2.1. Site Description and Experimental Design
The experiment was located at the Agricultural Complex for Advanced Research and Extension Systems (AG-CARES), a cooperative site between the Texas A&M AgriLife Research and Extension Center at Lubbock and the Lamesa Cotton Growers Association, located near Lamesa, TX, USA (N 32°46′22″, W 101°56′18″; 919 m a.s.l.). The site is classified as a semi-arid ecoregion with a mean (30-year average) annual temperature of 15 °C and mean annual precipitation of 450 mm [
13]. Additional environmental information from the experiment site is included in
Figure 1. The soil at this location is classified as an Amarillo (fine-loamy, mixed, superactive, thermic Aridic Paleustalfs) series, a benchmark series on the Southern High Plains of Texas. It is described as a fine sandy loam with a pH of 7.5 in the topsoil [
14].
The trial was arranged as a randomized complete block design with four replications in 2017 and three replications in 2018 through 2020. The trial area differed between years, but always followed continuous cotton with a terminated rye cover. The plot size was four rows (1 m row spacing) 15 m long in 2017 and eight rows (1 m row spacing) 15 m long in 2018 to 2020. Treatments included a conventionally tilled (CT) check and no-tillage (NT) with cover crop treatments, including two small-grain cover crop species: wheat (W) and rye (R). Cover crop species were seeded at rates of 34 kg ha
−1 (L) and 68 kg ha
−1 (H) and were terminated 6 to 8 weeks before cotton planting, which is considered the optimum timing for the region. The late termination timing was treated as optimum + two weeks (O + 2w). Cover crops were seeded using a no-till drill on 19 cm spacing (
Table 1) and chemically terminated using glyphosate at 1.12 kg ai ha
−1. Immediately before termination, cover crops were harvested from a 1 m
2 area and oven-dried at 65 °C for seven days to determine aboveground herbage mass on a dry matter (DM) basis.
Cotton was planted at 124,000 seeds ha−1 as either NexGen (NG) 4545 B2XF in 2017 or Deltapine (DP) 1646 B2XF from 2018 to 2020. The cotton stand establishment was determined by counting emerged cotton plants in four-row meters four weeks after planting. Cotton was mechanically harvested using a two-row John Deere 7445 plot stripper (Moline, IL, USA). Grab samples of seed cotton were collected during harvest and ginned at Texas A&M AgriLife in Lubbock, TX, USA, to calculate turnout for lint yield. A high-volume instrument (HVI) at Texas Tech University’s Fiber and Biopolymer Research Institute (Lubbock, TX, USA) was used to determine fiber quality to determine loan value (Cotton Incorporated, Cary, NC, USA).
Fertilizer applications were based on farm practice and were the same for all treatments. Nitrogen applications were split into four timings (one preplant and three in-season applications) and applied via fertigation throughout the growing season at 155 kg N ha−1, 129 kg N ha−1, 134 kg N ha−1, and 134 kg N ha−1 from 2017 to 2020, respectively. Phosphorous totals were 19.6 kg P ha−1 and 17 kg P ha−1 in 2017 and 2018, and based on soil tests, no phosphorous was applied in 2019 and 2020.
Following harvest each year, management varied between the CT and NT treatments. In the CT plots, cotton stalks were shredded and tilled using a chisel plow. Trifluralin was applied at 0.84 kg ai ha
−1 and incorporated to a depth of 5 cm using a spring tooth harrow in March, and beds were reformed using a lister. Before planting, beds were prepped using a rod weeder, and in-season cultivation was performed as needed to control weeds and establish dikes for low-energy precision application (LEPA) irrigation. In the NT plots, cover crops were planted into the stalks, and stalks were shredded after cover crop establishment. Pendimethalin was applied at 1.7 kg a.i. ha
−1 at the second termination date and incorporated using irrigation. In-season herbicide applications were the same for both CT and NT. Glyphosate was applied at 1.27 kg a.i. ha
−1 twice in-season to control emerged weeds, and S-metolachlor was added at the second in-season application for residual control at 1.6 kg a.i. ha
−1. The field management practices and farm equipment are thoroughly described in the work by Lewis et al. [
12].
2.2. Economic Budget Calculations
Economic budgets were created to calculate the variable costs associated with each management decision based on Lewis et al.’s study [
12]. Cotton lint prices were calculated using loan rates from Cotton Incorporated (Cary, NC, USA) to isolate production risk from market risk. Gross revenue (USD ha
−1) was calculated by multiplying the loan rate by crop yield. Variable costs (USD ha
−1) were originally obtained from the 2016 Texas Agricultural Custom Rate survey for the northern region [
15]. Operations for the conventionally tilled plots were estimated at USD 180 ha
−1 and included chisel plowing, sand-fighting (two events), cultivating (two events), rotary hoeing, rod-weeding, listing, and trifluralin herbicide incorporation. The cover crop system operations were estimated at USD 94 and USD 127 ha
−1 for the 34 and 68 kg ha
−1 seeding rates, respectively, and included seed, drilling, herbicide application, and termination. All other input costs were identical. Following Lewis et al. [
12], total variable costs were then subtracted from gross revenue to determine gross margin (USD ha
−1). Gross margin is a measure of relative profitability, assuming the operation’s fixed production costs are similar among alternative production activities.
2.3. Statistical Analysis
Data were analyzed using Proc GLIMMIX at a significance level of α = 0.05 (herbage mass, plant populations, lint yield, loan value, return, and margin) using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA). Differences in year were observed, so year was analyzed independently. Cover crop herbage mass was analyzed as a multivariate factorial, with the fixed effects being species, seeding rate, and termination timing. Plant populations, cotton lint yield, loan value, gross return, and gross margins were analyzed as a univariate with treatment as the fixed effect. Replication was treated as the random effect for both analyses.