Introduction

Managing the soil carbon cycle could help the bioenergy industry to deliver environmental benefits and mitigate the pace of climatic change. In addition to direct fossil fuel offsets, bioenergy cropping systems provide biogeochemical services such as the biological sequestration of atmospheric CO2 in soil carbon reservoirs and biophysical services such as reduced latent heating from evapotranspiration (Margaret S. [1–3]). Carbon sequestration occurs when soil organic carbon (SOC) accumulates more rapidly than it is respired (as CO2 or CH4) by soil heterotrophs. Deeply rooted perennial grasses offer high annual net primary productivity (NPP) and the potential to promote the accrual of SOC [4, 5].

Switchgrass is a perennial, warm-season C4 bunchgrass that is native to North America and is a promising bioenergy feedstock due to large aboveground yields and hardiness across climate zones, soil types, and landscapes [6–8]. Switchgrass is also suitable for marginal lands with low soil quality [8]. The extensive rooting system of switchgrass and its C4 photosystem efficiently use water and nutrients and reduce soil erosion [9, 10]. Switchgrass rooting depths >1 m may also promote the accrual of deep SOC pools in soils where SOC has been depleted by conventional row crop agriculture [11, 12].

The stability of SOC can be viewed as an ecosystem property with physical, chemical, and biological controls. For the purpose of estimating relative stability, SOC pools can be divided into protected and unprotected pools. Aggregate-protected and/or mineral-associated SOC can be isolated and quantified by size or density separation procedures ([13, 14]; von [15]; M. S. [16]). The unprotected or free-particulate organic matter in the low-density light fraction (LF, <1.8 g cm−3) predominantly contains plant necromass (leaf and root litter) with typical turnover times <10 years [17, 18]. The mineral-associated and aggregate-protected dense fraction (DF, >1.8 g cm−3) of SOC has mean residence times on the order of 10 to greater than 100 years ([19, 20]; von [21]).

Soil C storage in switchgrass plantations is a biogeochemical service that can be directly influenced through management practices, such as fertilization and harvesting rates. The responses of soil C and N pools to management practices are key indicators of the role that bioenergy landscapes can play in greenhouse gas abatement strategies [22]. Varied responses of SOC to switchgrass agriculture demonstrate the complexity in plant-soil interaction and the need to study mechanisms of SOC accrual and stability (Table 1). Both fertilizer application rate and harvesting frequency can affect the accrual and long-term stability of SOC by modifying the extent to which organic matter enters protected and unprotected C pools [23, 29]. In this study, we investigated soil C and N stocks in organic matter fractions of differing depth and stability (roots, LF, and DF) in response to two treatments: N fertilization rate and harvesting frequency, applied individually and in combination. We hypothesized that more frequent harvesting would reduce belowground C and N stocks due to preferential allocation of resources to aboveground biomass at the expense of root development, while applications of N fertilizer to the soil surface would reduce the growth of roots deep into the mineral soil profile and therefore attenuate the SOC and TN stocks in the unprotected and protected fractions (LF and DF).

Table 1 Summary of literature on soil C pool responses to N fertilizer in switchgrass plantations
Full size table

Materials and Methods

Field Site

The experiment was established at the W.K. Kellogg Biological Station (KBS) Long-term Ecological Research (LTER) site in southwest Michigan, USA (42°249 N, 85°249 W, elevation 288 m), as part of the Great Lakes Bioenergy Research Center (GLRBC). Mean annual temperature at KBS is 10.1 °C; mean annual precipitation is 1027 mm [30]. The soil is the Kalamazoo soil series, a mixed, mesic-Typic Hapludalf developed on glacial outwash with a fine and coarse loamy texture comprising 85% sand and silt [31]. Cropping history included corn-soybean and alfalfa rotations under conventional tillage prior to the planting of an upland switchgrass variety, “Cave-in-Rock,” on July 11, 2008, at a seeding rate of 7.84 kg/ha.

The experimental design was a randomized split-plot arrangement: four replicate blocks each containing eight plots measuring 4.6 m by 15.2 m. Each plot comprised one fertilization rate that was split into two harvest intensity treatments for a total of 64 plots, each with dimensions of 4.6 m by 7.6 m. Eight fertilization treatments were applied in 28 kg N/ha increments, from 0 to 196 kg N/ha once per year between 2009 and 2011. The recommended N application rates for warm season grass crops in this area are approximately 50–120 kg N/ha [32, 33]. Granular urea 46% N (w/w) was broadcast on June 17, 2009, 1 year after plant establishment. In subsequent years, liquid urea ammonium nitrate (40% NH4NO3, 30% CO(NH2)2, 30% H2O) was applied as a foliar spray at a concentration of 28% N (w/w) in May 2010 and 2011. The plots sampled for this study were those fertilized once annually at rates of 0, 84, and 196 kg N/ha. Harvest intensity treatments were once per year (in November, after a killing frost) or twice per year (July and November) (http://lter.kbs.msu.edu/datatables/375).

