Impacts of tillage practices on soil carbon stocks in the US corn-soybean cropping system during 1998 to 2016

The DLEM model was driven by spatially-explicit datasets, including daily climate conditions (average, minimum and maximum temperature, precipitation, shortwave solar radiation, and relative humidity), annual land use and cover change, monthly concentration of atmospheric CO₂, annual nitrogen (N) deposition, and agricultural management practices (such as N fertilizer use, irrigation, manure N application, tillage, and tile drainage).

2.1.1. Climate, CO₂ and nitrogen deposition data

2.1.2. Land use and land cover database developed

2.1.3. Tillage and other agricultural management practices

As we outlined in Yu et al (2018), we represented tillage impacts on C, nitrogen and hydrological cycles in DLEM following the mechanisms described in Fox and Bandel (1986), Lemunyon and Gross (2003), Li et al (1994), Linn and Doran (1984), and Gilley (2005). More specifically, the DLEM model assumes that, when compared with no till practice, (1) decomposition rate is increased by 1.5 and 3 times under plowing (conservation tillage) and disking (conventional tillage), respectively; (2) conservation tillage and conventional tillage reduce residue cover from 80% (no-till) to 30% and 15%, respectively; (3) soil moisture is reduced through enhancing evaporation by 1.207 and 1.448 times for conservation and conventional tillage, respectively; (4) denitrification is reduced by 15% and 30% for conservation and conventional tillage; and (5) soil erosion under conservation and conventional tillage is increased according to the residue type left from previous year (table 1). In DLEM, hydrological loss of C is the sum of C leaching in the forms of dissolved inorganic carbon (DIC), dissolved organic carbon (DOC) and particulate organic carbon (POC). Tillage indirectly affects DIC leaching through soil respiration and DOC leaching by altering the soil DOC amount and water flux (see supplementary file is available online at stacks.iop.org/ERL/15/014008/mmedia). Therefore, although DIC and DOC losses are primarily natural fluxes, their changes indirectly caused by tillage practice changes were accounted in this study. POC loss (mainly through soil erosion) is calculated with the Modified universal Soil Loss equation (Chaubey et al 2006, Williams and Berndt 1977). More details can be found in supplementary file and Tian et al (2015). Here we modified the way in which we calculate POC loss by considering the tillage-caused crop residue removal for different crop types as shown in table 1. According to observational studies (Miércio et al 2017, Omonode et al 2007, Ruan and Philip Robertson 2013), we assumed that tillage impacts can only last for a two-month period after implementation and the effects will linearly decrease through time.

As mentioned earlier, the DLEM model has been intensively calibrated at both site- and regional scales (Yu et al 2018). In this study, we revisited the calibration using the sites collected in a previous study (figures 1(a), (b)). Specifically, additional validations were performed using four individual studies to examine model capability in capturing tillage impacts on CO₂ emission and soil C loss in the US (figure 1(c)). The validation of model performance in simulating gross primary productivity (GPP) and crop yield can be found in Lu et al (2018a), Yu et al (2018), and Yu et al (2019).

Zoom In Zoom Out Reset image size

The calibrated and validated model was first run to equilibrium state with input drivers being set to the 1850 level or the earliest available year to obtain the initial status of C, N, and water pools for each biome in each grid-cell. The earliest climate data are from 1900, and we use 30 year (1901–1930) average to characterize climate condition before 1900 in the equilibrium run. Equilibrium state is achieved when the variations of net fluxes are less than 1 g C m⁻² yr⁻¹ for C, 1 g N m⁻² yr⁻¹ for N, and 1 mm m⁻² yr⁻¹ for water within a 20 year simulation cycle (Yu et al 2018). Before implementing transient runs using initial state information from the equilibrium run, we applied a 10 year spin-up run using climate data randomly selected between 1900 and 1930 to avoid sudden changes that can result from mode transition.

