Commentary on four publications testing the effectiveness of biostimulant products on pasture recovery from pasture dieback: an agronomy viewpoint

Paper 1, Whitton et al. (2022a): ‘Sea minerals reduce dysbiosis, improve pasture productivity and plant morphometrics in pasture dieback-affected soils’

Methods. Whitton et al. (2022a) examined the effect of a sea mineral biostimulant on pasture recovery from dieback compared with an untreated control treatment in a 38 ha paddock in Garnant, Queensland. The methods state there were three replicate plots (each 25 m²) of two treatments laid out in a ‘randomised complete block’ design. There is no description of when (date, month or even year) the experiment was established. The biostimulant product applied was ‘Olsson’s Liquid Sea Minerals’ diluted at 100 mL in 50 L of water. This 50 L was applied per treated plot, whereas control plots received 50 L water. This is equivalent to 40 L ha⁻¹ of product applied at a water rate of 20,000 L ha⁻¹.

Plant biomass was sampled in each plot prior to treatments being applied and again 20 weeks after application by cutting all plant material as close to ground as possible in two randomly thrown quadrats. No dimensions of the quadrat are provided. Plant biomass and root data collected 11 months after application are presented in the Results section of the paper, but no description of how these measurements were taken is given other than ‘Additional samples were taken 11 months after the trial, after the rainy season and cattle grazing. A new grass and sage sample from each plot were taken at Week 20 and at 11 months after the treatment to observe differences in plant morphometrics’ (p. 3). Similarly, there is no indication of the management of the plots over the experimental period (e.g. grazing intensity, duration and frequency), no details of the regrowth period following rainfall when the assessments were conducted, nor environmental conditions during the period the experiment was conducted. For statistical analysis of the data, the authors stated ‘Plant data, including dry matter and plant morphometrics, were analysed and presented using GraphPad Prism ver. 9’.

Results. The key agronomic results were that ‘The application of SM [sea minerals] significantly (P = 0.013) increased the total dry matter between week 0 and week 20. CTR [control] plots also increased during the first 20 weeks but not significantly’ (p. 5). Further, ‘at 20 weeks post-application, there was a significant increase (P = 0.044) in the total number of roots in the grass…’ and ‘11 months post-application, the total number of roots was still significantly higher (P = 0.027) in the SM [sea minerals] grass…’ (p. 6). The authors also reported that after 11 months, dry matter in the treated plots was reported as being significantly higher than control plots. Error bars were provided on figures but not defined.

Conclusion. Positive conclusions were made regarding the application of sea minerals and pasture growth, re-iterating the conclusions in the title of the paper ‘Sea minerals reduce dysbiosis, improve pasture productivity and plant morphometrics in pasture dieback affected soils’.

Funding, acknowledgements and data availability. Dr David Tomlinson was identified as providing funding for the scholarship of M. M. Whitton. Numerous people were then thanked in the Acknowledgements for offering various kinds of support with the project, whereas the authors declared no conflict of interest. The sequencing data have been made available.

Paper 2. Whitton et al. (2022b): ‘Remediation of pasture dieback using plant growth promotant’

Methods. Whitton et al. (2022b) examined the effect of biostimulant product ‘RC3’ (Green Earth Technology, Gin Gin, 4671, Australia) on pasture recovery from dieback compared with an untreated control treatment in a 38 ha paddock in Garnant, Queensland. The methods state that there were three replicate plots (25 m²) of each of the two treatments and the treatments were ‘randomised’. Like Whitton et al. (2022a), this paper does not provide important details such as when the treatments were applied and environmental conditions during the experiment. The biostimulant product RC3 was ‘diluted at the rate of 2 mL per litre of water and was applied 50 L per plot’ (p. 3), whereas control plots received 50 L of water. This is equivalent to 40 L ha⁻¹ of product and 20,000 L ha⁻¹ water. It was stated that the experiment was established in 2021.

Similar measurements to Whitton et al. (2022a) were described, as follows: ‘Plant samples for dry matter measurements were collected before the application (Week 0) and at Week 20, using a quadrat thrown twice per plot and cutting and collecting all plant material as close to the ground as possible using scissors.’ Also, the following further details of the grass and wild sage plant sampling at 20 weeks and 11 months after treatment application were provided: ‘A representative grass and a dominant weed (wild sage) plant were taken entirely from each plot and kept on ice until measurements were taken. The root widths were measured using a caliper, while root length, leaf size and plant height were measured with a ruler. The number of tillers, branches, roots, and seed heads were counted.’ There is no description how the sampled plants were selected and removed from the ground, nor the diameter and depth of the soil removed with each plant.

