Adaptive multi-paddock grazing management’s influence on soil food web community structure for: increasing pasture forage production, soil organic carbon, and reducing soil respiration rates in southeastern USA ranches

Grazing Strategies

Grazing ecosystem soils and biotic communities coevolved with herbaceous vegetation, grazer insects and other herbivores over the last 40 million years, contributing to the expansion of carbon-rich soils in semiarid to semi-humid grassland regions that cover approximately 40% of the earth’s land area (Frank, McNaughton & Tracy, 1998; Retallack, 2013). The replacement of free-ranging wild herbivores, with fenced-in livestock, has contributed to degradation of vegetation and soils, declines in productivity and biodiversity, and a reduction in ecosystem resilience, resulting in an overall decline in historic ecosystem services generated through evolved grazer/grassland relationships (West, 1993; Milchunas & Lauenroth, 1993; Knopf, 1994; Frank, McNaughton & Tracy, 1998; Peterson, Allen & Holling, 1998).

To ensure sustainability and resilience, grazed ecosystems should be managed using grazing management that avoids overstocking and overgrazing (Teague, 2018). Conventional grazing management allows grazing animals unrestricted access in large fenced pastures, leading to degradation of soil and plant function due to patch-selective overgrazing and stocking with no planned recovery from grazing at stocking rates in excess of carrying capacity (Teague, 2018). Soil function is also diminished by the application of inorganic fertilizers, herbicides, pesticides, and parasiticides (Bardgett & McAlister, 1999; Leake et al., 2004; de Vries et al., 2006; Birkhofer et al., 2008; LaCanne & Lundgren, 2018). As an alternative, adaptive multi-paddock (AMP) grazing is a regenerative management option that has proven to be more productive, give greater economic returns and more abundant ecosystem services by improving biological function (Teague et al., 2013; Jakoby et al., 2014; Jakoby et al., 2015; Teague & Kreuter, 2020).

AMP emulates ecosystem processes that evolved in response to intense but periodic herbivory by large herds of ungulate grazers that included post grazing recovery (Frank, McNaughton & Tracy, 1998; Retallack, 2013) and has been documented to avoid and reverse the damage caused by continuous grazing in a timely and cost-effective manner (Gerrish, 2004; Teague et al., 2011; Teague et al., 2013; Wang, Teague & Park, 2016). AMP grazing uses a management decision process, responding to weather and forage variability, to avoid overstocking of, and overgrazing by livestock in an adaptive manner (Jakoby et al., 2014; Jakoby et al., 2015; Teague, Grant & Wang, 2015; Wang et al., 2018). The grazing benefits are achieved by adaptively adjusting grazing frequency and duration on available forage resources, using short, but intense grazing intervals, allowing livestock to graze ∼40% of available SCB. The remaining ∼60% of plant residue provides photosynthetic surface area to aid plant regrowth and/or be trampled onto the soil surface through short-term hoof-impact of the grazing herd. The trampled plant residue provides soil cover, reduces surface water runoff and erosion and provides above and below ground plant biomass to support microbial activity (Thurow, 1991; Dowhower et al., 2019; Teague & Kreuter, 2020). Appropriate length recovery periods promote more efficient recovery of plant communities, the soil food web and ecosystem restoration. Most AMP ranchers find this grazing strategy allows them to avoid application of inputs that may compromise soil biotic function, such as synthetic fertilizers, herbicides and pesticides. In combination, these actions result in light to moderate grazing impact on herbaceous plants, the soil food web, and the ecological functions they perform (Teague et al., 2013; Jakoby et al., 2014; Dowhower et al., 2019; Teague & Kreuter, 2020).

Mixed-study results from grazing researchers have concluded that no improvements in ecological and livestock indicators occurred under rotational grazing when compared to continuous grazing (e.g., Briske et al., 2008). However, these conclusions have been criticized for their relatively long periods of grazing and short recovery periods, as well as the short-term and small-scale and fixed rather than adaptively managed research frameworks (Teague et al., 2013; Gosnell, Grimm & Goldstein, 2019; Gosnell, Charnley & Stanley, 2020; Teague & Kreuter, 2020). Other studies provide evidence that rotational grazing, with fewer paddocks per herd, more extended grazing periods, and shorter recovery periods, can result in limited plant and animal production advantages compared to continuous set-stocking at low stocking rates (Briske et al., 2008; Teague et al., 2013; Wang et al., 2022). Low stocking rates and improperly applied rotational grazing strategies do not facilitate degraded resource recovery or provide adequate economic returns over time (Jakoby et al., 2014; Jakoby et al., 2015; Teague & Kreuter, 2020). CG set-stocked at low stocking levels implementing long periods of grazing, paired with inadequate recovery, set-stocking rates, and low levels of soil surface forage residue promote a compounding of negative effects on ecological function and economic returns as illustrated by Augustine et al. (2020).

