Red Palm Oil versus Conventionally Refined Palm Oil: A Cradle-to-Gate Carbon Footprint on Mass, Carotene and Vitamin A Functional Units

2.1 Goal, Scope and Functional Units

The study is a cradle-to-gate product carbon footprint under ISO 14067 [42], following the inventory requirements of ISO 14044 [32]. It covers a single impact category, climate change, expressed as global warming potential over 100 years (GWP100), and it ends at the refinery gate. Distribution, use and end-of-life are excluded, so carotene losses during domestic heating [43] are not modeled. Throughout, RPO denotes the red palm oil produced by the pilot route, the oil that the source study calls neutralized deodorized palm oil (NDPO) [16], and NRPO denotes the neutralized intermediate.

Three functional units (FU) are carried through the same inventory. FU-1 is 1 kg of oil at the refinery gate, the basis of published palm oil LCA studies. FU-2 is 1 mg of total carotene delivered in the oil at the gate, the attribute for which RPO is produced. FU-3 is 1 mg of retinol activity equivalent (RAE) delivered at the gate, which places carotene and preformed vitamin A on one scale and admits fortified RBDPO as a comparator (Section 2.7).

RPO is compared with RBDPO made from the same CPO, and the comparison is indicative only. The comparator is represented by a published refinery-stage burden that embeds a substitution credit, whereas the RPO route is allocated (Section 2.5). Because the two systems do not use equivalent co-product rules, the study makes no comparative assertion in the sense of ISO 14044.

2.2 System Boundary and Data

The system runs from cultivation to oil at the refinery gate (Fig. 1) and is divided into a background and a foreground. The background covers cultivation, fruit transport and milling, ending with CPO at the mill gate. It is taken from published palm oil inventories (Section 2.6) and is common to both oils. CPO transport from the mill to the refinery is added to both, at 120 km, a Malaysian distance, by articulated truck [27, 44] and 0.095 kg CO₂-eq t−1 km−1 [45], and is varied over 50–300 km and 0.0775–0.22 kg CO₂-eq t−1 km−1 [44, 1].

The foreground comprises degumming, caustic deacidification and moderate-temperature deodorization. It is built here from two pilot-plant studies on Indonesian CPO [15, 16], and these studies are the reason the work is set in Indonesia: they are the only published process data for the RPO route on Indonesian CPO, and the Indonesian background is available at district resolution [3]. Neither source reports energy use, so all foreground energy is estimated.

Every inventory value is labeled measured (reported in the source), estimated (derived here by a stated calculation) or assumed (not reported, with a range). The inventory is summarized in Table 1; derivations are given in Supporting Information Sections S1–S3.

2.3 Foreground Inventory

2.4 Impact Assessment

Impacts are characterized as GWP100 with IPCC AR6 factors [53], using the non-fossil methane factor for combustion methane. Electricity is characterized with the Indonesian grid average of 0.681 kg CO₂-eq kWh−1 for 2024, a life cycle value including upstream fuel supply [54], varied in the sensitivity analysis of the deodorization energy model over 0.54–0.77, a bottom-up estimate for the 2024 Sumatra generation mix that uses direct combustion factors and omits upstream fuel supply [55]. Official Indonesian grid factors are CO2-only operating, build and combined margins intended for crediting mitigation projects [56]; they include no CH4 or N2O and are not average-mix factors, so they are not used here (Section S4).

Land use change (LUC) enters the background in its direct form, amortized over 20 years following PAS 2050 [57], the PalmGHG calculator [58] and the attributional background source [26]. Indirect LUC is excluded as a consequential construct [59]. Biogenic CO2 exchange in the crop cycle is also excluded, whereas methane from mill effluent is included.

2.5 Allocation

Multifunctionality is common in agricultural systems, and the way it is handled can change the results of food LCA [60, 61]. In this system the choice arises at deacidification, which yields NRPO together with 155.9 kg t−1 CPO of wet soapstock, a saleable stream. Allocation at this stage partitions both the upstream burden carried by the CPO and the burden of the stage itself, so three treatments are evaluated. The first treats soapstock as waste and assigns the whole burden to NRPO (factor 1.0). The second allocates by CPO-derived mass, 901.6 : 98.4 (0.9016), excluding the water and sodium added at the stage. The third allocates by the mass of wet soapstock (0.8525). The first two define the FU-1 result range, and the third bounds the sensitivity analysis. Economic allocation over a soapstock-to-oil price ratio of 0.15–0.60 falls within the same range. Acid splitting of soapstock is performed by its buyer and lies outside the boundary.

The RBDPO comparator handles its own co-product differently. Its published refinery stage embeds a substitution credit for palm fatty acid distillate [62]. Substitution and allocation are alternative answers to the same multifunctionality problem and need not agree [63], and this is the non-equivalence noted in Section 2.1.

