Site description and experimental design
The field experiment was conducted in central Maui, Hawaii (20.89°N, 156.41°W) on Hawaiian Commercial and Sugar Company (HC&S) land. At the time of the study, HC&S was the only remaining sugarcane plantation in Hawaii. The experimental plots were within a highly weathered, very-fine, kaolinitic, isohyperthermic Typic Eutrotorrox of the Molokai series in field #609, which is approximately 100 meters above sea level and has a total commercial area of 72 ha. The soil is well drained, rocky, and has deep, well-defined horizons below the plow layer . Soil pH was 7.97, and C concentration was 1.37% on average in the top 40 cm with a mean bulk density of 1.51 g cm3 as assessed by the baseline soil collection in 2011. During the trial period, average annual air temperature was 23.4 °C and annual precipitation was 241 mm, which are consistent with long-term averages for the area .
The experiment was a strip-plot, group-balanced design with two factors, irrigation and species with three replicates (blocks) (please see  for additional details). Irrigation was applied at the standard plantation rate (100%), and two deficit irrigation rates (75% and 50% of plantation standard). The original trial included four species, sugarcane (Saccharum officinarum), energycane (Saccharum officinarum x Saccharum spontaneum), napiergrass (Pennisetum purpureum Schumach.), and sweet sorghum (Sorghum bicolor (L.) Moench). For this study, two crops (sugarcane and napiergrass) were evaluated at two irrigation levels (50% and 100%). On June 26, 2011 the field plots were established in a recently harvested sugarcane field that had been in a cane-on-cane rotation for over 100 years. The sugarcane plots were planted with seed cane from an adjacent field within the HC&S plantation and napiergrass seed crop was supplied from a harvested population at the University of Hawaii’s research station in Waimanalo, Oahu.
Deficit irrigation treatments were applied to the field from November 13, 2011 depending on water availability across the plantation and were controlled with automated timers. From November 2011–October 2012, 1245 mm water ha−1 were applied to the 100% plots and 633 mm water ha−1 were applied to the 50% plots, for an actual deficit treatment of 50.8%. During the study period, the napiergrass plots were harvested every 6 months (a time interval that maximizes yield) on March 13, 2012 and September 25, 2012. The sugarcane crop was under a 2-year growth rotation and did not reach maturity during the scope of this study period and was not harvested.
Developing a baseline system for sugarcane
Sugarcane is a high-yielding, tropical C4 perennial grass of South Pacific origin. Tropical sugarcane biomass yields are up to 40 Mg dry wt ha−1 year−1 in Hawaii and 26 Mg dry wt ha−1 year−1 in Brazil [8, 17]. The species supports a drought resistant, robust root system that can improve soil structure and accumulate C on marginal lands [18, 19]. However, Hawaiian sugarcane has been grown on a 2-year crop cycle, reaching maturity after 24 months and then harvested after a low intensity burn followed by deep tillage and mechanized planting. Commercial sugarcane production existed on Maui for over 125 years. However, in January 2016 HC&S, the last remaining large-scale sugar producer, announced a wholesale transition on their 14,000 ha plantation to diversified agriculture including perennial grasses for forage, pasture, and bioenergy feedstock.
The development of a C budget or GHG analysis includes a detailed accounting of the agricultural inputs required to produce a specific crop. This includes quantifying the fossil fuel and non-fossil fuel based emissions. Fossil emissions are considered emissions resulting from fuel use during field preparation, planting, application of agrochemicals, harvesting, and maintenance [3, 5, 11, 12, 20]. Non-fossil emissions are considered biogenic GHG emissions that consider the production of carbon dioxide (CO2), methane (CH4), and nitrous oxide (N2O) as a result of the production system and are primarily a result of pre-harvest burn operations, soil GHG exchange, and residue management . This baseline system will be referred to as the sugarcane 100% scenario for future comparisons.
Fossil emissions from fuel consumption and agricultural inputs
Field operations (i.e., field preparation, harvest, fabrication and maintenance, seed propagation, and irrigation) often are considered to be hidden sources of emissions because of their indirect contribution to GHG flux. These emissions are caused by the burning of fossil fuel during equipment operation  and accounting can be challenging partly because of large variation in the descriptive energy units . A standard unit of kg Ceq was used to assess the contribution of field operations to the total C budget. Emissions factors (EF) were used to convert the fuel use requirements of each operation to Ceq. The EFs used in this analysis represent a synthesis of the best-available and current values found in the literature. Specific information pertaining to type of equipment and usage has been obtained from personal communication with plantation staff.
When EFs found in the literature were inadequate to describe the operation on Maui, they were generated independently to best reflect current plantation practices. For example, irrigation emissions were calculated based on the average energy required to pump water across the plantation at a rate specific to Maui, 0.0057 L/kWh (personal communication with L. Jakeway, Chemical Engineer at HC&S, 2013). Additionally, this value was adjusted to account for the total amount of water applied to a field in the 2011–2012 year to correctly reflect the amount of renewable energy used on the plantation for these operations (approximately 3:1 renewable to fossil energy ratio). An additional input to the baseline scenario was a calculation of fossil emissions related to seed cane production, which were adjusted for the weight of seed cane used in planting operations .
