Management impacts on organic carbon under continuous perennial grass, perennial grass-legume mixture, and annual cereals on a thick Black Chernozemic soil

Abstract Impacts of annual and perennial pasture management on soil organic carbon (SOC) and equivalent SOC stocks (equal soil mass basis) were investigated in two trials [CAESA (1994–1997) and BMP (2008–2012) trials] conducted on the same experimental paddocks at Lacombe, AB. The original site was broken from perennial grass in 1992, and the CAESA trial established in 1993. Between 1994 and 1997, half of the paddocks included winter triticale and a mixture of triticale and spring barley; half included smooth and meadow bromegrass; and each paddock was light, medium, or heavily grazed. The BMP trial (2008–2012) on the same paddocks included fertilized, direct seeded barley as silage; grazing and haying of unfertilized meadow bromegrass, fertilized meadow bromegrass, and meadow bromegrass and alfalfa mixture; and unfertilized oldgrass that was continuous since 1994. Between trials (1998–2007), all paddocks received no fertilizer. In the 0–15 cm depth, SOC under oldgrass was constant between 1994 and 2012 and averaged 88 Mg C ha−1. Under barley silage, SOC decreased from 89 to 72 Mg C ha−1 by 2012. Between 1994 and 2012, SOC decreased in all treatments re-established on original annual forage (1994–1998) but not to the level of barley silage. Light fraction carbon was the highest under oldgrass and the lowest under barley silage. Overall, oldgrass with no fertilizer inputs maintained a constant SOC, although annuals reduced SOC stocks. Re-establishment of perennial grass with grazing may therefore reduce SOC loss, whereas haying perennial grass may not reduce SOC loss.


Introduction
Approximately 30% of global soil carbon is stored in grasslands, with the potential to store even more depending on management practices (Le and Carlyle 2019).Land use and land use changes that cause soil disturbances can result in large losses of soil carbon.Conversion of natural land to growing crops can cause substantial loss of soil carbon, released into the atmosphere as a greenhouse gas, CO 2 .VandenBygaart et al. (2003) reported that conversion of native grassland to cropland on 72 farms in Alberta reduced soil organic carbon (SOC) by 24%.Using data from forest and grassland ecosystems, Ledo et al. (2020) reported that conversion from annual to perennial cropland increased SOC by 20% in the 0-30 cm depth interval over a 20-year period.This study also reported that SOC changes did not follow a linear trend over time, with soil under perennials showing an initial gain in carbon, followed by fluctuating gains and losses.
Management practices such as haying, grazing, nitrogen (N) fertilization, and changes to increased mixture of species including legumes are known to impact SOC (Conant et al. 2001).Increases in SOC under perennial grasses have been attributed to the increase in higher belowground inputs from turnover of perennial grass roots and the lack of tillage.A study comparing long-and short-term perennial and annual crops reported that SOC concentrations under perennial grasses such as switchgrass were high in the 0-5 cm depth interval but decreased with depth.However, under deep-rooted alfalfa, SOC increased over a 6-year period throughout the 33 cm of plough layer (Ferchaud et al. 2016).
Perennial grasses differ in their root distributions with depth, which can affect their contribution to SOC.For example, a study conducted in New Zealand reported that prairie grass had higher root counts in greater depths than smooth bromegrass (Bromus inermis Leyss.) and tall fescue (Festuca arundinacea Schreb.).At the 40 cm depth, root counts for prairie grass were 130% and 200% higher than those for smooth bromegrass and tall fescue, respectively (Shaffer et al. 1994).Biligetu and Coulman (2010) found that differences among Bromus species were evident in the amount of aboveground biomass produced.For example, meadow bromegrass is more adapted to grazing due to rapid regrowth after defoliation, whereas smooth bromegrass is intolerant to frequent defoliation.This difference may result in different aboveground and belowground contributions to the SOC (Biligetu and Coulman 2010).Wang et al. (2014) reported varying rates of SOC storage in grass mixtures in the Canadian Prairies, specifically for semiarid mixed grasses (0.05 Mg C ha −1 year −1 ), mixed grasses (0.16 Mg C ha −1 year −1 ), and fescue (0.22 Mg C ha −1 year −1 ).Such information is not available for Black Chernozemic soils that naturally have higher SOC nor for perennial grasses such as meadow bromegrass and smooth bromegrass.
Studies in Alberta have shown that meadow bromegrass tends to be higher yielding and has deeper roots and shorter rhizomes than smooth bromegrass.Meadow bromegrass is a recently introduced grass that has gained acceptance in Alberta due to its high productivity when grown in mixtures with legumes such as alfalfa (Medicago sativa L.) (Pearen and Baron 1996).Alfalfa plants are known to develop deep roots due to their extensive taproot system, and there may be a very strong relationship between SOC storage and the addition of soil N, whether from fertilizer, legumes, or manure from grazing cattle (Chang et al. 2012).
Relative to the effects of management practices on SOC, little is known about the impacts on the labile fraction of SOC, particularly on grasslands.Labile SOC has a rapid turnover rate and may be more sensitive to changes in plant species, soil temperature, soil moisture, and other factors (Song et al. 2014).Light fraction carbon (LFC) is a low-density labile fraction of organic carbon with low stability and is composed of undecomposed and partly decomposed particulate organic matter, which is available substrate for decomposition by microorganisms (Gregorich and Janzen 1996;Song et al. 2014).
While differences in the impacts of grazing and establishment of perennial versus annual crops are known to affect SOC, questions remain on the legacy effects of SOC changes that could occur when management is changed.For example, what happens to SOC when perennials are replaced with barley for silage, which removes almost all the aboveground biomass?What happens when the land is converted back to perennials and managed differently, such as application of N fertilizer, introduction of legumes into perennials, or addition of manure or grazing?Does SOC increase under these management practices, and if so, which treatments give the fastest rate of increase to achieve equilibrium?
