Introduction

The health of agricultural soils can be modified by cropping practices such as crop rotation and fertilization (Karlen et al. 2006; Bowles et al. 2020). These practices influence soil functions through their impact on soil properties such as organic matter, structure, porosity, pH, microbial biomass, and nutrient supply. The contribution of soil properties to soil health is complex due to dependence on agoecosystem, slow change in soil properties, and nonlinear relationships to soil functions. Due to these factors, long-term cropping system studies are of considerable value to quantify management impacts on soil properties and crop yield (Janzen 1995; Peterson et al. 2012).

A recent study relevant for the semiarid Canadian prairies was conducted by Smith et al. (2015) at Lethbridge, Alberta, in the Dark Brown soil zone. A cropping system experiment with 13 different crop management treatments that had been in place for 10–44 years was interrupted for 6 years during which all plots were uniformly maintained under a single “bioassay crop”. Compared to a fallow–wheat (FW) rotation, cropping systems with less frequent fallow and greater N inputs increased wheat yield over 4 years (excluding initial 2 transition years) by up to 85% when no N fertilizer was applied and by up to 19% when fertilizer N was applied. Wheat yield without applied N was positively correlated with soil organic C (SOC). The relationship of SOC to wheat yield was largely masked by the application of fertilizer N.

Soil properties that control available water will increase in importance with increasing aridity. Soil water-holding capacity and rate of water depletion are controlled primarily by texture, organic matter, and soil thickness (Saxton and Rawls 2006). Salinity also limits soil water availability. Cropping systems that use less water increase percolation and downward movement of salts that may otherwise contribute to salinity issues in other areas, while cropping systems that use more water may increase net upward movement of salts (Beke et al. 1994). The impact of the cropping system on salinity depends on subsoil hydrology and salinity levels, the dominant factors controlling salinity issues. Most of the agricultural land on the Canadian prairies has some risk of being salt-affected (Coote et al. 1981).

The termination of a long-term cropping study at Bow Island in the Brown soil zone in Alberta provided an opportunity to evaluate the impact of long-term crop management practices in the Brown soil zone, the driest soil zone in the prairies. The study was conducted from 1992 to 2015 by Alberta Agriculture to evaluate long-term impacts due to intensification of cropping practices. Prior to 1990, almost all nonirrigated cropland in the Brown soil zone of Alberta was managed as tilled FW, but as of 2016, summerfallow had been largely eliminated (<10% of dryland area fallowed in a given year; Alberta Agriculture and Forestry 2020) and no-till was adopted widely. Results from the first 18 years of the long-term cropping study showed that annual crop yields were highest in wheat-pulse and fertilized continuous wheat (CW) treatments, with increases in SOC of up to 4 Mg C ha⁻¹ (11%) under continuous cropping and up to 12 Mg C ha⁻¹ (30%) under perennial grass (Bremer et al. 2011). We conducted a crop bioassay for 5 years after the study had run for 24 growing seasons to determine the impact of prior cropping practices on soil properties and crop yield.

Materials and methods

The Bow Island long-term cropping study consisted of six crop rotation treatments with two to six fertilizer treatments in each rotation (Fig. 1). All rotation phases were present each year. The experiment was set up in split-plot design with six main plot treatments, six subplot treatments, and four blocks. However, subplot treatments were not randomized between each block. Plots were 4 m × 16 m in size.

Fig. 1.

Plot plan and treatments in the long-term cropping system experiment at Bow Island, AB, during rotation phase (1992–2015).

During the bioassay years, all plots were uniformly cropped to spring wheat in 2016, yellow mustard (Sinapis alba L.) in 2017, barley in 2018, and winter wheat in 2019–2020 (Table 1). Due to extremely dry conditions in the spring of 2019, a spring crop was not planted and the site was kept fallow until the fall when winter wheat was planted. Nitrogen fertilizer was applied to the same half of each plot (split lengthwise) at a rate of 80 kg N ha⁻¹ year⁻¹ in 2016, 2017, and 2018. No N fertilizer was applied to the other half of each plot or to winter wheat in 2020.

Table 1.

Cropping and weather conditions during crop bioassay (2016–2020).

