1 Introduction

Deforestation is a major factor that contributes to land degradation and soil loss within riparian zones with consequences for biodiversity loss, water quality deterioration, and soil nutrient limitations, particularly depletion of soil nitrogen (N) (Ye et al. 2012; Hale et al. 2018). Revegetation of riparian zones is a best management practice for reversing soil degradation, increasing soil N, and reducing sediment and nutrient delivery to waterways (Chen et al. 2012; Li et al. 2012; Olley et al. 2015). However, competitive biological and biochemical interactions between target restorer plants and weeds have the potential to hinder riparian revegetation projects (Bottrill et al. 2020). This may lead to revegetation failure, which can be costly (Schirmer and Field 2014). Therefore, it is essential to control the competing weed species to increase resource availability, ensure plant survival, and optimise the growth of the target revegetation plants (Campoe et al. 2014).

Nitrogen is a growth-limiting nutrient and plays an important role in plant productivity (Atkinson et al. 2010; Bai et al. 2012a, b). Mineralisation of organic N produces inorganic N, mainly as ammonium (NH4+-N) and nitrate (NO3−-N), which are the plant-available forms of N. A recent meta-analysis showed that revegetation increases soil N pools within riparian zones (Omidvar et al. 2021). However, N mineralisation depends on many environmental factors including soil types, vegetation types, environmental factors, and land management practices (Broadbent 1984; Groot and Houba 1995; Williams et al. 2007; Ros et al. 2011; Omidvar et al. 2021). Both short- and long-term soil N indicators must be assessed to measure the success of revegetation, including inorganic N concentrations and potentially mineralisable N (PMN) (Bai et al. 2014). Optimal land management practices need to sustain both short- and long-term N availability within riparian soils for revegetating plants. PMN is an indicator of biologically active soil N that is responsive to changes in land management practices (Mahal et al. 2019), which has a strong influence on the soil microbial diversity and composition (Wang et al. 2018).

Soil microbial composition (e.g. bacteria and fungi) and biomass are important measurable soil biotic components that are highly responsive to changes in land management practices (Falkowski et al. 2008; Urbanova et al. 2015; Nguyen et al. 2018). Land management practices typically alter the soil environment including soil pH, soil water content, and soil organic matter, which lead to shifts in soil microbial composition and abundance (Jangid et al. 2008; Lauber et al. 2008; Rousk et al. 2010). Herbicides used for weed control, and their residues, can be toxic to soil-borne microbes (Bai and Ogbourne 2016), and repeated applications may eventually alter soil microbial community (Helander et al. 2018).

Soil microbial biomass carbon (MBC) and N (MBN) are used as an early indicator of soil microbial community changes in response to land management practice and/or external environmental factors (Jordan et al. 1995; Trasar-Cepeda et al. 1998; Bending et al. 2004; Romaniuk et al. 2011). Soil MBC and MBN alteration may subsequently influence soil organic carbon (SOC) turnover and soil N availability (Yang et al. 2010). As an alternative to herbicides, an organically derived mulch such as wood chips can be applied to control weeds, which increases soil organic matter (SOM), preserves soil water, and reduces soil temperature fluctuations (Huang et al. 2008; Murungu et al. 2011; Bai et al. 2014). Mulch application may also improve both the physical and chemical properties of the soil (Huang et al. 2008) and increase soil N retention (Bai et al. 2014) due to high organic matter inputs. Although physiochemical changes can increase soil microbial biomass (MB) (Bai et al. 2014), the extent to which mulch or other weed control methods can alter the soil microbial composition, abundance, and biomass is largely unknown.

