Cover crop management practices to promote soil health and climate adaptation: Grappling with varied success from farmer and researcher observations

Abstract

Soil health is a pathway through which farm and environmental outcomes can be improved together on agricultural landscapes, and management to improve soil health is increasingly recognized as a strategy for agricultural producers to adapt to climate change–related impacts such as erosion and flooding. Many incentive programs exist or are in development to support the adoption of practices that promote soil health for these reasons, but few on-farm trials have compared the intersections of farmer versus researcher observations of soil health or of soil health compared with climate adaptation. The purpose of our study was to assess soil health outcomes and adaptation to simulated climate change in response to cover cropping on working farms enrolled in a Minnesota Natural Resources Conservation Service (NRCS)–funded Environmental Quality Incentives Program. This incentive program required the inclusion of diverse cover crop mixtures into existing farm crop rotations. We conducted farmer surveys, NRCS protocol field soil assessments, and NRCS-recommended laboratory assays on farms from across the state of Minnesota in a paired design comparing fields on the same or adjacent farms of the same soil mapping unit. Although 85% of farmers reported improvements in soil attributes or productivity, most field and laboratory assessments produced a high amount of variability in responsiveness to cover cropping. Despite this variability, we saw a significant decrease in bare ground and significant increases in earthworm counts, cellobiohydrolase microbial activity, and the Visual Evaluation of Soil Structure (VESS). Although researcher measurements did not show improvement in physical characteristics or infiltration other than the VESS field assessment, 67% of farmers reported improvements to the physical structure of the soil, associated with improved outcomes such as earlier planting dates and consistent crop growth across fields. When more than five species of cover crops were present, the percentage of reported improvements increased to >80%. We also found no significant improvement to climate change adaptation measured by nutrient or sediment loss after a simulated storm event. Together, our results suggest that adding a diverse annual cover crop mix to increase continuous cover can improve characteristics associated with soil health and that there needs to be a stronger focus in understanding variation in realized soil health outcomes on farms, including more co-creation of research with farmer partners.

Abbreviations

CC

cover crop

CI

confidence interval

EQIP

Environmental Quality Incentives Program

4-MUB

4-methylumbelliferone

LRR

logarithm of the response ratio

MUB

methylumbelliferone

NCC

no cover crop

NRCS

Natural Resources Conservation Service

SHI

Soil Health Initiative

VESS

Visual Evaluation of Soil Structure

1 INTRODUCTION

There is an increasing and collective motivation of farmers, land managers, researchers, agencies, and other stakeholders to improve soil health and soil-related ecosystem services, such as climate adaptation, in agricultural lands (Liu et al., 2020). This has resulted in an abundance of research, farmer incentive initiatives to improve soil health, and soil health data synthesis (Stewart et al., 2018), but the strategies to describe soil health remain inconsistent. This disparity could arise from the perspectives of those observing changes, measurement approaches in general, or inherent sources of variability but is a challenge worth understanding in order to further adoption and best management for improved soil conservation.

The Natural Resource Conservation Service (NRCS) has been at the forefront of developing incentive programs and metrics to support, promote, and validate soil health practices (Wander et al., 2019). The NRCS has created programs, such as the Environmental Quality Incentives Program (EQIP), that provide financial and technical assistance to producers to address resource concerns and deliver environmental benefits (USDA-NRCS, 2021). Conservation and restoration of the soil resource is one such environmental benefit that could see broader adoption under EQIP and similar programs. In 2013 the USDA-NRCS in Minnesota launched a 3-yr Soil Health Initiative (SHI) pilot program through EQIP. This program encouraged producers to sign up for a special cover crop (CC) initiative through EQIP. The initiative offered producers 5 yr of support to plant diverse CC mixtures and reduce fall tillage to any level they were able. From 2013 to 2015, 105 producers enrolled in the program, and approximately US$1 million of support were obligated. Many other programs, such as the Soil Health Institute, a public-private partnership, also seek to incentivize and monetize soil health (Wander et al., 2019).

Soil health is often paired in a narrative with soil conservation and broad land stewardship. How then is this spirit of conservation being conceptualized through soil health, and how do we know we are improving it? The NRCS defines soil health as the capacity of the soil to function as a vital living ecosystem that supports plants, animals, and humans (Stott, 2019). These functional outcomes are inherently based on a value system that is usually tied to a set of measurements used to determine or predict whether desired outcomes are being reached (Wander et al., 2019). Soil health is assessed by measuring multiple indicators. These may include a range of soil chemical, physical, and biological properties (Bunemann et al., 2018). Farmers and land managers typically characterize improvements in soil health in terms of field workability, crop productivity and vigor, disease and weed pressure, and presence of living organisms such as earthworms (Barbero-Sierra et al., 2018; Lobry De Bruyn & Abbey, 2003; Saleque et al., 2008; Yageta et al., 2019). Field evaluations that align with visible on-farm improvements are sometimes present (Valania et al., 2020) but are often excluded from researcher-led efforts. Instead, scientists focus more on measurements of physical properties (e.g., soil texture, aggregate stability, or infiltration), chemical analyses (e.g., percent carbon or nitrogen, micronutrient contents, pH, or organic matter composition), or biological assessments (e.g., soil respiration rates, microbial biomass, community composition, activity, or diversity). Researchers choose measurements for practical considerations (cost, throughput, low variability) or scientific norms (what measurements are regularly published) or to answer specific agroecological research questions. Stakeholder engagement is seldom the basis for analysis. The disconnect between the ways soil biological communities are assessed and perceptions of biological soil health outcomes is particularly striking (Fierer et al., 2021). Commonly used laboratory-based soil health assays align at times but also sometimes fail to align with farmer perceptions of soil health (O'Neill et al., 2021). Therefore, agricultural researchers may have different perceptions and draw different conclusions based on their preferred soil health measurements than farmers do with theirs (Barrios et al., 2006; Wade et al., 2021). Although researchers and farmers may have common goals, it is not yet common practice to synchronize farmer and researcher observations of soil health to more readily achieve desired outcomes.

