701, 775] 783 [746, 819] 0.075 Fig. 1. Bulk densities (BD, g cm −3 ), C%and N% in control and in BC-treated plots twoyears after the BC application at different depths (mean and 95% confidence intervals). Only BD and C% at the 0–10 cm layer differed significantly between the treatments. 6H. Soinne et al. / Science of the Total Environment 731 (2020) 138955 while in samples S2 and S3, the corresponding diameter range was 20–25 μm. 3.6. Aggregate stabilities and rainfall simulation BC application did not increase WSA%. In fact, the test indicated somewhat lower aggregate stability in BC-amended plots (WSA 79%, confidence interval 77–80%) compared to the control plots (WSA 82%, confidence interval 81-84%) (p= 0.007). It should be notet that for the BC treatment, the mass of soil was lower because of dilution by BC. Calculating WSA% based on the soil mass (BC subtracted from the mass weighted for the tests), there was no difference in the WSA% between the BC-amended and control soils (p= 0.10). Similarly, the detachment of colloidal particles during the wet-sieving procedure suggested that forest residue BC did not increase soil aggregate stability. Turbidity values measured for control and BC-amended plots were 30 and 36 NTU g −1 respectively (p= 0.045). Thus, a somewhat larger number of colloidal Fig. 2. Soil moisture characteristic curves of samples with and without (control) BC (estimated mean and 95% confidence intervals) and the effect of BC on soil porosity. Fig. 3. Near-saturated hydraulic conductivity (K unsat ) in the control and BC-amended plots (mean and 95% confidence intervals) as measured in field in 2018. Fig. 4. Left: Pore size distributions of the four analyzed BC samples (S1-S4). Right: Pore size distribution combining the four individual BC samples. 7H. Soinne et al. / Science of the Total Environment 731 (2020) 138955 size particles were detached from BC-treated soil than from the control soil. The data obtained in rainfall simulations was in line with the aggregate stability tests. In rainfall simulations, the parameters associated with aggregate stability, turbidity, and TP concentration had p-values of 0.167 and 0.059 respectively (Table 4). Hence, BC did not improve the stability of surface soil toward raindrop impact, but TP and the turbidity of percolation water were slightly higher for the monoliths cored from BC-treated plots than those of the control plots in both years. The only positive indication that BC would be beneficial for water protection wasthat the NO 3 -N concentration (and therefore also TN) of percolation water was lower than in water percolated through the control soil, but these effects also lacked statistical significance (Table 4). 3.7. Basal respiration and microbial community composition Measured in the second year of the study, microbiological activity (basal respiration) was higher in the spring in the BC-treated plots than in the control plots, but no differences in respiration were measured later in the fall. There were no differences in fungal ITS, bacterial, or archaeal 16S rRNA gene copy amounts between treatments (Supplementary material 4). The microbial biomass measures obtained from the PLFA analyses did not differ between the treatments (ANOVA) either: We measured 150, 60, and 4 nmol g −1 soil dw for PLFA tot , PLFA bact , and PLFA fung respectively (Table 5). The PCA analysis performed for the area% of each PLFA showed no treatment-related significant differences in the microbial community structure at the pb0.05 level (ANOVA) (Supplementary material 5). There were 12,652 and 21,462 assembled and quality filtered bacterial reads affiliating to 871 and 885 OTUs in samples from control and BC-treated plots respectively. More than half the bacterial reads for both control and BC amendment samples were clustered into Gammaproteobacteria (14%), Alphaproteobacteria (13%), Actinobacteria (9%), Gemmatimonadetes (8%), and Verrucomicrobiae (7%), while 11% remained unidentified at class level. At family level, the largest group of reads remained unidentified (33%), while the largest identified groups were Gemmatimonadaceae (8%), Chthoniobacteraceae (5%), Sphingomonadaceae (5%), Solibacteraceae (Subgroup 3) (3%), Xanthobacteraceae (3%), Intrasporangiaceae (3%), Burkholderiaceae (3%), Chitinophagaceae (2%), and Isosphaeraceae (2%) (Supplementary material 6). No trends toward shifts in bacterial community composition between control