This hypothesis was tested: Re-establishment of perennial grasses along with application of N fertilizer or addition of legumes to perennial-legume mixtures can increase SOC to equilibrium (a state where continuous gains are counterbalanced by continuous losses) on land where SOC loss had occurred from conversion to annual crops.Several objectives addressed in our research were to investigate (i) long-term SOC changes under grass-based perennial stands and cerealbased annual forage stands managed with minimal inputs of fertilizer; (ii) short-term effects of grazing intensity and plant species (perennials and annuals) on SOC after breaking land that had been under long-term, lowly managed perennial pasture (i.e., extensively grazed with no inputs added); (iii) SOC and LFC changes after re-establishing perennial grasses with added N fertilizer, legumes, and manure on land that previously had been under barley for silage for 5 years.

Site description and meteorological conditions
The study site was located at the Lacombe Research and Development Centre, Alberta (52 • 28'N; 113 • 45'W; 847 m), on an Orthic Black Chernozem of loam texture derived from glaciolacustrine parent material.Surface soil (0-15 cm) contains 15% clay, 34% silt, and 51% sand.Soil pH as determined using distilled water averaged 5.4, and the sodium adsorption ratio was 0.2 (Mapfumo et al. 2002).In 1992, the land was broken after ∼15 years of maintenance of an old mixture of extensively managed (no inputs) and moderately grazed perennial grasses.Before breaking, the mixture of perennial grasses included Kentucky bluegrass (Poa pratensis L.) and quackgrass (Agropyron repens L.), with common occurrences of dandelion (Taraxacum officinale L.).
Tracking of management effects on SOC and its characteristics was carried out on ∼1.4 ha beginning in 1993 and ending in 2012.Individual paddocks were maintained as described below and remained under minimal management between 1998 and 2007 and after 2012 until the present (2023).Experiments carried out between 1994 and 1997 (Baron et al. 1999) and 2008 to 2012 provided measurements used in this study.Relevant to the outcomes presented is the minimalist level of grassland and crop management and lack of fertilizer and tillage inputs which occurred between the two periods of intensive measurement.CAESA (1993CAESA ( -1997) ) experimental design, plant species, and grazing treatments The study area (Baron et al. 1999) was broken in 1992 and divided into paddocks of 9 m × 33 m.A total of 48 paddocks were established and divided into four replicate blocks.Within each block, a combination of plant species and grazing intensity treatments were applied.Three grazing treatments were heavy, medium, and light intensity defined by canopy height to initiate and exit grazing.Of four plant species used, two perennials were 'Carlton' smooth bromegrass (B.inermis L.) and 'Paddock' meadow bromegrass (Bromus riparius Rhem.), and two annuals were 'Pika' triticale (x Triticosecale Wittmack) and a mixture of triticale and barley (Hordeum vulgare L.).Perennials were seeded on 31 May 1993 at 11.2 kg ha −1 for smooth bromegrass and 16.8 kg ha −1 for meadow bromegrass.Establishment of annuals involved first rototilling each spring to a depth of 10 cm followed by seeding at 135 kg ha −1 for triticale plots only or a mixture of triticale and barley at 90 and 50 kg ha −1 , respectively.Application of 2-methyl-4-chlorophenoxyacetic acid was conducted on annual plots at 600 g a.i.ha −1 in 1994-1996 after crop emergence, whereas on perennial plots, no weed control was conducted.
Fertilizer was broadcast on the entire study area prior to seeding at 8, 14, 26, and 5 kg ha −1 for N (N), phosphorus (P), potassium (K), and sulfur (S), respectively.For the three years (1994)(1995)(1996) after seeding, fertilizer was broadcast at the same time for both perennials and annuals, and at 100, 22, and 41 kg ha −1 of N, P, and K, respectively.The fertilizer rates were common in intensively grazed pastures at the time.
The 3-year (1994-1996) average number of grazing events for perennials was 7, 5, and 3 for heavy, medium, and light grazing, respectively.The average grazing events per season for the annuals were 4, 4, and 2 for heavy, medium, and light grazing, respectively.Based on the assumption of a 450 kg animal unit, the 3-year average animal-unit months (AUM) for perennials were 45.2, 24.4, and 19.6 AUM ha −1 for heavy, medium, and light grazing, respectively.For the annuals, these were 23.5, 13.5, and 9.8 AUM ha −1 for heavy, medium, and light grazing, respectively.More details on the grazing systems and number of cow-days are described in several publications from the CAESA study (Baron et al. 1999;Mapfumo et al. 2000Mapfumo et al. , 2002)).CAESA (1994CAESA ( -1997) ) study: Bulk density, soil sampling, and organic carbon measurements Two Campbell Scientific Nuclear probes, MC1 and MC3-82, were used to measure bulk density and soil water content in the top 10 cm, respectively.These measurements were conducted in May 1994, October 1994, May 1995, October 1995, May 1996, and October 1996.Detailed descriptions of the procedure and summary table of the near-surface bulk density values under perennial and annual species are published in Twerdoff et al. (1999).Bulk densities for the 15-30 cm depth interval were obtained from the soil cores collected using a coring truck.
Soil samples for organic carbon determination were collected in September 1993 using a hand auger, and in October 1996 using a tractor-mounted hydraulic auger, both to a 60 cm depth.The whole experimental area was within the mid-slope landscape position.For both sampling times, three samples per plot were taken at random locations and separated into increments of 0-5, 5-15, 15-30, 30-45, and 45-60 cm.Samples were air dried and passed through a 2 mm sieve.Before conducting organic carbon analysis, the three samples from the same depth interval and within each plot were mixed to make a composite sample.For each depth interval, a total of four composite samples were collected from each species and grazing treatment combination.Total C in soil was determined for 0-5, 5-15, and 15-30 cm depth intervals using a Leco carbon analyzer (Model CN 2000;Leco Corp., St. Joseph, MI, USA).Initial soil tests did not indicate the presence of free carbonates; therefore, total C was considered primarily organic carbon.
For the calculation of unadjusted SOC for 1994, bulk densities determined in May 1994 were used, whereas for 1997 SOC, bulk densities determined in October 1996 were used.This involved determining the total mass of soil per hectare via multiplying bulk density for the depth interval by the volume of soil per hectare.The formula used to calculate actual SOC was as follows (Issah et al. 2021): where SOC concentration is the concentration of total SOC (kg C•Mg −1 soil), ρ b is the soil bulk density (Mg•m −3 ), T is the thickness of soil layer (m), and 0.001 Mg•kg −1 is the conversion factor.