We then set up simulation experiments (as shown in table 2) to distinguish and quantify the effects of tillage practices and tillage-intensity-change (TIC) on SOC changes. More specifically, the first simulation experiment (S1) was designed to produce our 'best estimate' of C stocks and their changes in the US, which was driven by historically varying tillage intensity and other input drivers (e.g. climate, N deposition, atmospheric CO₂, land conversion and crop rotation, crop technology improvement, fertilizer use, manure application, irrigation, and tile drainage). The second simulation experiment (S2) was a 'business as usual (BAU)-1998' reference case designed to keep intensity of conservation and conventional tillage 'fixed' since 1998 (we keep areas under these two tillage types unchanged, assuming that new cropland adopted no-till practice). The impacts of TIC on C storage during the period 1998–2007 are distinguished by comparing S1 with S2 (figure 2). Similarly, the third simulation experiment (S3) was designed to hold the two tillage intensities 'fixed' since 2008, which served as the 'BAU-2008' reference. By comparing S1 and S3, we can identify the impact of TIC on C storage during the 2008–2016 period (figure 2). Based on the above experiments, we estimated how TIC has affected C storage in the US. Moreover, we set up a fourth simulation experiment (S4), which assumed that the no-till practice was adopted in all croplands since 1998, and the fifth simulation experiment (S5) assuming that all conventional tillage land has been shifted to conservation tillage since 1998. Comparison of experiments S1 and S4 provides us with historical tillage impacts on cropland SOC (figure 2), while comparison of S1 and S5 implies potential SOC change of adopting conservation tillage in the US corn-soybean cropping system.

Zoom In Zoom Out Reset image size

Additional simulation experiments were designed to account for uncertainties of the TIC-induced C storage change. Three major uncertainty sources were quantified. The first uncertainty source is from the parameters used in LUCC-induced soil C and N loss in continuous cropland expansion area. Following Yu et al (2018, 2019), conversion-specific parameter values were adopted to describe instantaneous C loss during cropland expansion into different ecosystem types. Parameter-induced uncertainty from C storage change was derived from multiple simulation experiments using parameter values for average C/N loss percentage ±1 standard deviation. The second uncertainty comes from the harvest-related residual removal rate used in model simulations. The 'best estimate' simulation we performed adopted a coefficient of 0.5 for residue removal which was obtained from Allmaras et al (2000) and Perlack et al (2011) assuming that 50% of the residue was removed from cropland. The residue removal rate determines how much C is left on ground after crop harvesting, which is directly related to soil C storage change. This roughly estimated coefficient may vary by location and crop type, and may affect soil C accumulation/loss. We then adopted 40% and 60% residue removal coefficients as alternatives to the 50% coefficient when implementing our uncertainty analysis. The third uncertainty is from the timing of tillage implementation. Generally, fall tillage may be adopted before spring tillage in part of the US cropping areas (USDA Natural Resources Conservation Service and University of Wisconsin—Extension 2019). Due to the lack of timing information, we designed two types of experiments to quantify the impacts of with- and without-fall tillage practice. According to published research, fall tillage dates vary but generally occur within a month after harvest and spring tillage generally occurs between March to May around planting (Al-Kaisi and Yin 2010, Reicosky and Lindstrom 1993, Renner et al 1998, Vetsch and Randall 2004). In this study, fall tillage was implemented two weeks after harvest and spring-tillage was implemented at the beginning of planting. We used simulations from spring-tillage as our 'best-estimate', while the experiments with corn-soybean land tilled twice annually (i.e. fall and spring tillage) represent more intensive soil disturbance scenarios.

Corn and soybean areas increased by 18.9% and 15.6% in the US from 1998 to 2016, leading to a total area expansion of 10.8 Mha for the two crops during the period (figure 3). Based on the survey data and spatial maps, we found that the area of no-till increased by 71.6% from 1998 (16.9 Mha) to its peak year at 2008 (28.9 Mha), in which corn and soybean contributed 5.2 and 6.8 Mha, respectively. The total area under conservation tillage dropped by 5.6 Mha from 1998 to 2006 with corn and soybean contributing to 56.9% and 43.1% of the decrease, respectively. The lowest area of conventional tillage was recorded in 2008 with the decline mainly from soybean (−2.9 Mha reduction, figure 3(c)), while corn area under conventional tillage did not show a consistent trend during the period before 2008 (figure 3(a)).

Zoom In Zoom Out Reset image size

No-till land in corn and soybean declined by 2.6 Mha from 2008 to 2016, of which 2.4 Mha was from soybean planting area. In comparison, conservation and conventional tillage showed increasing trends in corn land from 2006 and in soybean land from 2007 (figure 3). More specifically, corn acreage under conservation and conventional tillage increased by 2.7 and 2.5 Mha from 2006 to 2016 (figure 3(a)). Soybean acreage under conservation and conventional tillage increased by 2.8 and 4.7 Mha from 2007 to 2016 (figure 3(c)).