Above and beyond who is measuring and what is being measured, external factors influence whether soil health–promoting management practices lead to desired outcomes or are “successful.” They include natural biophysical properties, such as local climate, soil type, soil texture, or landscape position (Amsili et al., 2021). Detailed assessments of spatio-temporal variability in soil health indicators across a range of soils and management systems are relatively lacking despite recognition that they influence soil health outcomes (but see Crookston et al. [2021]). Within the biophysical system, management decisions on a given farm might also influence soil health outcomes (Amsili et al., 2021). These could include the type and frequency of tillage events; CC species, species diversity, and performance; and cash crop management, including fertility and water needs, and strategies/inputs needed to control weeds and diseases.

Despite these obstacles to consistent understanding of how to manage for soil health and assess outcomes, shared understanding is critical to promote adoption of best practices and improve determination of incentive program efficacy. There is a need to understand how different perspectives might lead to varying perceptions of soil health outcomes, given the variability of soil health metrics on working farms, and to identify soil health metrics that can transcend farm-to-farm variability and serve as global yardsticks for the achievement of desired soil health outcomes. Toward this aim, agricultural researchers have made many attempts at developing assessment systems to measure changes in soil properties over time within a given management system or between systems (Bunemann et al., 2018). There are also national-scale efforts underway in the United States to assess soil health indicators across a broad range of soils and management systems (Norris et al., 2020; Wills et al., 2020; Wood & Bowman, 2021). Most of these widespread efforts have also been researcher led rather than farmer- or practitioner led.

Stakeholders, including farmers, land managers, and researchers, increasingly agree that soil health is a pathway to climate change resiliency (e.g., Carlisle, 2022; Lengnick, 2022; MN Pollution Control Agency, 2019; Paustian et al., 2016). The specific functional outcomes critical to this resiliency are the ability to grow crops with sustained yield and to preserve soil in the face of extreme flooding, drought, and inconsistent in-season rainfall that result from global climate change (Chou et al., 2015; Lal, 2014). However, it is difficult to relate these functions to commonly measured physical properties such as bulk density, texture, and aggregate stability (Stewart et al., 2018). High interannual variability of establishment of annual CCs due to a relatively short growing season is an additional factor in Minnesota (Strock et al., 2004). Minnesota, the fifth highest agricultural producer in the United States, has become approximately 20% wetter in the past 120 yr, with the top 10 warmest and wettest years all occurring in only the past 19 yr (Arguez et al., 2010; GLISA, 2018). Winters especially are expected to become wetter (U.S. Global Change Research Program, 2017), with consequently larger spring snow melts. Wet, cold, conditions on the fine-textured soils that dominate Minnesota's agricultural landscapes make spring field work difficult and shorten the window for planting fall CCs. In addition, Minnesota has very low no-till or conservation tillage adoption compared with other parts of the Midwest (Azzari et al., 2019) because farmers see tilled, black soil as the fastest path to a dry, warm seedbed for cash crops. Predictions of future growing season precipitation vary, but there is agreement that we may see more extremes in very wet or very dry growing seasons (MN-DNR, 2018; U.S. Global Change Research Program, 2017).

In light of the challenges and opportunities regarding management to promote soil health, our objectives were (a) to measure the response of common NRCS-recommended (USDA-NRCS, 2018) field and laboratory metrics of soil health to short-term (1–3 yr) implementation of CCs on working farms, (b) to assess CC performance and improvements to soil health as reported by Minnesota farmers participating in a CC incentive program, and (c) to evaluate the climate resiliency benefits of short-term CC implementation through rainfall simulation at a subset of sites.

To achieve these objectives, we conducted farmer surveys and field and laboratory assessments using a paired design of farm fields with and without EQIP-funded CC plantings. We then performed simulations of 50-yr storm events to explore climate adaptation benefits of CCs on a subset of six farms. Responsiveness of field and laboratory metrics (researcher observations) and farmer observations of soil changes are defined as two ways to evaluate the “success” of cover cropping to improve soil health. Our study included 15 pairs of fields across Minnesota, representing a range of soil types and specific management practices, so we could explore whether any measure of success in management for soil health transcended those variances.

Core Ideas

• The success of cover crops as a soil health promoting management practice varied across farms.
• Four of 42 soil health indicators had strong positive responses to soil health management.
• Only a small number of physical and biological measurements aligned with farmer observations.

2 MATERIALS AND METHODS

We present data from farmer surveys (n = 27 surveys) and field and laboratory soil assessments (n = 15 paired fields) collected from farms enrolled in the Minnesota SHI. All SHI participants agreed to plant at least two species of CC on the same acres for 5 yr and to retain cover over the winter. Fall tillage was prohibited, but CCs could be used for supplemental forage. The number of acres enrolled, CC species, seeding/termination methods, and planting preparation were decided by producers (Supplemental Tables S1 and S2; Figure 1)

FIGURE 1

Map of four current Natural Resources Conservation Service districts in Minnesota and Soil Health Initiative sampling locations. Yellow circles indicate the county within which farms were sampled. Numbers in each circle represent the number of farms sampled in the given county

2.1 Survey analysis

Minnesota SHI participants were contacted in the summer of 2016. Twenty-six participants, plus a nonparticipant whose farm met study criteria, granted permission to sample, for 27 total surveys. Permission to sample a neighboring non–cover cropped field was also obtained at one location. Pre-sampling surveys included questions about the qualities of healthy soil and past/current management. All survey questions were open ended and thematically analyzed (Braun & Clarke, 2012). For each question, free-text responses were reviewed, and relevant keywords were identified for coding. Keywords and clustering were verified for their relevance to soil health management practices. Questions on methods of CC planting and termination were straightforward. Termination using “herbicide” and RoundUp were combined as chemical termination; mentions of chisel, disc, and field cultivation were combined as mechanical termination. For CCs used in rotation, all mentions of radish (Raphanus sativus L.) (including tillage and oilseed radish) and of rye (including cereal rye; Secale cereale L.) were respectively combined. Responses on CC performance naturally corresponded to a five-point Likert scale from excellent to poor. Responses on noted changes to the ground were most variable. Most terms referred to the physical aspect of soil, including soil structure and infiltration. The mentions of organic matter were counted for both biological and chemical changes to the soil. There were no other mentions of chemical changes. The comments on changes to productivity were attributed to changes in the soil structure, but we chose to keep them separated. The complete list of soil health terms used, and our categorization appears in Table 1.