Data analysis was described as ‘Plant data, including dry matter and plant morphometrics, were analysed and presented using GraphPad Prism ver. 9. We used non-parametric Mann–Whitney tests for all morphometric measurements and dry matter calculations’ (p. 4) (Whitton et al. 2022b).

Results. Whitton et al. (2022b) reported that there was no significant difference between pasture biomass of the control and treated plots at 20 weeks. However, there was a significant biomass increase between 0 and 20 weeks after application in the RC3 treatment of 2100 kg ha⁻¹, but an insignificant increase of 1600 kg ha⁻¹ in the control treatment. Data presented in fig. 1 of Whitton et al. (2022b) indicate much larger variability in biomass in the control treatment at the start of the experiment, although no description of what the error bars represent is provided in the caption of the figure. The authors stated in the discussion that ‘The results of this study have shown that RC3 could be a candidate for short-term improvement in the productivity of dieback-affected pastures.’

Funding and acknowledgements. Dr David Tomlinson was thanked for providing a scholarship for M. M. Whitton, and the same acknowledgements were given as for Whitton et al. (2022a). No conflicts of interest were declared.

Paper 3, Whitton et al. (2023): ‘Humate application alters microbiota–mineral interactions and assists in pasture dieback recovery’

Methods. Whitton et al. (2023) investigated the effect of a potassium humate product compared with a control treatment on pasture recovery from dieback in a 38 ha paddock in Garnant, Queensland. Similar to the two earlier papers, there were three replicate 25 m² plots of control and potassium humate treatments laid out in a ‘randomised complete block design’ (p. 3). Potassium humate product was diluted using 6 mL per litre of water, with 50 L of diluted solution being applied to treated plots and 50 L water being applied to the control plots. Thus, the product was applied at 120 L ha⁻¹ with a water rate of 20,000 L ha⁻¹. Like for the two previously described papers, there were no details of when the treatments were applied.

Biomass samples were taken prior to applying the treatments and again 20 weeks and 11 months after application ‘using a 50 × 50 cm quadrat thrown at random twice per plot, cutting and collecting all plant material using scissors as close to the ground as possible’ (Whitton et al. 2023). Further details of plant collection at 11 months were also given: ‘…collecting whole plants with the root system intact, and keeping them on dry ice until measurements were conducted. Root width was measured using calipers, while plant height, root length, and size of leaves were measured using a ruler. Seed heads, the number of branches/tillers, and the total number of roots were also counted’ (p. 3).

Data analysis was described as ‘Plant data, including dry matter and plant morphometrics, was analysed and presented using GraphPad Prism ver. 9’ (p. 3) (Whitton et al. 2023).

Results. Data from fig. 5a (Whitton et al. 2023) shows ‘dry matter’ of the control and humate treatment 0 and 20 weeks after application. It is assumed that this is ‘total dry matter’. These data indicated that dry matter in the humate-treated plots at 20 weeks was significantly greater than biomass in the control treatment, and also significantly greater than biomass in the humate plots prior to treatment initiation. The error bars were not defined in any figures. Data from the control treatment prior to treatment application and after 20 weeks are shown in Table 1.

Data collected 11 months after treatment application indicated the humate application ‘… showed a statistically significant increase in the total number of roots in the grass, as well as a minor increment in plant height compared with the control. In the sage, there was a statistically significant increase in younger leaf length as well as the total number of roots. Therefore, even 11 months post-application Hum [humate] improved roots in both grass and sage’ (p. 7) (Whitton et al. 2023).

The authors then stated in the Discussion that ‘Morphometrics demonstrated significant improvements in root structure in both grass and dicots 11 months post Hum application, which supports further research about the capacity of humic acids to improve pasture in long-term PD [pasture dieback] recovery’ (p. 10).

Funding and acknowledgements. Dr David Tomlinson was thanked for providing a scholarship for M. M. Whitton, and the same acknowledgements were given as for Whitton et al. (2022a). No conflicts of interest were declared.