Chemical application is a globally utilised and relatively cheap weed control option and glyphosate is one of the most commonly used post-emergence, non-selective, and broad-spectrum herbicides (Bai and Ogbourne 2016). Glyphosate is applied on leaves but can exude through the roots within 24 h of application (Laitinen et al. 2007). Literature describing the potential glyphosate toxicity on soil microbial communities and diversity due to the compound’s mode of action remains contradictory. Some studies report that a single application of glyphosate changes soil microbial community structure (Puertolas et al. 2010; Banks et al. 2014). However, in a recent short-term field study, no significant impacts of glyphosate on soil microbial diversity and community structure were observed (Bottrill et al. 2020). Glyphosate may be directly toxic to soil microorganisms due to reduced amino acid biosynthesis via inhibition of the shikimate synthesis pathway (Anza et al. 2016; Nguyen et al. 2016) and initial longer-term studies have indicated some impacts on soil microbial communities (Bai and Ogbourne 2016; Lancaster et al. 2009; de Andrea et al. 2003). The half-life of glyphosate is reported to be within a few days to a few weeks (0.8–151 days) (Bai and Ogbourne 2016). However, the frequency of application may prolong the presence of glyphosate in soil (Primost et al. 2017) and later may have adverse consequences on the microbial community and abundance (Helander et al. 2018). Furthermore, the degradation of glyphosate may be delayed due to different environmental factors and soil properties (Bai and Ogbourne 2016). The impact of glyphosate on soil microbes is likely to be dependent on many factors including dosage, frequency of application, and soil characteristics (Bai et al. 2012a, b; Nguyen et al. 2016). Alternative organic chemical herbicides exist in the commercial retail marketplace and include the registered products BioWeed™, Slasher® (pelargonic acid), and horticultural vinegar (acetic acid). Currently, it is unknown what effects, if any, of these organic preparations have on the soil microbiota. This study aimed to assess and compare impacts of different weed control methods on (a) soil N dynamics, (b) soil microbial biomass C (MBC) and N (MBN), and (c) soil fungal and bacterial community composition in revegetated riparian zones over 2 years following repeated application of glyphosate and organic-based herbicides and 1 year after ceasing herbicide application (in total up to 3 years following revegetation establishment).

2 Materials and methods

2.1 Study sites description

The experimental sites were located in the Mary River basin in South East Queensland, Australia (map provided in Fig. 1a, b). These were the (a) Kandanga site (26° 23′ 14.22″ S, 152° 42′ 9.74″ E) and (b) Pinbarren site (26° 18′ 11.92″ S, 152° 51′ 21.39″ E), approximately 150 km and 170 km, north of Brisbane, respectively. Both the Kandanga and Pinbarren sites were bare land with some coverage of patchy weeds before revegetation. The soil type was loamy clay and brown sandy loam in the Kandanga and Pinbarren sites respectively. Temperature and rainfall data were collected from the nearest weather station to each site, being Gympie (Kandanga) and Traveston (Pinbarren) (Bureau of Meteorology 2020). The rainfall and average temperature at the Kandanga site during the study (August 2016 to August 2018) were approximately 1690.3 mm and 27.8 °C, respectively. Similarly, the rainfall and average temperature at the Pinbarren site during the study (September 2017 to September 2020) were approximately 2031.8 mm and 25.8 °C, respectively (Bureau of Meteorology 2020) (Fig. S1a, b). Kandanga site experienced a prolonged period of low rainfall after establishment, which significantly affected plant survival. Therefore, the site was not maintained or monitored after year 2. Tree survival rate was on average 25.8% by year 2 at the Kandanga and 58% by year 3 at the Pinbarren. Average tree height was 1228.50 mm ranging between 1072.74 and 1489.78 mm in the Kandanga site and 3871.52 mm ranging between 3410.20 and 4285.18 mm in the Pinbarren site by the end of the experiment.

Fig. 1
figure 1

Location of the experimental sites and treatment layout at (a) Kandanga and at (b) Pinbarren in the Mary River basin in South East Queensland

Full size image

2.2 Experimental design and treatments

Each experimental site was established as a randomised complete block design with five subplot replicates. The Kandanga site was established in 2016 and the Pinbarren site was established in 2017. Each site was flat or very close to flat and was divided into five blocks and then each block was separated into five plots, with 2 m distance between treatments to minimise cross contamination (30 m × 2 m). Each plot included a row of 20 different native subtropical trees, planted in the centre of the 2 m wide rows, 1.5 m apart) (Bottrill et al. 2020). The native plants were selected from the regional ecosystem as per Bottrill et al. (2020). The plant species included in this study are provided in Table S1. A slow-release fertilizer (10 g pelletised Agriform - total 20% N) was placed in all planting holes at the time of planting.