TABLE 1. Complete list of terms used in farmer's descriptions of changes in soil health. Column headings indicate our categorization of those terms

Improvements in soil health
Physical	Biological	Chemical	Improvement in productivity
firm, feels better underfoot infiltration less compaction less erosion mellow, soft moisture retention prevented soil loss root structure soil aggregate size soil appearance soil structure	darker less weed organic matter smell weed control worms	organic matter	crops looking more even earlier planting healthier cover crop, established earlier

All responses were coded for the presence or absence of the keywords included in the code. In cross-tabulation, responses were reviewed to ensure that only relevant combinations were counted. For example, a farmer may have reported using both broadcast and drill to plant CCS and yielding both fair and excellent crops. However, if the broadcasted crop was fair whereas the drilled crop was excellent, the cross-tabulation did not count the broadcast–excellent and drill–fair combinations. In Table 2, conditional distributions are calculated relative to the total number of farmers that met the criterion in a given row rather than the total number of mentions; thus, the row sums will equal or exceed 100%.

TABLE 2. Reported improvements in soil health and cash crop productivity, in total and coded by cover crop planting method, cover crop type, diversity, and quality of land used for the Environmental Quality Incentives Program

	Improvements in soil health
	Physical	Biological	Chemical	Improvement in productivity
Total (N = 27)
n	18	12	2	6
%	67	44	7	22
By cover crop planting method
Broadcast (N = 15)
n	9	8	2	3
%	60	53	13	20
Drill (N = 15)
n	11	8	1	2
%	73	53	7	13
Aerial (N = 9)
n	9	3	1	3
%	100	33	11	33
Interseed (N = 2)
n	1	1	0	1
%	50	50	0	50
By cover crop
Radish (N = 15)
n	12	8	0	3
%	80	53	0	20
Rye (N = 15)
n	12	9	0	4
%	80	60	0	27
Oats (N = 11)
n	9	6	2	2
%	82	55	18	18
By cover crop diversity
1–2 cover crops (N = 10)
n	7	4	2	1
%	70	40	20	10
3–4 cover crops (N = 6)
n	4	2	0	0
%	67	33	0	0
≥5 cover crops (N = 7)
n	6	6	0	3
%	86	86	0	43
By quality of land
High quality (N = 2)
n	1	1	1	0
%	50	50	50	0
Average quality (N = 22)
n	15	10	1	5
%	68	45	5	23
Low quality (N = 5)
n	4	2	1	2
%	80	40	20	40

• Note. Conditional distributions are calculated relative to the total number of farmers that met the criterion in a given row rather than the total number of mentions; thus, the row sums will equal or exceed 100%.

2.2 Soil sampling

We collected soil samples from 15 farms in the southern two-thirds of Minnesota, spanning much of the corn (Zea mays L.) and soybean [Glycine max (L.) Merr.]-growing area of the state (Figure 1). All had included CCs in their rotation for 1–3 yr before sampling (Supplemental Table S1). Fields with no cover crop (NCC) were provided by the same grower for sampling at 13 farms; a neighbor and a renter provided the NCC field at the other two sites. All NCC fields were selected on the same soil series except one instance at Site 21 (Supplemental Table S1). At Site 21, we paired the geographically associated soils, Port Byron (NCC) and Mt. Carroll (CC), based on the similarity in parent material, texture class, drainage, and slope.

Sampling began on 3 October and, due to record fall warmth, continued into the fourth week of November (MN-DNR, 2021). Each field included a central sampling point and two satellite points, 30 m to the east and west of the central point. Orientation and distance of the satellites were adjusted to stay in the same soil map unit or similar soil, preferably in the same landscape position. Data and samples were collected in-row if possible or from less disturbed areas in fall-tilled NCC fields. Soil samples were collected, processed, and stored appropriately for the given laboratory analysis described below at the central point and each satellite. The CC and NCC fields were sampled on the same day at each farm.

2.3 In-field assessments

Cover crop, crop, and residue covers were estimated from the vantage of the central point. Air temperature and soil temperature at 5 and 10 cm were also measured here. Percent exposed soil was estimated by examining the surface at 1-m increments along a measuring tape running from the central point to each satellite. Penetration resistance was assessed with a Dickey-John penetrometer at the central and satellite points (three replicates). Peak resistance (kPa) was measured in the top 15.2 cm, from 15.2 to 30.5 cm, and below 30.5 cm (to handle).

Infiltration rate (cm h⁻¹) was measured at the central point using two 15.2-cm-diameter, 15.2-cm rings inserted to 7.6 cm. The ring was lined with plastic film to protect the surface as water (2.5 cm precipitation, 464 ml) was added. Timing began as the film was removed and ended when all water was gone or other tasks in the field were complete. If timing was stopped, depth was measured at four points, and the average was deducted from 2.5 cm. Measurements in the first 2.5 cm were made in all fields. Data from the second 2.5 cm were obtained on 10 farms.

Biological activity was evaluated at the central point. The surface was examined for earthworm casts and middens, and worms were counted in a top 30.5-cm cube of soil centered on the row. Soil smell was scored on a three-point scale as follows: (1) no scent, or a sour, rotten-egg, putrid, or metallic/chemical odor; (2) slightly earthy scented; and (3) richly earthy/sweet scented (CA NRCS, 2015). To generate a Visual Evaluation of Soil Structure (VESS) score, one or two observers rated aggregated size and shape, porosity, and root distribution in a block of soil (20.3 cm by 20.3 cm by 15.2 cm) taken from the surface and broken apart by hand. If the block had horizontal layers, they were examined separately to produce a depth-weighted score. Based on Ball et al. (2007), VESS scoring was assigned a number from 1 to 5, with 1 representing a high degree of desirable structural quality for crop production and 5 representing a low degree of structural quality.

2.4 Laboratory analyses

Bulk density (g cm⁻³) was determined at the surface (0–5 cm) and plow pan (often 15.2–20.3 cm) of the soil description pit at the central point. A 7.6-cm-diameter, 5-cm ring was pounded flush into the pit wall and pried out, and the bottom was leveled. If the ring could not be fully inserted, interior height above soil was measured at four points, and the average was deducted from 5 cm. Air-dried cores were weighed, dried (105 °C, 24 h), and re-weighed.

A 15.2-cm cube of soil taken from the surface at the central point was used for water-holding capacity and wet aggregate stability assessment. To determine water-holding capacity, 10 ml of distilled water was added to 10 g of soil on pre-weighed filter paper in a funnel. The volume of water passing through was measured after 15 min. The soil was weighed, dried as above, and re-weighed. Retained volume was converted to grams and added to the difference in wet/dry soil weights. Water-holding capacity (g_H2O 10 g_soil⁻¹) was expressed as a percentage.

