Evaluating the Laboratory Performance of Pellet-fueled Semi-gasifier Cookstoves

Abstract

This study examines three representative semi-gasifier cookstove models each burning four types of pelletized-biomass fuel (hardwood, peanut hull, risk husk, wheat straw) using the International Organization for Standardization (ISO) 19867–1:2018 protocol. ISO tier ratings for fine particulate matter (PM_2.5) and carbon monoxide (CO) emissions ranged from 1–4 and 2–5 (where 5 = cleanest), respectively, suggesting that pellet-fueled cookstoves may provide substantial emissions reductions, dependent upon stove/fuel matching and operation, over other biomass-fueled cooking alternatives. PM_2.5 emission factors based on useful energy delivered (EF_d) varied by up to 25-fold, and organic and elemental carbon (OC and EC) EF_d values respectively varied by >200- and ~100-fold, reflecting complex variability in PM_2.5 composition. These semi-gasifier cookstoves showed higher ultrafine particle (UFP) emissions but lower bulk PM_2.5 emissions. Operation of pellet-fueled cookstoves at higher firepower resulted in higher PM_2.5 and UFP emission factors and higher EC to total carbon ratios; operation at lower firepower resulted in higher gaseous pollutant emission factors. Results of this work provide technical guidance for stove developers, users, and policy-makers. These ISO-protocol based emission factors are also pertinent to health and climate modeling efforts.

1. Introduction

Exposure to air pollution results in an estimated 6.7–8.9 million deaths annually,^{1, 2} with ~2.3 million deaths attributed to household air pollution from solid fuel use.³ Globally, 43% of the world’s population is rural,⁴ and an estimated 60% of rural homes still rely on unprocessed biomass for cooking.⁵ With roughly three billion people globally, including 75% of the population (~80% of households) in sub-Saharan Africa, reliant on solid fuels to meet their energy needs,^6–8 the wider implementation of cleaner-burning solid fuel cookstoves remains a timely and pressing concern.

Transitions to cleaner energy [e.g., liquified petroleum gas (LPG), electricity] inherently reduce exposures to household air pollution. However, clean energy adoption in rural areas is typically slow⁹ due in part to free access to biomass¹⁰ and lack of available and affordable modern energy solutions.¹¹ Additionally, the implementation (i.e., integration into communities) of semi-gasifier (wherein the gasification and combustion processes are closely coupled) pellet cookstoves can be challenging to achieve due to the perceived difficulty of operating them compared to traditional stove types,^{12, 13} lack of stove support infrastructure (e.g., advice, repairs, parts),¹⁴ and a variety of pellet feedstock supply limitations.^14–16

However, pellet-fueled semi-gasifier cookstoves provide an opportunity to bridge the emissions-performance gap between solid fuel burning stoves and cleaner gas burning and electric stoves, and promote an overall cleaner “stack” (i.e., combination of cooking options) for those reliant on solid fuels.¹⁷ Compared to traditional and even improved wood and charcoal stoves, pellet stoves perform well in both laboratory^18–20 and field emissions testing.^21–23 For example, in Rwanda the Mimi Moto pellet stove was consistent with International Organization for Standardization (ISO) Tiers 5 and 4 (i.e., “best” and “second best”) for carbon monoxide (CO) and fine particulate matter (PM_2.5), respectively, and approached gas stove performance.^{21, 24} In terms of environmental (climate) impacts, pellet stoves tend to be comparable to LPG or even negligible (depending on biomass renewability and upstream emissions),²¹ though more recent research has suggested reduced climate benefits from pellet stove use.²⁵

Performance for pellet stoves may be less sensitive to user behavior than for other biomass stove types, but performance is nevertheless dependent on operating variables and emissions can sometimes be much higher than for other cleaner energy carriers (e.g., LPG, electricity). For example, our research previously showed that pollutant- and mutagenicity-emission factors for different pellet stove and fuel combinations vary more than two orders of magnitude on a fuel energy basis, highlighting the importance of proper matching of fuel to stove (or vice versa).²⁶ Additionally, the performance of top-lit updraft semi-gasifiers (TLUDs, a common design for pellet-fueled cookstoves) is sensitive to fuel type,^{27, 28} fuel moisture,^{29, 30} fuel size, ^{28, 30, 31} primary and secondary air flow rates,^{32, 33} and operating mode (e.g., rapid changes from flaming to smoldering).²⁷