Soil ecosystem

Intensification of agricultural production has promoted management practices that have reduced soil carbon and the biomass and diversity of soil microbiota (Stoate et al., 2001; Postma-Blaauw et al., 2010) including shifts to more bacterial-dominated microbial communities with increased soil nitrogen use (de Vries et al., 2006), and reduced soil carbon stocks and accrual rates (Six et al., 2006). Soil microbes are instrumental in C and N cycling; however, microbial biomass and activity are strongly influenced by higher trophic-level organisms of the soil food web. The upper trophic-level organisms (protozoa and nematodes) consume soil microbiota and stimulate microbial turnover through “grazing” of lower trophic level organisms (Postma-Blaauw et al., 2005) providing nutrient mineralization processes that promote plant productivity (Setälä & Huhta, 1991).

Soil organic matter (SOM), and the carbon and nutrients it contains, are key components for supporting fundamental bio-geochemical processes for: plant carbon assimilation and growth; soil respiration; and carbon-climate feedbacks (Kallenbach, Frey & Grandy, 2016). Soil food web population, structure, diversity, and biological functionality facilitate these bio-geochemical processes and contribute substantially to: nutrient generation, nutrient cycling, nutrient capture, soil fertility development, and SOM formation and turnover (Schloter, Dilly & Munch, 2003; Van der Heijden, Bardgett & Van Straalen, 2008; Murray et al., 2009; García-Orenes et al., 2013). A shift towards fungal dominance in the soil food web has been observed to enhance C accumulation and reduce SOM turnover rates (Six et al., 2006). More efficient microbial biomass production and the accumulation of SOM are now considered to be driven by distinct soil food web structures, where microbial-derived SOM is greatest in soils that contain higher fungal abundances (Kallenbach, Frey & Grandy, 2016; Johnson, 2017).

Predicting the effects of soil food web physiological regulation on soil C processes and their interaction with plants and livestock management, is critical if we are to improve the performance of our agroecosystems, project future global warming offset potentials, and enhance atmospheric CO₂ reduction (Billings & Ballantyne, 2013). Despite this expectation, many studies have concluded there is no direct evidence that soil fungal-to-bacterial ratios (F:B) characterize the turnover of soil organic matter (Rousk & Frey, 2015), soil nutrient content or growth of vegetation (Wong et al., 2015), or that fungi are capable of enhancing soil carbon storage mechanisms (Thiet, Freya & Six, 2006; Johnson, 2017).

The influence of AMP grazing management on: in-situ field measurements for the soil food web population, structure, diversity and metabolic functionality, and the soil food web’s influence on standing crop biomass (SCB) production, soil organic carbon stocks (SOC) and soil microbial respiration efficiency is not well documented. The experimental hypothesis in this research is that AMP grazing methodologies will promote beneficial changes in soil food web population and structural composition (bacterial and fungal biomass, F:B ratio, and protozoa enumeration) and function, and that these shifts in the soil food web population structure will positively influence grazing system photosynthetic capacity, soil microbial respiration efficiency and the storage of carbon in AMP grazed rangeland soils when compared to CG systems. In this paper we evaluate the effects of two grazing management strategies on soil food web system functions and vegetation interactions measuring:

1. early season standing crop biomass production (g dry biomass m⁻²);

2. soil food web biomass (µg g⁻¹ dry soil) for bacteria, actinobacteria, fungi; and population enumeration for total protozoa- (flagellates, ciliates, amoeba) (number g⁻¹ dry soil);

3. a fungal-to-bacterial (F:B) structural biomass ratio;

4. a soil food web proxy (summation of all normalized soil food web constituents);

5. soil organic carbon (SOC%); and,

6. in-situ, diurnal soil CO₂ respiration kinetics (g C m⁻² day⁻¹) of the soil food web,

in an effort to ascertain what can be learned about the soil food web’s contribution to the larger research team question: Can Adaptive Multi-Paddock (AMP) grazing contribute to sequestering carbon in soils and improve delivery of ecosystem services and socio-ecological resilience in grazing ecosystems?