Treatments including chemical and organic-based herbicides and mulch were applied within each plot as a spot application around the base of each plant in a circular area with a radius of 0.5 m from the base of each plant. Treatments included Roundup® (glyphosate) at the rate of 10 mL/L, BioWeed™ (derived from pine oil) at 200 mL/L, Slasher® (pelargonic acid) at 700 mL/L, and acetic acid 90% at 125 mL/L, and wood chip mulch (Table S2). At the Kandanga site, the liquid herbicide treatments were applied by a hand sprayer directly onto the ground to cover weeds at the time of planting (month 0) followed by six further applications of each herbicide at months 4, 7, 10, 12, 17, and 22 following revegetation establishment (Table S3). At the Pinbarren site, two pre-establishment sprays were applied, 1 to 2 months before planting directly onto the ground cover weeds and one at the time of planting (with the exception of Roundup®, which was only applied 2 months prior to planting) followed by six maintenance applications with all of the herbicides at 1, 2, 6, 12, 16, and 20 months following revegetation establishment. Mulch was applied only once at the time of planting at both the Kandanga and the PinbarrenSoil sample collection (Table S3).

Soil samples were collected three times following revegetation at each site; at months 2 (August 2016), 14 (August 2017), and 26 (August 2018) at the Kandanga site and months 10 (September 2018), 22 (September 2019), and 34 (September 2020) at the Pinbarren site (Table S3). All soil samples were collected within 2 to 4 months of herbicide application with the exception of the last sampling at the Pinbarren site (months 34), which occurred 12 months after herbicide application. Three soil cores at 0–10 cm depth, 20 cm from the seedling stems were collected from each plot using an augur (60-mm internal diameter) and were combined to provide one sample per plot. All samples were kept on an ice container during transportation to the laboratory. A sub-sample of approximately 20 g of the soil from each plot was stored at − 20 °C for genomics analyses and the remainder was air-dried and passed through a 2-mm sieve for chemical analysis.

2.3 Soil chemical analyses

Soil pH was measured using deionised water with a 1:5 (soil:water) ratio. Soil microbial biomass carbon (MBC) and soil microbial biomass N (MBN) were determined using chloroform fumigation extraction (Jenkinson and Powlson 1976) and the extracts were then analysed using a Shimadzu total organic carbon (TOC) and total nitrogen (TN) analyser (Shimadzu Corp., Kyoto, Japan) (Chen and Xu 2005; Bai et al. 2012a, b). MBC and MBN were derived from the equations as described in Vance et al. (1987) and Brookes et al. (1985), respectively. The K2SO4 extractions from the NF samples were also used to determine dissolved organic carbon (DOC) and dissolved nitrogen (DN).

To measure potentially mineralisable N (PMN), three sub-samples (8 g) from each air-dried sample were weighed. Two sub-samples were incubated at 30 °C for 7 and 14 days. After incubation, 40 mL of 2 M KCl was added to the samples, and the suspension was shaken with an end-over-end shaker for 60 min, centrifuged for 20 min at 2000 rpm followed by filtration through a Whatman No. 42 filter paper. The third sub-sample of soil (namely day 0) was added to 40 mL of 2 M KCl processed as above, without incubation. The 2 M KCl extraction was used to determine the concentration of mineral-N using a SmartChem 200, Discrete Chemistry Analyser (DCA). PMN was measured as the difference between the mineral-N concentration before and after incubation. The KCl extractions from day 0 (non-incubated) were also used to determine NH4+-N and NO3−-N concentrations using a Shimadzu TOC/TN analyser (Shimadzu Corp., Kyoto, Japan).

2.4 Soil DNA extraction

Soil DNA was extracted from 0.25 g soil samples using a Tiangen Soil DNA Extraction Kit (Tiangen, Beijing, China) according to the manufacturer’s instruction. Extracted DNA was visualised on a 1% agarose gel and quantified with a Qubit 2.0 Fluorometer (Invitrogen, Carlsbad, CA, USA). Soil DNA samples were extracted for each treatment from each of the five blocks. The five DNA samples from each treatment were pooled to constitute one composite sample per treatment at each sampling occasion. In total, 10 composite samples from the Kandanga site (collected at months 14 and 26) and 15 composite samples from the Pinbarren site (collected at months 10, 22, and 34) were analysed.