Paper 4, Ren et al. (2023): application of phytogenic liquid supplementation in soil microbiome restoration in Queensland pasture dieback

Methods. Ren et al. (2023) investigated the effect of a biostimulant, phytogenic liquid, which the authors reported to contain cinnamaldehyde, carvacrol, and citric acid, and an untreated control on pasture recovery from dieback. As per the studies of Whitton et al. (2022a, 2022b, 2023), the study of Ren et al. (2023) comprised three 25 m² replicate plots of each of the two treatments, which were ‘randomised’. The authors state in the Methods that ‘Phytogenic liquid solution diluted at a ratio of 0.27 mL L⁻¹ was sprayed twice, 1 week apart, and each 25 m² plot received 50 L of phytogenic liquid diluted solution’ (p. 3), with control plots again receiving 50 L of water. At each application, 5.4 L ha⁻¹ product was applied (total 10.8 L ha⁻¹) with a water rate of 20,000 L ha⁻¹.

Similar to the three Whitton et al. (2022a, 2022b, 2023) papers, ‘Two random samples were collected from each plot before phytogenic liquid was applied and again in Week 20 using a 50 cm × 50 cm square quadrat to randomly throw twice per plot, then everything above the ground was collected within the quadrat’ (p. 4). Mistakenly placed with Results rather than in the Methods section, the authors stated an important management procedure: ‘Prior to PHY [phytogenic liquid] application at week zero, CTR [control] and PHY plots had 2800 kg ha⁻¹ and 2300 kg ha⁻¹ of dry matter, respectively. This dry matter was then slashed so that the growth on all plots could restart evenly’ (p. 14).

Measurements were taken for plant morphometrics: ‘In addition to dry matter, plant properties were recorded at Week 20 and after 11 months of application (Week 48), including the length and width of the longest leaf, length and width of the youngest leaf, the number of tillers, plant height, the total number of roots, length of the longest root, root thickness, and seed heads (if applicable). The measurements were taken for grass and for the most dominant weed, wild sage’ (p. 4) (Ren et al. 2023).

The results indicated that a biomass assessment was conducted 18 months after treatment application, but details are not provided in the Methods. A second experiment was mentioned but details were scant, including the date the experiment commenced and its duration.

For statistical analysis, the authors stated the following: ‘All plant data, including biomass and morphometrics, were analysed for significance using the Mann–Whitney test performed in GraphPad Prizm ver. 9’ (p. 4).

Results. As per the Whitton et al. (2022a, 2022b) papers, there was no difference in biomass between the control and treatment plots 20 weeks after application; however, the authors stated that ‘Phytogenic liquid-treated plots significantly (P = 0.0116) harvested more biomass 20 weeks post-application compared to the dry matter prior to application…; this growth was not significant in CTR’ (p. 14) (Ren et al. 2023). Data then indicated no significant difference in biomass between the control and treated plots at 11 months after treatment application, but a significant difference in biomass was present at 18 months after application (error bars on figures are not defined). There were no details on the management of the pasture between 11 months and 18 months after treatment.

The authors concluded that their ‘study provides the first evidence of the long-term recovery of pasture dieback affected paddocks in Australia’ (p. 17) and in the abstract they stated that ‘Phytogenic liquid application produced plant morphology improvements and a consistent enhancement of pasture productivity extending beyond 18 months post-application’ (p. 1).

Funding, data availability and acknowledgements. The Fitzroy Basin Association was identified as providing the scholarship for X. Ren. Similar acknowledgements were made as in the Whitton papers described above. No conflicts of interest were declared, with the statement explicitly stating that ‘The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.’

Publication of agronomic results from the same field experiment

Based on the site details, methodologies and the mean values of the controls across the four papers, we speculate that the agronomic data from all four papers arose from the same field experiment. Across the three studies by Whitton et al. (2022a, 2022b, 2023), the site is described as a 38 ha paddock in Garnant, Queensland. Ren et al. (2023) described their study site in more vague terms as being ‘near Rockhampton’ (Table 1). Garnant is located approximately 40 km north-west of Rockhampton. Whitton et al. (2023) and Ren et al. (2023) described the study properties as being 1400 and 1500 ha in size respectively, and both studies reported that ~100 ha of the property has been affected by pasture dieback. No other locational information, such as coordinates for the experimental sites, nor mean climate details, were provided for any of the four studies. Additionally, the start date of the experiments was not provided in any of the papers. All papers also referred to the treatments being arranged in a randomised-block design, suggesting that more than two treatments were included in the experiment.

