Impact of Sugarcane Cultivation on C Cycling in Southeastern United States Following Conversion From Grazed Pastures

Abstract

ABSTRACT

The expansion of sugarcane, a tropical high‐yielding feedstock, will likely reshape the Southeastern United States (SE US) bioenergy landscape. However, the sustainability of sugarcane, particularly as it displaces grazed pastures, is highly uncertain. Here, we investigated how pasture conversion to sugarcane in subtropical Florida impacts net ecosystem CO₂ exchange (NEE) and net ecosystem carbon (C) balance (NECB). Measurements were made over three full growth cycles (> 3 years) in sugarcane—plant cane, PC; first ratoon cane, FRC; second ratoon cane, SRC—and in improved (IM) and semi‐native (SN) pastures, which make up ca. 37% of agricultural land in the region. Immediately following conversion, PC was a stronger net source of CO₂ than pastures, indicating the importance of CO₂ losses related to land disturbance. Sugarcane, however, shifted to a strong net sink of CO₂ after first regrowth, and overall sugarcane was a stronger net CO₂ sink than pastures. Both stand age and low water availability during cane emergence and tillering substantially decreased its potential gross CO₂ uptake. Accounting for all C gains and removals (i.e., NECB), greater frequency of burn events and repeated harvest increased removals and overall made sugarcane a stronger C source relative to pastures despite substantial C inputs from the previous land use and a stronger CO₂ sink strength. Time since conversion substantially reduced C losses from sugarcane, and the NECB of SRC was similar to that of IM pasture but lower than that of SN pasture, indicating a rapid shift in the NECB of cane. We conclude that the C‐balance implications following conversion will depend on the proportion of IM versus SN pastures converted to sugarcane. Furthermore, our findings suggest that no‐burn harvest management strategies will be critical to the development of a sustainable bioenergy landscape in SE US.

Introduction

Bioenergy is a central mitigation strategy to limit global warming to below 2 carbon (C; Fuss et al. 2014). The Southeastern United States (SE US) has the capacity to produce nearly a third of the 36 billion gallon target established by the Energy Independence and Security Act (EISA 2022). Endemic to tropical and subtropical regions, sugarcane (Saccharum officinarum) offers an excellent alternative for bioenergy production over extensive areas at low latitudes where other, more common, dedicated bioenergy crops tend to underperform (Blanc-Betes et al. 2023), and is, therefore, likely to reshape the US bioenergy landscape (US Department of Energy 2011). However, the role of a cane-based biofuel industry in climate mitigation portfolios is contingent upon its long-term sustainability, largely driven by the ability of canes to sustain yields and retain carbon (C) following expansion. Direct measurements of ecosystem CO₂ fluxes upon land conversion to cane cropping systems and over multiple years with assessments of imports and exports of C are urgently needed.

With high yields, critical coproducts, and a range of hybrids for optimal allocation, canes represent an ideal candidate for biofuel production (Formann et al. 2020; Matsuoka et al. 2015). Unsurprisingly, sugarcane provides 96% of bioethanol in Central and South American countries (FAOSTAT 2024). In the United States, however, despite a still modest contribution to bioethanol production, sugarcane production has increased by 35% in the last two decades (FAOSTAT 2024).

The impact of cane cultivation on the ecosystem C cycle has been previously investigated with most studies focusing on sugarcane plantations several years after land use conversion. Studies on cropland with a long history of sugarcane cultivation consistently report sufficient gross primary productivity (GPP) to offset ecosystem respiration (R_eco), indicating that established sugarcane ecosystems are net sinks of CO₂ (Anderson et al. 2015; Cabral et al. 2013, 2020). The net ecosystem C balance (NECB) of sugarcane, which reflects the C stored in an ecosystem or lost from the ecosystem after accounting for all C inputs and removals, however, can be highly variable over time. For example, Cabral et al. (2020) found that in the first growth cycle, the system was a net sink of C, but the C-sink strength capacity decreased as the plantation aged. A second growth cycle sugarcane plantation in Brazil had a NECB close to zero and shifted to a source of C in the third growth cycle (Cabral et al. 2013). This high variability in NECB estimates from sugarcane implies the need for multiyear C-flux measurements coupled with measurements of processes that pinpoint underlying NECB dynamics.

Grasslands are a typical agricultural landscape in subtropical and tropical regions (Dixon et al. 2014; Ramankutty et al. 2008). How land conversion from grassland to sugarcane will affect the CO₂ cycle is highly uncertain (Jaiswal et al. 2017; Lisboa et al. 2011). Previous studies suggested that it takes several years for other high-yielding perennial crops to recover the C lost upon conversion and that over those initial years, R_eco tends to outweigh GPP due to C losses associated with land disturbance. Abraha et al. (2018) showed that a 22-year-old minimally managed perennial grassland lost substantial C during the first 8 years after conversion to switchgrass. In this study, a similar timeframe for C losses was observed when the minimally managed grassland was converted to a restored prairie harvested for bioenergy purposes. During that timeframe, both converted systems were consistently stronger net C sources than the minimally managed grassland. In this study, R_eco was larger than GPP in the systems for 2–3 years following conversion due to soil disturbance associated with land use change. This trend shifted gradually with time since conversion. In a different study, a grassland in the SE US established decades ago for hay production and converted to switchgrass was a net source of CO₂ to the atmosphere for 7 years following conversion (Ahlswede et al. 2022). Similarly, in a modeling exercise, it was after 13–15 years following conversion from minimally managed grassland to perennial bioenergy crops across the rainfed US region that miscanthus and switchgrass plantations returned to soil organic C-content values similar to preconverted grassland (Blanc-Betes et al. 2023). Overall, these studies suggest that land conversion from grassland to perennial bioenergy crops is associated with C losses that can span over a decade when accounting for C inputs (i.e., GPP) and outputs (i.e., R_eco and C lost by harvest).