and BC-amended plots could be detected. After assembly and quality filtering, there were 22,416 and 21,877 reads affiliated to 287 and 286 fungal OTUs in samples from control and BC-amended plots respectively. More than 60% of the fungal reads for both control and BC-amended samples were clustered into Ascomycota –21% to Basidiomycota, and 0.1% to Glomeromycota. At family level, the largest group of reads were clustered into Thelebolaceae (20%), Piskurozymaceae (13%), Mortierellaceae (10%), and Pyrenomataceae (8%), while 8% remained unidentified (Supplementary material 6). Even at species level, there were no major shifts in fungal communities between control and BC plots (data not shown). Root symbiotic arbuscular mycorrhizal fungi were equally infrequent in both BC and control samples. Sequences from Claroideoglomeraceae, Paraglomeraceae, and Glomeraceae families comprised b0.1% of all reads. However, Glomus sp. Glomeraceae was three times more abundant in BC than in the control. 4. Discussion 4.1. Forest residue biochar and soil productivity Forest residue BC did not affect the oat yields in the two years following the fall application. Similar results have been reported in other short-term biochar field experiments in Northern Europe (Tammeorg et al., 2014b Nelissen et al., 2015;O'Toole et al., 2018), as well as in mesocosm experiments (Borchard et al., 2014). However, when observed, it has been suggested increases in yields result from an increase in pH (Major et al., 2010;Vaccari et al., 2011), increased nutrient availability (Major et al., 2010), and relieved water deficit (Sohi et al., 2010; Jeffery et al., 2011). The major elements in the oatseedswereunaffected by the BC addition. The BC treatment slightly increased the Zn content and decreased the Ni content, but the differences between the treatments were significantly smaller than the variation between years. The increase in Mo content was significant in the first harvest after the BC treatment, supporting the findings of Rondon et al. (2007), which show a higher Mo uptake of non-N-fixing beans growing on a biochar-treated soil. Acidic soils commonly benefit from lime application and an increase in pH. Though BC used in this experiment had pH of 8.4 it did not clearly increase the soil pH and no beneficial effects on yields could have been expected. The relatively high C and clay content of the experimental field indicates high buffering capacity against changes in pH which might explain the lack of effectofBC on pH.Ofthe measured plant available nutrients, BC only increased Ca and P in the surface soil. Major et al. (2010) found maize yields to be higher in biochar-amended plots after the first year and attributed the higher yields to the greater availability of Ca and Mg in soil where biochar was applied. However, it is unlikely that a higher Ca in BC-amended soils would have had the potential to Table 4 Means of pH, turbidity (NTU), and nutrient concentrations (mg l −1 ) (dissolved organic carbon, DOC; ammonium, NH 4 -N; nitrate, NO 3 -N; total nitrogen, TN; dissolved phosphate, DRP; total phosphorus, TP) in drainage water collected in the rain simulation experiment from forest residue biochar (BC)-amended and control soils in 2017 and 2018. TP and turbidity were analyzed with gamma and lognormal distributions respectively. CI = confidence interval. 2017 2018 pvalues of F-test Control CI BC CI Control CI BC CI Treatment Year Interaction pH 7.2 [7.1, 7.3] 7.3 [7.1, 7.4] 7.3 [7.2, 7.4] 7.4 [7.3, 7.5] 0.089 0.077 0.881 DOC 21 [15, 26] 23 [17, 29] 30 [28, 32] 28 [26, 30] 0.845 0.008 0.397 NH 4 0.15 [0.07, 0.24] 0.16 [0.06, 0.26] 0.30 [0.21, 0.38] 0.31 [0.23, 0.39] 0.820 0.009 0.941 NO 3 9.0 [4.6, 13.4] 6.3 [1.5, 11.1] 4.5 [3.3, 5.7] 3.4 [2.2, 4.7] 0.251 0.02 0.53 TN 11 [7.3, 15.2] 8.5 [4.2, 12.8] 7.4 [6.3, 8.4] 6.7 [5.6, 7.8] 0.253 0.036 0.375 DRP 0.06 [0.04, 0.09] 0.08 [0.05, 0.11] 0.16 [0.08, 0.23] 0.12 [0.04, 0.20] 0.725 0.076 0.363 TP 1.4 [1.2, 1.8] 1.8 [1.4, 2.3] 1.5 [1.2, 1.8] 1.8 [1.4, 2.2] 0.059 0.986 0.776 Turbidity 1009 [800, 1271] 1180 [852, 1467] 665 [527, 838] 827 [656, 1043] 0.167 0.019 0.644 Table 5 Total microbial PLFAs (nmol g −1 soil) in control and in forest residue biochar- (BC) treated soils. CI = confidence interval. Control CI BC CI p Total PLFAs 159 [123, 196] 158 [121, 195] 0.955 Fungi 4.1 [3.1, 5.0] 4.2 [3.2, 5.2] 0.800 Bacteria 66 [50, 81] 64 [49, 80] 0.905 B:F ratio 16 [14, 19] 15 [13, 18] 