Emissions from agricultural inputs (i.e., fertilizer, herbicide, and lime) are a result of the energy required to produce, transport, and distribute these items [11, 20, 22]. A pre-emergence herbicide mix containing atrazine (1-chloro-3-ethylamino-5-isopropylamino-2, 4, 6-triazine), 2, 4-D (2, 4-dichlorophenoxyacetic acid), Prowl ((N-1- ethylpropyl)-3, 4-dimethyl-2, 6 dinitrobenzenamine), Rifle (3, 6-dichloro-2-methoxybenzoic acid), and Velpar (3-cyclohexyl- 6-dimethylamino-1-methyl-1, 3, 5-triazine-2, 4(1H,3H)-dione) was applied once three weeks after planting. Each plot received 345 kg N ha−1 (as liquid urea: 46-0-0) applied through the drip irrigation system. The fertilizer was applied monthly once the crops were established and concluded after 10 months. The timing and rate of urea application were optimized for the 2-year sugarcane crop and were based on current HC&S plantation practices. Deficit irrigation treatments were postponed during all fertilizer application events.
Quantification of the C emissions resulting from agricultural inputs in Maui sugarcane production was based on application rate and converted to Ceq with reported EF [11, 20]. An EF of 0.97 was used to convert the fertilizer application rate to Ceq. Lime (CaCO3) was applied at a rate of 2569 kg ha−1 prior to field planting and converted to Ceq using an EF of 0.12. Individual emission factors were identified for each chemical used in the herbicide mix reported by HC&S. These factors were averaged and a new EF (5.64) was developed for herbicide application specific to Maui.
Emissions from litter decomposition is a function of residue management. For example, N2O emissions can increase due to decomposition of leaf material following harvest and are greater in intact compared to burned fields . The goal on Maui sugarcane fields has been to maintain 15% of total field biomass for crop residue. Using this percentage, litter Ceq were calculated based on an average biomass production of 80.4 Mg ha−1 year−1 reported by HC&S. Emission factors for litter decomposition are based on the amount of N in crop residues following harvest [3, 11, 12].
Pre-harvest burn emissions
Conventional cultivation of sugarcane in Hawaii included a pre-harvest, low intensity burn to remove unwanted leafy material prior to harvesting. Pre-harvest burning significantly increases GHG emissions through the production of CH4 and N2O and the release of black carbon (BC) to the environment [3, 12]. The values used for the cropping scenarios in Maui were based on a 15% residue retention rate of total field biomass. Current IPCC values outlined in the Guidelines for National Greenhouse Gas Inventories assessment in 2006 suggest an EF of 0.07 kg N2O per ton of dry matter burnt and an EF of 2.7 kg CH4 per ton of dry matter burnt. Black C has a GWP that is 500 times greater than CO2 on a 100-year time horizon and it is estimated that 1 kg of BC is created for every kg of trash burnt [12, 23].
Soil N2O and CH4 emissions
Field measurement of soil fluxes in CH4, and N2O began in October of 2011 and were sampled at least monthly until October of 2012 following the GRACEnet sampling protocols , as previously reported in detail (Pawlowski et al. in review). Mid-morning measurements were collected from sealed static PVC chambers affixed to permanent collars installed in the sugarcane and napiergrass rows and inter-rows. Samples were collected by sealing each chamber and using a 10 mL polypropylene syringe and extracting 8 mL of headspace air through a septum on the styrene lid at 0, 15, 30, 45, and 60 min after chamber closure. Each gas sample was immediately injected into an evacuated Exetainer® (Labco Limited, UK) fitted with a Doubled Wadded Teflon/Silicon septa (Labco Limited, UK) for short-term storage. Samples were analyzed using a Shimadzu GC-2014 Gas Chromatograph (Shimadzu Scientific Instruments, Inc.), which used a flame ionization detector to measure the concentration of CH4 and CO2 after methanization, and an electrical conductivity detector for N2O analysis. Flux rates were calculated by assuming a linear change in gas concentration over time [25, 26]. Cumulative annual emissions of N2O and CH4 were interpolated from daily fluxes and summer over the first year and reported in terms of kg CH4 ha−1 year−1 and kg N2O ha−1 year−1 [2, 13]. To discuss N2O and CH4 emissions in terms of a C balance, annual rates were converted into CO2eq using the IPCC 100-year horizon factors for calculating GWP. Therefore, when CO2 = 1 on a 100 year−1 time scale, then the GWP for N2O and CH4 are 298 and 25 respectively [26,27,28]. For the purposes of making direct comparisons with soil C storage on these plots, the GWP values were converted to Ceq relative to C in CO2.