Equivalent (or adjusted) SOC was determined by multiplying actual SOC (Mg ha −1 ) by the average soil mass across all treatments divided by the unadjusted soil mass calculated for that depth interval.The average soil masses per hectare (±standard errors in parentheses) used for SOC calculations at depth were as follows: 0-5 cm, 559 (±2.88)Mg ha −1 ; 5-15 cm, 1117 (±5.75)Mg ha −1 ; 15-30 cm, 1412 (±6.82)Mg ha −1 ; 0-15 cm, 1676 (±8.63)Mg ha −1 ; and 0-30 cm, 3088 (±10.18)Mg ha −1 .These soil masses were average values based on the bulk density collected at the site during the CAESA trial (Twerdoff et al. 1999) and during the BMP trial period (2008)(2009)(2010)(2011)(2012) to standardize SOC on an equal mass basis.The overall variabilities of the average bulk densities used in the SOC calculations were very low as shown by the small standard error values given above (in parentheses) and the coefficient of variation (%) values that ranged between 3.61% and 5.64%.CAESA (1994CAESA ( -1997) ) study: Statistical analyses for SOC (%) distribution with depth and SOC stocks Regression analyses were conducted for the SOC % against soil depth for each species and grazing treatment combination using the SPSS Software version 26.The slopes of regression were compared to determine if differences among treatments occurred.A significance level (critical p-value) of 0.05 was used to determine if the slope of each regression was greater than zero, and a t-test was then used to compare each pair of slopes.
Statistical analyses of SOC data to compare plant species, grazing intensity, and years were conducted using the generalized linear model (GLM) procedure for the three-way analysis of variance (ANOVA) (F-test).To evaluate significance among plant species, grazing intensities, years, and their interactions, the critical significance level of 0.05 was applied.When the F-test showed a significant difference among treatments, a pairwise comparison was conducted using the least significant difference (LSD) test (Steel et al. 1997).BMP (2008BMP ( -2012) ) study: Establishment, plant species, haying, and grazing management Between 1998 and2006, all the annual plots that previously had been under either winter triticale only or a mixture of winter triticale and barley were used to grow barley for silage.During that period, there were no inputs of fertilizer or manure.Cereals were direct seeded at 300 seeds m −2 in late May of each year using a John Deere (Model 750) zero till drill equipped with disk openers, coulter in front, and packing wheels behind.Glyphosate [N-(phosphonomethyl) glycine] was applied as a preseeding weed treatment at 1.33 kg a.i.ha −1 .No fertilizer was applied over this period to ensure both perennial and annual stands received similar minimal inputs.Harvest occurred at a stage for silage or cereal hay and all except stubble removed; yields and residue were not recorded.
The BMP study was established in 2007 on 20 of the original 48 paddocks from the CAESA study.Two paddocks previously used for barley-triticale and two paddocks under triticale only in the CAESA study were each divided into two smaller plots to establish meadow bromegrass, which was either fertilized with N (broadcast 100 kg ha −1 of actual urea-N), planted in a mixture with alfalfa, or unfertilized control.Hayed treatments were cut twice during the growing season (between May and August) with a conventional farm swather at a height of 7.5 cm, and all harvested biomass was removed from the plots.
Another two paddocks previously under barley-triticale and two paddocks under triticale only in the CAESA study were subdivided into smaller plots seeded to meadow bromegrass, which was either unfertilized, fertilized (broadcast 100 kg ha −1 of actual urea-N) with N, or planted in a mixture with alfalfa.These plots were grazed three times during the growing season.There were two meadow bromegrass controls: control #1 was a pure meadow bromegrass stand established in the same paddock as control #2 on the side close to the fertilized meadow bromegrass treatment and control #2 was a pure meadow bromegrass stand on the side very close to the meadow bromegrass-alfalfa mixture.
Another two paddocks previously under barley-triticale and two paddocks under triticale were used to establish barley for silage, and therefore were under annual forage treatment since 1993.These paddocks were seeded as described above with zero tillage practices, except they were now fertilized by broadcasting 100 kg N ha −1 , which was similar to that applied to hayed and grazed meadow bromegrass treatments.All aboveground material except for stubble or residue was removed to simulate silage production.This management was imposed intentionally to reduce aboveground organic-C inputs, which also occurred in the oldgrass treatment.Four paddocks seeded with smooth bromegrass (Bromus inermis L.) during the CAESA study remained under perennial grass and in the BMP study, referred to as perennial oldgrass.This treatment did not receive fertilizer and was managed as hay.
Data of equivalent SOC and equivalent LFC collected from the BMP trial were analyzed using the GLM model with treatments and year as fixed factors and replicate blocks as a random factor.Post-hoc analyses were conducted using the LSD test when the global F-test indicated there were significant differences at the p < 0.05 significance level.Another comparison was made for a group of hayed treatments versus grazed treatments using the LSD test with unequal samples sizes (n = 48 for hayed treatments; n = 64 for grazed treatments for SOC; n = 24 for hayed treatments; and n = 24 for grazed treatments for LFC).

Results
Total precipitation and mean air temperatures between 1992 and 2012 are presented in Table 1.Precipitation totals for April-October were less than the long-term normal mean precipitation (100-year average) of 366.6 mm in 10 out of the 20 years of the study, specifically in 1993, 1997, 2001-2005, 2008-2009, and 2012.The driest years with April-October precipitation total of less than 300 mm were 2001-2003 and 2009.Mean air temperature between April and October was the lowest (9.3 • C) in 2002 and the highest (12.5 • C) in 1998.
There were very strong negative and highly significant linear relationships between SOC concentration and soil depth (Fig. 1a-1c).For all species (Fig. 1a), annuals versus perennials (Fig. 1b) and grazing intensities (Fig. 1c), SOC (%) between 0 and 60 cm depth decreased with depth at ∼0.09% per centimeter, as shown by the slopes of the regression equations.There were no significant differences between annuals and perennials or among grazing intensities (p > 0.05).Consequently, the regression equation for mean SOC concentration across species and years during the 1994-1997 period versus soil depth was as follows: SOC% = 6.0931 − 0.0928 × depth; R 2 = 0.9843; P = 0.000839 * * where depth is in centimeters and * * indicates a highly significant negative linear relationship (p < 0.001) between SOC concentration (as %) and soil depth.The R 2 of 0.9843 indicates a very strong linear relationship using the criteria by Fowler et al. (1998).