2.6 Background System

No commercial database was used. Instead, four background scenarios span the published cradle-to-mill-gate burden of CPO (Table 2), from a single best-performing company through the industry averages for plantations certified and not certified by the Roundtable on Sustainable Palm Oil (RSPO) to the Indonesian average. Three of these scenarios, drawn from two sources, report their burden per kg of refined oil. These burdens are converted to a CPO basis by subtracting the published refinery stage, 0.31 kg CO₂-eq kg−1, and applying the refinery yield of each source, 0.926 for United Plantations and 0.953 for the industry averages [26, 62]. The Indonesian average and the consequential bound report per kg CPO and are used directly [3, 2]. The Indonesian district-level inventory [3] provides the reference case and the spatial bounds. The consequential value without land use change [2] serves as a bound on the literature rather than as a result scenario. The RBDPO comparator is built from the burden of the same CPO divided by the refinery yield of 0.953, plus the published refinery stage, which already contains its chemicals, water and energy [62].

2.7 Vitamin A Basis and Fortified Comparator

For FU-2 the comparator is RBDPO with its residual carotene, which is small but measurable: 1.85 mg kg−1 after standard refining [64], and 3.24–21.02 mg kg−1 on one refined sample measured by two instruments under ISO 17932 [65]. The trading specification for refined palm oil regulates acidity, moisture, iodine value, melting point and color, not carotene [66].

For FU-3, carotene is converted at 12 µg β-carotene per µg RAE and preformed vitamin A at 0.300 µg retinol per international unit (IU), the conventions of both the Indonesian regulation [24] and the dietary reference intakes [67]. Treating all carotene as β-carotene favors RPO, so a 38.9% α-carotene share [68] converted at 24:1 is tested separately. The comparator is RBDPO fortified with retinyl palmitate to the Indonesian minimum of 45 IU g−1 [69], 13.5 mg RAE kg−1 or 24.75 mg retinyl palmitate per kg oil [70]. No published carbon footprint of synthetic vitamin A was found, so its burden is set to zero, the bound most favorable to the comparator. Two break-even quantities are reported instead: the fortificant burden and the fortification level at which fortified RBDPO would equal RPO per mg RAE.

2.8 Calculation and Sensitivity Analysis

The inventory is computed by six chained Python scripts. Every parameter carries a value, unit, status, source and range, and each script reads the outputs of its predecessors rather than restating their numbers. Result ranges are the minimum and maximum over the scenario grid. For FU-1 this grid combines the four background scenarios, the five deodorization scenarios and the first two allocation treatments. For FU-2 and FU-3 it comprises all 540 combinations of the four backgrounds, three allocation treatments, five deodorization scenarios, three carotene retentions and three residual carotene levels in RBDPO. At system level, nine axes are varied one at a time from the reference case (Indonesian average background, pilot deodorization, mass allocation): background source; plantation location within Indonesia; allocation basis; deacidification yield; CPO transport; deodorization scale and heat source; deacidification energy; the refinery-stage figure subtracted in background conversion; and carotene retention (51.5–71.1%, as measured at 130–150 ∘C and 1–2 h [16], against 69.8% at the reference case, with the carotene content of RPO scaled in proportion). The refinery-stage axis moves only the scenarios converted from refined oil, so it leaves the reference case unchanged. The RPO–RBDPO difference is analyzed over the same axes.

3.1 Refining Inventory and Stage Burdens

Table 1 and Fig. 2 trace the refining chain from 1 kg of CPO to 0.897 kg of RPO. The finished oil still carries 375.33 ± 22.87 mg kg−1 of total carotene, the property that sets it apart from conventionally refined palm oil.

Of the two refining stages, deodorization is the one whose burden depends most on how the plant is run. In the pilot scenario it emits 64.7 kg CO₂-eq t−1 RPO. Scaling the same process to an electrically heated industrial deodorizer lowers this to 24.5, and replacing the electric heater with boiler steam lowers it further, to 19.2, 15.8 and 11.2 kg CO₂-eq t−1 for coal, natural gas and biomass firing. Steam heating comes out lower because boiler heat avoids the conversion losses embedded in grid electricity.

Degumming and deacidification together emit 36.8 kg CO₂-eq t−1 NRPO. Thermal energy accounts for 27.1 of this, and chemicals, process water and methane from the wash water for the remaining 9.7. Among the chemical inputs NaOH supply is the largest, at 6.8 kg CO₂-eq t−1 CPO (Table S3). Transporting CPO from the mill to the refinery adds 0.011 kg CO₂-eq kg−1 CPO, a term shared by both products because both start from the same oil.