Here, we investigated how the cultivation of sugarcane in subtropical Florida (FL) impacts CO₂ fluxes compared to grazed improved and semi-native pastures over three full growth cycles (i.e., > 3 years). Grazing lands are a typical landscape of the SE US, and particularly in FL improved and semi-native pastures make up for ca. 37% of agricultural land (Farmland Information Center 2022). We hypothesize that following conversion: (1) R_eco will be consistently greater than GPP in sugarcane due to typical C losses associated with land disturbance (e.g., Abraha et al. 2018), and hence, net ecosystem CO₂ exchange (NEE) will reflect that the system is a net source of CO₂ compared to pastures; and (2) when accounting for C losses derived from harvest and preharvest burn, the conversion of pasture to sugarcane will release C to the atmosphere, and sugarcane will consistently lose more C—or will store less C—than pastures. Furthermore, by examining the mechanisms driving responses to conversion, we identify sustainability traits and propose alternative management practices that could reduce environmental impacts from the expansion of the cane industry in subtropical and tropical regions.

Methods

Site Description

The study was conducted from 2019 to 2022 at Archbold Biological Station's Buck Island Ranch (BIR), a 4290-ha commercial cow–calf ranch that operates as an ecological field station (Lake Placid, Florida; 27°09′ N, 81°11′ W) and is a site in the USDA's Long-Term Agroecosystem Research (LTAR) network. The area has a typical, humid, subtropical climate with distinct seasons, namely a wet season (mid-May through mid-October) and a dry season (late-October through early-May). The area receives mean annual precipitation of ca. 1300 mm, with ca. two-thirds of total annual precipitation falling from June to September. The area has a mean high summer air temperature of ca. 33°C (July–August) and mean minimum winter temperature of ca. 11°C (December–January) for the 1980–2017 period (Thornton et al. 2017; DayMet database).

The experimental design consisted of three plots: two plots of 16 ha, one on semi-native (SN), and one on improved (IM) pasture were established in 1950; and, one 34-ha plot of IM pasture that was cleared for sugarcane establishment at the beginning of 2019. Semi-native and IM pastures are two main pasture types that dominate the agricultural landscape in the SE US (Guo et al. 2023; Paudel et al. 2023; Swain et al. 2013). Our SN pasture is dominated by native grasses (Schizachyrium stoloniferum) with patches of non-native bahiagrass (Paspalum notatum; Gomez-Casanovas, DeLucia, Bernacchi, et al. 2018), and our IM pasture is dominated by Argentine bahiagrass (Paspalum notatum Flugge; Chamberlain et al. 2017). Historically, both SN and IM pastures have been grazed by cattle (Bos taurus L.). Over the course of the study, the stocking rate for SN pasture was 0.17 AU ha⁻¹ (0.19 AU ha⁻¹ in 2019; 0.15 AU ha⁻¹ in 2020; 0.05 AU ha⁻¹ in 2021; and 0.3 AU ha⁻¹ in 2022); and for improved pastures, it was 0.67 AU ha⁻¹ (0.89 AU ha⁻¹ in 2019; 0.65 AU ha⁻¹ in 2020; 0.63 AU ha⁻¹ in 2021; and 0.52 AU ha⁻¹ in 2022). IM pastures were extensively ditched and historically fertilized (nitrogen, ca. 26 kg ha⁻¹ every 2 years) and limed (ca. 229 kg ha⁻¹ every 5 years; Swain et al. 2013). Fertilization and lime amendments were discontinued for the duration of this study. SN pastures did not contain drainage ditches and have never been fertilized. Pastures are managed following traditional management practices for the region, including prescribed fire (Boughton et al. 2022). Both IM and SN pastures were burned at the beginning of 2020 (on 2/12/2020 for SN pasture and 3/19/2020 for IM pasture) as it is typical for grazing lands in the area (Boughton et al. 2022). Soils under both pastures were Spodosol and have similar bulk density and soil organic matter. Root biomass was larger in IM than in SN pasture (Figure S1).

Our sugarcane plot was established in 2019 on an IM pasture (Table 1). Prior to land conversion, the plot was grazed by cattle (average grazing intensity of 1.1 AU ha⁻¹ year⁻¹), and the plot was regularly fertilized (nitrogen, ca. 26 kg ha⁻¹ every 2 years) and limed (ca. 229 kg ha⁻¹ every 5 years; Swain et al. 2013). Before land conversion to sugarcane, root biomass, soil organic matter, and soil bulk density were similar for IM pastures (Figure S1). Before planting cane, bahiagrass was terminated using herbicide (Table 1), and the field was tilled immediately after (Table 1). Sugarcane stalks were machine planted on 2/6/2019 with mean row spacing of 1.3 m. Sugarcane was fertilized with solid N-P-K fertilizer (ammonium nitrate for N; P₂O₅; Table 1). Weed management included the application of herbicides (Table 1). Preharvest practices included the burning of leaves and tops of the plant according to common practices in the region. In the Southern United States, 99% of sugarcane is burned (Baucum and Rice 2009). In our study, sugarcane was burned 1–2 days prior to harvesting, and harvest occurred on 2/16/2020, 2/8/2021, and 4/06/2022. Our sugarcane field contains drainage ditches, and therefore it was seepage irrigated. Vegetation was cut approximately 10 cm above the ground and transported to the mill. Soils under our sugarcane plot were Spodosol.

TABLE 1 Agronomic management for the sugarcane field.