Sample Collection and Analysis

Soil samples for this study were collected in July and November of 2011, immediately following the biomass harvest. In 2011, the mean annual temperature and total annual precipitation were 9.6 °C and 1125 mm (http://lter.kbs.msu.edu/datatables/7). Two soil cores from each plot were collected by first removing the litter layer and then pushing a 5-cm steel tube (5 cm diameter with plastic liner) to a soil depth of 60 cm using a hydraulic GeoProbe™. A total of eight cores per treatment (two cores per each of four replicate blocks) were extracted and capped in the field. The liners were split on site, sectioned into four depth intervals (0–5, 5–15, 15–30, 30–60 cm), and sealed in separate plastic bags before being packed with ice in coolers and shipped to Baylor University where they were stored at −20 °C until processed. Each soil sample bag was allowed to warm to room temperature and then weighed as an initial step before handling. Each depth interval for all bulk soil cores were individually homogenized before being processed and analyzed separately. An initial subsample (50–100 g) was oven dried at 50 °C for at least 24 h (to constant mass) to determine soil dry weight for bulk density calculations. A subset of the soils were also oven dried at 105 °C to quantify any potential bias in soil masses obtained at 50 °C (Table S6). Soil bulk density was calculated by dividing the oven-dried weight by the soil core volume for each depth interval after correcting for the mass of the gravel fraction (>2 mm) (http://lter.kbs.msu.edu/datatables/308).

The remaining soil used to calculate SOC and TN stocks was air dried, picked for roots, and sieved to 2 mm. Roots were hand-picked with tweezers, lightly brushed of any adhered soil, and placed in an aluminum tray for drying. Roots and a subsample of the sieved soil were placed in the drying oven at 50 °C for at least 24 h, weighed, and stored for further analysis. Approximately 20 g of the soil subsample was placed in a 50-mL centrifuge tube with approximately 30 mL of sodium iodide (NaI) solution (density = 1.8 g/cm3). After shaking for 30 s by hand, the tubes were centrifuged at 82×g for 20 min. The solution was then allowed to settle before the floating LF was decanted onto glass fiber filters (Whatman, GFF) under vacuum. The LF was rinsed with deionized water to remove residual NaI then dried in the oven at 50 °C for 24 h before being transferred to a glass vial for storage until C and N elemental analysis. The DF (>1.8 g cm−3) remaining in the centrifuge tube was drained and rinsed of residual NaI solution, dried, and stored for future analysis.

The remaining subsample of root-free, oven-dried soil (<2 mm) was homogenized in a planetary ball mill before determining weight percent C and N. The roots were pulverized and homogenized using dry ice and a Scienceware™ Micro-Mill grinder. An initial group of soils treated with 10% hydrochloric acid (HCl) to remove inorganic C produced no detectable carbonate at any sampled depth interval. Therefore, HCl pretreatment was deemed unnecessary for the remaining samples. The soil, root, and LF samples were weighed into tin capsules and combusted in a Thermo Scientific Flash EA 1112 Series NC Soil Analyzer to obtain total organic C and total N concentrations. SOC and TN stocks (kg m−2) were calculated from the elemental concentration, soil layer bulk density, and soil layer depth (stock = concentration (g/g) × soil density (g/cm3) × depth interval (cm)). The C and N stocks in the mineral-associated, dense fraction (CDF and NDF, respectively) were calculated as the difference between whole soil and the free light fraction (CLF and NLF, respectively) stocks: CDF = (SOC − CLF) and NDF = (total N − NLF).

The aboveground switchgrass C and N stocks were estimated as the product of biomass yield and C and N concentrations obtained from KBS LTER datatables (KBS LTER Datatables: Costech Elemental Combustion System CHNS-O, 2004; Total Soil Carbon and Nitrogen, 2009; Plant Carbon and Nitrogen, 2012). Total ecosystem carbon stocks were calculated from the sum of aboveground and belowground stocks as the following: Total ecosystem C stock = (total aboveground biomass C + standing root biomass C + soil CLF + soil CDF). For plots harvested twice annually, the total aboveground biomass C was calculated from the sum of the July and November biomass C yields.