Conventional zero-tillage practices for the region were followed. Glyphosate was applied in fall 2015 and spring 2016 to all plots for weed control and termination of the grass stand. Plots were seeded without tillage and P fertilizer (triple superphosphate, 0-45-0) was seed-placed at 9 kg P ha⁻¹ on the entire cropped area, while N fertilizer (urea, 46-0-0) was side-banded at 80 kg N ha⁻¹ in half of each plot in 2016, 2017, and 2018, as described above. In-crop weeds were controlled with herbicide applications. Seed yields were determined by harvesting all rows within subplots (25 m²) by using a small plot combine. Seed yields were adjusted to standard moisture content: 13.5% for wheat and barley and 8.5% for mustard.

In 2017, mustard was seeded on 4 May following soil application of ethalfluralin (Edge) and sulfentrazone (Authority). Herbicide interaction caused variable emergence, with almost no emergence in compost treatments, and therefore the original crop was sprayed out and re-seeded on 9 June. Due to high temperatures and negligible precipitation after re-seeding (Table 1), 50 mm of irrigation was applied on 4 July to ensure survival.

To monitor drought stress in 2018, crop canopy temperatures in the mid-afternoon of clear days were obtained weekly from 6 June until 18 July using a narrow field-of-view infrared radiometer (MI-220, Apogee Instruments, Logan, UT). Crop canopy temperature increase under drought stress due to lower rates of evapotranspiration (Jackson 1982). Soil moisture depletion over the growing season was also determined each year in one or two subplot treatments of each rotation by determining gravimetric soil moisture on soil cores obtained to a depth of 0.9 m within a few days of seeding and harvest.

Soil samples were collected from all plots in the fall of 2015 and analyzed for total and organic C and N using the same methodology used previously (Bremer et al. 2011). In brief, four cores with a diameter of 6.7 cm were taken with a hydraulic soil corer and composited into 0–7.5, 7.5–15, and 15–30 cm depth increments. Crop residues on the soil surface were not included in the sample. Whole soil samples, including root and crop residue fragments present, were ground to <2 mm on a rotating sieve. Representative subsamples were then ground to <0.15 mm and analyzed for total C and N with an automated combustion analyzer (CE Elantech, Lakewood, NJ). Soil organic C was determined in the same way after removal of carbonates with the addition of excess hydrochloric acid to the combustion capsule (Ellert and Rock 2008). Soil inorganic N was determined by colorimetric determination of nitrate and ammonium after KCl extraction (Keeney and Nelson 1982). Total SOC and N were calculated on an equivalent mass basis (4348 Mg ha⁻¹ to 30 cm) (Ellert and Bettany 1995). Mineralization of soil C and N was quantified by determining respired CO₂ and the increase in inorganic N during a 10-week incubation of a 75 g subsample maintained at 25 °C and 80% of field capacity (Bremer et al. 1994). Soil pH was determined in 0.01 mol/L CaCl₂ (2:1 solution:soil suspension) (Hendershot et al. 2008).

Soil nutrient supply was determined in situ in 2016 through 2018 in one to three subplot treatments of each rotation. Three pairs (anion and cation) of Plant Root Simulator (PRS^®) probes (Western Ag Innovations, Saskatoon, SK) were buried in each plot for 4 weeks starting 10–14 days after seeding. Nutrients adsorbed by the ion-exchange membranes over the burial period were determined by eluting the ions with 0.5 mol/L HCl and analyzing the eluant for NO₃–N and NH₄–N using an automated continuous-flow analyzer with a colorimetric detector (Skalar Inc., Netherlands) and other nutrients (P, K, S, Ca, Mg, Fe, Mn, Cu, Zn, and B) by using an inductively coupled plasma optical emission spectrometer (Optima ICP-OES 8300, PerkinElmer Inc., USA).

Inductive sensing of electrical conductivity (EC) was determined in all plots on three dates in 2017 and 2018 using an inductive electromagnetic sensor (EM38-MK2, Geonics, Mississauga, ON). Lab-based measurements of soil conductivity (2:1 water:soil suspensions) to a depth of 0.9 m were determined on all soil samples collected for gravimetric soil moisture in April 2018 (two subplot treatments in each rotation, ±N fertilizer).