It appears that the key measurements from the four papers were taken in a similar fashion and timeframe. First, dry-matter cuts from each plot were taken prior to the treatments being imposed by using two cuts from a randomly thrown 0.5 × 0.5 m quadrat (i.e. 2 × 0.25 m²) by cutting biomass to ground level. Note that we have assumed the quadrat dimensions for the Whitton et al. (2022a) study, as none was provided in the methodology. In all papers, further biomass cuts were then taken using the same method 20 weeks after application. At 20 weeks and 11 months after application of treatments, a single grass plant and a single wild sage plant were removed from each plot for assessment of root and shoot parameters. In each paper, roots were counted, although the method used was not described, and root length was measured with a ruler. Cattle grazed the experimental site between the 20-week and 11-month post-application assessments, although no details were provided of the grazing duration and frequency. Further biomass and morphometric measurements were taken after 11 months by using the methods described above.

The biomass production provided for the control treatment in each of the four papers, and their associated errors, is remarkably consistent across the four studies. We determined the means and errors of the control treatment at the pre- (Week 0) and Week 20 post-application biomass assessments from figures in the four papers by using DataThief III (ver. 1.7, Tummers 2006) because values were not reported in the text. These estimates are provided in Table 1. Inspection of these data suggested that the same control is presented in each paper. The estimated presented biomass for the control at Week 0 ranged from 2675 to 2769 kg ha⁻¹, with errors between 879 and 1150 kg ha⁻¹ (Table 1). At Week 20, estimated presented biomass for the control ranged from 4308 to 4397 kg ha⁻¹, with errors ranging from 770 to 884 kg ha⁻¹ (Table 1). Although there appears to be some variability, this is likely to be due to differences in the estimates produced using image analysis in DataThief, with all means within error of each other (Table 1).

The similarities and omissions in site details, key measurement methodology, and biomass values of the control treatment suggest that these are different treatments from the same experiment. That is, the four separate studies may be leveraging the same experiment.

Although there may be reasons to present data from the same experiment as separate studies, none was provided. However, their approach has implications for the weight of evidence presented that supports the use of biostimulants as a treatment for pasture dieback. Presenting the results across separate studies may be used when treatments are imposed after establishment, different analyses are used to answer different aims or as part of long-term evaluation (Urbanowicz and Reinke 2018). Without dates or more information being provided, it is not clear that these studies meet any of these reasons. The studies therefore represent four contributions that draw positive conclusions about the use of biostimulants to treat pasture dieback. Regardless of the conclusions supported by the data itself, the weight of evidence provided by four publications is more likely to influence land managers and their advisors who may not be able to interrogate the work through a scientific lens than is a single paper comparing four biostimulants with a control. As a result, the presentation of these results may influence decision-making. Anecdotal evidence on the northern coast of NSW is that these findings have influenced some graziers, with more willing to trial the biostimulant products (N. Jennings, NSW Local Land Services, pers. comm.). This is particularly concerning, considering the following analyses of the conclusions drawn in these individual studies.

Conclusions based on biomass production assessed 20 weeks after application

One of the major arguments made by the authors for success of the biostimulant products was that biomass production in treated plots 20 weeks after application was significantly (P < 0.05) greater than was the initial biomass before treatments were imposed. This is in contrast to the not significant (P > 0.05) increase in biomass in control plots over the same period. First, we question the validity of this comparison, because it largely reflects variability in the initial assessment rather than the effect of the biostimulant (Table 1). Using a power analysis, we calculated the sample size required to detect significant differences between the initial and 20 week biomass values. We used the reported α of 0.05, an assumed β of 0.2 (power of 80%), plus our smallest estimated median and error for the control in week 0 (2675 ± 879 kg ha⁻¹; Table 1). We used the median in line with the Mann–Whitney test specified in Whitton et al. (2022b) and Ren et al. (2023), and assumed the error, which the authors had not defined, to be the standard deviation. This was contrasted against the largest estimated control biomass 20 weeks after application (4397 kg ha⁻¹; Table 1). Using this analysis, five replicates would be required to detect significant differences between the initial and 20 week biomass values. This is a conservative estimate because the number would increase if the error presented in the papers was the standard error, given that it is defined as the standard deviation divided by the square root of the sample size (regardless of whether the sample size was three or six). Therefore, this result may indicate a Type II error owing to the power of the methodology rather than the effect of biostimulants.