Site screening, selection and soil type

A list of potential AMP and CG grazing partners was obtained from the Natural Resources Conservation Service (NRCS) technical staff, grazing consultants and organizers, in each of four states (Mississippi, Alabama, Tennessee and Kentucky) included in this study (Table 1). Potential candidates were invited to participate in an online survey to help in the selection of suitable paired ranches to conduct an across-the-fence comparison of adjacent AMP grazed and well-managed CG paired neighbors. Preliminary scouting was conducted on the prospective paired-ranches to confirm that each ranch pair had: (1) the same soil types, to reduce the potential that variable soil types may have on primary productivity, (2) similar topography, and (3) land use history with the primary deviation being the conversion of a former CG managed ranch to AMP grazing management that had been in AMP management for a minimum of one decade (Table 1, Table S1). Scouting included both soil typing using NRCS Soil Survey website (https://websoilsurvey.sc.egov.usda.gov/App/WebSoilSurvey.aspx) and pre-selection field soil core sampling and analysis to confirm accuracy.

Research area precipitation

Regional precipitation records, for the 6 months (January 2018–June 2018) prior to field sampling were obtained through NOAA’s Advanced Hydrological Prediction Service online data site (https://water.weather.gov/precip/). Heatmap images were obtained for the monthly percent-of-normal and April and May departure-from-normal precipitation maps for the Southeastern US corridor to confirm uniformity of precipitation events on research ranch pairs to verify similar growing conditions for assessing accurate SCB growth potential (Fig. S1).

Soil organic carbon analyses

Soil sample cores for SOC analyses were collected as described in USDA/NRCS Soil Survey Laboratory Methods Manual (Burt, 2004) specifically, harvested in a three-week sampling period, between the last two weeks of May 2018 to the first week of June 2018. Ten catenas were located on the five AMP/CG ranch pairs. Twelve soil core samples were taken, at evenly spaced locations, across each of three pre-determined linear transects within each catena (sloped or flat), having rectangular dimensions of ∼100 m by ∼50 m (120 total soil core samples). The soil cores for SOC analyses were comprised of a soil core four cm in diameter and 10 cm deep taken at each sample site (12 sites at each AMP and CG farm), placed in 1-quart plastic bags and shipped to Ward Laboratory Inc., (4007 Cherry Ave, Kearney, NE 68847) for SOC (%) carbon analysis (LECO, dry combustion). Laboratory testing included a preliminary acidification treatment to remove soil carbonate content and provide an accurate SOC (%) assessment. Each of the sample cores were sampled in triplicate to insure SOC (%) sampling accuracy.

Standing crop biomass (SCB)

Standing crop biomass was assessed (g dry biomass m⁻²) by clipping living/standing plant biomass 2.54 cm above ground height, assessing from 15–20 individual quadrats (quadrat = 1,000 cm²) taken inside multiple fenced livestock-exclusion areas, in each catena, on AMP and CG pastures. All plant materials were paper bagged, oven dried and weighed for dry biomass. The sample sites (sloped or flat) corresponded with the microbiological sampling sites for each ranch pair (Table S1) for the January to early June pasture forage growth period in each treatment plot. Each sample site was harvested, in a three-week sampling period, between the last two weeks of May 2018 to the first week of June 2018 corresponding with the sample timing for all microbiological and respiration metrics.

Soil food web analyses

Soil samples for soil food web analyses were collected as recommended by Earthfort (https://earthfort.com/wp-content/uploads/2021/08/Sample-Instructions-Combined_81221.pdf), specifically sampling at each location from a soil core four cm in diameter and 10 cm deep of the soil profile (n = 12 for each ranch). These samples were immediately placed on ice and mailed overnight to Environment Celebration Institute Inc., (Berry Creek, CA, USA) for soil food web analyses to quantify soil food web biomass (µg g⁻¹ dry soil) for bacteria, actinobacteria, fungi; and in number of organisms gram⁻¹ of dry soil for: total protozoa (flagellates, ciliates and amoeba) and nematodes. This data provided microbial biomass data to determine a structural fungal and bacterial biomass ratio (F:B) and a predator/prey assessment of the different trophic levels. Sample preparation and biomass quantification implemented direct observation microscopy (visible light, differential interference contrast) and other methodologies (Ingham, 1995; Ekelund, 1998; Stamatiadis, Doran & Ingham, 1990).