3.2 Carbon Footprint per Kilogram of Oil

At the reference case (Indonesian average background, pilot deodorization and mass allocation), the footprint of RPO is 5.84 kg CO₂-eq kg−1. Almost all of it arrives with the CPO. Cultivation, milling and transport contribute 5.740 kg CO₂-eq kg−1 (98.3%), whereas deacidification adds 0.033 (0.6%) and deodorization 0.065 (1.1%). Varying the background, the deodorization scenario and the allocation treatment widens the result to 1.28–6.47 kg CO₂-eq kg−1 (Table 3), and across this scenario grid the foreground built here accounts for 0.8–7.4% of the total. The spatial bounds show how much further the background can move the result: with the lowest- and highest-emitting Indonesian districts as background, the footprint runs from 0.76 to 29.10 kg CO₂-eq kg−1 over the same deodorization and allocation scenarios.

Comparison with published footprints is limited in two ways. No life cycle inventory of RPO exists, so the foreground cannot be checked against an independent one. Placement against refined palm oil is also less independent than it appears, because three of the background scenarios were themselves derived from those publications. What can be compared is the structure of the result, and on that the agreement is clear: the stage-resolved assessments available place most of the burden upstream of the refinery [25, 71, 28]. Minimal-refining routes for carotene-bearing palm fractions, such as oil from palm-pressed fiber, are being developed without such data [72].

The absolute values are consistent with the literature as far as the literature allows. A recent footprint of Indonesian refined palm oil, 2.20 kg CO₂-eq kg−1 [1], falls within the range reported here. The RPO footprint on the Indonesian reference background, 5.78–6.47 kg CO₂-eq kg−1, lies above the 3.42 and 5.32 kg CO₂-eq kg−1 reported for RSPO-certified and non-certified refined oil averaged over Indonesia and Malaysia [26]. The two estimates differ in reference year, boundary detail and producer population, and the published data do not allow the difference to be assigned to any one of these.

3.3 Sensitivity and Where a Lower Footprint Would Come From

Fig. 3 ranks the sensitivity axes by how far each moves the footprint. Plantation location within Indonesia dominates: it moves FU-1 by 25.43 kg CO₂-eq kg−1, more than five times the next axis, the choice of background source (4.51). An accounting choice and a data-reading choice follow, the allocation basis at 0.94 and the deacidification yield at 0.25. The refinery parameters come last. CPO transport moves FU-1 by 0.062, deodorization scale and heat source by 0.054 and the deacidification energy model by 0.049, each 400 to 500 times smaller than the plantation-location swing. The refinery-stage axis leaves both units unchanged at the reference case, whose background needs no conversion. Carotene retention is the only axis that leaves FU-1 unchanged but moves FU-2, by 5.81 g CO₂-eq mg−1, because it alters only the carotene content of the oil.

This ranking means that where the CPO is grown outweighs every refinery parameter by two to three orders of magnitude, because 93–99% of the footprint is fixed before the oil reaches the refinery. For a producer seeking a lower-carbon RPO, the leverage therefore lies in sourcing rather than in refinery design. That leverage is plausible in practice: expansion that avoids forest and peat is compatible with continued growth in Indonesian output [73], and certification has reduced deforestation on certified plantations [74].

3.4 Per-Kilogram Difference from RBDPO

At the reference case RBDPO comes to 6.30 kg CO₂-eq kg−1 against 5.84 for RPO, a difference (RPO minus RBDPO) of −0.46 kg CO₂-eq kg−1, or 8.0% of the RPO footprint. Taken alone, this figure would favor RPO, but it does not survive the sensitivity analysis (Fig. 4). One axis reverses its sign, and it is an accounting choice rather than a physical parameter: with soapstock treated as waste the difference becomes +0.17, and with mass allocation on wet soapstock it widens to −0.78. Across these allocation bases the RPO footprint moves from 5.52 to 6.47 kg CO₂-eq kg−1, while RBDPO stays at 6.30. Plantation location, the largest axis for the footprint itself, moves the difference from −0.24 to −1.36 but never brings it to zero.

The reason lies in how much CPO each product draws. One kilogram of RPO requires 1.115 kg of CPO. With soapstock treated as waste it carries the burden of all 1.115 kg, whereas under mass allocation it is charged with the burden of 1.005 kg, against 1.049 kg for RBDPO. The burden per kilogram of CPO is identical for the two oils, so only the shared part of that burden cancels in the difference. The remainder scales with the background, which varies 37-fold across districts. The mass balance alone suggests a yield penalty for the caustic route, but that penalty appears only when soapstock is treated as waste. It is therefore an artifact of the allocation convention rather than a property of the route.