2.5 Next-generation sequencing (NGS) library preparation and Illumina MiSeq sequencing

NGS library preparation and Illumina MiSeq sequencing were conducted by GENEWIZ® (GENEWIZ, Biotechnology Co., Ltd, Suzhou, China) according to standard protocols (Huang et al. 2019). In brief, a total of 30–50 ng of each DNA sample was used to generate amplicons using a MetaVx™ Library Preparation kit (GENEWIZ, Inc., South Plainfield, NJ, USA) (Han et al. 2018). The V3 and V4 hypervariable regions of prokaryotic 16S rDNA were selected and amplified using the forward primer 5′CCTACGGRRBGCASCAGKVRVGAAT-3′ and the reverse primer 5′ GGACTACNVGGGTWTCTAATCC-3′ (You et al. 2016). For the fungal community, 50–100 ng DNA was used to generate amplicons using a panel of primers designed by Genewiz (Huang et al. 2019). Oligonucleotide primers were designed to anneal to the relatively conserved sequences spanning fungal ITS regions (Huang et al. 2019). The hypervariable regions of ITS2 were selected and amplified using the forward primer sequence 5′GTGAATCATCGARTC3′ and the reverse primer was 5′-TCCTCCGCTTATTGAT-3′ (Huang et al. 2019; Cai et al. 2018). At the same time, a linker with index was added to the ends of the PCR product of 16S rDNA amplicons to generate indexed libraries ready for downstream NGS sequencing on the Illumina Miseq platform (Qiao et al. 2018). DNA libraries were validated using an Agilent 2100 bioanalyzer (Agilent Technologies, Palo Alto, CA), and quantified with a Qubit 3.0 Fluorometer (Applied Biosystems, Carlsbad, CA, USA), and then multiplexed and loaded on an Illumina MiSeq instrument according to the manufacturer’s instructions (Illumina, San Diego, CA, USA) (Cai et al. 2018; Fu et al. 2016). Next-generation sequencing was performed (paired-end, 2 × 300 bp) on the MiSeq instrument (Fu et al. 2016; Cai et al. 2018; Huang et al. 2019).

2.6 Statistical analyses

Two-way analysis of variance (ANOVA) tests was conducted where treatments and sampling times were assumed as the main effects, for soil pH, DOC, DN, NH4+-N, NO3−-N, PMN, MBC, and MBN. Where interaction between treatment and sampling time was found to be significant, a series of one-way ANOVA tests were performed to explore differences among specific factors at each sampling time. Stepwise regression using a linear model was developed to ascertain which soil variables explained Shannon index and soil chemical properties. SPSS Statistics (IBM, version 26) was used for all statistical analyses.

The Quantitative Insights into Microbial Ecology (QIIME) package was used for 16S rRNA and ITS rRNA data analysis whereby any low-quality reads were filtered out, and any sequences that did not fulfil the following criteria were discarded: sequence length < 200 bp, no ambiguous bases, and no mean quality scores ≤ 20 (Han et al. 2018; Huang et al. 2019). The sequences were compared with the reference database (RDP Gold database) using the UCHIME algorithm to detect and remove chimeric sequences (Han et al. 2018). The effective remaining sequences were used in the final analysis. The operational taxonomic units (OTUs) with 97% similarity were grouped using the clustering program VSEARCH (v.1.9.6) against the SILVA119 and UNITE ITS database (https:/unite.ut.ee/) (Han et al. 2018; Huang et al. 2019). The Ribosomal Database Program (RDP) classifier was applied to assign a taxonomic category to all OTUs at a confidence threshold of 0.8 (Han et al. 2018). The bacterial and fungal alpha diversity indices were calculated in QIIME from rarefied samples using the Shannon index for diversity analysis and the ACE and Chao1 indices for richness analysis (Cai et al. 2018; Han et al. 2018; Huang et al. 2019). Non-metric multidimensional scaling (NMDS) was performed to visualise similarities among bacterial and fungal communities using R software (v.3.3.1). Heatmap analysis was performed using MeV (v.4.2) to visualise similarities within and between the bacterial and fungal species under different weed control methods. A bipartite interaction network using the R package was created to visualise the bacterial and fungal communities in the different treatments (Dormann et al. 2009).

3 Results

3.1 Effect of different treatments on soil pH, C, and N dynamics

During the term of the study and across all treatments, soil pH ranged from 4.8 to 6.2 at the Kandanga site and from 5.2 to 6.4 at the Pinbarren site (Table 1). Soil pH did not differ significantly among the treatments at any sampling time at the Kandanga site. However, at the Pinbarren site, soil pH was significantly higher in the mulch treatment group compared with the other herbicide treatment groups at months 10, 22, and 34 after revegetation, with the exception of the acetic acid treatment group at month 34 (Table 1).

