Replacing bare fallow periods with cover crops in dryland agroecosystems can help reverse soil degradation and control erosion but may also result in cash crop yield penalties due to water limitations. Two field trials were conducted on the Colorado Plateau to quantify the impact of cover cropping on crop production, multiple ecosystem services, and economic trade-offs in this semi-arid region. No-till and different cover crop planting windows (fall and spring) were explored as strategies to optimize ecosystem service provision and productivity trade-offs. After three full cover crop cycles (6 years), fall-planted cover crops improved soil structure and erosion control, but associated wheat yield penalties (48% decrease on average) and additional costs caused a 176 USD ha^-1 cycle^-1 average decrease in net returns. However, including the hypothetical sale of forage (based on 50% cover crop biomass removal) more than offset these costs in fall-planted treatments; with both forage and wheat revenue, cover cropping increased net returns as compared to the fallow control by 92 USD ha^-1 cycle^-1. Spring-planted cover crops presented a lower productivity trade-off (24% average wheat yield penalty) but did not provide clear ecosystem service benefits and did not produce enough biomass to offset costs of cover cropping. Our findings indicate that fall-planted cover crops have the potential to reverse soil degradation and control erosion in dryland systems globally, but productivity trade-offs and decreased economic returns must be compensated for by alternative revenue sources, conservation payments, or other incentives to ensure their feasibility. 

Experimental Design

Field trials were implemented at the Colorado State University Southwestern Colorado Research Center near Yellow Jacket, Colorado (37°32′ N latitude; 108°44′ W longitude; 2100 m.a.s.l.) on the Colorado Plateau. This region is semi-arid with an average rainfall of 370 mm yr-1 (yearly average from 1991-2020) and mean monthly high temperatures ranging from 3 °C in January to 31 °C in July (PRISM Climate Group, 2014). The dominant soil type is a Wetherill loam (fine-silty, mesic Aridic Haplustalfs; 36% sand, 41% silt, and 22% clay) (Soil Survey Staff, 2020) with low organic matter content (0.81% SOC) and neutral pH of 6.9. Prior to trial establishment, fields were in dryland, conventionally tilled production since 2010. Additional details regarding field history can be found in Eash et al. (2021), which presented the short-term productivity results of these trials. 

Two field trials were established to compare the common local two-year winter wheat-fallow rotation with a winter wheat-CC (cover crop) rotation. The first trial (T1) was established in 2015 and the second (T2) was established in 2016 in an adjacent field to ensure that both crop phases were represented each year. T1 was under no-till management and followed a randomized complete block design with three replicate blocks. Plots in T1 were 6 m wide x 61 m long, or 372 m² each. T2 followed a split-plot design to also evaluate tillage effects (no-till vs. conventional tillage) and CC planting window (fall- vs. spring-planted). Tillage treatments were randomly allocated to three replicate blocks and CC treatments were randomly allocated to subplots (3.7 m x 30.5 m, or 113 m²).

Conversation with local producers and stakeholders informed CC mixture selection, all of which were planted at a target seeding rate of 34 kg ha^-1. While various CC mixtures were evaluated within the trials, dry conditions caused only grasses and winter pea to dominate, similar to observations by Holman et al. (2020). Since stand expression was very similar among CC treatments of the same planting window, this analysis focuses on planting window rather than CC mixture, though a full list of CC mixtures and analysis among mixture treatments are presented in the "Cover Crop Mixtures" tab in the attached dataset.

Winter wheat (hard red winter wheat variety ‘Fairview’) was planted at a rate of 56 kg ha^-1 in September and harvested in July. Control treatments remained fallow in between winter wheat crops. In fall CC treatments, CC were planted in September and terminated in June, while spring CC were planted in April or May and terminated in June or July.

In no-till plots, glyphosate (N-(phosphonomethyl)glycine) and 2,4-D amine (2,4-dichlorophenoxyacetic acid) were used to control weeds and terminate CC. In tilled plots, this was done mechanically using a tandem disk or field cultivator. 

Wheat and Forage Yields

Wheat was harvested in July of each year from subplots in the center of each plot to avoid edge effects using a Hege plot combine (1.2 m width; 6 rows). Wheat grain was then cleaned using an electric winnower, weighed and tested for moisture and density. All wheat yields are reported at 11% moisture content.  