Year	Tillage event date and type	Fertilizer event date, rate (kg/ha)	Herbicide event date, type, rate (kg/ha)	Lime event date, type, rate (Mg/ha)
N (kg/ha)	P (kg/ha)	K (kg/ha)
2018	12/25/2018, plowing				12/19/2018, Roundup, 7.9
12/26/2018, disking
2019	5/1/2019, Aerate	2/6/2019, 110	2/6/2019, 124	2/6/2019, 207	2/6/2019, Thimet 20-G, 21
	4/14/2019, 50	4/14/2019, 0	4/14/2019, 101	3/22/2019, Atrazine 4F, 2.7; Hydrate Plus, 0.01; Latigo, 1.04; Grounded, 0.2; Beseige, 0.7, Callisto, 0.2; Armezon 0.16	1/21/2019, Calcium silicate, 7.06
	5/11/2019, 130	5/11/2019, 0	5/11/2019, 25	4/11/2019, Hydrate Plus, 0.01; Beseige, 0.7	1/21/2019, Dolomite, 2.24
				4/25/2019, Asulam 4F, 8.3; Envoke, 0.02; Grounded, 0.21
2020	4/14/2020, Aerate	3/12/2020, 75	3/12/2020, 35	3/12/2020, 130	3/16/2020, Asulam 4F, 8.3; Hydrate Plus, 0.01; Beseige, 0.79; Callisto, 0.21; Armezon, 0.16
	4/23/2020, 131	4/23/2020, 25	4/23/2020, 87
2021	5/5/2021, Aerate	3/12/2021, 59	3/12/2021, 27	3/12/2021, 102
	3/22/2021, 93	3/22/2021, 43	3/22/2021, 160

In our study, we followed standard definitions in cane literature and defined plant cane (PC) as cane crop from planting until first harvest (here, from 2/6/2019 to 2/16/2020), first ratoon cane (FRC) as cane crop from first to second harvest (here, from 2/17/2020 to 2/8/2021), and second ratoon cane (SRC) as cane crop from second to third harvest (here, from 2/9/2021 to 4/6/2022).

Maximum tiller production in cane is generally achieved between 90 and 120 days after planting cane (Santos and Diola 2015). In our study, emergence and tillering of cane covered the period from planting—in PC—or after harvest—in FRC and SRC—until May and extended on average over 105 days (data not shown).

Agronomic and Biometric Measurements

For sugarcane, peak aboveground and litter biomass was collected annually when GPP reached a plateau (end of October to beginning of November) indicating biomass was no longer increasing in the system. Peak season was defined based on GPP instead of plant developmental stage to normalize procedures among years to avoid heterogeneity inherent to weather variability or stand age. Peak aboveground biomass as well as plant height, stalk density, and crown diameter were measured at 12 randomly selected locations within the sugarcane plot. Plant height, stalk density, and crown diameter were measured in 1-m-row sections within the randomly selected locations. Biomass was collected near ground level with 1 m² quadrats.

For sugarcane, harvested biomass (harvest; i.e., yield) was estimated by mechanized harvest. Total aboveground biomass measurements (i.e. standing dead/live and litter biomass) were also collected 1–2 days prior (i.e., AB_{pre-harvest&pre-burn}), and 1–2 days after harvest and burn (AB_{post-harvest&post-burn}) to estimate aboveground burned biomass (AB_burn). These samples were collected near ground level by hand at 12 randomly selected locations within the sugarcane plot with 1-m² quadrants, and AB_burn was estimated as follows:1$$$ {\mathrm{AB}}_{\mathrm{burn}}={\mathrm{AB}}_{\mathrm{pre}-\mathrm{harvest}\&\mathrm{pre}-\mathrm{burn}}\hbox{--} {\mathrm{AB}}_{\mathrm{post}-\mathrm{harvest}\&\mathrm{post}-\mathrm{burn}}\hbox{--} \mathrm{Harvest} $$$

For pastures, biomass consumed by cattle was estimated using the moveable-cage method (McNaughton 1983). Six 0.25 m² cages were distributed evenly throughout the pasture plots and were moved every other month during the wet and dry seasons. Before each cage was moved, paired 0.25 m² plots were clipped within and adjacent to the cage. Consumed aboveground biomass (i.e., standing dead/live and litter biomass) was estimated as the difference in plant biomass within (ungrazed) and outside (grazed) the cages. Consumed biomass values obtained in this study were within the range reported elsewhere (Gomez-Casanovas, DeLucia, Bernacchi, et al. 2018). Aboveground biomass was also collected prior (AB_pre-burn) and after (AB_post-burn) the burn event, and burned biomass (AB_burn) was estimated as follows:2$$$ {\mathrm{AB}}_{\mathrm{burn}}={\mathrm{AB}}_{\mathrm{pre}-\mathrm{burn}}\hbox{--} {\mathrm{AB}}_{\mathrm{post}-\mathrm{burn}} $$$

Prior to land use conversion (LU), we measured inputs from existing total grass biomass to sugarcane (TB_priorLU) as follows:3$$$ {\mathrm{TB}}_{\mathrm{priorLU}}={\mathrm{AB}}_{\mathrm{priorLU}}+{\mathrm{BB}}_{\mathrm{priorLU}} $$$where AB_priorLU is the standing aboveground and litter biomass and BB_priorLU is belowground biomass prior to sugarcane conversion. These measurements were taken in September 2018, and samples were collected from six randomly selected locations within the plot. Aboveground biomass was collected from 0.25 m² quadrants. A total of six soil cores (55 cm depth) were retrieved from each location, and roots were separated from soil and rinsed with deionized water.