Three features of the comparison prevent the reference difference from being read as a ranking. First, its sign depends on a co-product convention, not on an uncertainty in the system itself. Convention choices have reordered food comparisons before [75], and they do so here: the convention alone reverses the sign of the result. Second, the two systems are not treated alike: RPO is allocated, whereas the comparator carries an embedded substitution credit, so the comparison does not meet the equivalence that ISO 14044 requires of a comparative assertion. Third, the convention alone moves the difference by 0.94 kg CO₂-eq kg−1, nearly as much as the 1.12 by which plantation location moves it across districts, so a methodological choice weighs about as much as the largest physical axis. On a mass basis, therefore, the two oils cannot be ranked.

3.5 Carotene and Vitamin A Functional Units

The comparison changes once the functional unit reflects what RPO is produced for. On FU-2 the RPO reference case is 15.55 g CO₂-eq mg−1 carotene. It spans 3.40–17.23 across the four result backgrounds of Table 3, and 3.16–23.36 when carotene retention and the wet-soapstock bound are also varied (Table 4). RBDPO is not carotene-free. With its measured residual carotene of 1.85 mg kg−1 the comparator is defined on FU-2, but its value is large: 3407 g CO₂-eq mg−1 at the reference case, and 76–3407 over the residual range and the full scenario grid, 13–273 times the RPO value (Table 4; Fig. 5). This ratio is governed almost entirely by the residual carotene, a quantity that conventional refining is designed to drive toward zero. FU-2 therefore shows less about the carbon efficiency of RBDPO than about the fact that it delivers carotene only incidentally.

FU-3 compares the two oils on the function that Indonesian regulation actually sets for cooking oil, vitamin A activity, to which carotene and preformed vitamin A both contribute [24]. RPO delivers 31.3 mg RAE kg−1 at a conversion of 12:1, and RBDPO fortified to the national minimum delivers 13.7. Per mg RAE, the fortified comparator carries 2.47 times the burden of RPO at the reference case and 1.47–3.08 times across the scenario grid, even though its fortificant is assigned no burden at all. Because the comparator already carries the larger burden with a burden-free fortificant, no positive fortificant footprint can close the gap at the national minimum, and the break-even fortificant burden is negative at every point of the grid. Parity would instead require fortifying RBDPO at 69–140 IU g−1, 2.5 times the minimum at the reference case. The result does not depend on treating all carotene as β-carotene: counting 38.9% of it as α-carotene at 24:1 lowers the reference ratio to 2.00 and the parity level to 90 IU g−1, but leaves the direction unchanged.

The oil also supplies fat, and a multifunctional unit that recognizes this would not change the comparison. Both oils deliver the same fat per kilogram, so any partition of the burden between fat and vitamin A that is applied equally to both scales their FU-3 values by the same factor and leaves the ratio unchanged. The fat function is, in effect, the per-kilogram comparison of Section 3.4, which is indeterminate. The mass basis and the vitamin A basis therefore answer different questions, and only the second separates the two oils.

Two limits apply to this result. FU-3 measures vitamin A activity delivered at the refinery gate, not the amount absorbed. The 12:1 factor for dietary carotene is conservative for carotene dissolved in oil [67], so it is more likely to understate than to overstate the vitamin A value of RPO, while losses during storage and frying, of either carotene or retinyl palmitate, fall outside the boundary. The result also rests on fortification at the regulatory minimum, and oil fortified well above it would narrow the gap or bring the ratio below 1. Measured levels suggest that oil is commonly fortified near or below the minimum. In a national program evaluation, fortified oil left the factory at 43.6 IU g−1 [69], and 32% of regulatory samples taken in 2021–2023 fell below 45 IU g−1 [76]; both levels lie below the parity range.

3.6 Limitations

The least certain input in this study is foreground energy, which is estimated from equipment specifications rather than measured, although Section 3.3 shows its effect on the result to be small. The pilot vessel is specified at 100 L but was charged with about 112 L of oil, so heat loss is likely to be understated. The foreground is also pilot-scale, and the industrial scenarios rest on scaling laws, an assumed boiler efficiency of 0.80 and a Malaysian transport distance.

The inventory itself joins two experiments. Their intermediate NRPO differ in carotene by a factor of 1.15, so carotene retention across the chain is the product of two stage ratios rather than a single measurement. The comparator, in contrast, is a published refinery stage rather than an inventory built here, and it follows a different co-product convention. Reference years are not aligned either: the backgrounds refer to 2015–2024, the process data to 2012–2016 and the electricity factor to 2024. For the vitamin A comparison, the residual carotene of RBDPO rests on two sources, and no footprint of synthetic vitamin A was available. Finally, only climate change is assessed, and other impact categories, eutrophication from mill effluent and fertilizer among them, are not addressed.