Forage production was estimated to be 50% of CC biomass per Natural Resources Conservation Services (NRCS) guidelines (NRCS, 2023). Cover crop biomass was measured at CC termination using a 0.25 m² quadrat with three subsamples randomly distributed per plot. Biomass was carefully cut at the soil surface (~1 cm height) and the three subsamples were combined. Biomass was then dried at 60°C and weighed.

Dried CC biomass from all treatments was finely ground and analyzed for forage quality using near-infrared reflectance spectroscopy. 

Soil Sampling

Soil samples were collected at wheat planting and CC termination to evaluate supporting services (N supply, soil water availability, and soil structure). Approximately one week prior to wheat planting and one week after CC termination, two subsamples per plot were taken 30 cm increments using a tractor-mounted Giddings probe (5 cm diameter) to a depth of at least 90 cm and separated into 0-15 cm, 15-30 cm, 30-60 cm, and 60-90 cm depth increments. The target sampling depth was 180 cm, but the probe could not always penetrate to this depth due to dry conditions in the sub-surface soils. Subsamples were composited by depth increment for each plot and a 50 g subsample of each depth increment was weighed, dried at 105°C and reweighed to determine gravimetric moisture content. Gravimetric moisture content was then converted to volumetric moisture content by multiplying by the bulk density that corresponded to the sampling depth (1.35 g cm^-3 for the 0-30 cm depth, 1.40 g cm^-3 for the 30-60 cm depth, and 1.45 g cm^-3 for the 60-90 cm depth; based on unpublished data averaged across the study site prior to trial establishment). A subsample from the 0-30 cm depth increment sampled at wheat planting was air-dried and sent to Ward Laboratories in Kearney, NE for nitrate analysis using a flow injection analyzer (Keeney and Nelson, 1982).

To assess wet aggregate stability, soil was sampled to a depth of 5 cm using a 6.7 cm diameter core at two sampling points per plot at termination of the third CC cycle (2020 in T1 and 2021 in T2). Subsamples were combined, gently passed through an 8 mm sieve, and air-dried to further analyze aggregate stability by wet sieving, following methods adapted from Elliott (1986). In summary, a 40 g sample was placed on a 2 mm sieve and submerged in deionized water for 5 minutes for slaking. The sieve was then oscillated up and down (in and out of the water) 50 times during a 2 min period. Soil passing through the sieve was subsequently transferred to a 250 mm, and a 53 mm sieve. Soils remaining on each sieve were rinsed into a pan and were dried at 60 °C. The proportions of whole soil in each aggregate size fraction and average diameter for each fraction were used to calculate mean weight diameter according to van Bavel (1950). 

Soils sampled to 15 cm at CC termination were frozen and processed for 16S rRNA amplicon sequencing using an Illuma Miseq platform. Soil DNA was extracted from 0.50 g of samples using the Powersoil® DNA Isolation Kit (Mo Bio Laboratories, Carlsbad, CA, USA) according to manufacturer instructions. To characterize prokaryotic diversity and community composition we used the primer sets 515F/806R (Caporaso et al., 2012) that amplifies a portion of the bacterial 16S rRNA gene.  A combination of USEARCH (Edgar, 2010) and UNOISE3 (Edgar, 2016) was used for bioinformatics processing. The USEARCH pipeline was used to generate Amplicon Sequence Variant (ASV) tables based on 97% sequence similarity. Sequencing run quality was assessed using fastQC (Andrews, 2010), and raw sequences were discarded if they were short in length (<100 bp), had a low quality score (Q < 20), or contained ambiguous nucleotides. Adapters and primers were removed using cutadapt (Martin, 2011). To demultiplex samples, paired-end reads were merged, and an initial quality check test was performed. The representative set database was created using the UCLUST and UPARSE algorithm (Edgar, 2013). Unique sequences were located and identified as unique ASVs, which were then clustered using DADA2 and DeNoised using uNoise3 (Xiong et al., 2021a, 2021b). ASV tables at the sample level were generated by mapping reads to the the Silva database (Pruesse et al., 2007). Bacterial sequences that match host mitochondria and chloroplast were removed. ASV abundance tables were rarefied to the median count per sample to ensure relatively equal sampling depth. Alpha diversity (Shannon index) and species richness were calculated using the phyloseq package in R (McMurdie and Holmes, 2013).