All vegetation samples collected were dried at 60°C until constant mass and weighed. Subsamples were analyzed to determine C content using a flash combustion chromatographic separation elemental analyzer (Costech 4010 CHNSO Analyzer, Costech Analytical Technologies Inc. Valencia, California, USA). For sugarcane, row distance was also measured to account for planting separation.

Eddy Covariance Measurements

Eddy covariance towers were established at the center of each plot to measure fluxes from February 2019 to April 2022. This method provides high-frequency (10 Hz) continuous measurements of net CO₂, latent and sensible heat, and CH₄ fluxes between each land use and the atmosphere, aggregated in half-hour intervals. Each tower consisted of a 3D sonic anemometer (81,000 V; R. M. Young Company, Traverse City, Michigan, USA), an enclosed CO₂/H₂O infrared gas analyzer (LI-7200; LI-COR Biosciences, Lincoln, Nebraska, USA), and an open-path CH₄ analyzer (LI-7700; LI-COR Biosciences) operating at 10 Hz. The height of the EC system was adjusted over time to keep the sensors at optimal distance above the canopy and to minimize occasions when the flux footprint extended beyond the plot's edge. In this study, the minimum height was 2.5 m, and the instruments were always higher than 1.34 times the average plant height (Raupach 1994).

Auxiliary instrumentation consisted of temperature (Young 43,502 with 4347-L probe; R. M. Young Company) and relative humidity sensors (HMP155; Vaisala Oyj, Helsinki, Finland); a barometer (CS105; Campbell Scientific, Logan, Utah, USA); radiometer for up- and down-welling short- and longwave radiation (CNR4; Kipp & Zonen, Delft, The Netherlands); quantum sensors for up- and down-welling photosynthetic active radiation (PAR; LI-190; LI-COR Biosciences); soil heat flux plates (three per plot; H.FP1, Hukseflux Thermal Sensors, New York, The Netherlands); a tipping bucket rain gauge (TR-525 M, metric heated; Texas Electronics, Texas, Houston, USA); and soil moisture and temperature sensors (Stevens HydraProbe, US; and PR2 type profile probe, Delta T Devices, Cambridge, UK). Soil moisture was measured at depths of 10, 20, 50, and 70 cm, and soil temperature at 10 cm depth. Auxiliary measurements were logged to Campbell Scientific dataloggers (CR3000 and CR1000X) and compiled at 30 min intervals.

High-frequency data were processed with EddyPro 7 (LI-COR Biosciences, US) following (Gomez-Casanovas et al. 2020). In brief, a double-rotation scheme was used to align the coordinate system to the main wind direction, and crosswind correction of sonic temperature was implemented by the firmware (R. M. Young Company). Lagged covariances between vertical wind velocity and each flux scalar were computed and applied to account for lag times between the sonic anemometer and the flux sensors. Humidity and spectral corrections for high-pass and low-pass filtering were implemented. Spikes in raw flux data were eliminated and the quality of each half-hour average was determined following the 1–9 method (Foken et al. 2004) and low-quality data were eliminated (flags > 7). Fluxes of CO₂ were also discarded when they were < −70 or > 30 μmol m⁻² s⁻¹, and when the standard deviation of the CO₂ concentration exceeded the mean ± 3.5 standard deviation of a moving window of 200 records. Fluxes of CO₂ that corresponded to an area outside the edges of each plot were also eliminated using Hsieh cross-wind–integrated flux footprint model (Hsieh, Katul, and Chi 2000). Half-hour CO₂ fluxes were filtered by u* threshold. On average, 48% of half-hour CO₂ flux data were discarded (40% for sugarcane, 52% for IM pasture, and 51% for SN pasture). On average, 13% of half-hour CO₂ flux data were discarded as fluxes originated from an area outside the footprint.

Visualization and QA/QC of data were performed using PyFluxPro (Isaac et al. 2017). Missing half-hour net ecosystem CO₂ exchange (NEE) data were gap filled and fluxes were partitioned into ecosystem respiration (R_eco) and GPP using the eddy covariance gap-filling and flux partitioning online tool (Reichstein et al. 2005). Total uncertainty in fluxes was estimated by error propagation accounting for both random measurement error and gap-filling uncertainty (Richardson and Hollinger 2007).

Surface energy balance closure was calculated as in Zeri et al. (2011a) using the following equation:4$$$ \mathrm{H}+\mathrm{LE}+\mathrm{G}+{\mathrm{S}}_{\mathrm{ph}}=\mathrm{Rn} $$$where H is sensible heat flux, L is latent heat flux, G is soil heat flux, S_ph is the energy stored as photosynthesis, and Rn is net radiation. The S_ph was calculated using the conversion of 28 Wm⁻² for each 56.7 μmol m⁻² s⁻¹ (Zeri et al. 2011a). The slopes of the relationships varied among the years studied, and ranged from 0.58 to 0.78 in sugarcane, from 0.66 to 0.99 in IM pasture, and from 0.66 to 0.78 in SN pasture. The intercepts of the relationships varied among the years studied and ranged from −10.4 to +10.6 W m⁻² in sugarcane, ranged from 64.6 to 19.0 W m⁻² in IM pasture, and from −22.7 to −10.4 W m⁻² in SN pasture. The coefficients of determination (R²) of the relationships varied among the years studied, and ranged from 0.62 to 0.78 in sugarcane, from 0.63 to 0.91 in IM pasture, and from 0.78 to 0.93 in SN pasture. The slopes, intercepts, and R² obtained in this study were well within the range reported by most flux sites (Gomez-Casanovas, DeLucia, Bernacchi, et al. 2018; Wilson et al. 2002; Zeri et al. 2011b).