Statistical analysis of crop and soil variables was conducted with the MIXED procedure (SAS Institute Inc. 2002), with prior cropping treatment (main plot and subplot) as fixed effects and block as a random effect. Initial analysis indicated that prior P fertilization and rotation phase had negligible impacts on soil organic matter or pH. Thus, soil variables were also analyzed with the main plot as fixed effects and block and subplot treatment as random effects for treatments with a similar fertilizer history: 0N ± P, N-fertilized treatments ± P and compost. Grass and annual legume–wheat (LW) rotations were included in this analysis as separate main plot treatments. In situ soil nutrient supply was analyzed for 2017 and 2018 with year included as a random effect; 2016 was excluded due to different treatments being monitored. Post-rotation N fertilizer rates were included as a fixed effect for crop yield in each year, but crop yields were not analyzed across years due to strong dependence of treatment effects on growing conditions. Crop yields were normalized across years by expressing as a fraction of the average yield of all FW treatments at the same post-rotation N rate each year. Statistical significance of treatment differences was evaluated with the Tukey–Kramer's test (p = 0.05). The correlation of crop yield with soil properties was determined with the CORR procedure.

Results and discussion

Soil properties

Cropping practices from 1992 to 2015 had a large impact on SOC (Table 2). Application of N fertilizer did not increase SOC in annual crop rotations, but increased SOC in the perennial grass treatment by 9 Mg C ha⁻¹. Application of compost in annual crop rotations increased SOC by an average of 6 Mg C ha⁻¹. Compared to FW, continuously cropped rotations (CW and LW) increased SOC by an average of 4 Mg C ha⁻¹. Compared to annual crop rotations, perennial grass increased SOC by an average of 11 Mg C ha⁻¹ when unfertilized and 19 Mg C ha⁻¹ when fertilized. These differences are similar to those observed previously, but with continued increases of SOC in the perennial grass treatment (Bremer et al. 2011).

Table 2.

Effect of 24 years of cropping management on soil properties.

The increase in SOC in the perennial grass treatment was largely composed of undecomposed roots and other organic materials with a wide C:N ratio (Table 2). Soil C mineralization and mineralizable C:N ratio were also much greater in perennial grass than annual cropping systems. This gain was greater than observed under perennial crops in other long-term cropping studies in Canada (3–14 Mg C ha⁻¹) determined with the same sampling and analytical methods and comparable stand age (VandenBygaart et al. 2010). The greater impact of perennial grass in this study may be due to greater proliferation of roots in the surface 0.3 m due to elevated subsoil salinity (Fig. 2, discussed below).

Fig. 2.

Salinity and water relations: (a) subsoil salinity in spring 2018 was elevated in prior rotations with grass or less frequent fallow, (b) subsoil moisture after barley harvest in 2018 was not depleted following grass, and (c) grain yields in 2018 declined with increasing mid-afternoon canopy temperature on 11 July. See Fig. 1 for treatment descriptions.

Soil C mineralization over a 10-week incubation period was influenced similarly by rotation treatment as SOC, although with a larger relative gain due to perennial grass (Table 2). Nitrogen mineralization was highest in perennial grass and cropping systems with an annual legume (LW and fallow-winter wheat–legume (FWL)) and least in FW and CW treatments. Soil N supply determined in situ with PRS probes in 2017 and 2018 were also influenced similarly by prior cropping system (Table 2). Application of compost did not increase lab mineralization or in situ NO₃ supply, but increased P, K, and Zn supply (data not presented). The organic N in feedlot compost is only very slowly mineralized (Helgason et al. 2007).

Soil pH in the surface 0.15 m ranged from 5.7 to 7.0 (Table 2). Soil pH was reduced by an average of 0.5 with long-term application of N fertilizer, consistent with other long-term studies on the Canadian prairies (Bouman et al. 1995). Application of compost increased soil pH by an average of 0.8 due to the addition of CaCO₃ in feed rations (Eghball 1999).

Inductive sensing of soil EC using an EM38 in 2017 and 2018 revealed that grass plots consistently had the highest conductivity and FW treatments had the lowest conductivity. Lab measurements of EC in soil samples obtained in April 2018 were greatest in grass and least in FW treatments in both the 0.3–0.6 and 0.6–0.9 m depths (Fig. 2a). Soil EC below 0.3 m was elevated in all rotations except FW compared to 1991 measurements. Thus, subsoil salinity had been altered by the prior cropping system. Crops can induce subsoil salinity by removing water from the root zone by evaporation and transpiration in the presence of a perched or rising water table that is high in soluble salts (Rengasamy 2002). Transient subsoil salinity occurs without the influence of groundwater when leaching is limited by low permeability of deeper soil layers and low rainfall, while seepage salinity occurs due to the presence of a shallow and saline water table.