Soil organic carbon content was measured on samples taken to a depth of 15 cm at the third CC termination. Samples were air-dried, passed through a 2 mm sieve, finely ground and analyzed for total C content using a LECO combustion analyzer (LECO Tru-SPEC, St. Joseph, MI).  

Erosion (yearly soil loss) was estimated for all treatments using the Revised Universal Soil Loss Equation (RUSLE2) for water erosion and Wind Erosion Prediction System (WEPS) for wind erosion, both available through the NRCS (Widman, 2004; Agricultural Research Service, 2020). We ran the simulation on a Wetherill loam, with data populated from the Soil Survey Geographic (SSURGO) database with low organic matter and an average slope of 2%. Modelling was based on historical average daily weather data for Dove Creek, CO (NOAA). The model was run based on a custom management plan and crop rotation that reflected all herbicide, tillage, planting, and harvest events for each treatment. Yearly average CC biomass and wheat productivity data were included in the simulation for each tillage and CC treatment. Soil loss estimates from wind and water erosion were summed together for total estimated soil loss over the six-year study period.

Economic Analysis

The CC-wheat and fallow-wheat cropping cycles spanned 2-year periods and Net Present Value (NPV) methods were used to assess returns to investments in the CC both with and without the sale of 50% of CC biomass as forage. First, enterprise budgets were developed to determine the annual net returns (in USD ha^-1 cycle^-1) for each tillage and planting window combination in each experiment. Gross revenues included wheat revenue as the primary cash crop. Revenue from the sale of CC biomass as forage was included as appropriate to compare net returns with and without forage sales. Wheat revenues were calculated using average annual wheat yields from the trials and wheat grain prices from Colorado State University (CSU) Extension Crop Enterprise Budgets for dryland wheat production in Western Colorado (Colorado State University Extension, 2016; Colorado State University Extension, 2018; Colorado State University Extension, 2021). Operating expenses included seed costs (wheat seed and CC seed), chemical costs (herbicides), variable machinery costs (fuel, oil, repairs, labor) for field operations, and six months of interest expenses on operating capital. Inputs were assigned the same prices as the actual inputs used in the experiments and interest expenses were estimated using an interest rate of 7.5% based on the CSU Extension enterprise budgets. Ownership expenses (depreciation, interest, taxes, insurance, housing) were included for tillage, planting, weed control, harvest, and hauling field operations for the wheat cash crop and for planting and termination field operations for the CC. The operating and ownership expenses for machinery and equipment are from annual reports published by Iowa State University (Plastina, 2016; Plastina, 2018; Plastina, 2020). Other economic factors of production (e.g., overhead, land, unpaid labor and management) were assumed to remain constant across treatments and are therefore not included in the economic analysis.

Second, NPV for a given cropping cycle was determined using a discount rate of 7.5% from the CSU Extension budget because it is a short-term financial analysis. Prior to discounting, revenues and costs were adjusted for inflation using the producer price index (U.S. Bureau of Labor Statistics, 2022). The average NPV per cycle was obtained for each treatment and trial combination by dividing total NPV over the three cycles. Changes in net returns for each combination of CC planting window and tillage treatment were calculated using the net returns of fallow and corresponding tillage treatment as reference treatments. Positive values therefore represent an increase in net returns and negative values indicating a decrease in net returns due to CC as compared to the reference treatments.

When including revenue from forage sales, we estimated forage yields to be 50% of total CC biomass production to ensure a level of soil cover consistent with CC best management practices (NRCS, 2023). Yields were assumed to contain 15% moisture and priced as hay on a USD ton^-1 basis (Ponderosa Partnership, 2018). Hay prices were based on reported grass hay prices from the Agricultural Market Service (Agricultural Marketing Service, 2017; Agricultural Marketing Service, 2019; Agricultural Marketing Service, 2021). All RFQ values were in Grade 1 or 2 quality standard ranges [on the American Forage and Grassland Council scale of 1 to 5 (1 being best; Marsalis et al., 2009)] and prices corresponded to “Good” quality grass hay in the Agricultural Market Service reports (Agricultural Marketing Service, 2017; Agricultural Marketing Service, 2019; Agricultural Marketing Service, 2021). To verify that the prices were reasonable for the study area, local hay farmers corroborated similar hay prices during the years of study. Additional operating and ownership expenses for raking, mowing, and baling were included and are also sourced from annual reports published by Iowa State University (Plastina, 2016; Plastina, 2018; Plastina, 2020).