Wet aggregate stability was determined using a 50-g subsample of the cube. Samples were placed in the top of a five-sieve shaker stack on an elliptical oscillator, contact-wetted in a tube of water for 10 min, and washed for an additional 10 min (31 oscillations min⁻¹). Sieve contents were rinsed into pre-weighed beakers, dried as above, and weighed. Dry soil weights in the >4- and 2-to-4-mm (stxvvable mega-aggregate) fractions and in the 1-to-2-, 0.5-to-1-, and 0.25-to-0.5-mm (stable macro-aggregate) fractions were measured; the <0.25-mm fraction was determined by subtraction. Air-dried moisture content was measured in a second 50-g subsample and used to correct dry weights. An aggregate stability index (geometric mean diameter) was calculated from the dry weight percentages.

Additional analyses were conducted on bulk samples collected from 0 to 20 cm and from 20 to 50 cm at the central and satellite points of the field. Two spades of soil were dug and mixed at each depth; 400 g was bagged as a fresh sample, and 150 g was placed on ice and frozen as soon as possible. A fresh subsample of 10 g from each depth was weighed, dried as described above, and re-weighed to determine gravimetric soil moisture content.

Fresh samples were air-dried. Shallow (0–20 cm) subsamples were sieved to 2 mm for analysis at the University of Minnesota Research Analytical Lab. Soil organic matter (%) was measured by loss on ignition (University of Minnesota Research Analytical Laboratory, 2021a). Extractable phosphorus (mg kg⁻¹) was measured with the Bray method (University of Minnesota Research Analytical Laboratory, 2021b) and the Olsen method in calcareous soils (University of Minnesota Research Analytical Laboratory, 2021c). Available potassium was determined by ammonium acetate extraction (University of Minnesota Research Analytical Laboratory, 2021d). Nitrate was determined by calcium sulfate extraction (University of Minnesota Research Analytical Laboratory, 2021e). Shallow and deep (20–50 cm) subsamples were ground and analyzed for total percent carbon and nitrogen using a combustion analyzer (Elementar, Americas, Inc.). The soil C/N ratio was then calculated using total carbon and nitrogen data.

Soil respiration (mg CO₂–C kg⁻¹) was assessed with the Solvita Burst-CO₂ test, 50% water-filled pore space (WFPS) method. The volume of 40 g of fresh soil (0–20 cm), sieved to 4 mm and dried (40 °C, 24 h), was used to determine the amount of deionized water needed for 50% WFPS. The soil was wetted in a Solvita sample jar. A CO₂ probe was added, and the sealed jar was incubated at 24 °C for 24 h. Probes were read with a Solvita Multi-Mode Color Reader.

Organic nitrogen (mg NH₃–N kg⁻¹) was assessed with the Solvita Labile Amino Nitrogen test. Four grams of soil (0–20 cm), sieved to 4 mm, was put in a 50-ml plastic beaker in a Solvita sample jar, and 10 ml of 2 N NaOH was applied to the soil. A NH₃ probe was placed inside the beaker, and the sealed jar was incubated at 20 °C for 24 h. Probes were read with a Solvita Multi-Mode Color Reader. One deep fresh sample served as a standard for both tests.

Extracellular hydrolytic and oxidative enzyme activity in frozen soil samples (0–20 cm) was measured by fluorimetric assay with 4-methylumbelliferone (4-MUB)–linked substrates (German et al., 2011; Sinsabaugh et al., 2003). Beta-glucosidase and cellobiohydrolase (cellulose degradation), N-acetylglucosaminidase (nitrogen acquisition), and phosphatase (phosphorus acquisition) were tested with 4-MUB–β-glucopyranoside (300 μmol L⁻¹), 4-MUB–cellobioside (200 μmol L⁻¹), 4-MUB–N-acetyl–β-glucosamide (200 μmol L⁻¹), and 4-MUB–phosphate (300 μmol L⁻¹), respectively. Substrate stocks (Sigma-Aldrich) were diluted with Milli-Q water.

Each soil sample was tested on a black 96-well plate with four substrate blanks (four replicates, 50 μl), four enzymes (eight replicates, 50 μl of each substrate), methylumbelliferone (MUB) quench coefficient dilution series (two replicates), soil homogenate blank (16 replicates), and 50 mM Tris buffer blank (eight replicates). One MUB emission dilution series (two replicates) plus Tris buffer blank (eight replicates) plate was run per day. A 200-μM MUB stock solution was diluted with Milli-Q water to 2.5, 1.25, 0.625, and 0.16 μmol L⁻¹ (to 50 μl) for the quench and emission coefficient series. Emission, substrate (50 μl), and homogenate blank wells were brought to 250 μl with Tris buffer.

To prepare the homogenate, 0.5 g of soil was sonicated for 5 min in 50 ml of Tris buffer, and 200 μl was dispensed into the assay, MUB, and homogenate blank wells of the assay plate. Plates were covered and incubated at room temperature for 1 h, after which 10 μl of 1 M NaOH was added to each well. Plates were immediately analyzed on a Biotek Synergy HT plate reader using Gen 5 software (Biotek Instruments, Inc.), with excitation at 360 nm and emission at 460 nm. Enzyme activity was expressed in nmol (h × g_soil) ⁻¹.

Soil particle size analysis of the <2-mm fraction was determined by wet sieving (sand fraction) and hydrometer (clay fraction) methods. Separation and quantification of the sand fraction by wet sieving followed method 3A1a1 described by Soil Survey Staff (2014). Percent clay was determined by the hydrometer method described in Grigal (1973). Percent silt was determined by difference (% silt = 100% − % sand − % clay). The ssc_to_texcl() function in the R package aqp (Beaudette et al., 2021) was used to assign each soil to a USDA textural class (Supplemental Table S1).