Table 1 Soil pH, DOC, and DN in response to different weed control methods at months 2, 14, and 26 at the Kandanga site and at months 10, 22, and 34 at the Pinbarren site following revegetation establishment
Full size table

At the Kandanga site, at month 14, soil DOC was significantly higher in the mulch treatment group compared with the acetic acid treatment group (Table 1). However, at month 26, soil DOC was significantly higher in the mulch treatment group compared with all other treatments, with the exception of the Slasher® treatment group where soil DOC did not differ from the mulch (Table 1). Soil DOC ranged from 42.1 to 103.7 μg g−1 during the period of study at the Kandanga site (Table 1). At the Pinbarren site, at month 10, soil DOC was significantly higher in the mulch treatment group compared with all other treatments (Table 1). Soil DOC ranged from 77.1 to 169.8 μg g−1 over the study and across all treatments at the Pinbarren site (Table 1).

Soil DN was not significantly different following any treatment at any sampling time at the Kandanga site, which ranged from 54.6 to 158.4 μg g−1 over the study period (Table 1). At the Pinbarren site, soil DN ranged from 61.5 to 177.7 μg g−1 during the study period and was significantly higher in the mulch treatment group compared with all other treatment groups only at month 34 (Table 1).

Neither soil NH4+-N nor NO3−-N concentrations were significantly different among treatments at the Kandanga site regardless of sampling time (Table 2). However, at the Pinbarren site, at month 22, treatment with mulch significantly lowered soil NH4+-N compared with all other treatments (Table 2). Also, NO3−-N was the dominant inorganic N in the mulch treatment group at the Pinbarren site at months 22 and 34, at a level of 78.5 and 72.4 mg kg−1, respectively (Table 2). Furthermore, at the Pinbarren site, at months 22 and 34, soil NO3−-N was significantly higher in the mulch treatment group compared to the Roundup® treatment group (Table 2).

Table 2 Soil NH4+-N, NO3−-N, PMN-7 days, and PMN-14 days in response to different weed control methods at months 2, 14, and 26 at the Kandanga site and at months 10, 22, and 34 at the Pinbarren site following revegetation establishment
Full size table

Total N mineralised in days 7 and 14 of incubation (PMN-7 days and PMN-14 days) was not significantly different between the treatment groups regardless of sampling times at the Kandanga site (Table 2). However, at the Pinbarren site, at month 22, soil PMN-7 days and PMN-14 days were significantly higher in the Roundup® treatment group compared with the acetic acid and BioWeed™ treatment groups (Table 2).

At the Kandanga site, MBC was not significantly different among treatments at months 2 and 14 following revegetation (Fig. 2a). However, by month 26, MBC was significantly higher in the mulch treatment group compared to the Roundup® or organic-based herbicide groups (Fig. 2a). Similarly, at the Pinbarren site, MBC was significantly higher in the mulch treatment group compared to the Roundup® or organic-based herbicide treatment groups at month 10. However, at month 34 at the Pinbarren site, soil MBC was only significantly higher in the mulch treatment group compared with the Roundup® and BioWeed™ treatment groups (Fig. 2b). Soil MBN did not differ significantly among the treatments at all sampling times at both Kandanga and Pinbarren sites (Fig. 2c, d).

Fig. 2
figure 2

Soil microbial biomass C and N (MBC and MBN) for each treatment at the Kandanga site (a, c) at months 2, 14, and 26 and at the Pinbarren site (b, d) at months 10, 22, and 34 following revegetation establishment. Different lower-case letters at each bar indicate significant differences among treatments at P < 0.05. No letters indicate no significant difference among treatments

Full size image

3.2 Soil bacterial and fungal community composition

A total of 5,345,488 high-quality sequences, in a range of 101,676–297,692 sequences per sample, were detected from the 16S rDNA gene region. These clustered into 6272 OTUs at a 97% sequence similarity threshold. Taxonomic classification indicated these were representative of 28 bacterial phyla, comprising 530 genera and 583 species. Of these, the dominant bacterial phyla were Proteobacteria (38.16%), Actinobacteria (21.42%), and Acidobacteria (19.64%) (Fig. 3). The bacterial Shannon diversity index ranged from 9.53 to 10.16 (Table 3). Non-metric multidimensional scaling (NMDS) ordinations based on the Bray–Curtis similarity matrices showed no segregation of treatments in the community structure of bacteria at the Kandanga site at either sampling time (Fig. 4a). However, clear segregation of the bacterial community structure was evident between the mulch treatment group and the other treatment groups at the Pinbarren site, at months 10, 22, and 34 (Fig. 4b).