Deep soil core samples were collected immediately prior to switchgrass establishment in June 2008 by KBS staff and sectioned at depth intervals of 0–10, 10–25, 25–50, and 50–100 cm. These samples were passed through a 2-mm sieve, oven dried at 60 °C, and stored in air-tight glass jars at room temperature. Subsamples were sent to Baylor University in 2016 for C and N elemental analysis. Soil C and N stocks were calculated, as described above, using elemental concentration values measured at Baylor and KBS bulk soil density values from the GLBRC Sustainability Data Catalog (KBS LTER Datatables: Soil Bulk Density, 2013). The initial (pre-switchgrass) soil C and N stocks provide a meaningful baseline against which to evaluate the switchgrass treatment effects. However, differences in sampling depth intervals preclude direct quantitative comparisons of initial soil C and N stocks to those for switchgrass treatments using statistical analysis methods.

Statistical Analyses

To test for treatment effects on C and N stocks, we used a three-way analysis of variance (ANOVA) general linear model univariate. The fixed factors in this analysis were fertilization rate, harvest frequency, and depth intervals. Homoscedasticity of data was checked by Levene’s test prior to ANOVA. The p value <0.05 was chosen as the significance level in testing for differences between experimental treatments. The 84 kg N/ha fertilization rate was omitted from the ANOVA due to a lack of data for the November sampling of the twice-annual harvest treatment. Analyses were performed with IBM SPSS statistics 21.0 software (SPSS Inc., Chicago, IL).

Results

Ecosystem Carbon and Nitrogen Stocks Were Highest in Unfertilized Switchgrass Treatments

Treatment plots with the combination of twice-annual harvesting and high rates of N fertilization generated the largest aboveground biomass C and N stocks; however, the root C stock in the annually harvested treatments were significantly larger than twice-annually harvested plots (p = 0.018) (Fig. 1, Table S1). The SOC and TN stocks were highest in unfertilized plots (Fig. 2). The SOC stocks were 13% higher in unfertilized plots than in plots fertilized at a rate of 196 kg N ha−1 (p = 0.004, Fig. 2a), and 85% of the change in SOC stocks occurred below 15 cm between these treatments. The soil TN stocks were also higher in unfertilized plots both in annually harvested (p = 0.006, Fig. 2b) and twice-annually harvested treatments (p = 0.055).

Fig. 1
figure 1

Carbon (a) and nitrogen (b) stocks in total annual aboveground (twice annual: sum of 2011 July and November harvests) and belowground (root) biomass after three full growing seasons under different harvesting and fertilizer treatments. Standing root biomass C and N were measured in November

Full size image
Fig. 2
figure 2

Soil C (a) and TN (b) stocks (roots, LF, DF) at different fertilization rates (open symbols) in Fall 2011. Initial soil C and TN stocks (closed symbols, n = 4) were sampled adjacent to the experimental plots at time of switchgrass establishment. Plotted values are averages across harvest treatments for 0 and 196 kg N/ha (n = 8) and the single annual harvest data for the 84 kg N/ha (n = 4) fertilization rate at each soil depth interval. Horizontal bars are standard error for replicated field plots

Full size image

In our accounting of the total ecosystem C stock, Fig. 3, the CDF was the largest contributor to SOC stocks. Most notably, high N fertilization rates attenuated the total ecosystem C stocks (Fig. 3) due to smaller soil CDF stocks.

Fig. 3
figure 3

Total ecosystem C stocks for switchgrass cropping systems after the 3rd full growing season under fertilizer and harvest intensity treatments. Total ecosystem C stock = (total aboveground biomass + root C stock + soil C stock (light + dense fraction)). Uppercase letters represent significant differences (p < 0.10) between total ecosystem C stocks

Full size image

Treatment Effects on Soil C and N Pools

Fertilization Reduced C and N in the Dense Fraction

The addition of N fertilizer reduced CDF (p = 0.003) and NDF (p = 0.005) stocks by 14% relative to unfertilized controls through the 60-cm soil profile (Fig. 4). The fertilizer treatments did not significantly affect CLF and NLF stocks (p = 0.725 and p = 0.261, respectively) or the root C and N stocks (p = 0.253 and p = 0.225, respectively).