Spatial analysis revealed that cropland under the no-till practice increased from 1998 to 2008 in the US with Eastern Nebraska as a hot spot for no-till increase (figure 4(a)). In contrast, conservation and conventional tillage have decreased in most other cropland areas during this period (figures 4(c) and (e)). Declines in conservation and conventional tillage were mainly found in the region where no-till practice increased, such as the Nebraska, Wisconsin, and South Carolina (figures 3(a), 4(c), and (e)).

Zoom In Zoom Out Reset image size

For the 2008–2016 period, however, tillage trends were reversed in most of the regions when compared with the earlier period. For example, the no-till practice declined in the mid- and Lower-Mississippi River Basin, and especially the Mississippi Alluvial Plain, during the period (figure 4(b)). Mixed trends of tillage intensity were found in the US with 12.1% and 9.0% of the cropped area showing increased and decreased conservation tillage (figure 4(d)), respectively. By contrast with 1998–2008 period, conventional tillage was found to have generally increased from 2008 to 2016, and especially so in California, Kansas, Northern North Dakota, Illinois and Minnesota (figure 4(f)).

The best-estimate simulation (S1) showed that tillage-induced soil C loss was 10.3 ± 1.9 Tg C yr⁻¹ (mean ± std, standard deviation derived from multiple uncertainty experiments) in the US during the entire 1998–2016 period (figure 5, table 3). However, interannual variations are evident. The tillage-induced SOC loss decreased from 14.7 Tg C yr⁻¹ in 1998 to 7.1 Tg C yr⁻¹ in 2008 then increased to 10.3 Tg C yr⁻¹ in 2016. Approximate 88% of the tillage-induced SOC reduction was attributed to gaseous loss, while the remaining 12% (1.25 Tg C yr⁻¹, table 3) was due to hydrological processes and was dominated by water erosion (>99%) (figure 5).

Zoom In Zoom Out Reset image size

In comparison to the tillage impacts (difference between experiments S1 and S4), the TIC-induced SOC loss (differences between experiments S1, S2 and S3) was relatively small at 0.50 ± 0.16 Tg C yr⁻¹ during this period (table 3). Indicated by our model estimates, TIC led to a SOC accumulation of 1.0 Tg C yr⁻¹ during the 1998–2008 period, while it caused a SOC loss of 2.4 Tg C yr⁻¹ during the period of 2008 to 2016 (table 3). We partitioned TIC-induced SOC loss into two pathways and found that gaseous- and hydrological-C losses contributed to approximate 85% and 15% of the loss for the study period, respectively (table 3). However, although the TIC-induced hydrological-C loss was relatively small, contrasting impacts of reduced (−0.13 Tg C yr⁻¹) and enhanced (0.31 Tg C yr⁻¹) hydrological loss were also found in the periods of 1998 to 2008 and 2008 to 2016, respectively (table 3).

For the period of 1998 to 2008, TIC-induced soil C accumulation were detected in most cropland areas, especially in Nebraska, Illinois and Alabama (figure 6(a)). In contrast, during the period of 2008 to 2016, TIC-induced soil C loss was found in most of the Corn-belt region, except for the central Iowa and the south of Minnesota (figure 6(b)). Total soil C change during the periods 1998–2008 and 2008–2016 were −10.0 Tg (C accumulation) and 19.2 Tg (C loss), respectively. The estimate range of TIC-induced soil C change, represented by standard deviation of all uncertainty-related simulation experiments (described in section 2.4) divided by the multi-simulation average, were about 5%–15% for the study area (figures 6(c) and (d)). Nonetheless, it should be noted that uncertainties may also come from under-represented various soil C responses to tillage along depth, and the lack of tillage information before 1998.

Zoom In Zoom Out Reset image size

People learned from the catastrophe of Dust Bowl that management practices are required to replace excessive tillage (Baumhardt 2003). Reduced tillage intensity practices, such as conversions from conventional tillage to conservation tillage and no-till have been advocated for the purpose of environmental protection (Lal 2015). Boosted by technology development and elevated fuel prices, conservation and no-till farming almost tripled between the 1980s and 1996 (Reagan 2012, Uri 1998). Such reduced tillage intensity in the US has been reported in various studies (Banerjee et al 2010, Fawcett and Towery 2003, Uri 1998). More recently, reduced tillage intensity has been reported to arise primarily from an increase in no-till cropland—from 15.7 Mha in 1994 to 25.3 Mha in 2004 (Conservation Technology Information Center 2018, Banerjee et al 2010), which can be partly attributed to encouragement from government subsidy and other protection programs (Huggins and Reganold 2008). Consistent with the above, we found an increase of no-till land in corn-soybean rotation system by 71.6% from 1998 (16.9 Mha) to its peak year at 2008 (28.9 Mha). Besides, technology improvements, such as no-till machines and herbicide-resistant crops are essential drivers of reduced tillage intensity in the US. For example, USDA survey data revealed that the acreage percent of herbicide-tolerant corn and soybean rose from less than 17% in 1997 to over 90% in 2018 (Perry et al 2016, USDA ERS 2018).