CAESA study: Equivalent SOC Equivalent (or adjusted) SOC (Mg ha −1 ) for each soil depth interval did not significantly differ among species and grazing intensities (p > 0.05; data not shown).Averaged across species, SOC in the 0-15 cm depth interval accounted for 50%-53% of the total SOC in the 0-60 cm profile.SOC in the 0-30 depth interval was 127-137 Mg ha −1 , which accounted for 76%-79% of total SOC in the 0-60 cm depth interval.Cumulative SOC (addition of SOC in successive soil depth increments) was not significantly different among plant species, annuals versus perennials plant groups, or grazing intensities (p > 0.05; data not shown), and contributions to total SOC in the profile were largest from 0-5, 5-15, and 15-30 cm depth intervals.Overall, the results from the CAESA study showed that there were no differences in SOC between annuals versus perennials, and therefore in that period, they had similar SOC.Retrospective statistical power analysis was not conducted for our study, but it is possible that the lack of SOC differences among treatments may be partly due to low statistical power to detect small SOC changes.This is a common challenge for field-based studies with large plot sizes because it is difficult and impractical to have many replications of the treatments needed to detect small SOC changes.
BMP study: Equivalent SOC stocks Among grouped treatments, the general trend for equivalent SOC was oldgrass > grazed treatments = hayed treatments > barley for silage.The largest equivalent SOC in the 0-5 cm depth interval was 31.8Mg ha −1 under oldgrass (Table 2).Grazed meadow bromegrass control #1 was significantly larger than barley silage (24.4 Mg ha −1 ), while all other treatments were similar, and SOC was intermediate.For all other soil depth intervals, equivalent SOC was larger under oldgrass than barley for silage.Mean equivalent SOC for all hayed treatments was not significantly different from the mean of all corresponding grazed treatments in all depth intervals.However, in three out of five depth intervals (i.e., 5-15, 15-30, and 0-30 cm), the grazed-fertilized meadow bromegrass had significantly larger equivalent SOC than the grazed alfalfa-meadow bromegrass mixture.The grazedfertilized meadow bromegrass treatment had similar equivalent SOC to hayed-fertilized meadow bromegrass in all depth intervals.Total equivalent SOC in the 0-30 cm depth interval was largest under oldgrass and grazed meadow bromegrass control (#1) and smallest under barley-silage treatment.
For each depth interval, there were no significant differences (p > 0.05) in mean equivalent SOC across years.Equivalent SOC in the 0-5 cm depth interval was 25.8 to 27.6 Mg ha −1 over the 4 years of the BMP study.Cumulative equivalent SOC in 0-15 cm and 0-30 cm depth intervals was 76.2 to 78.0 to 118.7 to 122.9 Mg ha −1 , respectively.Comparison of equivalent SOC across years indicated that there were nonsignificant differences over four years for all the depth intervals.
BMP study: Light fraction carbon (LFC) Significant differences among treatments for light fraction carbon (LFC) occurred at all depth intervals (Table 3).For 0-5 cm depth interval LFC stocks for oldgrass and grazed fertilized meadow bromegrass treatments were more than double that for barley silage, and meadow bromegrass controls were 182% of barley silage.Among grouped treatments, grazed meadow bromegrass treatments had significantly higher LFC than hayed treatments (Table 3).Years were not significantly different (p > 0.05) for this depth interval.
For 5-15 and 15-30 cm depth intervals, oldgrass had the highest LFC, and barley silage had the lowest LFC.Grazed versus hayed treatments did not differ significantly in LFC fractions in 5-15 and 15-30 cm depth intervals.Cumulative LFC for 0-15 and 0-30 cm depths showed that oldgrass had the highest LFC, which was more than double that under barley silage.For both depth intervals, grazed treatments had 11.2% higher LFC than hayed treatments.Comparison between years for 5-15, 15-30, 0-15, and 0-30 cm depth in-  -year period (1994 and 1997).SOC, soil organic carbon.tervals indicated that LFC values were significantly higher in 2011 than in 2012, from 1.4 to 2 times greater.LFC is sensitive to environmental conditions; therefore, possible differences between years may have been due to dry conditions and lack of productivity in 1 year versus the other.April to October precipitation during 2010, 2011, and 2012 was 550, 411, and 317 mm, respectively.These precipitation amounts were respectively 152%, 114%, and 88% of long-term normal precipitation for that period.Yearly organic residues, which precede LFC, would be proportional to productivity (yield) of preceding years, and soil microbial biomass would be large initially, consuming organic matter causing respiratory loss in concert with the preceding year biomass and residue size.
Equivalent SOC stocks: Trends in plots used in CAESA and BMP studies Trends in equivalent SOC (0-15 cm depth) for the 20 paddocks that were part of the CAESA study (1994)(1995)(1996)(1997), barley silage (1998( -2007( ) and BMP study (2008( -2012) ) are shown in Figs. 2 and 3. Equivalent SOC (0-15 cm) under continuous oldgrass remained relatively constant over time from 1994 to 1997 through to 2008 to 2012, but all others decreased as they continued in annual forage management for a period prior to perennial re-establishment.Barley silage, hayed alfalfa-meadow bromegrass, and hayed fertilized meadow bromegrass treatments decreased linearly in equivalent SOC at ∼1.0 Mg ha −1 year −1 between 1994 and 2012.However, the rate of equivalent SOC decline was smaller under hayed meadow bromegrass control at 0.75 Mg ha −1 year −1 .Results from regression analyses indicated highly significant negative linear relationships between time and equivalent SOC in the 0-15 cm depth interval for the barley silage (Fig. 2a), hayed meadow bromegrass control (Fig. 2b), hayed alfalfameadow bromegrass mixture (Fig. 2c), and hayed fertilized meadow bromegrass (Fig. 2d).