Fig. 3
figure 3

Soil bacterial community structure at the phylum level at the Kandanga site at months 14 and 26 and at the Pinbarren site at months 10, 22, and 34 following revegetation establishment. Bacteria phylum identified are shown as individual colors and listed in the legend

Full size image
Table 3 Alpha diversity statistics of soil bacteria and fungi in response to different weed control methods at the Kandanga and Pinbarren sites
Full size table
Fig. 4
figure 4

NMDS ordinations derived from the Bray–Curtis dissimilarity matrices indicating the changes in the community structure of soil bacteria in different weed control methods at the (a) Kandanga and (b) Pinbarren sites, and soil fungi at the (c) Kandanga and (d) Pinbarren sites. Treatments including mulch (circle symbols), BioWeed™ (square symbol), Slasher® (diamond symbols), Roundup® (right-side-up triangle symbols), acetic acid (upside-down triangle symbols) at the Kandanga site at months 14 (green) and 26 (purple), and the Pinbarren site at months 10 (grey), 22 (brown), and 34 (blue) following revegetation establishment. Treatments from the same sampling time were marked in the same color

Full size image

A total of 17 fungal phyla, comprising 310 genera and 532 species, were identified from hypervariable regions of ITS2. The three most abundant fungal phyla identified were Ascomycota (40.13%), Basidiomycota (34.25%), and unclassified fungi (11.87%) (Fig. 5). Interestingly, at months 34 at the Pinbarren site, the fungal community shifted in composition from Basidiomycota (41.28%) towards Ascomycota (15.98%) domination (Fig. 5). The fungal Shannon diversity index ranged from 5.34 to 7.49 (Table 3). Similarly, the NMDS ordinations based on the Bray–Curtis similarity matrices showed that there was clear segregation of the fungal community structure between the mulch treatment group and the other treatment groups at both Kandanga and Pinbarren sites at all sampling times (Fig. 4c, d). Soil pH and DN explained 44% and 25% variations of Shannon index bacteria and Shannon index fungi, respectively (Table 4).

Fig. 5
figure 5

Soil fungal community structure at the phylum level at the Kandanga site at months 14 and 26 and the Pinbarren site at months 10, 22, and 34 following revegetation establishment. Fungal phylum identified are shown as individual colors and listed in the legend

Full size image
Table 4 Stepwise regression to identify factors explaining Shannon index of bacteria and fungi, NH4+-N, NO3−-N, MBC, and MBN using soil variables as independent parameters
Full size table

At the bacterial species level, C. Solibacter, unclassified Acidothermus, Bacillus, and uncultured bacterium were most abundant at both the Kandanga and Pinbarren sites after all the treatments, regardless of the sampling times (Fig. 6). The bacterial-focussed heatmap indicated that unclassified Bradyrhizobium, Mycobacterium, and Acidibacter bacterial species were more abundant in the mulch-treated samples compared with the other treatments at months 22 and 34 at the Pinbarren site (Fig. 6). Similarly, there was a lower abundance of C. solibacter and C. koribacter in the mulch treatment group compared with the other treatment groups including at both sites and at all sampling times (Fig. 6). The 30 most abundant fungal species detected among all samples were plotted in the heatmap with the most abundant being Mortierella Elongata at the Kandanga site and unclassified Mortierella at the Pinbarren site (Fig. 7).