Fig. 4
figure 4

Averaged densefraction C (a) and N (b) stocks by depth in 0 and 196 kg N/ha (open symbols) treatments sampled in November 2011 with harvest intensities of annual and bi-annual pooled by depth interval (n = 8). Initial stocks (closed symbols, n = 4) were sampled adjacent to the experimental plots at time of switchgrass establishment at different depth intervals. Horizontal bars are standard errors for replicated field plots

Full size image

Twice-Annual Harvesting Increased C and N in the Dense Fraction

Soil NDF stocks were 12% larger in the twice-annually harvested plots (p = 0.037). The CLF stocks were 32% larger, and NLF stocks were 18% larger in twice-annually harvested plots (p = 0.049 and p = 0.073, respectively), compared to annually harvested plots (Fig. 5a, b). No major differences were observed between harvest treatments for overall LF mass. The CLF and NLF stocks declined significantly with depth in all treatments (p ≤ 0.01), and on average, 70% of these stocks were located in the upper 15 cm (Fig. 5a, b). The root C and N stocks were considerably more variable between treatments than other C and N pools. Nevertheless, twice-annual harvesting significantly reduced standing root biomass and root C stocks through the 60-cm soil profile (p = 0.026, p = 0.018, respectively; Table S1; Fig. 5c).

Fig. 5
figure 5

End of season distribution of light fraction (a, b) and root (c, d) C and N stocks for annual (open triangles) and twice-annual (open diamonds) harvest frequencies with initial LF stocks shown where measured (closed circles). Horizontal bars are standard error for replicated field plots (n = 8, annual and twice-annual harvest; n = 4, time zero)

Full size image

Soil C and N Pools Changed Seasonally

The SOC and TN stocks declined by 9% from July to November, and SOC stocks were also significantly smaller with N fertilization for both seasons (p = 0.025, Table S3). The late season decline in SOC and TN was driven by a reduction in CDF and NDF stocks, which occurred between the July and November harvests (Table S4). The LF mass was 28% larger with N fertilizer application (p = 0.043, Table S3); however, the CLF and NLF stocks showed no significant seasonal changes between July and November harvest dates. Root N stocks increased from July to November (p = 0.008, Table S4), but no other significant changes were apparent between harvest dates and among fertilization treatments for root biomass, root C stocks, and root N stocks.

Discussion

A review of recent publications on switchgrass agriculture shows substantial variability in the response of SOC stocks to N fertilizer applications (Table 1). The complex interplay of substrate quality (plant residue chemistry), nutrient availability, soil redox gradients, microbial enzyme capacity/activity and community structure, soil mineralogy, and available surface area may contribute to disparate responses of SOC and the effects of N fertilization across switchgrass field trials. In this study, we found several important changes in soil C and N with harvesting and fertilizer treatments. The SOC and TN stocks were significantly larger in unfertilized switchgrass stands. Approximately half of the SOC and TN stocks are found at depths >15 cm (Fig. 2) and predominantly in the mineral-associated dense fraction (Figs. 3 and 4). Additionally, twice-annual harvesting caused a reduction in the root C and free-particulate CLF stocks.

Changes in Soil Carbon and Nitrogen Stocks

The unfertilized SOC stocks to 60 cm depth measured 0.78 kg C m−2 larger than the fertilized treatment over the course of the study (3.7 years), corresponding to steady-state change of 0.21 kg C m−2 year−1. The annualized rate of 0.21 kg C m−2 year−1 to 60 cm depth is similar to those reviewed by [34], where the average SOC accrual was 0.1 kg C m−2 year−1 to 30 cm for fertilized sites. None of the perennial grass sites they reviewed were unfertilized. Follet et al. [24] also observed an accrual rate of 0.2 kg C m−2 year−1 to 150 cm, where half of the SOC accumulated at depths below 30 cm. These relative rates of SOC change are relatively modest, and we note that [27] found no significant changes in SOC at the KBS GLBRC site, but took fewer samples and did not fractionate nor include root biomass. Nevertheless, modest SOC accrual rates can lead to significant C sequestration if the accrual occurs within protected soil pools with potential for long-term stability. The N fertilizer treatment may be detrimental to long-term sequestration potential by affecting both the accrual depth and mineral association of C and N stocks [5, 35].

Nitrogen Fertilizer Reduced Soil CDF and NDF Stocks

The N fertilizer treatment plots had significantly lower CDF and NDF stocks compared to the unfertilized control, mainly at depths >15 cm (Figs. 2 and 3). This result is important because deeper soil C pools have longer mean residence times, which can be attributed to lower O2 availability and slower rates of decomposition and mineralization [3, 36, 37]. The residence time (radiocarbon age) and the thermodynamic stability of SOM typically increase with soil depth [38–40]. Radiocarbon dating and laboratory incubation studies indicate that SOM associated with soil minerals (both mineral-bound and aggregate-occluded) has greater stability against biodegradation than free-particulate SOM (Margaret S. [1–3]).