0.562 8H. Soinne et al. / Science of the Total Environment 731 (2020) 138955 increase yields in the current study, because according to the Finnish fertility assessment, the Ca status of the soil was “satisfactory”,andthe Mg status of the soils was “good”or even “high”. Furthermore, no differences were detected in Ca content in oat seeds between the BCamended and control plots. Similarly, STP was “satisfactory”and at a level at which no yield increase due to P fertilization would be expected (Valkama et al., 2011). One of the mechanisms behind the biochar-induced yield increases is the positive effect of biochar on soil water holding capacity (Jeffery et al., 2011). However, the reported effects vary, depending on the soil type, and study design or method. According to Tammeorg et al. (2014a), studies using undisturbed samples from field experiments show mixed results, whereas when repacked soil columns have been used, the results have often shown a biochar-induced increase in water holding capacity. This difference likely originates from more uniform distribution of biochar in experiments conducted in the laboratory whereas in the field a full range of interactions with soil, plants and soil biota takes place and thus, higher biochar application rates may be needed for some of its effects to become evident (Tammeorg et al., 2014a). The pore structure of the forest residue BC revealed potential for water retention, and the porosity was in the same range as reported by Hyväluoma et al. (2018) for various biochars (0.34–0.68), Jeffery et al., 2015 for herbaceous biochar (0.48–0.57) and Berhanu et al. (2018) for wood-based biochar (0.50–0.58). The inherent heterogeneity of forest residue containing an unknown mix of of different tree species is reflected in BC pore structure as a variation. When the pore size distributions of the four imaged samples were combined, the resulting distribution had a single clear peak at a pore diameter of 5–10 μm with a tail of up to 50 μm.This combined distribution suggests that BC derived from heterogeneous feedstock not only affects thewaterretention properties at a certain matric potential value (cf. e.g. Rasa et al., 2018) but at a wider range of potential. However, the water characteristic curve showed no clear difference between the BCamended and control plots, but only a slight increase in the share of pores smaller than 10 μm and pores with a diameter around 25–30 μm occurred. These pore size regimes correspond to the pore size distribution of three of the four analyzed BC samples. This supports the findings suggesting that biochar affects soil moisture characteristics directly via its internal porosity (Rasa et al., 2018). In addition to the micrometer-scale porosity considered here, biochar also contains pyrogenic nanometer-scale porosity. This porosity contributes only minimally to the total porosity (Grayetal.,2014)and,in addition, pores smaller than 200 nm are not able to store plant available water. Therefore, the effect of nanometer-sale pores on water retention is minor. Amending topsoil with biochar is known to reduce soil bulk density at the soil surface (Zhang et al., 2012;Mukherjee and Lal, 2013). The lower bulk densities in the 0–10 cm layer point toward increased porosity in BC-amended plots, but this was not supported by the water retention properties, because the water retention curve of BC-amended plots did not significantly differ from the control plots. Indeed, the total porosity of the BC was close to that of the unamended field soil and therefore the porosity of the mixture of soil and BC would not significantly differ from pure soil. The significantly lower bulk densities probably originate from mixing the lower density material in the clayey soil, with no significant changes in aggregate structure. However, it is reasonable to assert that changes in the soil's physical properties due to BC addition depend on soil type, while BC's capacity to store water in its internal structure is immutable (as long as BC quality remains the same). 