Nevertheless, we found that trend toward lower tillage intensity has reversed in the most recent decade. No-till land in corn and soybean declined by 2.6 Mha from 2008 to 2016, while conservation tillage increased by 5.5 Mha from 2006 to 2016 and conventional tillage increased by 7.2 Mha from 2007 to 2016, respectively. Such a tillage intensity reversal from reducing to increasing coincides with a reversal in Conservation Reserve Program (CRP) land area, from increasing to decreasing. The US CRP land expanded by 2.7 Mha from 1998 to 2007, but dropped by 5.2 Mha from 2007 to 2016 (USDA, Farm Service Agency 2019). Since the corn-soybean cropping system is a primary use of exited CRP land and plowing (conventional tillage) CRP land is an effective approach to kill grasses and weeds which allows crop to be planted in the first year of conversion, we surmise that the rise of intensive tillage area may greatly contributed from CRP land released since 2007. However, it should be noted that the potential linkage between CRP land change and tillage intensity may not be dominant, and needs to be further explored. For example, wheat was also one of the most common first crop to be planted after conversion from CRP land to cropland (Lark et al 2015).

The other possible explanation for rising tillage intensity after 2008 is due to increasing resistance of weeds to herbicides (Cerdeira and Duke 2006, Perry et al 2016). The number of weed species resistant to glyphosate has increased rapidly since 2001 in the US (van Deynze et al 2018, Price et al 2011, Wechsler et al 2018). This increased weed resistance may have compelled farmers return to plowing over the most recent decade. Although more evidence is required to identify the dominant driver, the undisputed rising tillage intensity for the last decade in US corn-soybean cropping system implies an urgent need to assess its environmental impacts, including the changes in biodiversity, water pollution, GHG emission, and soil erosion.

Previous studies have reported varied magnitudes of SOC change induced by tillage practices in different regions. By simply extrapolating from randomly selected sites to the entire US Corn-belt, Lee et al (1993) projected a soil C loss at 1.0–3.2 Tg C yr⁻¹ under various tillage intensity scenarios for the next century. Similarly, Bernacchi et al (2005) extrapolated from a single site measurement to the entire US corn-soybean agricultural system and estimated that no-till sequestered 2.2 Tg C yr⁻¹ and conventional tillage released 9.4 Tg C yr⁻¹. However, such simplified estimations may be largely biased in upscaling. It should be noted that none of these studies quantified the spatially TIC-induced SOC changes or attributed them to different SOC loss ways as we did in this study. The results reported in this study indicate the SOC difference between the scenarios of historical tillage and fixed-intensity tillage (assuming new cropland adopting no-till practice). We found that, although the increasing adoption of the no-till practice in the 1998–2008 period enhanced soil C storage by 10.0 Tg (1.0 Tg C yr⁻¹), the SOC accumulation was totally offset by the rising tillage intensity during the 2008–2016 period, leading to a net loss of 9.0 Tg (0.50 Tg C yr⁻¹) soil C for the entire 1998–2016 study period.

By comparing with the size of C accumulated in other protection programs, we found that the escalated plowing in the recent decade increased C loss by 2.4 Tg C yr⁻¹, which offset 31% of SOC annually sequestered in CRP land (7.78 Tg C yr⁻¹, calculated from CRP land area in 2009 and the SOC accumulation rate of 57.0 g C m⁻² yr⁻¹ reported by Piñeiro et al 2009), or 45% of the C annually sequestered in the US protected lands (5.4 Tg C yr⁻¹ sequestered in public and private lands under some level of conservation, Lu et al 2018b). By considering both spring and fall tillage, the simulated net loss of soil C could be even larger at 11.5 Tg (0.64 Tg C yr⁻¹) during the entire study period.