The causal mechanism for the rapid response of CDF to N fertilizer remains unclear, but we consider two likely mechanisms. First, molecular level studies of grassland SOM suggest that roots and microbial biomass are the predominant sources of organic matter in the dense fraction (or humin fractions) [41–43]. Our measurements at KBS indicate that root biomass C is ∼30% lower in the fertilized plots (196 kg N ha−1) than the unfertilized plots, though the effect was not statistically significant in the 2011 samples (p = 0.25, Table S1). Nevertheless, a reduction in root C inputs may have contributed to lower CDF and NDF over the 3.7-year duration of the study. Second, N fertilization may reduce SOM accrual in the dense fraction by indirect effects on SOM decomposition rates, caused by changes to SOM chemical composition and/or microbial activity. For instance, high rates of N fertilization can increase root decomposability through the reduction of root C/N ratios [44]. Furthermore, soil nutrient availability can affect microbial community structure and activity and promote or retard the decomposition of SOM [45, 46]. Chen et al. [45] demonstrate that N fertilizer added to soil in combination with fresh plant residues tends to accelerate the mineralization of organic matter. Acceleration of the decomposition rate may reduce the accrual of SOC and TN.

Twice-Annual Harvesting Reduced LF and Root C and N Stocks

Mechanisms for the reduction in CLF and NLF pools with twice-annual harvesting (Fig. 5a, b) could be due to a more efficient removal of aboveground biomass and therefore less incorporation into the soil C and N pools, or the increased exposure at the soil surface favoring increased erosion (physical transport) and aerobic (biotic) or photic (chemical) decomposition of surface residues and associated LF organic matter. In the present study, root C stocks below 15 cm represented 30–45% of total root C to 60 cm for all samples collected in November. The smaller root C and N stocks observed in the twice-annually harvested treatment (Fig. 5c, d; Table S1) may be from the mid-season harvesting disturbance which could modify resource allocation to aboveground biomass. The 12% reduction in root C stocks with fertilization at the deepest depth (30–60 cm) may be a function of nutrient availability at the surface. The reduced root C and N inputs may also have contributed to the lower CLF and NLF pools in the twice-annual harvesting treatments, as root biomass can be transformed into LF SOM [28].

Soil Dense Fraction C and N Declined Rapidly Between Summer and Fall Harvests

The rapid decline of the CDF and NDF pools over the intervening months between July and November harvests is surprising, given the presumed stability of this fraction (Table S3 and S4). There are several mechanisms that might explain such a rapid reduction of CDF and NDF stocks between harvests. (1) Seasonal soil aggregate stability could diminish between seasons as a function of increased autumn precipitation and cooler temperatures [47, 48]. (2) The priming of microorganisms by surface residue additions during the mid-season harvesting and the soil disturbance associated with that harvest could accelerate the mineralization CDF [49]. (3) Alternatively (or additionally), mid-season harvesting could cause a reallocation of photosynthate from root growth to shoot growth, leading to a decline in the substrates supporting mineral-associated microbial biomass, thus diminishing CDF and NDF between harvests (De [50]). The reduction in CDF was larger in the unfertilized treatments between harvests; however, the unfertilized plots had significantly larger CDF and NDF stocks at both harvest dates. This implies that high rates of N fertilization and harvesting, which reduce the production of roots, CLF, and NLF stocks, may also affect inputs to the CDF and NDF pools [51].

Summary

Although a primary objective in bioenergy production is maximizing aboveground biomass for use as feedstock for energy and fuel, energy conservation and soil C storage are also valuable biogeochemical services [52] that can further reduce the carbon intensity of bioenergy systems. Our results show that the largest total ecosystem C stocks (aboveground + belowground) were achieved with the least energy-intensive agricultural practices: no N fertilizer and a single postseason harvest. Harvest intensity and N fertilizer rates affected the magnitude of soil C and N storage, as well as the depth and relative stability of the C and N pools. The changes in SOC occurred primarily at depths greater than 15 cm and in the dense fraction of the SOC pool where organo-mineral associations provide a mechanism for long-term soil C storage. The N fertilizer treatments caused a reduction in soil C stocks, particularly in the mineral-associated fraction, while the combination of annual harvesting and N fertilization reduced soil N stocks in the mineral-associated fraction. The twice-annual harvest treatment reduced LF and root C pools. Unfertilized switchgrass plots contained 15% more SOC, on average, 4 years after planting than did plots under high fertilization rates. Ruan et al. [27] recently demonstrated the high carbon cost of fertilizing biomass crops such as switchgrass. Our findings demonstrate that management practices that minimized carbon emissions from N fertilization and mechanical harvesting also enhanced the magnitude and longevity of soil carbon storage.