4.2. Forest residue biochar and surface water quality Low permeability of surface soil increases the risk of surface runoff and losses of nutrients to surface waters. BC addition has been reported to reduce the saturated hydraulic conductivity in sandy soil, whereas in clay soils, conductivity increases (Barnes et al., 2014). It is suggested that the increase in water infiltration of clayey soils originates in the lower bulk density resulting from increased porosity in BC-treated soils (Barnes et al., 2014). BC reduced the surface soil bulk density of the BC-treated plots in the current study, but near-saturated hydraulic conductivity did not significantly increase, suggesting that the BC addition's potential to reduce the risk of surface runoff was not great. The aggregate stability of the clay soil in the field experiment was not improved over the 1.5-year study period, which is in line with previous studies that also failed to find positive effects of biochar amendments on aggregate stability (O'Toole et al., 2018;Heikkinen et al., 2019;Zhang et al., 2015). The relatively high charring degree of BC makes biochar a chemically inactive material, providing only a few sites for cation bridging and ligand exchange (Heikkinen et al., 2019), known to be important abiotic mechanisms for aggregate formation (Six et al., 2004). Furthermore, BC did not significantly affect soil bacteria or fungi, and thus, the widely reported microbe-induced increase in aggregate stability (Six et al., 2004;Cosentino et al., 2006) could not be expected. Instead, the detachment of colloid-size particles during wet sieving increased in samples from the BC-treated plots. The export of TP was shown to increase slightly after the BC addition, and as there was no change in the leaching of DRP, these results point toward the increased mobility of particle-bound P. The results suggest that BC does not have the potential to reduce erosion and the loss of particle bound Ptosurfacewaters. Variable results on the effectof biochar on P solubility and on the risk of P loss to surface waters has been reported (Novak et al., 2009;Saarnio et al., 2018). BC has been suggested to act as sorptive surface for P (Novak et al., 2009)whereasSaarnio et al. (2018) have reported a biochar-induced increase in DRP concentration and loss in surface runoff. The increased P solubility in biochar-amended soils probably results from an increase in soil pH due to the biochar addition (Soinne et al., 2014;Saarnio et al., 2018) or from water-soluble P included in the biochar (Schnell et al., 2012). Although the pH increase was not significant in this field experiment, the solubility of P (P AAAc )increasedintheBCtreated plots. However, the increased solubility of P did not enhance theleaching of DRP in the rain simulation experiment. In alysimeter experiment by Novak et al. (2009), the BC application increased P solubility (Mehlich1-P) in soil, but the P concentrations in leaching water decreasedin the BC-treated plots. The effect of BC on DRP export is likely to differ, dependingon the transportation route. Increasedsolubility of P in the surface soil originating from anincrease in pH increases the risk of greater P loss through surface runoff. However, while water percolates through the soil profile toward artificial drains, the P solubilized from the surface soil is more likely to be sorbed into mineral or BC particles. However, the ability of biochar to participate in sorption reactions strongly depends on the raw material and pyrolysis process (Heikkinen et al., 2019). In contrast with previous findings (Barnes et al., 2014;Bu et al., 2017;Saarnio et al., 2018;Borchard et al., 2019), the rain simulation experiments showed no changes in N leaching in two sequential springs after a single application of the BC in the fall. It has been suggested that the ability of biochar to reduce N leachingoriginates in the sorption of NH 4 + and/or NO 3 − into BC particles (Wanget al., 2015) or increased biological activity leading to the immobilization of N (Bruun et al., 2012). Indeed, the respiration measurement showed higher microbiological activity in the BC-amended soils in the spring, which may suggest increased nitrogen immobilization. However, the higher respiration is also related to the increased mineralization of natural organic C and N, which could lead to increase in dissolved N (NH 4 + ,NO 3 - ) concentrations in soil water. Possible effects of BC on N leaching might have been masked by large variation in soil monoliths and mobilization of soluble N below the uppermost 10 cm of soil where the BC was mixed. Thus, BC mixed in the soil surface did not significantly reduce the N concentrations in percolating water, but this experimental setup does not allow 9H. Soinne et al. / Science of the Total Environment 731 (2020) 138955