Ecosystem Estimates of Maximum GPP, Half Saturation Constant of Gross Photosynthesis, and Quantum Efficiency

To calculate maximum GPP (GPP_max) and half saturation constant of GPP (Ks), we used light response curves of GPP as done previously (Street et al. 2007). These parameters were estimated using data during the growing season (from May 15th to October 15th; wet season) for all land uses (Bao, Li, and Xie 2019; Xu et al. 2022). Parameters were estimated as follows:5$$$ \mathrm{GPP}=\frac{\mathrm{GPPmax}\times I}{K\mathrm{s}+I} $$$where GPP_max is the rate of light-saturated photosynthesis and I is incident radiation.

The E₀, which is the initial slope of the light response curve and reflects how efficiently an ecosystem uses incident radiation, was calculated as follows (Street et al. 2007):6$$$ E0=\frac{\mathrm{GPPmax}}{K\mathrm{s}} $$$

Estimates of Net Ecosystem Carbon Balance

The net ecosystem C balance (NECB), which represents the net storage or loss of C at the ecosystem scale (negative reflects C gain), was estimated as follows:7$$$ \mathrm{NECB}=\mathrm{NEE}\hbox{--} {\mathrm{C}}_{\mathrm{dung}}+{\mathrm{C}}_{\mathrm{harvest}}+{\mathrm{C}}_{\mathrm{burn}}\hbox{--} {\mathrm{C}}_{\mathrm{priorLU}} $$$where C_priorLU reflects the C input from existing biomass before land conversion to sugarcane, and thus, it is not applicable to our studied pastures; C_dung input reflects the C in dung deposited on the pastures; C_harvest reflects the C lost by either harvest or grazers; and C_burned reflects the C lost due to burn. In our study, C_dung was estimated assuming that grazing cattle do not digest 35% of their daily intake and a C content in dung of 56% (Bai et al. 2020; IPCC 2006).

We acknowledge that not including C-CH₄ losses derived from enteric ruminant fermentation and the underlying ecosystem could increase the uncertainty in our NECB estimates. However, this uncertainty is negligible given that previous studies in the area showed that the lost C-CH₄ is small (between 4.2 and 25 g C m⁻² year⁻¹ in pastures; and <1 g C m⁻² year⁻¹ in sugarcane) and that the contribution of CH₄ to the NECB of pastures and sugarcane is negligible (Chamberlain et al. 2017; Gomez-Casanovas, DeLucia, Bernacchi, et al. 2018; Gomez-Casanovas, DeLucia, Hudiburg, et al. 2018).

Overall uncertainty for NECB was calculated by propagating the sum of the uncertainty in NEE, biomass (harvest/grazed and burn biomass, and biomass before land conversion to sugarcane), and C input in dung deposited on pastures. Given that C_dung values were estimated according to Bai et al. (2020), we assumed a conservative uncertainty at 95% confidence interval (Chang et al. 2019; Gomez-Casanovas et al. 2016; IPCC 2006).

Statistics

Differences in biomass (peak aboveground biomass, consumed biomass by grazers, and harvest/burn), plant height, stalk density, and crown diameter were tested with complete block repeated-measures ANOVA with land use as fixed factor. Differences in soil temperature and moisture between land uses were tested with a Student t-test. Tests were run after transforming the data to ensure normality and homogeneity of variances. All statistical tests were conducted using R version 4.0.3. Because of the difficulty of replicating eddy covariance measurements and because this technique provides continuous measurements of fluxes integrated over large spatial scales, this study—as well as the majority of studies using this method—provides uncertainty in C fluxes to test differences in NEE and NECB between land uses as well as the uncertainty in photosynthetic parameters to test differences between land uses (Aubinet, Vesala, and Papale 2012; Baldocchi 2003, 2020; Bao, Li, and Xie 2019; Xu et al. 2022).

Results

Weather and Soil Conditions

Air temperature was similar between years, but large seasonal precipitation differences were observed between years (Figure 1). In 2019, precipitation was higher during sugarcane emergence and tillering stages than during 2020 and 2021 (from mid-February to May; 264 mm in 2019; 140 mm in 2020; and 217 mm in 2021; Figure 1b). Compared over the same time periods, air temperature was similar from February to May of 2019 (average of 23.1°C), 2020 (average of 23.2°C), and 2021 (average of 21.7°C; Figure 1a).

[IMAGE OMITTED. SEE PDF]

Soils were consistently wetter under pastures than sugarcane, and they were slightly wetter in semi-native (SN) than in improved (IM) pastures (Figure 2). Following trends in precipitation, soils under sugarcane were wetter in 2019 during sugarcane emergence and tillering stages than in 2020 and 2021 (from mid-February to May; 31% and 11% higher soil moisture during 2019 than in 2020 and 2021 for those time periods; Figure 2a–d). Similarly, soils were wetter in pastures in 2019 than in 2020 and 2021 for those time periods (Figure 2a–d; 29% wetter on average in IM pasture, and 36% wetter on average in SN pasture compared to 2019). Typical of this region, soil moisture was greater during the wet season than during the dry season.

[IMAGE OMITTED. SEE PDF]

Soils were slightly cooler under SN pasture than under sugarcane and IM pasture (0.7 and 0.4°C, respectively), and soil temperature varied among the years studied (Figure 2). Compared over the same time periods, soils under all land uses were slightly cooler in 2019 than in 2020, 2021, and 2022. Soils were slightly cooler under cane during emergence and tillering of cane in 2019 than during the same time period in 2020 and 2021 (Figure 2e–h).

Agronomic Measurements in Sugarcane

Yields, peak aboveground biomass, and plant height varied with land use. These parameters were greater in plant cane (PC; from 2/6/2019 to 2/16/2020) than in first ratoon cane (FRC; from 2/17/2020 to 2/8/2021), further decreasing in second ratoon cane (SRC; from 2/9/2021 to 4/6/2022; Table 2). Stalk density was higher in PC followed by SRC and it was the lowest in FRC (Table 2). Crown diameter was similar throughout the duration of the study regardless of stand age (Table 2).