Soil C quantification
Ten baseline soil cores were collected in June 2011 using 20-cm depth increments up to a vertical depth of 2.4 m. Cores were extracted using a standard wet core diamond tipped drill bit with an internal diameter of 7 cm (Diamond Products Core Borer, Elyria, Ohio, USA). Each core barrel was inserted into the soil by a rotating hydraulic drill to minimize compaction within the barrel and to ensure accurate depth measurements. Soil samples were frozen at field moisture conditions until laboratory analysis. The cores were sieved at < 2 mm and dried for 48 h at 105 °C. Subsamples were ground to pass through a 250 micron sieve for heterogeneity, weighed, and analyzed for C concentration by combustion using a Costech ECS 4010 CNH Analyzer (Costech Analytical Technologies, Inc., Valencia, CA, USA). Soil C stock was determined with the equivalent soil mass method in increments of 3600 Mg ha−1 and a mean value for the baseline cores was determined at the 7200 and 18,000 Mg ha−1 reference masses, which represent the surface layer (25–40 cm) and deep profile (1–1.4 m) of soil .
Three soil cores (65-mm inner diameter bucket auger) were collected annually to a depth of 120 cm in 20 cm increments from each of the experimental plots. Samples were processed as noted above for the baseline cores for C concentration and soil C stock determination at the 18,000 Mg ha−1 reference mass. The change (∆) in soil C stock was made for each plot by subtracting the mean baseline value from the mean of the three cores for each plot and was previously reported . The mean ∆ values for experimental years 1 and 2 were reported here as an annualized ∆.
With local and national mandates for cellulosic energy sources increasing, there is opportunity to consider alternative-cropping scenarios to sugar production that may increase financial security and ameliorate environmental impacts caused by conventional management practices. One potential candidate species is napiergrass, a warm-season perennial C4 grass of African origin that produces high biomass yields under tropical conditions. In recent studies, napiergrass produced more than 45 Mg dry wt ha−1 year−1 in Florida  and between 40 and 53 Mg dry wt ha−1 year−1 in Hawaii [31, 32]. In contrast to sugarcane production in Hawaii, napiergrass is a ratoon-harvested crop that maintains high yields with zero-tillage management and no pre-harvest burning. Regional water shortages from drought and limitations to water access from litigation impose uncertainty on Maui’s agricultural communities. As a result, experimental scenarios include deficit irrigation to better anticipate changes in crop yields and GWP brought on by limited water. For this experiment, the napiergrass plots received the same amount of fertilizer as the sugarcane plots.
For the purpose of creating a C budget for these alternative-cropping scenarios the following assumptions were made: (1) napiergrass fossil-based emissions during the field preparation, harvest, and maintenance operations were considered equal to baseline sugarcane operations due to the similarities in production requirements between the species, (2) seed emissions for napiergrass were calculated based on actual planting rate (seed pieces per plot) relative to the sugarcane seed rate (6500 kg ha−1), because a total weight of napiergrass seed was not known, (3) both the lime and herbicide application rate were considered equal to the baseline sugarcane system, regardless of the species or irrigation treatment, (4) deficit irrigation values were adjusted to the known amount of water applied to each of the 100% and 50% treatments, (5) total biomass for each scenario was adjusted based on a 44% reduction in yield found under the deficit irrigation treatments measured in the field at the time, (6) litter decomposition rates and emissions were reduced for the deficit treatments based on residue management and napiergrass decomposition was calculated based on unburned residue N content after harvest, (7) emissions resulting from pre-harvest burn operations were adjusted for the sugarcane 50% scenario by accounting for the reduction in yield under the deficit irrigation, and (8) no SOC accumulates under the baseline scenario due to the incorporation of the pre-burn harvest followed by the associated soil disturbance of conventional tillage operations after harvest.
Calculating GHG intensity
Greenhouse gas index (GHGI) relates net GWP to crop production by dividing net GWP values by total crop biomass . Net GWP was calculated based on the net C balance for each scenario (described above) and reported in kg Ceq ha−1 year−1. In the first 2 years of the trial, yields at 100% irrigation were 73.9 Mg ha−1 year−1 for sugarcane and 47.7 Mg ha−1 year−1 for napiergrass . These values were similar to 2011 sugarcane crop yields data reported from HC&S as 80.4 Mg ha−1 year−1. Similarly, best-case Hawaiian napiergrass biomass averages reported from Kinoshita et al.  and Kinoshita and Zhou  were approximately 45 Mg ha−1 year−1. The measured reductions in yields under the deficit irrigation treatment during the first 2 years of the field trial were 60% for sugarcane (29.4 Mg ha−1 year−1) and 31% (32.9 Mg ha−1 year−1) for napiergrass . This metric is useful for comparing scenarios with different yield potentials relative to their impact on GWP and C balance: more positive GHGI values are a stronger source of atmospheric GHGs per unit crop yield, whereas, more negative values are a stronger sink.