A soil-food web proxy, for each of the ranch pairs, was assessed as a composite summation of all normalized soil food web measurements for each sampled ranch catena, offering an internal, normalized full soil food web metric to compare aggregate soil food web population dynamics between ranches.

Soil CO₂ respiration rates

Soil-respiration C (g C m⁻² day⁻¹) was measured with in-situ static alkali reactors (Gupta & Singh, 1981), using a 50 mL plastic centrifuge tube, containing 15 mL of standardized 1 M KOH, with a cross-sectional area of about 25% of soil surface sampled. Diurnal soil respiration measurements were conducted for the 24-hour period prior to, and at the identical location that soil samples were taken to assess SOC and the soil food web. Reactors were covered with a ∼1-liter glass cover (canning jar) screwed into the soil profile approximately two cm deep. Prior to placing the jar over the 50 ml centrifuge tube, all vegetation was removed from the surface for accurate evaluation of CO₂ respired from the soil. Reactors measured soil CO₂ emissions, over a diurnal time period, remaining in the field for 24 h (starting times and ending times were recorded to assess total time to 1-minute accuracy). After the 24-hour period, the 50 ml tubes were then removed from the reactors, capped, and taken to a laboratory for quantification of CO₂ absorption by the 1 M KOH solution. A non-destructive analysis technique was implemented assessing reduction in reactor KOH solution conductivity as correlated to the quantity of CO₂ absorbed, as described in Methodology S1. Respiration results were expressed in g C m⁻² day⁻¹. A small number of static-alkali reactors were disturbed by grazing livestock resulting in a different quantity of “n” comparisons.

The static alkali reactor methodology was chosen to insure identical atmospheric conditions (temperature, solar insolation, etc.) and field moisture conditions for simultaneous side-by-side same-day comparison of all 12 sampling locations for each AMP and CG farm.

Statistical analyses

Univariate statistical analyses were used to test differences in means between paired and pooled measurements of AMP and CG treatments for: soil organic carbon percent (SOC%), standing crop biomass (SCB) (g dry biomass m⁻²), soil food web (µg g⁻¹ dry soil and microorganism counts g⁻¹ dry soil), and soil-respiration C (g C m⁻² day⁻¹), in the five ranch pairs. Statistical data analysis was conducted with Statistics Kingdom 2017 (https://www.statskingdom.com/). All data components were initially examined for normal distribution with Shapiro–Wilk (if n < 50) and/or Shapiro-Francia, and Anderson-Darling (if n > 50); F-tests were used to determine data variance. Means testing methodologies were selected based on the equality of variances and normality of data. Pairwise means comparisons were conducted with Two sample t-test (Welch’s T-test) if data had a normal distribution and equal variances and a Mann Whitney U test (Wilcoxon rank-sum) means tests was implemented if data was not distributed normally, had unequal variances or outliers. Simple linear regression analyses were conducted to test the fit of a linear model to the data. Means results are reported as mean values,  ± standard error (mean ± SE), and a significance value α ≤ 0.05 threshold was used to determine statistical significance. Soil food web proxy measurements involved multiple data types (µg microbial biomass g⁻¹ dry soil) and entity data counts (number of microbes g⁻¹ dry soil) requiring normalization to permit data to be interrogated as a soil food web proxy, using a “universal data sub-language” (Codd, 1970).