TABLE 2 Aboveground peak biomass, harvest, plant height, stalk density, and crown diameter in plant cane (PC), first ratoon cane (FRC), and second ratoon cane (SRC).

	Aboveground peak biomass (ton DW ha−1)	Harvest (ton DW ha−1)	Plant height (m)	Stalk density (stalk m−2)	Crown diameter (cm)
PC	44.2 ± 1.2a	31.2 ± 0.9a	5.1 ± 0.4a	17.3 ± 1.0a	12.1 ± 2.9a
FRC	34.5 ± 4.2b	25.2 ± 3.1b	4.3 ± 0.1b	11.2 ± 1.2b	8.8 ± 1.5a
SRC	20.7 ± 2.1c	17.8 ± 2.5c	3.8 ± 0.3c	14.3 ± 1.3c	8.3 ± 0.5a

Net Ecosystem CO₂ Exchange, Gross Primary Productivity (GPP), and Ecosystem Respiration (R_eco)

While subject to substantial interannual variability, sugarcane overall was a stronger net CO₂ sink than pastures over the years following conversion (Figure 3a–c; Table 3). Immediately after conversion, PC was a net source of CO₂ while both pastures were sinks. However, both FRC and SRC were stronger net sinks of CO₂ than pastures, increasing net C uptake by 873 more g C m⁻² than IM pasture and 253 more g C m⁻² than SN pasture on average.

[IMAGE OMITTED. SEE PDF]

TABLE 3 Cumulative net ecosystem CO₂ exchange (NEE) for sugarcane (plant cane, PC; first ratoon cane, FRC; second ratoon cane, SRC), improved (IM) and semi-native (SN) pastures.

Time period (from–to)	Land use	Cumulative NEE (g C-CO2/m2)
2/6/2019–2/16/2020	PC	+51 ± 14
IMP	−301 ± 18
SN	−254 ± 12
2/17/2020–2/8/2021	FRC	−673 ± 23
IMP	−165 ± 20
SN	−319 ± 12
2/9/2021–4/6/2022	SRC	−319 ± 16
IMP	+46 ± 15
SN	−168 ± 6

Gross primary productivity was in general higher in sugarcane than in pastures (Figure 3; Table 3). Ecosystem respiration consistently was higher in sugarcane than in SN pasture, and it was only larger in PC compared to IM pasture (Figure 3; Table 3). In sugarcane, differences in GPP and R_eco varied greatly over the years measured. Compared over the same time periods, GPP was greater in PC than in FRC and SRC, and it was slightly greater in SRC than in FRC. R_eco was substantially larger in PC compared to FRC and SRC. Both GPP and R_eco were larger in IM than in SN pastures.

Photosynthetic Parameters

Photosynthetic parameters including maximum GPP (GPP_max), half-saturation constant (K_s), and quantum efficiency (E₀) were in general greater in sugarcane than in pastures with the exception of FRC displaying similar GPP_max, K_s, and E₀ than IM pasture at equivalent time period (i.e., from 2/6/2019 to 2/16/2020; Figure 4; Table S1). Averaged over the years studied, E₀ was 0.12 ± 0.009 for sugarcane, 0.1 ± 0.004 for IM pasture, and 0.07 ± 0.004 for SN pasture (Table S1). In sugarcane, differences in photosynthetic parameters varied greatly with time since conversion. Specifically, GPP_max, K_s, and E₀ were greater in PC than in FRC and SRC, and they were greater in SRC than in FRC (Figure 4; Table S1).

[IMAGE OMITTED. SEE PDF]

C Removals and Net Ecosystem C Balance (NECB)

Biomass removal due to harvest and burn was higher in sugarcane than in pastures, and the amount of C burned from the sugarcane field varied over years (Figure 5). In sugarcane, C removal with burn accounted for 23%–32% of yields. Burning removed 198 g C m⁻² from PC, 169 g C m⁻² from FRC, and 152 g C m⁻² from SCR (Figure 5). The removal of C due to grazers was larger in IM pasture than in SN pasture, and it was similar among years within pasture type (Figure 5).

[IMAGE OMITTED. SEE PDF]

Overall, sugarcane was a stronger net source of C than pastures, but C storage varied greatly with time since conversion (Figure 5). Following conversion, PC was a net source of C and both pastures were C sinks (i.e., from 2/6/2019 to 2/16/2020). FRC was a slightly stronger net source of C than IM pasture, whereas SN pasture was a net C sink in the same time period (i.e., from 2/17/2020 to 2/8/2021). The net C balance of SRC and IM pasture was similar and they were both net C sources, whereas SN pasture was a net sink of C in the same time period (i.e., from 2/9/2021 to 4/6/2022).

Discussion

Immediately following conversion from pastures, plant cane (PC) was a net annual CO₂ source (i.e., positive net ecosystem CO₂ exchange or NEE; Figure 3a–c). Contrary to our hypothesis, sugarcane switched to a net annual CO₂ sink after first regrowth (i.e., first and second ratoon canes, FRC and SRC, respectively), overall becoming a stronger net CO₂ sink than pastures 1 year after land use conversion (Figure 3a–c). Accounting for all ecosystem carbon (C) gains and losses (i.e., NECB), sugarcane was a sustained net C source (Figure 5). Biomass removals through grazing, harvest, and fire reduced net C gains of both canes and pastures substantially (Figures 3 and 5). However, greater frequency of burn events and repeated harvest increased C removals and made sugarcane an overall stronger C source relative to pastures despite substantial C inputs from the previous land use and a favorable NEE (Figure 5). Time since conversion reduced net C losses from cane cultivation (Figure 3). Contrary to our hypothesis, the NECB after second regrowth (i.e., SRC) was similar to that of the improved (IM) pasture suggesting rapid shifts in the net C balance of sugarcane after initial C losses associated with land conversion (Figures 3 and 5; Table 3). Moreover, the rapid shift in the NECB of sugarcane toward that of IM pasture along with enhanced C storage in semi-native (SN) pasture than in IM pasture (Figure 5) indicates that the impacts of sugarcane expansion on C storage capacity will depend on the proportion of IM versus SN pastures converted to this bioenergy crop.