By contrast all grazed meadow bromegrass treatments re-established on CAESA annual forage paddocks declined in negative quadratic relationships.The results of regression analyses indicated that there was no significant decline (p > 0.05) in equivalent SOC over time under oldgrass between 1994 and 2012, as shown by a relatively flat regression line (Fig. 3a).The other four treatments (two grazed controls, alfalfa-grass mixture, and fertilizer-N applied) showed significant curvilinear relationships between equivalent Note: Within each depth interval equivalent SOC means followed by the same letter indicate non-significant difference among the treatments at 0.05 significance level; * Equivalent SOC based on equal soil mass for each depth interval.The average soil masses (standard errors in parentheses) used were 559 (±2.88)Mg ha −1 for 0-5 cm; 1117 (±5.75)Mg ha −1 for 5-15 cm; 1412 (±6.82)Mg ha −1 for 15-30 cm; 1676 (±8.63)Mg ha −1 for 0-15 cm and 3088 (±10.18)Mg ha −1 for 0-30 cm.LSD stands for the least significance difference test at 0.05 significance level.ξ Grazed meadow bromegrass control was a meadow bromegrass stand that was close to meadow bromegrass fertilized with N; Grazed meadow bromegrass control #2 was a meadow bromegrass stand that was close to the meadow bromegrass-alfalfa mixture.
SOC and time (Fig. 3b-3e).The grazed meadow bromegrass treatments were established on land previously used for barley silage until initiation of the BMP, and therefore the greatest portion of decline in equivalent SOC was between 1994 and 2008.Between 2008 and 2012, the equivalent SOC in the 0-15 depth interval increased by 1.2, 0.48, and 0.10 Mg ha −1 year −1 for the grazed meadow bromegrass control (#1), grazed alfalfa-meadow bromegrass mixture, and grazed meadow bromegrass control (#2), respectively.

Anticipated management impacts
The CAESA and BMP studies were intentionally set up to track SOC under two contrasting forage systems (perennial grass and annual forage pasture or hay) in a Black Chernozemic soil with a high SOC stock.During the 1950s and 1960s, this soil was not considered suitable for crop production (Doran et al. 1963).Some researchers found that soils of similar SOC at Ellerslie, AB (Nyborg et al. 1995), and Melfort, SK, (Malhi et al. 2010) did not increase significantly in SOC stocks with N-fertilizer application or with forages added as hay to cereals in rotation.
The net gain or loss (change) of SOC due to management is the difference between rates of C input and soil respiration (Janzen et al. 1998).Because of the asymptotic nature of car-bon accumulation in soils, those which are already high in SOC, such as the Orthic Black Chernozem used in our studies, may gain organic C at lower rates or have less capacity to accumulate C because respiration rate soon overcomes C inputs after management change (Janzen et al. 1998).In the initial phase the CAESA study, intensive grazing treatments were implemented on perennial and annual pastures with fertilizer-N applied (1993)(1994)(1995)(1996)(1997), whereas after 1997, all annual and perennial paddocks were subjected to zero-fertilizer inputs and annual forages direct seeded into the prior annual pastures for a period of years (1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006).The intent was to emulate management of old pasture and hay stands which consist mainly of grasses and are not fertilized in the region.To treat annual forage land similarly, it was direct seeded but did not receive N-fertilizer to reflect a similar level of management as the perennial.Forage was removed to emulate barley silage production in which grain was removed from the paddocks; therefore, C inputs were limited.It was hypothesized that both oldgrass and barley treatments would lose SOC over this period as vegetative residue-C inputs would be lower than respiration rates.However, this hypothesis was retained for barley treatments which indeed lost SOC but rejected for the oldgrass treatment because that did not lose SOC.
ξ Grazed meadow bromegrass control was a meadow bromegrass stand that was close to meadow bromegrass fertilized with N. ξ Grazed meadow bromegrass control #2 was a meadow bromegrass stand alone that was close to the meadow bromegrass-alfalfa mixture.
in this study was used to build SOC loss that had occurred during the 10 years of barley silage years (1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007) before the re-establishment of perennial grass.Over that period, the SOC levels remained constant under oldgrass, increased the most under N-fertilized re-established perennial grass, and continued to decrease under barley silage.

CAESA study: SOC changes over time and percentage distribution with depth
Factors that contributed to the lack of significant differences in SOC during the CAESA study (1994)(1995)(1996)(1997) were mainly the initially high SOC level as well as large amounts of root mass, return of aboveground residues following grazing, and return of animal excreta (e.g., Baron et al. 2002), which maintained organic C relatively stable.Previously published research on root and litter mass data collected on the same paddocks indicated that root masses under perennial treatments, seeded in 1993, were two to five times that under annual treatments, and the litter mass under perennials was up to twice the amount under annual treatments (Mapfumo et al. 2002).In other experiments conducted in central Alberta, root mass of unfertilized meadow bromegrass (0-15 cm) was ∼3400 kg ha −1 compared to ∼500 kg ha −1 for barley (Malhi and Gill 2002).Several researchers have reported that root biomass is the most important contributor to SOC storage, especially for perennial grasses that tend to develop deep roots and have a high belowground biomass-to-aboveground biomass ratio estimated at 2:1 (Sommer et al. 2000;Lemus and Lal 2005;Monti et al. 2012).
The lack of differences in SOC % and equivalent SOC between 1994 and 1997 was likely due to an initially high SOC in the Black Chernozemic soil.Our previously published data from the same CAESA study site indicated that concentrations in all treatments averaged 56.8 g kg −1 (or 5.68%) for the 0-5 cm depth interval (Baron et al. 1999).Post and Kwon (2000) reported that the maximum annual rate of SOC storage is usually less than 1 Mg C ha −1 year −1 , although that varies depending on geographic locations, soils, and meteorological conditions.Agostini et al. (2015) reported gross C inputs to soil via litter between 1.04 and 2.57 Mg C ha −1 year −1 .They also reported total SOC input from roots within the 120 cm depth were 0.65-1.8Mg C ha −1 year −1 .On average, C inputs from herbaceous crops and grasses were <1 and 3 Mg C ha −1 year −1 , respectively.Given that the Fig. 2. Equivalent (or adjusted) soil organic carbon (Mg ha −1 ) trends for the 0-15 cm depth interval through land use and management changes.SOC, soil organic carbon.* Note: treatments shown were applied between 2008 and 2012, whereas previously (2002-2006) these were all under barley silage.initial SOC in the 0-30 cm depth in our study was approximately 130 Mg ha −1 at the beginning of the CAESA trial, it is unlikely changes of the magnitude indicated above would produce significant differences in only 4 years.However, it is possible that the initial nonsignificant difference in SOC observed at the beginning of the CAESA trial may be in part due to a large type II error because of fewer replications than needed to achieve a very high statistical power.Our study used large plots of 33 m by 9 m in size (292 m 2 ) and therefore operationally difficult to have many replications that would be needed to achieve a high statistical power, especially if the detectable SOC difference between treatments is small.Type II error (or low statistical power) to detect small SOC differences is a common challenge in field-based SOC management studies (Kravchenko and Robertson 2011;VandenByGaart and Allen 2011).However, in our study, the use of blocking as replications and the collection of at least three samples within each plot was a way of reducing type II error.