Fuel ignition and burnout periods specifically may result in outsized emissions contributions.^{21, 34} For this reason, Tryner et al.³⁵ noted that some test protocols that exclude burnout and refueling (e.g., Water Boiling Test) are poorly suited for characterizing pellet cookstoves. Generally, laboratory-based cookstove testing can underestimate emissions when compared to field testing owing to many variables including ignition procedure, heterogenous fuel composition/moisture, and non-optimized stove operation.³⁶ Additionally, compared with other improved cookstove types, gasifier-type stoves, including some pellet stoves, tend to emit higher ultrafine particle (UFP) number concentrations.^{19, 37} Compared with larger particles, these UFPs are known to elicit greater toxicological response^{38, 39} due to their increased surface area per mass and ability to deposit deeper in the respiratory tract. Given the potential for pellet cookstoves to provide relatively clean energy, and the variable findings from limited evaluations of this technology, we aim to systematically assess pollutant emissions from several typical pellet stove and fuel types. In the present study we: (1) characterize laboratory-based emissions and performance of twelve pellet-fueled stove/fuel combinations using the most recent and relevant laboratory-based testing standard, (2) investigate the influence of stove operation and fuel matching/composition on performance and emissions metrics, and (3) provide recommendations for pellet stove designers and end-users, as well as for household energy policymakers.

2. Methods

2.1. Cookstove/fuel combinations

A total of twelve semi-gasifier cookstoves (with forced-draft air injected below and above the combustion chamber providing primary and secondary air, respectively) and fuel combinations were assessed. Three semi-gasifier stoves were tested: the African Clean Energy (ACE) One stove, the Mimi Moto stove, and the Philips HD4012 stove, herein referred to as ACE, MM, and Phlp, respectively. Images and details of these stoves are provided in Section S1 of the Supplemental Information (SI).

Four pellet fuel types were examined: hardwood, peanut hull, rice husk, and wheat straw, herein referred to as HW, PH, RH, and WS, respectively. Fuels were sourced commercially (HW), from a regional (southeastern US) manufacturer (PH), and from a marketplace in China (RH, WS); these fuels represent a range of feedstock types and are not intended to represent specific user-groups. Fuel compositions are tabulated in Table S2 of the SI. Fuel moisture content was measured for each testing day using ASTM Standard Method D4442–20.⁴⁰ Fuel heat of combustion and proximate and ultimate analyses were measured using ASTM Standard Method D5865–19;⁴¹ per ISO stove test protocol specifications, the lower heating value of each fuel was used in calculations. Stove and fuel combinations are referred to using combined abbreviations (e.g., MMHW is the Mimi Moto stove burning hardwood pellets).

2.2. Test protocol

The ISO 19867–1:2018 test protocol was employed in the present study.⁴² This protocol specifies three distinct power levels (high, medium, and low), each consisting of a start-up period, an approximately 30-minute fuel-burning period, and a shutdown period. The shutdown period concludes once either: 1) water temperature drops 5°C from maximum value observed during fuel-burning period, or 2) five minutes have elapsed following conclusion of fuel-burning period (i.e., shutdown period may not last more than 5 minutes). High- and low-power operation, respectively, refers to reasonable (i.e., consistently achievable as identified through shakedown testing) upper and lower limits of the stove’s fuel-burning rate, and subsequently, firepower (i.e., fuel energy released per time). We have previously compared data using the ISO protocol to those using the Water Boiling Test (WBT, the former widely used protocol for laboratory cookstove testing) and the Firepower Sweep Test (FST, a protocol intended to more accurately represent real-world conditions), and found emissions metrics using the ISO protocol to generally rank between them.³⁷ Specific to forced-draft pellet-fueled semi-gasifier stoves (i.e., the stoves tested here), use of the ISO vs. WBT protocol resulted in significantly higher emissions metrics for CO, due largely to the inclusion of the burnout/shutdown period in the ISO protocol. In the present study, five replicates each of the ISO high-, medium-, and low-power test phases were conducted for each of the 12 stove/fuel combinations, for a total of 180 test phases, or 60 complete ISO test replicates.