Impact of Land Use on NECB

Our results indicate that the conversion of pastures to sugarcane turned the system into a sustained net C source (i.e., positive NECB). Previous research reported similar C losses during up to 5 years after conversion of minimally disturbed grasslands to high-yielding bioenergy perennials (e.g., Panicum virgatum; Abraha et al. 2018). These results imply that strategies targeting reductions in C losses or C offsets will, therefore, contribute to the incorporation of sugarcane into a sustainable bioeconomy.

Impacts of changes in land use on NECB can be traced back to contributions from the major components that govern the C balance of ecosystems (Chapin et al. 2006). In our study, these major components were NEE, and C removals through harvest and fire, which we discuss in detail below.

Impact of Land Use on NEE, GPP, and R_eco

Sugarcane experienced substantial net CO₂ losses upon conversion from pasture, with ecosystem respiration (R_eco) largely offsetting gross primary productivity (GPP). This brought the system to a net CO₂ source (i.e., positive NEE) over the planting year, while both pastures were net sinks of CO₂. Significant CO₂ losses following the conversion of grasslands to cropland have been widely reported in the literature (Abraha, Gelfand, et al. 2018; Blanc-Betes et al. 2023; Gelfand et al. 2011) and are explained by a combination of factors. First, large inputs of relatively labile plant material from the previous land use, incorporated into the soil by herbicide application and tillage common in preplanting conditioning practices, are likely to accelerate SOC mineralization rates (Moore et al. 2020). Gelfand et al. (2011) estimated that 33% of total above- and belowground biomass from previous land use is decomposed in the first year following conversion. Second, soil conditioning and tillage disturb soil structure and destabilize soil aggregates. These practices increase the vulnerability of physically protected soil organic carbon, making this C susceptible to decomposition (Haddaway et al. 2017; Reicosky, Dugas, and Torbert 1997; Six, Elliott, and Paustian 1999), especially when the converted land use stores vast amounts of C (Conant, Paustian, and Elliott 2001; Gomez-Casanovas et al. 2021) as is the case of relatively undisturbed soils under both pastures (Adewopo et al. 2014). Cane cultivation, which includes pre-establishment tillage, as well as annual shallow tillage, events to aerate the soil likely contributed to large R_eco following land disturbance relative to pastures. Third, GPP was larger over the planting year than afterward and increases in autotrophic respiration, which largely contributes to R_eco, could be driven by enhanced GPP (Gomez-Casanovas et al. 2012).

While we observed initial CO₂ losses following conversion, contrary to our hypothesis of sustained CO₂ losses, sugarcane became a stronger net CO₂ sink (i.e., negative NEE) after first regrowth (i.e., FRC and SRC) compared to pastures. This was likely a consequence of decreases in R_eco, implying that initial net CO₂ losses in sugarcane derived primarily from disturbance rather than physiological limitations on GPP. Accordingly, sugarcane exhibited overall higher potential gross CO₂ uptake (i.e., GPPmax; Figure 4) than pastures translated into greater aboveground biomass of sugarcane than pastures (Table 1).

Semi-native pasture was an overall stronger net sink of CO₂ than IM pasture because in relative terms R_eco was larger than GPP in IM than in SN pasture (Figure 3). This supports the view that while pasture intensification enhances C inputs—as shown by greater GPP—soil microbial decomposition was likely higher in IM than in SN pasture as previously shown in these systems (Adewopo et al. 2015).

Impact of Stand Age and Early-Season Water Availability on Potential CO₂ Uptake of Sugarcane

Plant biomass, height, and GPP_max of sugarcane decreased with sugarcane maturity, suggesting that stand age was a dominant factor driving potential C fluxes (Figure 4; Table 2). Juvenile plants tend to have higher plant productivity and GPP than more mature plants and an age-related decline in photosynthetic capacity has been observed in trees as well as perennial grass plantations (Drake et al. 2011; Kantola et al. 2022; Tejera et al. 2022). However, while potential gross CO₂ uptake in sugarcane decreased with stand age, this metric was higher in SRC than in FRC, indicating that stand age alone did not explain observed trends.

Low soil water availability during emergence and early developmental stages of sugarcane has often been reported to decrease stalk density (Inman-Bamber, Lakshmanan, and Park 2012; Misra et al. 2020; Silva et al. 2008). In our study, we also observed that low early-season soil water availability—which was lower in FRC than in SRC and lower in SRC than in PC—was in close correspondence with low stalk density (Table 2; Figure 2; Figure S2). The lower stalk density likely resulted in a more sparse canopy in FRC compared to SRC given the strong positive relationship between stalk—or stand—density and canopy structure (Sprintsin et al. 2009). Lower stalk density due to early low water availability could result in less dense canopies potentially affecting C cycling (Baldocchi, Wilson, and Gu 2002).