Over the longer term, comparing SOC in the CAESA trial with those in the BMP trial, SOC declined under annual crop management, whereby grazed triticale or barley-triticale mixture was grown in 1994-1997, followed by growth of bar-ley for silage from 1998 to 2012.However, SOC remained nearly constant under the oldgrass.
The general decrease in SOC with depth observed during the CAESA trial is expected and has been observed in other studies (e.g., Sinoga et al. 2012;Wells et al. 2012;Li et al. 2013;James et al. 2014;Lawrence et al. 2015).The usual relationship generally tends to be an exponential decrease with depth (Sharifi et al. 2018;Sulman et al. 2020;Sun et al. 2020).However, similar to the current work, some studies found the decline in the top 60 cm to be largely linear for organic rich soils (Périé and Ouimet 2008;Sulman et al. 2020).

CAESA and BMP studies: Annuals versus perennials effects on SOC stocks
The maintenance of SOC under oldgrass between 1994 and 2012 suggests that there was a trade-off between C inputs and respiratory losses that occurred even though management change occurred after 1994-1997 CAESA study, due to elimination of fertilizer-N inputs, reduced residue, and drought (2001-2003 and 2012).This constant SOC level for soil under oldgrass provided stability in SOC storage for almost Fig. 3. Equivalent (or adjusted) soil organic carbon (Mg ha −1 ) trends for the 0-15 cm depth interval through land use and management changes.* Note: all treatments shown were applied between 2008 and 2012, whereas previously (2002-2006) these were all under barley silage, except for the continuous perennial grass (oldgrass) which was always under perennials since 1994.SOC, soil organic carbon.two decades, although productivity was not necessarily high.The continuous annual cropping from 1994 to 2012 resulted in a linear decrease in SOC over time.However, from 2008 to 2012, increased SOC occurred in the paddocks converted from cropland to perennial grass that received fertilizer-N application and was intensively grazed.This finding agrees with studies showing effects of increasing the frequency of perennials in annual crop rotations on SOC as simulated by the CENTURY model (Environment and Climate Change Canada 2022), although different species were used.The addition of fertilizer-N to barley silage over the same period did not reduce the rate of SOC loss.
Losses in SOC in barley silage treatments in our study were larger than in oldgrass even though direct seeding was used.No till on black soil in western Canada resulted in saving ∼123 kg ha −1 yr −1 SOC (VandenBygaart et al. 2003).The SOC loss in the barley silage treatment was also exacerbated by the removal of aboveground production as silage.A comparable study (Campbell et al. 1991c) removed wheat (Triticum aestivum L.) straw by baling, although fertilizer was applied, and saw no decline in SOC on a thin black soil.In another study on semi-arid brown soil by Campbell et al. (2000), the fertilized continuous wheat under zero tillage and grain production significantly increased SOC (0-15 cm) over a 10-year period, while SOC under a fertilized crested wheatgrass (Agropyron cristatum L.) remained constant.
The results from our study are consistent with findings from other researchers who reported a decrease in SOC when perennial grassland was converted to annual cropland, and vice-versa.Several studies reported increases in SOC of 0.33-1.0Mg C ha −1 year −1 when annual cropland was converted to perennial grasslands (Smith et al. 2001;Conant et al. 2017;Crews and Rumsey 2017;McGowan et al. 2019).A study comparing experiments in Canada (Alberta, Saskatchewan, and Ontario) over a 17-year period reported mean SOC in the 0-30 cm depth interval that were 9.0 Mg C ha −1 higher under perennial grasses compared to annual crops (VandenBygaart et al. 2010).The replacement of annual crops with perennial grass species increased SOC by 0.6 Mg C ha −1 year −1 , which was close to that observed in our BMP study.For example, in our study, when grazed perennial meadow bromegrassalfalfa mixture replaced barley grown for silage from 2008 to 2012, SOC increased by 0.48 Mg C ha −1 year −1 .The increase was even larger, at 1.2 Mg C ha −1 year −1 , when grazed N-fertilized meadow bromegrass followed barley for silage.
Other studies reported SOC storage under perennial crops between 0.6 and 3.0 Mg C ha −1 year −1 (Lemus and Lal 2005;Sartori et al. 2006).More recently, Dheri et al. (2022) reported SOC gain of 0.99 Mg C ha −1 year −1 over an 8-year period for soil under perennial bioenergy crops (switchgrass and Miscanthus) but SOC loss of 1.03 Mg C ha −1 year −1 from soil under annual crops such as corn and sorghum.A metaanalysis conducted by Martani et al. (2022) indicated that 1 year after arable land was returned to a perennial crop, SOC increased by 15% in the top 0-30 cm depth.They attributed this increase to incorporation of belowground biomass into soil after reversion to perennial crop.Means et al. (2022) showed that intermediate wheatgrass had more SOC storage than annual monoculture and differences in SOC under different plant species were more pronounced in shallow depths compared to deeper soil samples.
CAESA and BMP studies: Grazing versus haying impacts on SOC Grazing intensity or plant species in the CAESA trial had no significant effects on SOC.The fact that the SOC was initially high, the return of litter residues and animal excreta from grazing livestock are the main factors that likely contributed to keeping SOC steady during the 4 years of the CAESA trial, despite paddocks having different grazing intensities and forage species.Few studies in Canada are conducted that include N-fertilization (100 kg N ha −1 ) applications of this magnitude with intensive grazing of this nature.Baron et al. (2002) estimated that the quantity of fecal C was similar between heavy and light grazed meadow bromegrass and represented 68% and 42% of the fecal and residue-C of respective treatments compensating for a lower vegetative residue in the heavily grazed treatment.In agreement, Franzluebbers et al. (2001) also did not find significant differences in SOC mass among high-to-low grazing intensities over a 5-year period in Georgia, USA.