2.5 Responses to an extreme weather event

We simulated a 50-yr storm on six farms, using the NOAA 14 Atlas (https://hdsc.nws.noaa.gov/hdsc/pfds/) to estimate the intensity and volume of a 50-yr rainfall in a 1-h duration. The range was 14.1–16.7 L of rain in a 0.25-m² area, which was the area of mini-runoff plots (Seitz et al., 2015, 2016) used to determine the amount and composition of runoff from each event. To perform the rainfall simulation, a 100-L liter tank was outfitted with an adjustable sprinkler head and calibrated to deliver the rate of water desired in each field. Mini-runoff plots were placed 5 cm into the soil, and a hole was dug in order to place an 18-L bucket below the outflow tubing. Runoff was collected as possible during and after each event until the runoff stopped flowing. The total volume of runoff was then recorded, followed by filtering the runoff (Whatman #1, 0.45-μm pores) to determine the sediment weight. The concentrations of total organic carbon, total nitrogen, and total phosphorus were measured in the solution. Total organic carbon and nitrogen were determined on a Shimadzu liquid analyzer. Total phosphorus was determined at the University of Minnesota Research Analysis Laboratory using standard methods.

2.6 Statistical analysis

We evaluated 15 pairs of fields across Minnesota that varied in soil texture and farm management (Supplemental Table S1). For biological and chemical laboratory analyses, assays were repeated on three subsamples from each field, whereas physical analyses and field assessments were conducted once per field (Table 3). The rainfall simulation was conducted (n = 1) on six pairs of fields. To evaluate differences in all metrics between CC and NCC fields, we calculated effect size as the natural logarithm of the response ratio (LRR, ln CC/NCC) and report this with a 95% confidence interval (CI) across the 15 paired fields (six for rainfall simulation metrics). The LRR variance and 95% CI were calculated with Equations 1 and 2 from Hedges et al. (1999). If the 95% CI did not overlap with 0, we considered the effect of CC significant and report both LRR and 95% CI alongside the back-transformed percentage change between CC and NCC, calculated by taking the antilog of the LRR estimate and CI limits (Hedges et al., 1999). The ordinal measure VESS (1–5 scale) was treated as a continuous variable for the purposes of the analysis (Johnson & Creech, 1983; Norman, 2010). Differences in soil smell scores (1–3 scale) between CC and NCC fields were evaluated with a Fisher's exact test of independence. In order to explore reasons for variability across sites, we include a two-way ANOVA including CC treatment and soil texture in Supplemental Table S3. Although texture and the texture × cover crop interaction were significant for a few metrics, these effects primarily represented differences between two sandy loam pairs (035 and 024) and all other farms without illuminating the important control of texture on effects of CCs. Given the unbalanced representation of different textures, we focused only on the LRR of CC treatment, which was slightly more conservative in classifying significant effects of CC than the ANOVA. Unfortunately, management data such as long-term rotation and tillage intensity were not collected from NCC fields. This limits exploration of the factors expected to alter soil properties but does allow us to address the question of whether and which soil properties may be altered by CCs irrespective of other factors. The LRR is a conservative but sufficient approach to detect any broad effect of CCs on dynamic soil properties across this diverse range of sites.

TABLE 3. Summary statistics for soil health metrics assessed across sites

Measurement	Cover crop	No cover crop	Overall LRR (95% CI)
Physical laboratory analysis (n = 1)
Aggregate index	1.29 (0.743)	1.37 (0.617)	−0.042 (−0.448 to 0.365)
Microaggregates, % soil mass <0.25 mm	18.4 (5.95)	17.5 (4.9)	0.051 (−0.191 to 0.292)
Macroaggregates, % soil mass 0.25–1 mm	41.1 (15.2)	38.2 (13.4)	0.047 (−0.233 to) 0.327)
Mega-aggregates, % soil mass >1 mm	40.5 (18.9)	44.4 (16.3)	−0.062 (−0.38 to 0.256)
Bulk density at 0–5 cm, g cm^−3	1.27 (0.191)	1.23 (0.201)	0.029 (−0.095 to 0.154)
Bulk density at plow pan (∼15–10 cm), g cm^−3	1.4 (0.135)	1.39 (0.183)	0.001 (−0.091 to 0.094)
Soil water content at 0–5 cm, g g^−1	0.138 (0.0525)	0.106 (0.0568)	0.251 (−0.126 to 0.628)
Soil water content at plow pan (∼10–15 cm), g g^−1	0.118 (0.0723)	0.119 (0.0792)	0 (−0.524 to 0.524)
Total water-holding capacity, %	58.4 (10.7)	59.8 (6.7)	−0.018 (−0.141 to 0.105)
Biological laboratory analysis (n = 3)
β-glucosidase activity, nmol h^−1 g dry soil^−1	87 (41.7)	63.7 (39.4)	0.267 (−0.069 to 0.604)
Cellobiohydrolase activity, nmol h^−1 g dry soil^−1	9.69 (5.11)	6.08 (4.13)	0.452 (0.056 to 0.849)
N-acetylglucosaminidase activity, nmol h^−1 g dry soil^−1	23 (10.3)	20.5 (11)	0.05 (−0.294 to 0.394)
Phosphatase activity, nmol h^−1 g dry soil^−1	378 (191)	295 (166)	0.209 (−0.195 to 0.614)
Soil organic matter, %	4.01 (1.39)	3.78 (1.3)	0.059 (−0.209 to 0.327)
Total soil C, %	2.07 (0.892)	2.17 (0.741)	−0.02 (−0.309 to 0.269)
Total soil N, %	0.155 (0.0907)	0.189 (0.0802)	−0.117 (−0.467 to 0.232)
C/N	25.5 (27.8)	18.2 (20.5)	0.232 (−0.625 to 1.089)
Total soil C at 20–50 cm, %	1.43 (0.956)	1.65 (0.796)	−0.099 (−0.525 to 0.328)
Total soil N at 20–50 cm, %	0.104 (0.0713)	0.133 (0.0771)	−0.188 ([−0.637 to 0.26)
C/N at 20–50 cm	23 (23.3)	20.7 (21.1)	0.117 (−0.532 to 0.767)
Solvita C (CO2–C), mg kg^−1 soil	92.4 (29.6)	95.8 (23.1)	−0.01 (−0.24 to 0.22)
Solvita N (NH3–N, mg kg^−1 soil	160 (45.8)	156 (38.7)	0.014 (−0.242 to 0.269)
Chemical laboratory analysis (n = 3)
pH	6.68 (0.861)	6.54 (0.853)	0.016 (−0.082 to 0.114)
Nitrate (mg L^−1)	21.4 (11.6)	19.1 (9.17)	0.126 (−0.244 to 0.496)
P (mg L^−1)^a	39.8 (48)	30.1 (26.9)	0.374 (−0.463 to 1.211)
K (mg L^−1)	156 (65.7)	146 (58)	0.078 (−0.228 to 0.383)
Field assessments (n = 1)
Bare soil, %	0.182 (0.126)	0.325 (0.247)	−0.62 (−1.215 to −0.025)
Soil temperature at 5 cm	48.2 (6.2)	51.3 (5.82)	−0.072 (−0.172 to 0.028)
Soil temperature at 10 cm	47.4 (5.28)	48.7 (5.26)	−0.03 (−0.122 to 0.063)
Penetration resistance at 15 cm, PSI	161 (43.2)	168 (48.9)	−0.022 (−0.245 to 0.201)
Penetration resistance at 30 cm, PSI	227 (70.9)	227 (62.8)	0.014 (−0.203 to 0.232)
Earthworm count, n	21.1 (14.7)	8.48 (6.93)	0.693 (0.127 to 1.259)
VESS, 1–5 scale	2.7 (0.8)	3.5 (0.6)	−0.26 (−0.45 to −0.069)
First 2.5 cm of infiltration, cm h^−1	38 (53.3)	31.7 (37.6)	0.321 (−0.732 to 1.374)
Second 2.5 cm of infiltration, cm h^−1	10.5 (7.17)	7.98 (6.47)	0.318 (−0.363 to 1)
Rainfall simulation (n = 1 per site at six sites)
Runoff volume, ml 0.25 m^−2	4,680 (2,530)	4,730 (3,770)	−0.01 (−1.021 to 1)
Runoff sediment, mg ml^−1	5.38 (2.94)	10.7 (10.4)	−0.691 (−1.862 to 0.479)
Total sediment per 0.25-m^2 plot, g	25.2 (20)	26.9 (13.3)	−0.067 (−1.047 to 0.913)
Runoff total N, mg L^−1	214 (139)	201 (67.3)	0.063 (−0.705 to 0.831)
Runoff total C, mg L^−1	66.5 (23.5)	73.9 (41.5)	−0.104 (−0.801 to 0.592)
Runoff total P, mg L^−1	0.503 (0.933)	0.102 (0.0571)	1.6 (−0.432 to 3.632)