Crop bioassay yield

Drought conditions during the crop bioassay, particularly from 2017 through 2019, limited yield potential and increased crop dependence on stored soil moisture (Tables 1 and 3). Growing season precipitation was considerably below normal in 2017, 2018, and 2019. Supplemental irrigation in 2017 increased precipitation to normal, but mustard yields were still low due to high evapotranspiration and temperatures following re-seeding required due to herbicide injury. In 2018, barley effectively depleted soil water during the growing season, thus supporting moderately high yields. In 2020, above-average precipitation in May and June supported high winter wheat yields.

Table 3.

Average crop yield (Mg ha⁻¹) following 24 years of continuous wheat (CW) and fallow-wheat (FW) under six fertility treatments.

Application of N fertilizer during the crop bioassay period (2016–2018) increased crop yield by 34%–61% (p < 0.0001) in former CW and FW rotations (Table 3). Residual impacts of N fertilizer applied from 2016 to 2018 increased winter wheat yield in 2020 by 9% (p < 0.0001). Nitrogen mineralized during the fallow period in 2019 contributed to N uptake of winter wheat in 2020.

The impact of prior rotation on subsequent crop yield was greater in 2016 than in subsequent years of the bioassay (Fig. 3). This was partly due to short-term impacts of fallow in the previous year as evident by the higher yield following the fallow phase of the FW rotation (FW II) in 2016 but similar crop yields between phases in subsequent years (Table 3). The most notable impact of prior rotation in 2016 was the low yield following perennial grass that we initially attributed to soil N immobilization from decomposing grass residues. However, spring measurements of soil N supply in 2016 in the perennial grass treatment were not different from those in the CW and FW rotations (not presented) and the protein concentration of wheat in 2016 without N was the same following perennial grass and FW (110 vs. 112 g kg⁻¹). Elevated subsoil salinity in the perennial grass treatment may have limited available water during a critical period for grain yield, even though soil moisture depletion over the growing season was similar among treatments.

Fig. 3.

Impact of prior rotation (1992–2015) at Bow Island, AB, on crop yields during bioassay period, expressed as a fraction of average FW yield: (a) without N fertilizer and (b) with 80 kg N ha⁻¹ (no N fertilizer applied in 2020 (residual impacts)). See Fig. 1 for treatment descriptions. Error bars are standard errors.

Differences in crop productivity in 2017 were also evident (Fig. 3 and Table 3). However, low crop yields following herbicide injury that was partly related to soil pH complicated interpretation of yield differences and reduced their reliability as an indicator of long-term impacts.

Differences in barley yield in 2018 and winter wheat yield in 2020 were the most reliable indicators of long-term management impacts because short-term factors had dissipated and crop establishment was satisfactory. Similarly, Smith et al. (2015) used total wheat yield from 3–6 years after initiation of uniform cropping as the indicator of long-term crop management impacts.

The most notable impact of prior rotation on crop yield in 2018 was again the low yield following perennial grass, particularly in the N-fertilized treatment (Fig. 3). This was in contrast to most previous studies (Entz et al. 2002; Franco et al. 2018 and citations therein). Negative or neutral impacts may occur for 1 or 2 years after breaking of perennial crops in semiarid regions due to reduced water availability (Cutforth et al. 2010). Perennial grass had clearly increased rather than decreased soil N supply by 2018: in situ spring measurements of soil N supply in grass treatments were 2.2-fold higher than FW treatments and 1.5-fold higher than CW and LW treatments (Fig. 4c) and grain protein concentrations were the highest in the perennial grass treatment (135 g kg⁻¹, compared to 94–111 g kg⁻¹ in other N-fertilized treatments). The negative impact of perennial grass on crop yields was not associated with deficiencies in nutrients other than N as in situ supply rates of P, K, and other nutrients were similar for grass to those in other rotations with equivalent prior fertilization. The negative impact of perennial grass on crop yields was also not associated with low soil pH (Table 2). Instead, subsoil salinity reduced barley yield by reducing the availability of soil water in a year where subsoil moisture was critical for yield. Barley was unable to deplete soil moisture below 0.3 m as effectively in the perennial grass treatment as other treatments (Fig. 2b) and N-fertilized barley had greater mid-afternoon canopy temperatures on 11 July than other treatments (Fig. 2c; also observed on 20 June, 28 June, and 18 July), indicative of increased drought stress (Jackson 1982). Mid-afternoon canopy temperatures were negatively correlated with final barley grain yield on all of these dates (r ranged from −0.89 to −0.94, p < 0.001). The application of N fertilizer exacerbated the negative impact of subsoil salinity by further stimulating vegetative growth and depletion of soil moisture early in the growing season and increasing drought stress during periods critical for grain yield later in the growing season (van Herwaarden et al. 1998).