Grazing in the BMP study (2008)(2009)(2010)(2011)(2012) was used on reestablished perennial grass after 10 years of SOC loss due to growth of barley with no fertilizer inputs.We speculate that livestock fecal C, increased belowground root mass, and residue C contributions likely contributed to slightly higher SOC under grazed perennial grass relative to hayed perennial grass treatments, and both treatments had much higher SOC than barley silage.This speculation is based on evidence obtained from studies in Canada and the United States.For example, a meta-analysis on studies of grasslands in the Northern Great Plains indicated that grazing increased SOC by an average of 5.2% and stimulated decomposition of litter 26.8% higher (Wang et al. 2016).North American grasslands have been reported to sequester SOC of 0.07-0.30Mg C ha −1 year −1 annually (Derner et al. 1997;Wang et al. 2016).
As reviewed by Schnabel et al. (2001) less residue is left on the soil surface after hay harvest than after grazing due to dry matter removal.The lower residue yield has a direct relationship on the rate of SOC storage between hay and pasture systems.Studying soils under Bermudagrass [Cynodon dactylon (L.) Pers] with low SOC, Franzluebbers et al. (2001) found that grazing at various levels of intensity sequestered significantly more SOC than under unharvested or hayed treatments.This finding was unexpected, and no clear explanation was provided.After 15 years of side-by-side hay versus grazing trials, SOC in the 0-20 cm depth under grazing was 5.6 Mg ha −1 more than under hay (Franzluebbers et al. 2000) with the greatest difference in the 0-5 cm depth interval.A study by Bremer et al. (2002) on a Brown Chernozemic soil in southeastern Alberta indicated that continuous perennial grass (pubescent wheatgrass--Agropyron trichophorum) harvested for hay resulted in a higher SOC in the 0-15 cm depth by 1.5 Mg ha −1 relative to continuous wheat over a 6-year period.Consequently, our study and other previous studies (e.g., Campbell et al. 1991a) show that for soils with high SOC, residue inputs likely result in relatively small SOC changes than in soils that initially have lower SOC.
Other studies have reported negative effects of grazing on SOC (An and Li 2015), while numerous studies have shown grazing to have a positive effect and, in some cases, no effects on SOC (Derner et al. 2006;Golluscio et al. 2009;Wu et al. 2010;An and Li 2015).In a study in six different climate subregions of Alberta, Hewins et al. (2018) concluded that longterm grazing may enhance SOC in the 0-15 cm depth interval.However, these studies involved extensive grazing with no fertilizer inputs, which is different from intensive grazing with fertilizer inputs applied during the CAESA study and fertilizer inputs applied at the start of the BMP study.
BMP study: Impact of fertilizer N application versus inclusion of alfalfa in mixtures with perennials on SOC During the CAESA study, all paddocks maintained high SOC, which was above the average of 82 Mg ha −1 in the 0-15 cm depth interval for all CAESA and BMP studies treat-ments and years.However, when the paddock was converted to barley silage without fertilizer-N between 1998 and 2006, SOC decreased significantly.Re-establishment of meadow bromegrass on some of the annual paddocks in the BMP trial (2008)(2009)(2010)(2011)(2012) resulted in increasing SOC.The rate of increase was greater for fertilized than unfertilized meadow bromegrass or meadow bromegrass-alfalfa mixture treatments.
Generally, fertilizer-N application will increase productivity and SOC storage in soils of lower nutrient status, but the response of SOC storage to fertilization ranges from positive (Campbell et al. 1991b;Congreves et al. 2014;Karimi et al. 2018) to none (Campbell et al. 1991a;Bremer et al. 2002).In central Alberta, Malhi et al. (1997) reported increased SOC (0-30 cm) by 23.4 Mg ha −1 after 27 years of applying 112 kg ha −1 of fertilizer-N annually to smooth bromegrass on thin black soil.The application of fertilizer N at 100 kg ha −1 to smooth bromegrass managed as hay increased root mass by 57% compared to no fertilizer-N at Lacombe, AB (Malhi and Gill et al. 2002).
Following a decrease in SOC stocks between 1998 and 2006 on paddocks sown to barley for silage, re-establishment of a perennial mixture, meadow bromegrass-alfalfa mixture, increased SOC steadily, but at a lower rate than when meadow bromegrass was N-fertilized.Legumes contribute to SOC storage by supplying aboveground and belowground dry matter supported by N 2 fixation.For example, Malhi et al. (2002) showed that alfalfa in mixtures with smooth bromegrass could economically replace ∼100 kg ha −1 year −1 of fertilizer N applied to smooth bromegrass for hay.Smooth bromegrass fertilized at 100 kg N ha −1 year −1 , alfalfa monocrop and a mixture of alfalfa and smooth bromegrass had similar but significantly greater root mass than unfertilized smooth bromegrass (Malhi and Gill 2002).However, alfalfa content in forage stands may be variable and reduced by stand age, winter kill, and grazing, and legume-dominated stands may not yield as much as fertilized grasslands (Schnabel et al. 2001).
Research conducted by Mortenson et al. (2004) in South Dakota on a Mollisol with low SOC (41.8 Mg C ha -1 ) indicated that inter-seeding alfalfa in northern mixed-grass rangelands increased SOC in the 0-30 cm depth by up to 17% compared to native rangeland control sites that were not inter-seeded with alfalfa.The study did not mention whether fertilizers were applied to inter-seeded or control sites.In a recent study in Europe, inclusion of legumes in grass-legume mixtures increased SOC storage due to higher plant carbon inputs, and belowground biomass decreased nonlinearly with increased legume proportion in legume-grass mixtures (Barneze et al. 2020).
BMP study: Management impacts on LFC LFC represents the intermediate organic carbon pool between undecomposed plant residues and organic matter that has been humified in the soil and is also the main driver for soil respiration (Tan et al. 2007).The LFC is therefore prone to high carbon mineralization rates (Ramirez et al. 2020).Increased residue inputs, establishment of perennial grasses, and reduced disturbance of soils by minimizing tillage prac-tices, reportedly increased the soil's LFC (Malhi et al. 2003).However, a more recent study has revealed that legumes such as alfalfa, milk vetch, and sainfoin grown in a mixture with meadow bromegrass did not influence LFC, but landscape position did have an effect (Issah et al. 2021).