• Note. Means and SDs are expressed for cover cropped (CC) and non-cover cropped (NCC) fields. Overall log response ratios (LRRs) are reported (ln CC/NCC) with 95% confidence intervals in parenthses. Samples were taken from 0–20 cm unless otherwise noted. Number of samples (n) per site is noted for each category of analyses. VESS, Visual Evaluation of Soil Structure.
• ^a Bray P was used for pH <7.4, Olsen P for pH >7.4

3 RESULTS

3.1 Farm survey responses

Responses showed a range of spring tillage in CC fields (conventional, reduced, strip-till, no-till) (Table 2). Tillage was not always reported and so was left out of our analysis. Most (81%) participants described their enrolled CC acres as of average quality. Five enrolled less productive, eroded or sandy land; two chose high-quality fields.

Farmers described experimentation with cover crop planting methods (Supplemental Table S4). The two most popular methods were broadcast and drill (each mentioned by 65% of respondents with CC acres), followed by aerial application (39%). All planting methods yielded a range of CC performance. Nearly 80% of farmers who drilled reported “good” or better outcomes (29% outcomes rated as “excellent”), compared with broadcast or aerial applications that were associated with “fair” or “poor” outcomes 42 and 33% of the time, respectively. Regarding methods of CC termination, 61% indicated chemical treatment in spring prior to cash crop planting, followed by winter kill (39%) and mechanical termination in spring (26%) (Supplemental Table S5).

Corn was the most popular crop in rotation, planted by 81% of farmers, followed by soybeans (59%) and wheat (22%) (Supplemental Table S2). Twenty-three farmers with CC acreage planted on average 3.2 CC species, with 10 reporting one or two species and 13 reporting five or more species. The most common species were radish (65%), rye (65%), and oat (Avena sativa L.; 48%), followed by rape (Brassica napus L.; 26%), wheat as a cover crop (Triticum aestivum L.; 26%), crimson clover (Trifolium incarnatum L.; 22%), peas (Pisum sativum L.; 22%), and turnip (Brassica rapa L. ssp. rapa; 22%).

Farmers in the program were asked to report changes in CC fields. Twenty-three farmers, including eight 1-yr participants, noted changes associated with better soil health or yield (Table 2). Two-thirds noted improvements in physical aspects of soil health, ranging from appearance to observable structural changes. Almost half (44%) reported biological changes, which are detectable by inspection, earthworm presence, smell, or changes in weed pressure. About one-quarter (22%) reported improved productivity, mostly related to their ability to plant cash crops earlier, allowing crops to establish better. Cross-tabulations of observed improvements by CC planting method and CCs show that productivity improvements were not associated with any particular planting method or CC species. Regardless of planting methods or CC species, the majority of farmers perceived improvements in physical aspects of soil, and roughly half reported changes in biological aspects of soil. Cross-tabulation with CC variety shows that farmers who adopted a greater variety of CC species were more likely to report improvements in both physical and biological nature of soil. Cross-tabulation with self-reported quality of land shows that farmers noticed improvements even on low-quality land (Table 2).

3.2 Field and laboratory assessments of soil health

The effect of cover cropping was significant for only 4 of 42 metrics assessed, including cellobiohydrolase activity, earthworm counts, VESS scores, and percentage bare ground (Figures 2 and 3; Table 3). Soil smell scores, which were evaluated with Fisher's exact test for independence, did not differ between CC and NCC fields (p = .1) (Supplemental Figure S1; Supplemental Table S6).