The tight control of the environment on canopy structure and the strong regulation of potential gross CO₂ uptake have been observed in other systems (Migliavacca et al. 2017; Shaver et al. 2007; Street et al. 2007). For instance, Street et al. (2007) showed that in an ecosystem dominated by sedges and shrubs, changes in canopy structure explained seasonal variations in GPP. Our results suggest that changes in canopy structure in sugarcane primarily mediated by stalk density and shaped by early-season water availability regulated the potential gross CO₂ uptake of sugarcane (Figure 4; Figure S2). These results suggest that plant breeding strategies targeting improved tolerance to water stress in sugarcane could be important to enhance GPP, and they might be crucial for enhancing C sequestration in soils, particularly if carbon use efficiency increases and enhanced GPP does not stimulate soil microbial respiration.

Impact of Land Use on C Removals Including Harvest and Fire

In addition to NEE, C removals through harvest and prescribed burns can have a major impact on the NECB of terrestrial ecosystems. The greater removal of C from grazing in IM than in SN pasture along with an overall weaker net CO₂ sink strength of IM versus SN pastures explained that SN pasture has greater C storage potential (Figures 3 and 5). This is likely explained by a combination of driving factors including higher biodiversity and lower grazing intensity in SN than in IM pasture. In this context, higher biodiversity and lower grazing intensity have been previously linked to higher C storage (McSherry and Ritchie 2013; Weiskopf et al. 2024).

Proportionally, most of the NECB in cane was allocated to biomass yield and harvested for energy production, and burn was the second C removal pathway contributing to bringing the cane system to a net C source. Preharvest burn is the primary management practice in the United States, where 100% of sugarcane is burned prior to harvest to enhance the economic profit of the productive process (Baucum and Rice 2009). In this study, preharvest burns removed 2 t C ha⁻¹ from PC, 1.7 t C ha⁻¹ from FRC, and 1.5 t C ha⁻¹ from SCR (23%–32% of yields; Figure 5), suggesting that eliminating burning in the second- and third-growth cycles may bring sugarcane to either neutral (i.e., FRC) or to a slight net source of C (i.e., SRC). Therefore, a no-burn harvest (a.k.a. green harvest) management could substantially reduce C losses observed upon and following conversion, particularly when plant residue, that otherwise would be burnt, is left in the field.

Conclusions

In summary, we showed rapid shifts in the net ecosystem C balance of sugarcane after initial C losses typical of land conversion disturbance, and that the regional C balance implications following conversion across the SE US will depend on the proportion of improved versus semi-native pastures converted to sugarcane. We also showed that both sugarcane maturity and low early-season water availability can decrease potential gross C uptake of sugarcane, which could have an impact on C sequestration. Overall, sugarcane was a net source of C to the atmosphere and C lost through fire substantially contributed to C losses. This suggests that no-burn management strategies could substantially alleviate C losses following conversion. The SE US has the potential to help achieve US energy targets, and bioenergy from sugarcane production is considered a key mitigation strategy for global warming. Future empirical and in silico experiments should focus on management and optimized breeding strategies to enhance the environmental sustainability of sugarcane.

Author Contributions

Nuria Gomez-Casanovas: conceptualization, formal analysis, funding acquisition, investigation, writing – original draft, writing – review and editing. Elena Blanc-Betes: conceptualization, formal analysis, funding acquisition, investigation, writing – original draft, writing – review and editing. Carl J. Bernacchi: conceptualization, funding acquisition, investigation, writing – review and editing. Elizabeth H. Boughton: funding acquisition, investigation, writing – review and editing. Wendy Yang: conceptualization, funding acquisition, writing – review and editing. Caitlin Moore: formal analysis, investigation, writing – review and editing. Taylor L. Pederson: formal analysis, writing – review and editing. Amartya Saha: formal analysis, writing – review and editing. Evan H. DeLucia: conceptualization, funding acquisition, investigation, writing – review and editing.

Acknowledgments

This work was funded by the DOE Center for Advanced Bioenergy and Bioproducts Innovation (US Department of Energy, Office of Science, Office of Biological and Environmental Research under Award Number DE-SC0018420), the US Department of Agriculture NIFA (Projects No. 2016-67019-24988 and 2021-67019-33431), and Arizona State University (AZ, USA; Award No. ASU092762). The authors would like to acknowledge the funding support from the Global Change and Photosynthesis Research Unit of the USDA Agricultural Research Service, Urbana, IL. Any opinions, findings, conclusions, or recommendations expressed in this publication are those of the author(s) and do not necessarily reflect the views of the US Department of Energy. Mention of trade names or commercial products in this publication is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the USDA. USDA is an equal opportunity provider and employer. Archbold's Buck Island Ranch is a site within the Long-Term Agroecosystem Research (LTAR) network and therefore this research is a contribution from LTAR. LTAR is supported by the United States Department of Agriculture. The authors would also like to thank research staff at the University of Illinois, at Archbold's Buck Island Ranch, and at Texas A&M that assisted with this research including Katie Bowman, Evan Dracup, Dr. Nicholas DeLucia, Tess Rutstein, Steffan Pierre, Alan Rivero, Hannah Van Zant, Nate Spicer, Carly Tolle, and Rohit Fenn, and many other undergraduate students, research assistants, and interns. We acknowledge Gene Lollis, Laurent Lollis, and Mary Margaret Hardee for leading sugarcane production. We would also like to thank Dr. Sneha Bandyopadhay for her comments on a draft of this manuscript.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

The data that support the findings of this study are openly available in Figshare at

References

Abraha, M., I. Gelfand, S. K. Hamilton, J. Chen, and G. P. Robertson. 2018a. “Legacy Effects of Land Use on Soil Nitrous Oxide Emissions in Annual Crop and Perennial Grassland Ecosystems.” Ecological Applications 28, no. 5: 1362–1369. [DOI: https://dx.doi.org/10.1002/eap.1745].