Ranch pair selection and antecedent precipitation

The grazing-methodology adoption period in these ranch pairs ranged from 10 to 29 years in the AMP ranches (average 17.6 years) and from 6 years to 40 years in the CG ranches (average 25.6 years) providing no significant difference in years of management between AMP and CG ranches (Two sample t-test; T = −1.052996; DF = 6.3342; p = 0.3308; Effect size = 0.67) (Table 1). Total 2018 annual net primary productivity (g dry biomass m⁻²) and recovery goal (days) demonstrated no significant difference (Two sample t-test; T = 0.9018; DF = 6.7518; p = 0.3981; Effect size = 0.57) (Two sample Mann–Whitney; Z = −0.1054; no DF in MW; p = 0.9161; Effect size = 0.033) respectively (Table 1). A statistical analysis of the grazing management practices indicated significant differences between AMP vs. CG grazing management in: Average Animal Units (AU acre⁻¹) (Two sample Mann–Whitney; Z = 2.5377; no DF in MW; p = 0.011; Effect size 0.8); Average Paddock Size (ha), (Two-sample t-test; T = −10.8336; DF = 4.3721; p = 0.00025; Effect size = 6.85); Average Number of Paddocks per Herd, (Two sample Mann–Whitney; Z = 2.0953; no DF in MW; p = 0.036; Effect size = 0.66); Graze Period Goal (days), (Two sample Mann–Whitney; Z = −2.5536; no DF in MW; p = 0.01066; Effect size = 0.81) and Rest vs. Graze Period (ratio), (Two sample Mann–Whitney; Z = 23.5143; no DF in MW; p = 0.01193; Effect size = 0.8) (Table 1), supporting the validity of the AMP vs. CG ranch-pair management selection criteria in this research project.

Precipitation maps, of the AMP/CG ranch locations (Fig. S1), available at https://water.weather.gov/precip/, indicated similar antecedent precipitation prior to the May-June sampling event, reducing the potential impact of differing rainfall amounts on ranching system SCB production. Monthly Percent-of-Normal rainfall, for January–June of 2018 and Departure-from-Normal rainfall, for April to May 2018 precipitation maps (Fig. S1) indicate that precipitation amounts supplied the historical monthly average rainfall or better at each sampling site, to ensure comparable precipitation quantities were available at each research site to support equivalent growth potential of SCB on ranch pairs.

Data from one-meter core samples, taken at the CG, and AMP area locations, on the same day by other researchers on this project, concurrent with the timing and location of microbiological sampling, supported sufficient soil water content for accurate soil carbon respiration and the soil bulk density assessment. The analysis of the core samples relayed the average soil moisture content on the AMP, and CG sites was ∼17.56% ± 1.1%. Soil bulk density averaged 1.21 g cm⁻³ ± 0.03 g cm⁻³ in the top ten centimeters of both AMP and CG soil profiles (Mosier et al., 2021).

Soil organic carbon

Soil organic carbon percentages, SOC (%), from AMP management strategies were significantly greater when compared to CG ranches in three of the five pairwise comparisons: AMP-3/CG-3, (Two sample t-test; T = 6.3614; DF = 20.99; p = 2.62 ×10⁻⁶; Effect size = 2.6); AMP-4/CG-4, (Two sample Mann–Whitney; Z = 2.3393; no DF in MW; p = 0.0193; Effect size= 0.48); and AMP-5/CG-5, (Two sample t-test; T = 6.1855; DF = 21.5438 ; p = 3.474 ×10⁻⁶; Effect size = 2.52) (Table 2). There was no significant difference of SOC (%) in AMP-1/CG-1 and AMP-2/CG-2.

Pooled AMP soil organic carbon exhibited mean cumulative ranch SOC (%) of 2.83 ± 0.123%C, or 21% greater than mean cumulative CG SOC (%) of 2.354 ± 0.09%C (Two sample Mann–Whitney; Z = 2.6507; no DF in MW; p = 0.008; Effect size = 0.24) (Table 2).

Standing crop biomass (SCB)

Pairwise comparisons of standing crop biomass (SCB) were significantly greater in AMP systems in three of the five pairwise comparisons AMP-3/CG-3, (Two sample t-test Mann–Whitney; Z = −3.2627; no DF in MW; p = 0.0011; Effect size = 0.42); AMP-4/CG-4, (Two sample t-test; DF = 31.4136; T = −5.7858p = 2.1 ×10⁻⁶; Effect size = 1.49) and AMP-5/CG-5, (Two sample t-test; T = 5.31036; DF = 51.0642; p = 2.007 ×10⁻⁶; Effect size = 1.38) (Table 2); and reversed in the AMP-2/CG-2 comparison (Two sample Mann–Whitney; Z = −3.2627; no DF in MW; p = 0.0011; Effect size = 0.42) and not statistically significant in the AMP-1/CG-1 ranch pair.