FIGURE 2

Physical and field soil health indicators. Log response ratios (LRR) for laboratory soil health indicators where the number of replicates per field were 1 (the central location). See Table 3 for LRR point estimates and 95% confidence intervals for each variable. VESS, Visual Evaluation of Soil Structure

FIGURE 3

Laboratory soil health indicators (SHIs). Log response ratios (LRR) for laboratory SHIs with three replicates per field (one central and two satellite locations). BG, β-glucosidase; CBH, cellobiohydrolase; NAG, N-acetylglucosaminidase; Pase, phosphatase. See Table 3 for written log response ratios and 95% confidence intervals for each variable

Cellobiohydrolase activity was greater in CC fields, with LRR 0.452 (95% CI, 0.056–0.849), representing an estimated 57% higher activity in CC fields compared with NCC fields (95% CI, 0.06–1.34). Earthworm counts were higher in CC fields with LRR 0.693 (95% CI, 0.127–1.259), representing 100% more earthworms (95% CI, 0.14–2.52). The VESS scores, for which lower values indicate better structure, were lower in CC fields, with LRR −0.26 (95% CI, −0.45 to −0.069), representing a 23% decrease in score for CC fields (95% CI, 0.36 to −0.07). The percentage of bare soil was significantly lower in CC fields, representing an estimated 46% decrease in bare soil (95% CI, −0.02 to −7.0). Variability was too high to establish significant effects of cover cropping for measurements including infiltration, soil water content, β-glucosidase activity, N-acetylglucosaminidase activity, P enzyme activity, and smell test, where trends of increase were observed (Figures 2 and 3; Supplemental Figure S1; Table 3; Supplemental Table S7). For many measurements, there were not even numeric trends toward improvement with CCs (e.g., Solvita, bulk density, total C and N).

Rainfall simulation data were too variable to detect any effect of CCs, but sediment loss trended lower with CCs, whereas P in runoff trended higher (Figure 4).

FIGURE 4

Measurements taken as part of rainfall simulations. Log response ratios (LRR) are shown for each variable. See Table 3 for written LRRs and 95% confidence intervals for each variable

4 DISCUSSION

In this study we conducted a broad survey of field and laboratory soil health assessments, including both NRCS standard and widely available commercial tests (e.g., Solvita) across farms in Minnesota using a paired field design. We found few consistent results across farms using an effect size approach but also found that metrics with statistically significant responses to CCs aligned well with farmer observations of soil health. Here we discuss the potential to align farmer and researcher observations, the variability in results across farms, and the significant and positive effects of management for soil health that did transcend the variability we observed.

Where we saw few significant effects from a researcher perspective, our survey results demonstrated that a majority of farmers observed improvements in soil biological and physical characteristics, including smell, earthworms, soil structure (including the soil being more “mellow”), and infiltration, among others. These positive results were observed regardless of CC seeding or termination method, suggesting that simply having multiple species of continuous cover can improve farmer-observable soil health outcomes. The specific practice of having more (5+) species varieties present was related to a greater likeliness of reporting positive soil health outcomes. Although monocultures of CCs are often higher biomass, suppressing weeds and sequestering nitrogen, diverse cover crop mixes have been found to supply nitrogen, and diverse crop rotations have been shown to build soil health (Finney et al., 2017; McDaniel et al., 2014; Saleem et al., 2020).

Our field and laboratory assessments aligned somewhat with farmer observations, where we found similar improvements in earthworm counts, soil structure (VESS), and one measure of soil enzyme activity (cellobiohydrolase). We found marginal improvements in soil smell and infiltration and highly variable results of soil aggregation, which were all improvements specifically mentioned in farm surveys when describing soil health improvements. Limitations with the soil smell test protocol may have led to less significant results. Distinguishing “how earthy” two samples smelled, when appraised hours apart, could lead to inconsistent results. Performing this test on both or many samples at once would have been more robust. Farmers have the closest knowledge of how their land is changing, relying on sensory observations (Romig et al., 1995), whereas scientists focus more on the mechanisms or reasons behind those changes (Carlisle, 2016). Farmers also must make decisions based on farm economics and other socioeconomic factors (Carlisle, 2016). Had we begun with a more detailed account of farmer observations and information about how the soil on their farms is changing, this could have guided and informed our analysis approach to better understand what farmers are seeing in terms of improved soil qualities. Farmers, researchers, and agricultural professionals could work together more effectively to create positive outcomes on farming landscapes by integrating sensory observations and analytical methods, both of which should tie back to farmers’ goals and decision-making processes.

Our results show a range of positive and negative effect sizes for most soil health metrics, illustrating a wide range of outcomes on farms using CCs in the short term (1–3 yr) (Figures 2 and 3). There are many reasons that we could have observed this wide range in outcomes, including variability in management practices not captured in our data collection (tillage intensity, CC planting date and establishment, prior use of CCs in the rotation; Sprunger et al. [2021]). Incomplete information about management practices across our 15 participating farms is one limitation in this study. Because many soil properties are expected to change slowly, it is likely that a history of cover cropping would yield stronger positive results. Measurable improvement in soil organic matter, for example, may not be seen for 3–5 yr after soil health management practices are implemented (Myers et al., 2019; Sullivan et al., 2019).

We did not specifically seek information on tillage intensity in NCC fields, preventing statistical analysis of tillage effects, but suspect that widely varied tillage practices in the CC fields are a primary reason for varied results. As noted above, the SHI program only prohibited fall tillage in CC fields because the main focus of this program was on CC implementation. Participants used a range of tillage practices, largely depending on their existing farm cropping systems and goals and the cash crop the year we sampled. Increased tillage intensity has been found to reduce biological soil indicators but does not cause uniform effects (Nunes et al., 2020). In addition, tillage practices on working farms may use similar equipment but cause different levels of disturbance (varying tillage depth or speed), and research to date does not describe the effects of these nuances (Derpsch et al., 2014).

In northern climates like that of Minnesota, the window of opportunity to plant CCs is small and variable (Strock et al., 2004). As a result, it can be challenging to establish a significant amount of CC biomass, which has ramifications not only for soil properties but also for the ecosystem services expected from CC systems (Finney et al., 2016). The farmers in our study reported mostly good or very good CC performance, with planting method being a driver (Supplemental Table S4, where better performance was reported with drill planting). Earlier harvested cash crops such as sweet corn generally allow for earlier CC planting and more options for planting methods, leading to greater opportunities for CC growth and cover (Finney et al., 2016). Although the paired fields we sampled generally included the same cash crop, in five cases they did not (Supplemental Table S1). Among the metrics this could have influenced was earthworm populations because fresh soybean residue has been found to be an attractive short-term food source for earthworms, whereas corn residue supported larger populations (Abail & Whalen, 2018).