The C lost from soil is only partially released to the atmosphere as GHG (e.g. CO₂, CH₄). A fraction of C was transported from a site through hydrological processes (i.e. leaching and water erosion) and deposited at another site, or in streams, rivers, and ocean sediment, where it could remain sequestered indefinitely (Lee et al 1993). The water erosion-associated C loss was 21.7 Tg C yr⁻¹ in the US in 1990, as reported in Van Oost et al (2007), which is slightly higher than 17.3 ± 0.60 Tg C yr⁻¹ during the 1998–2016 period as estimated in this study (standard deviation derived from historical tillage change experiments with different uncertainty sources). This difference is reasonable as the former study covers both the US and Canada croplands. We also found that reduction of hydrological-C loss contributed to 13.0% of the soil C increase for the 1998–2008 period, while the hydrological-C loss was accelerated by increasing tillage intensity and contributed to 12.9% TIC-induced soil C loss for the 2008–2016 period.

Aside from the accelerated C loss from soil, CO₂ emission from use of tillage machinery is boosted. Fossil fuel consumption from conventional tillage and conservation tillage increased CO₂ emission by 4.9 g C m⁻² and 2.2 g C m⁻² in corn land and 1.7 g C m⁻² and 4.4 g C m⁻² in soybean land when comparing with the no-till practice (West and Marland 2002). Simply extrapolated to the entire US, adopting reduced tillage from 1998 to 2008 reduced CO₂ emission by 0.24 Tg C from machinery energy use, which is 2.4% of the magnitude of TIC-induced SOC accumulation or 18% of the TIC-induced hydrological-C loss reduction. Nevertheless, increased tillage intensity for the later 2008–2016 period enhanced fossil fuel CO₂ emission by 0.36 Tg C, which amounts to 8% of the TIC-induced SOC loss and 12% of the hydrological-C loss.

Limitations in this study should be addressed in future studies. The major limitation was from the data available for analysis. Potential biased in model estimation can be reduced if tillage information before 1998 and tillage data of other crop types were available. In addition, uncertainties may come from under-represented tillage responses of soil C in different depth. Like many other terrestrial ecosystem models, DLEM has multiple C compartments in the soil to represent C substrates with different decomposability and responses to environmental drivers, but these C pools have no depth-related feature in model. Improvements can be achieved by stratifying soil C pools into different layers to better capture tillage impacts on soil profiles according to tillage type used. More discussions regarding uncertainties can be found in the supplementary file.

Benefits from adoption of no-till practices have been elaborated in many previous studies (Van Oost et al 2007, Six 2013, Ussiri et al 2009). Although unrealistic, we estimated that converting all cropland from conservation and conventional tillage to no-till (difference between experiments S1 and S4 in table 2) could potentially accumulat SOC by 10.3 Tg C yr⁻¹ (if historical till practices were implemented in spring only) to 15.2 Tg C yr⁻¹ (historical till implemented in both spring and fall) during the 1998–2016 period (table 3). There are several reasons preventing the change from conventional tillage to conservation tillage. First, the no-till practice often leads to heavier use of herbicides and pesticides, rendering environmental concerns and negative effects on investments and products (Plumer 2013). Second, reduced tillage increases soil wetness, which retards soil warming up in spring, leading to a potentially shorter growing season and therefore lower crop yield (Soane et al 2012, Turmel et al 2015). Due to higher water retention, farmers using no-till practice have to wait until the field dries naturally before planting can be implemented, while traditional conventional tillage dries out field quickly and therefore enables earlier planting opportunity (Reagan 2012).

Conservation till, by leaving at least 30% residue cover on surface after planting, provides compromised environmental and economic benefits when comparing with no-till and conventional tillage. For example, it reduces soil erosion and compaction, increases soil organic matter, and improve wildlife habitat (e.g. crop residue provides food and shelter) (Afzalinia and Zabihi 2014, Allen and Vandever 2012). We found that, by shifting all the conventional tillage area to conservation tillage, a SOC accumulation potential of between 8.7 and 13.2 Tg C yr⁻¹ can be achieved for the 1998–2016 period (the range indicating an estimated spread between the spring-till only scenario and that in which both spring and fall tillage occur; difference between experiments S1 and S5 in table 2). Besides, conservation till also reduces fuel consumption and maintains or even enhances crop yield (Busari et al 2015, Farooq et al 2011, Mileusnić et al 2009). For environmental benefit, we expect expansion of no-till and conservation tillage will be the dominant trend in US cropland. Nevertheless, the drivers of recent increasing tillage intensity should be explored and identified to project future tillage trends, perform accurate assessments, and promote reduced tillage intensity practices while not harming crop production.