AMP management strategies exhibited a pooled SCB mean of 294.1  ± 11.43 g dry biomass m⁻², ∼92.7 g dry biomass or ∼46% more SCB than pooled CG systems SCB mean of 201.5  ± 10.74 g dry biomass m⁻² (Two sample Mann–Whitney; Z = 6.1836; no DF in MW; p = 6.26 ×10⁻¹⁰; Effect size = 0.35) (Table 2).

Soil food web

Pairwise comparisons of AMP and CG soil food web analyses provided only one farm pair, AMP-3/CG-3, that demonstrated statistically significant differences in the means of soil food web- bacteria, (Two sample t-test; T = −2.8435; DF = 16.003; p = 0.0117; Effect size = 1.16); F:B ratio, (Two sample Mann–Whitney; Z = 2.1096; no DF in MW; p = 0.03489; Effect size = 0.43); soil food web proxy, (Two sample Mann–Whitney; Z = 2.5094; no DF in MW; p = 0.01209; Effect size = 0.5); and protozoa enumeration, (Two sample Mann–Whitney; Z = 2.4858; no DF in MW: p = 0.01292; Effect size = 0.51) (Table S2).

Pooled comparison of AMP and CG systems indicated a significant difference in the means of the bacterial population (Two sample Mann–Whitney; Z = −2.1766; no Df in MW; p = 0.02951; Effect size = 0.2) and protozoa enumeration (Two sample Mann–Whitney; Z = 2.3522; no DF in MW; p = 0.01866; Effect size = 0.21) (Table S2) but no statistically significant differences were observed in fungi, F:B ratio or soil foodweb proxy.

Simple linear regression analyses of mean AMP soil food web variables with SCB provided very strong direct relationships with fungal biomass, (R = 0.9915; F(1,3) = 175.35; p = 0.015); F:B ratio (R = 0.9371; F(1,3) = 21.56; p = 0.019); and soil food web proxy (R = 0.9616; F(1,3) = 36.75; p = 0.009) (Figs. 1B–1D), and a very strong inverse relationship with bacterial biomass (R = −0.946; F(1,3) = 25.56; p = 0.015) (Fig. 1A).

Simple linear regression analyses of CG soil food web variables with SCB for bacteria biomass, fungal biomass, F:B ratio and soil food web proxy were not statistically significant; (R = −0.3061, F(1,3) = 0.31, p = 0.6165); (R = −0.6605, F(1,3) = 2.32, p = 0.225); (R = −0. 2386; F(1,3) = 0.18, p = 0.6691); (R = 0.02491, F(1,3) = 0.00016, p = .991) respectively (Figs. 1B–1D) (trendlines not shown).

Predator/prey relationships, comparing protozoa populations (# g⁻¹ dry soil) and bacteria biomass (µg g⁻¹ dry soil) with standing crop biomass in AMP systems, provided a very strong positive relationship to total protozoa (R = 0.9826; F(1,3) = 83.68; p = 0.003), and a strong inverse relationship to bacterial biomass (R = −0.946; F(1,3) = 25.56; p = 0.015) (Fig. 2A); where CG systems provided no significant comparisons (Fig. 2B).

Soil respiration

Pairwise comparisons of AMP and CG soil respiration provided two farm pairs, AMP-3/CG-3 (Two sample t-test; T = −2.8734; DF = 9.4854; p = 0.01744; Effect size = 1.54) and AMP-5, CG-5 (Two sample t-test; T = −3.797; DF = 12.7065; p = 0.0023; Effect size = 1.78) that demonstrated statistically significant differences in respiration (Table 2).

Pooled AMP management soil carbon respiration exhibited 19.52% less mean cumulative soil respiration (2.237 ± 0.131 g C m⁻² day⁻¹) when compared to CG soil respiration (2.779 ± 0.1890 g C m⁻² day⁻¹) (Two sample t-test; T = −2.3581; DF = 52.3541; p = 0.0221; Effect size = 0.59) (Table 2). Carbon dioxide flux towers have been installed by associated researchers on this project and are currently tracking soil respiration on two of the farm pairs AMP-4/CG-4 and AMP-5/CG-5. Data is still being collected but their initial data is also observing a commensurate ∼20% reduction in soil C (CO₂) respiration comparing AMP to CG management (R. Clement, 2022, Pers. Comm.).