Based on the factorial design alluded to in the methods, the biomass in the treatment should be compared with the control at each assessment, not compared between assessments (Reuter and Moffet 2016). Again, utilising a power analysis using the lowest Week 20 estimated biomass for the control (4308 kg ha⁻¹; Table 1), the highest estimated Week 20 biomass (5920 kg ha⁻¹; Ren et al. 2023), and the lowest estimated error (770 kg ha⁻¹; Table 1), a sample size of four would be required to detect significant differences between the treatments. However, for a comparison of multiple treatments, using the difference between these values, we estimate the required number of replicates to be three. This would not detect differences between treatments where there has been lower reported biomass values for the biostimulant treatments in Whitton et al. (2022a, 2022b, 2023). Here, we have used the term mean for the estimates because the papers did not state what the bars in the figures represent.

Our analysis of the data and sample size above provided evidence that the four studies are limited by the large variability in pasture growth observed. The experimental plots were 5 × 5 m (25 m²), yet biomass was sampled only with two 0.5 × 0.5 m (0.25 m²) quadrat cuts (Whitton et al. 2022a, 2023; Ren et al. 2023). This sampled area represents only 2% of the total experimental plot area. A number of destructive or non-destructive techniques could have been used to assess biomass (‘t Mannetje 1978). On the basis of the description provided in these papers, pasture biomass may have been more appropriately assessed using a calibrated visual assessment, either by multiple random quadrat assessments per plots (e.g. Lodge et al. 2003a, 2003b) or by dividing the plot into sections of equal area (e.g. Li et al. 2008, 2010). Botanical composition could have been assessed at the same time by using the dry-rank method (’t Mannetje and Haydock 2006), providing the proportion of biomass that was grass and broadleaf. These biomass techniques would have increased the plot area assessed to 16% (visual assessment of 10 quadrats plot⁻¹, each 0.25 m²) and 100% (visual assessment dividing plot into four sections) respectively.

Finally, in the results section of Ren et al. (2023), the authors stated the following: ‘Prior to PHY application at week zero, CTR and PHY plots had 2800 kg ha⁻¹ and 2300 kg ha⁻¹ of dry matter, respectively. This dry matter was then slashed so that the growth on all plots could restart evenly’ (p. 14). Slashing after conducting the biomass assessment renders comparisons between Weeks 0 and 20 biomass values irrelevant. Further, given the above question whether data for all four studies arose from the same experiment, this detail should have been provided in the three Whitton et al. (2022a, 2022b, 2023) papers as well. It is critical information and indicates than any comparison made in the four papers between biomass at 0 and 20 weeks is non-sensical.

The statistical power limitations and method ambiguities that we have raised, therefore, bring into question the strength of the conclusion that biostimulants do aid in pasture dieback recovery. This is somewhat misleading, given the perceived weight of evidence that is presented in these four separate studies.

Further methodological limitations

Attribution of funding, data availability and acknowledgements

There are aspects of acknowledging funding sources and conflicts, as well as data availability, which may have been overlooked, that are important for clarity and transparency. Dr David Tomlinson is identified as providing funding for M. M. Whitton. Dr Tomlinson has a known association with the Green Earth Technology and the RC3 product used in Whitton et al. (2022b) (https://greenearthtechnology.com.au/about-us/). Robert Alder of Geo Leak Solutions is also thanked in the acknowledgements of all studies, with Geo Leak Solutions having an association with Green Earth Technology (https://greenearthtechnology.com.au/geo-leak-solutions-research/). Across all four studies, no conflict of interest was declared, despite trialling products under funding from and working with associates with close industry ties. This could easily be perceived as vested interest in the positive effect of the biostimulants. In addition, only sequence data are available in accessible repositories, with no indication that other data are available on request. Therefore, the data do not align with the findable, accessible, interoperable and re-usable (FAIR) principles (Wilkinson et al. 2016). We are not implying that malpractice has occurred; however, we feel transparency in acknowledgements, funding and data need to be fully upheld to the highest standard surrounding studies that use a commercial product.