Fig. 6
figure 6

Heatmap of bacterial distribution of the top 30 abundant species present in the microbial community of samples. The heatmap plot depicted the relative abundance of bacteria in the soil after different treatments at the Kandanga site at months 14 and 26 and the Pinbarren site at months 10, 22, and 34 following revegetation establishment. The relative value for bacterial species is indicated by color intensity

Full size image
Fig. 7
figure 7

Heatmap of fungal distribution of the top 30 abundant species present in the microbial community of samples. The heatmap plot depicted the relative abundance of fungi in the soil after different treatments at the Kandanga site at months 14 and 26 and at the Pinbarren site at months 10, 22, and 34 following revegetation establishment. The relative value for fungal species is indicated by color intensity

Full size image

The bacterial co-occurrence patterns and network analysis at the phylum level showed no major differences in the communities among the treatments at either site. However, Proteobacteria showed higher abundance with all treatments at month 26 at the Kandanga site and at month 34 at the Pinbarren site (Fig. S2a, b) compared to the earlier sampling times. The soil fungal phyla Mortierellomycota and unclassified fungi increased in abundance at month 34 in Pinbarren in all treatments (Fig. S3a, b) compared with earlier sampling times.

4 Discussion

Soil microbial biomass carbon (MBC) was significantly lower in soil samples after treatment with Roundup® and BioWeed™ compared with mulch, 12 months after treatments had ceased (month 34 at Pinbarren). In general, an attainable soil MBC limit for normal functioning of soil ecosystems in forest soil is 520 mg kg−1 and soils with MBC as low as 160 mg kg−1 have soil constraints (Gonzalez-Quiñones et al. 2011). None of the soils treated with mulch and herbicides groups had a MBC level lower than 160 mg kg−1 up to 2 years in this study. However, 12 months after cessation of herbicide application at Pinbarren (month 34), soil MBC was lower than 160 mg kg−1 in the Bioweed™ and Roundup® treatment groups, and therefore MBC could be a soil constraining factor for plant growth at this site in the longer term. The relatively high soil MBC in the mulch group was likely the result of increased SOM inputs and enhanced soil moisture retention in the soil layer (Li et al. 2004; Grigg et al. 2006; Bai et al. 2014; Kader et al. 2016). No effect of herbicide was found on soil MBN.

Inorganic N concentrations were responsive to the treatments in the Pinbarren site, where soil NO3−-N was significantly higher in the mulch treatment group compared to the Roundup® treatment group at months 22 and 34, but not in comparison to the organic-based herbicides. Decreased denitrifying bacteria is one of the driving factors to increase soil NO3−-N. Both C. solibacter and C. koribacter are denitrifying bacteria involved in N cycling and are associated with the reduction of nitrate, nitrite, and possibly nitric oxide (Ward et al. 2009). We observed a lower abundance of C. solibacter and C. koribacter in the mulch treatment group compared with the other treatment groups at the Pinbarren site, which could partly explain the higher NO3−-N through decreased denitrification in the mulch treatment group compared with the Roundup® treatment group. However, there were no differences in soil NO3−-N between the mulch treatment group and the other organic-based herbicides despite the fact that C. solibacter and C. koribacter were still lower in the mulch treatment than those other organic-based herbicides. Differences in soil NO3−-N were only observed between the mulch and Roundup® treatment groups, which may suggest that glyphosate may have affected soil nitrification and denitrification differently from the other organic-based herbicides.

Interestingly, segregation of bacterial and fungal communities in soil treated with herbicide compared with mulch was still observable 12 months after ceasing herbicide application at the Pinbarren site. Bacteria and fungi are sensitive to changes in soil physical and chemical properties and different environmental conditions such as substrate availabilities, soil moisture content, climate conditions, and management practices (Dong et al. 2017; Zhao et al. 2018). Soil pH and moisture content changes can also influence soil microbial composition (Schimel et al. 1999; Bååth and Anderson 2003). We observed a higher pH in the soil of the mulch treatment group compared to the herbicide treatment groups at months 10, 22, and 34 at the Pinbarren site, except for acetic acid at month 34. It is likely that both bacterial and fungal diversity segregation between the mulch treatment group and the herbicide treatment groups was partly driven by differences in pH. Therefore, our study highlights the importance of soil pH as one of the driving factors in the bacterial and fungal communities and that the application of mulch can catalyse these changes.