The higher LFC in soil under perennial oldgrass compared to all other treatments was a consequence of minimal soil disturbances, and greater residue inputs have been reported in other studies involving annuals and perennial grasses (Whalen et al. 2000).In our study, the perennial oldgrass treatment had the least disturbance having been seeded in 1993 and not disturbed until 2012.The barley silage treatment was tilled from 1994 to 1997, then direct seeded without tillage between 1998 and 2012, and was therefore more disturbed than the oldgrass treatment.Unfertilized smooth bromegrass had a root mass of ∼4000 kg ha −1 and fertilized barley ∼640 kg ha −1 in the 0-15 cm depth interval on black soil in central Alberta in previous research (Malhi and Gill 2002).Consequently, the highest LFC observed for perennial (continuous) oldgrass, the lowest LFC observed for barley for silage treatments, and intermediate LFC for re-established perennial grass treatments, either hayed or grazed, reflected the spectrum in the range of disturbance (low to high) and residue inputs (high to low).
Higher LFC of soil under grazed treatments versus hayed treatments is attributed to the return of plant residue and fecal C under grazing, while little residue remained after the removal of hay.These results differ from those by Cao et al. (2013) who reported that LFC decreased in the surface horizon due to grazing, with larger decreases as grazing intensity increased.The general decrease in LFC with depth observed in our study for all treatments agreed with the findings of Tan et al. (2007) and reflects higher contributions from aboveground plant residue and roots to the surface layers compared to deeper layers.

Overall implications
Alberta tame hay and pasture areas decreased from 2.26 to 1.55 million ha and from 2.46 to 2.2 million ha, respectively, from 2010 to 2020 (Alberta Statistics Factsheet (2011, 2021).Pasture and hay area decreased by 1.15 million ha from 2006 to 2016 in the subhumid region in western Canada (Liang et al. 2020).Rates of SOC storage or loss may be considered similar on conversion from annual to perennial or the reverse (Liang et al. 2020) but in fact are heavily impacted by conversion management including fertilization (Malhi et al. 2010).The assumption is that this breaking of perennial forage land resulted in an increase in cropland (Liang et al. 2020) and release of SOC (Liang et al. 2020) until another equilibrium has been attained (Janzen et al. 1998).The SOC loss may occur rapidly for some years (5-20) after breaking.On an area basis, re-establishment of perennial tame hay and pasture will have to occur at a faster rate than land broken to regain the SOC regionally and nationally and is dependent on beef cattle numbers increasing (Liang et al. 2020).This implies that efforts and policies to increase SOC storage in agricultural soils will have to promote stability in perennial tame hay and pastureland uses.In practice, land that remains unbroken as perennial may be less productive, receiving no fertilizer-N and P inputs.Our study showed that annual forage such as barley silage as a replacement for perennial forage when managed in a similar low input manner resulted in the loss of SOC at high rates.
An often unrecognized value of older predominantly grass pasture and hay land is the capacity to store carbon by preventing SOC loss at minimal cost of management.Based on the work of Doran et al. (1998), Schuman et al. (2001) estimated that not converting extensive rangeland to cropland in Nebraska prevented SOC loss of 0.3 Mg ha −1 year −1 .Doran et al. (1998) reported SOC declines in the 0-7.6 cm depth of 25.6-20.5Mg ha −1 after breaking native sod followed by notill cropping compared to a decline from 14.2 to 13.7 Mg ha −1 in previously cropped land managed with no-till systems over a period of 11 years.Gain or loss of SOC may be rapid during the first decade of management change (Janzen et al. 1998).However, the loss of SOC from Black Chernozemic soils with relatively high SOC such as those in our study (>1.0 Mg C ha −1 year −1 ) indicates that grassland of this nature may be effective in preventing larger SOC losses.Further re-establishment of perennial forages into annual stands and managed under grazing may be sufficient to slow or halt the SOC loss on cropland managed for low returns of vegetative-C through residues and roots.However, perennial grass harvested as hay may not be as effective in this regard as grazing due to the removal of productivity.

Conclusions
Over 4 years between 1994 and 1997, following the establishment of perennial and annual grasses immediately after breaking of predominantly bromegrass-bluegrass sod with little management, SOC decreased linearly with depth between 0 and 60 cm but was not affected by grazing intensity or plant species.Conversion from perennial grass to continuous annual barley for silage without fertilizer significantly decreased SOC.However, the re-establishment of perennial grass with application of fertilizer resulted in SOC increasing between 2008 and 2012 but did not reach 1997 SOC levels under perennial grass before conversion or SOC under continuous perennial grass (oldgrass) throughout CAESA andBMP study periods (1994-2012).Establishment of barley for silage continuously decreased SOC over time, whereas steady and high SOC occurred in soils under continuous perennial pasture (oldgrass).While the inclusion of a legume (alfalfa) in a grass-legume mixture without fertilizer resulted in a general increase in SOC over the 4-year study, this increase was lower than that for perennial grass with fertilizer N applied.Hay removal of perennial grass with or without fertilizer, or of perennial grass and legume mixtures, resulted in generally decreased SOC throughout the CAESA and BMP study periods.LFC was the highest under continuous perennial grass (oldgrass) with minimal soil disturbance, the lowest with annually tilled barley silage, and intermediate under hayed or grazed perennial grasses either in pure stands or in mixtures with legumes.

Table 1 .
Seasonal and annual mean air temperatures ( • C) and precipitation totals (mm) between 1992 and 2012.
Note: Bold values indicate average temperatures and precipitation totals that were below the long-term normal values.

Table 2 .
Equivalent soil organic carbon (SOC) stock (Mg ha −1 ) under different treatments and depth intervals averaged over the 4 years of the BMP study.Treatment * Equivalent (or adjusted) SOC (Mg ha −1 ) for each soil depth interval

Table 3 .
Equivalent (or adjusted) light fraction carbon (LFC) under different treatments, years, and depth intervals of the BMP study.