Fig. 4.

Relationship of crop yields to soil properties: (a) 2018 barley yields vs. soil organic C, (b) 2020 winter wheat yields vs. soil organic C, and (c) 2018 barley yields vs. soil NO₃–N supply. Yields following grass were excluded from correlations. See Fig. 1 for rotation descriptions.

Winter wheat yields in 2020 were not lower following perennial grass but were still lower than expected based on the relationship with SOC (Fig. 4b). Drought stress and crop dependence on subsoil moisture were considerably less in 2020 than 2018 due to greater precipitation that was better synchronized with crop use (Table 1).

Relative to FW, prior rotations with continuous cropping or pulse legumes (CW, LW, FWL) increased barley yields in 2018 by 13%–19% in the non-N-fertilized treatment and increased winter wheat yields in 2020 by 10%–20% (Fig. 3). Prior rotations other than perennial grass did not impact barley yields in bioassay years with N fertilizer. In comparison, Smith et al. (2015) reported that cropping systems with less frequent fallow and greater N inputs increased wheat yield by up to 85% when no N fertilizer was applied and by up to 19% when fertilizer N was applied. The smaller benefit in our study primarily reflects reduced N fertility benefits following low crop yields in 2017 and fallow in 2019.

Crop yields in 2018 and 2020 were not strongly impacted by long-term fertilization treatments (Table 3). In 2018, prior fertilization treatments had no impact on barley yield. In 2020, compost increased winter wheat yield by 7% compared to prior 0N treatments, which is attributable to improved moisture conditions and increased soil organic matter and/or nutrient supply (P, K, and Zn, data not presented).

Models relating bioassay crop yields to soil properties are useful for interpretation and extrapolation due to the multiple factors controlling yield. Compared to FW rotation, the SimPLE.ca model (Bremer et al. 2008) estimated that crop yield in the perennial grass treatment would be reduced by 30% in 2018 due to the reduction in available water. The model also estimated yield benefits of up to 20% due to continuous cropping or inclusion of annual legumes when unfertilized but none when fertilized, similar to observed (Fig. 3). Simulation of winter wheat yields without the fallow period increased yield differences to similar levels reported by Smith et al. (2015). Semiarid regions are characterized by large annual variation in available moisture, which strongly interacts with nutrient supply and soil organic matter to impact crop yields.

Conclusions

The bioassay approach, whereby the changes in soil health influencing crop yields accruing to rival long-term cropping systems are inferred from crop yields deployed uniformly over the entire study, was influenced by drought (2017, 2018, and 2019) and by herbicide damage (2017) in this study. Benefits of the keystone property for soil health, SOC, were counteracted by increased subsoil salinity under drought conditions. This may not apply to years with excessive precipitation as SOC may support increased infiltration, more stable aggregation, and greater resistance to erosion.

The benefits of annual cropping systems with reduced fallow frequency or greater crop diversity to crop yields persisted to the fifth and final year of the crop bioassay. These benefits were positively correlated with SOC.

The long-term perennial grass forage was an outlier in this bioassay. Despite substantially increased SOC after 24 years, the expectation that this would support greater productivity was trumped by elevated subsoil salinity. Subsoil salinity primarily depends on hydrology and groundwater salinity and may have been exacerbated by the surrounding land management in this study. Nevertheless, these results highlight the importance of including subsoil salinity in soil health assessment in arid and semiarid regions. The presence of subsoils that limit crop yield is often not recognized as they are not readily observable and their impact varies with precipitation and crop dependence on subsoil moisture reserves.