We also observed changes in the abundance of bacteria at the species level in response to the treatments. For example, we found that an unclassified Bradyrhizobium species was less abundant in the Roundup® treatment group, compared with the mulch treatment group and the organic-based herbicide treatment groups at month 34, 1 year after ceasing application of the herbicides. Bacteria of the genus Bradyrhizobium (Proteobacteria) are biologically important in soil due to their role in N fixation (Yao et al. 2014; Wongdee et al. 2018; Praeg et al. 2020). The lower abundance of Bradyrhizobium in the Roundup® treatment group indicated that glyphosate had a negative influence on the relative abundance of Bradyrhizobium species. The toxicity of glyphosate on the abundance of some strains of Bradyrhizobium species has also been previously reported (Zablotowicz and Reddy 2004; dos Santos et al. 2005; Malty et al. 2006), which may negatively impact the nodulation process by Bradyrhizobium, thus affecting soil N cycling. The lower abundance of Bradyrhizobium (Proteobacteria) species was also observed in the organic-based herbicide treatment groups at month 22 at the Pinbarren site, but not at month 34. Our study suggested that although the acute effect of herbicides on soil bacterial communities is relatively low, glyphosate may have negative effects on the abundance of some bacterial species such as Bradyrhizobium even 12 months after cessation of herbicide application, which may have long-term implications for soil N cycling.

Interestingly, at the Pinbarren site, fungal community composition shifted at month 34, which was not associated with treatments. For example, Ascomycota was initially the dominant fungal phyla followed by Basidiomycota at months 10 and 22 at the Pinbarren site. However, Basidiomycota became the most abundant phylum at month 34. Both Ascomycota and Basidiomycota metabolise organic substrates (Hanson et al. 2008) and are sensitive to changes in soil physical and chemical properties such as soil pH, moisture, temperature, and soil nutrient content (Zhao et al. 2018). However, our data do not suggest that this change over time was treatment-dependent because it occurred across all treatments. Microbial shifts can also result from changes in environmental factors such as vegetation cover and revegetation age (Guo et al. 2018). It has been previously reported that there is a distinct shift in fungal communities from Ascomycota in stands of young trees to Basidiomycota in stands of older trees due to changes in litter quality and quantity (Zumsteg et al. 2012; Zhang et al. 2018a) and changes in vegetation cover (Zhang et al. 2018b). Furthermore, in our study, the abundance of Mortierellomycota had increased at month 34 at the Pinbarren site compared with months 10 and 22. Mortierellomycota decomposes plant litter and degrades aromatic hydrocarbons (Osono 2005; Ellegaard-Jensen et al. 2013), and increased abundance of Mortierellomycota is used as an indicator of healthy soil (Zhang et al. 2020). In this study, we did not assess the leaf litter quality, but it is likely that the characteristics of the litter changed as the vegetation matured (Zhang et al. 2018a, b), and may to some extent explain the alteration in the abundance of some communities in our study. We did not observe any major differences in the soil microbial communities among the herbicide treatment groups.

Our study indicated that overall soil bacterial and fungal composition may not be significantly affected by the application of herbicides in the short term. However, 12 months after cessation of treatment, soil MBC levels were lower than the acceptable threshold of 160 mg kg−1 in the Roundup® and BioWeed™ treatment groups and were significantly lower than in the mulch treatment group. Furthermore, the abundance of a small number of microbial species was lower than in the Roundup® treatment group compared with the mulch treatment group, 12 months after cessation of herbicide application. Our data is therefore suggestive that the application of herbicides may have long-term implications on soil microbes and soil health.

5 Conclusion

This study examined the effects of different weed control methods on soil N cycling and microbial communities in two revegetated riparian zones. No major alteration of soil inorganic N was found among the different herbicides. However, we observed an effect of glyphosate on soil nitrification and denitrification because soil NO3−-N was lower in soil following application of Roundup® compared with soil treated with mulch at months 22 and 34 at the Pinbarren site. We did not observe any significant differences in the microbial communities or structure in the soil after the application of Roundup®, BioWeed™, Slasher®, and acetic acid. However, the abundance of a small number of microbial species was lower in the Roundup® treatment group compared with the mulch treatment group,12 months after cessation of herbicide application. Most significantly, we observed that the level of soil MBC after application of Roundup® and BioWeed™ was lower than acceptable thresholds (160 mg kg−1) at month 34 at the Pinbarren site. Furthermore, we observed segregation in soil bacterial and fungal communities between all herbicides and mulch, which persisted 1 year after cessation of herbicide application. This was only observed at one field site, however, which suggested that microbial responses to mulch and herbicides were site-specific. Our study suggested that the application of mulch to assist with riparian revegetation would be beneficial for soil microbial functionality and soil health and that the use of herbicides may have long-lasting effects on soil microbial biomass and microbial communities and should be used with caution as part of an integrated land management plan.