We did not have sufficient data to capture local edaphic factors (Amsili et al., 2021), but soil texture and the composition of past cropping systems could also have had an influence on observed differences in soil health measurements (Tu et al., 2021). Accounting for differences in soil forming factors is also relevant on a regional scale, where differences in parent material, native vegetation, and other initial conditions influence how soils respond to management changes (Zuber et al., 2020).

Given these and other factors that could influence the degree to which soil health metrics change on a given farm, on-farm soil health monitoring should be designed to take them into consideration. For example, sampling could be taken not only with consideration of soil mapping units, as we did, but with consideration of slope and other spatial patterning to separate management effects from inherent soil variability caused by differences in soil forming factors (Adhikari et al., 2021; Tu et al., 2021; Zebarth et al., 2019). The CC and NCC sampling sites in our study were usually in similar landscape positions, but the soil mapping unit had precedence. Temporal variability in soil health indicators is another area that requires further research and is of particular interest for soil biological characteristics, which are growing in popularity and commercial test availability (Bunemann et al., 2018). Soil biological characteristics can vary on multiple spatial and temporal scales, potentially complicating their interpretation unless these variations are constrained and accounted for (Leitner et al., 2021; Wickings et al., 2016). This could especially be the case for the Solvita Burst soil respiration test, which can vary even with small changes in conditions over time or with how the test is conducted in the laboratory (Brinton et al., 2019; Tu, 2016). Despite its commercial success, wide availability, and low cost, Solvita is highly variable and inconsistent in detecting CC treatment (e.g., Chahal & Van Eerd, 2018) and is perhaps not ideal for field use.

Although our limited rainfall simulation was not able to detect significant CC effects, this remains an important arena of research. Response to rainfall was of indirect importance to farmers because they cited earlier planting windows as drivers of increased productivity, and in Minnesota spring planting is often delayed by waiting for soil to dry after rain. Future research controlling for tillage could help illuminate responses to rainfall because intense tillage increases soil particle liberation (Hou et al., 2018), but tillage was not a clear driver of response to rainfall in our dataset.

Although most soil health measurements undertaken produced variable responses to soil health management, we did observe significantly higher biological activity in the form of earthworm counts and one of the carbon acquisition extracellular enzymes and saw positive trends for other extracellular enzyme activities. Biological indicators could be correlated to total plant inputs; for example, Strickland et al. (2019) found that several soil and soil microbial characteristics were related to CC biomass in their study of four working farms in Virginia. Other studies have found similar positive results of cover cropping (Johnson et al., 2010) and reduced tillage (Nunes et al., 2020). Understanding how soil biological communities mediate soil health is an actively discussed topic in need of more attention (Fierer et al., 2021; Wander et al., 2019). We also observed a reduction in two other field assessments: bare soil and the soil structure (VESS) scores. The reduction in bare ground is logical and is in agreement with the CC success that farmers reported. It is interesting that we observed better soil structure from the field assessment but found no significant improvement in other physical properties such as infiltration or wet aggregate stability. Simple ring infiltration measurements are notoriously variable, and it has been reported that at least 10 measurements should be conducted to provide representative measurements (Haws et al., 2004). Our aggregate stability protocol applied a uniform disruptive force to all soils, which likely contributed to high variability in aggregate distribution because different soil types lead to stronger aggregates (Jastrow & Miller, 1991). Although the success of VESS is subjective, it was performed on a block determined to be representative of the field by an experienced soil scientist, perhaps more similar to the experienced farmer evaluating field-scale response to a new practice.

5 CONCLUSIONS

We characterized field and laboratory assessments of soil health on 15 field pairs and analyzed survey data collected from 27 farms enrolled in a Minnesota NRCS soil health EQIP incentive program. Our data showcase the need to acknowledge that the range of management decisions involved in implementing a CC system can lead to varied soil health outcomes, including tillage, cash crop rotation, and planting timing. In addition, site-specific edaphic factors (e.g., soil texture, topography, and local climate) interact with management factors to produce varied outcomes. We did find that biological metrics including earthworm abundance and microbial extracellular enzyme activity (cellobiohydrolase) and the field metrics VESS and % bare ground transcended this variance, suggesting that these metrics could be early signals of improved soil health. We suspect that these improvements are tied to successful CC implementation and that multiple seasons of successful CC growth would expand the list of responsive metrics. We recommend using these indicators and co-creating sampling regimes with farmer input and observations in order to assess the effectiveness of incentive programs such as the NRCS EQIP program. Climate resiliency should be assessed in more controlled field settings as well as through farmer surveying because farm logistics are tightly tied to annual climate fluctuations. By working together, researchers and farmers can use more effective metrics and take appropriate consideration of spatiotemporal variability across farms to bring us closer to understanding soil health.

ACKNOWLEDGMENTS

The authors thank partners, including Kristin Brennan, NRCS Assistant State Soil Scientist; Sienna Nesser, who provided references and a summary of farmer perceptions of soil health; Carol Loopstra, who supported laboratory analysis; and NRCS Area and Field Office staff who assisted with sampling and field assessments (Area Resource Soil Scientists Dan Nath, Jennifer Hahn, Jennifer Smith, Mike Walcynzski and Brandon DeFoe; District Conservationists Shannon Rasinski, Julie Salmon and Josh Bork; Donna Walters, Soil Conservation Technician, and Blayne Doty, ACES Soil Health Coordinator). Mari Mumford coded all farmer surveys for qualitative analysis. We also thank two anonymous reviewers whose suggestions greatly improved this manuscript.

AUTHOR CONTRIBUTIONS

Jessica Gutknecht: Conceptualization; Funding acquisition; Investigation; Methodology; Project administration; Supervision; Writing – original draft; Writing – review & editing. Hava Blair: Data curation; Formal analysis; Visualization; Writing – original draft; Writing – review & editing. Ann Journey: Data curation; Formal analysis; Investigation; Methodology; Project administration; Visualization; Writing – original draft; Writing – review & editing. Hikaru Peterson: Formal analysis; Methodology; Writing – review & editing. Anna Cates: Data curation; Formal analysis; Visualization; Writing – original draft; Writing – review & editing.

CONFLICT OF INTEREST

There are no conflict of interest reported by any author associated with this manuscript.

Open Research

DATA AVAILABILITY STATEMENT

Laboratory analyses performed at the University of Minnesota are being prepared to become available publicly with acceptance of this manuscript. NRCS collected farmer surveys and field soil assessments are under their jurisdiction.