2.3. Test facility and emissions characterization

Testing was conducted at the US EPA’s Household Energy Research Facility located in Research Triangle Park, NC, USA. A schematic of the testing system and a description of the facility were provided previously.¹⁸ Briefly, emissions were collected into a full-capture stainless-steel hood connected to a dilution tunnel with constant flow (~4.2 m³ min⁻¹) from which pollutants were sampled. An induced-draft blower maintained negative pressure in the entire system and provided hood air and filtered dilution air flows; a second stage of dilution (9.9:1) was provided by a dilution sampling system, required for aerosol instrumentation.

CO, carbon dioxide (CO₂), total hydrocarbons (THCs), and methane (CH₄) were monitored continuously (0.2 Hz) with nondispersive infrared (NDIR) and flame ionization detector (FID) analyzers (Models 600, California Analytical, Orange, CA). Nitrogen oxides (NO_x) were monitored (0.2 Hz) with a chemiluminescence analyzer (Model 600 HCLD, CAI Inc., Orange, CA). PM_2.5 was sampled isokinetically onto both polytetrafluoroethylene (PTFE) membrane and quartz fiber filters (QFF) positioned downstream of PM_2.5 cyclones (URG, Chapel Hill, NC) at a flow rate of 16.7 lpm each. A second QFF downstream of the PTFE filter was used to compensate for gas-phase artifacts.⁴³ PM_2.5 mass was measured gravimetrically with a microbalance (AH225–7, MTL, Minneapolis, MN), and the PTFE filters were equilibrated at 35% relative humidity and 23°C in an environmental chamber prior to weighing. QFF were analyzed for organic and elemental carbon (OC and EC) content using an OC-EC aerosol analyzer (Sunset Laboratory, Tigard, OR) and a modified NIOSH protocol.⁴⁴ An engine exhaust particle sizer (EEPS; Model 3090; TSI; Shoreview, MN) measured particles of mobility diameter range 6.0–523 nm in real-time (1 Hz) to report ultrafine particle (UFP) emissions over the range of 6.0–100 nm as factors/rates on a particle number basis. UFP data were unavailable for MM-HW for this study, but UFP data for this stove and fuel combination tested using the ISO protocol in the same lab are available elsewhere.³⁷

Emissions of all pollutants were quantified using the total-capture mass-flow method which required continuous monitoring (0.2 Hz sampling rate) of the dilution tunnel air velocity, temperature, pressure, and relative humidity over the test period. Emissions metrics reported include mass (or number for UFP) of pollutant emitted per: (a) useful energy delivered to the cooking pot (energy emission factor, EF_d), (b) dry fuel mass consumed (mass emission factor, EF_m), and (c) time (emission rate, ER). These calculations are defined in the ISO 19867–1:2018 protocol. Internal quality assurance comparing total capture and carbon balance methods⁴⁵ was also employed. Test phase-specific detection limits (DLs) were calculated as previously described,⁴⁶ and for the purpose of calculating central tendencies (e.g., mean, median), those emissions metrics that were below the calculated DL, but above background, were imputed with the DL itself, while those values at or below background were considered to be zero. This approach, compared to imputation with DL/√2 for example, provides conservative estimates of emissions for the relatively clean-burning stove types tested here.

3. Results and Discussion

3.1. Emission factors

The SI summary spreadsheet reports sample size, mean, and standard deviations for all pollutants measured (PM_2.5, UFP, OC, EC, CO, THC, CH₄, NO_x, and CO₂), aggregated by stove/fuel combination and ISO test phase (including overall, defined as average of high-, medium-, and low-power phases), on the basis of useful energy delivered, dry fuel mass, and time (EF_d, EF_m, ER, respectively). The overall average metric is further discussed in the SI Section S2 and Tables S3 and S4. Discussion of emission factors in this manuscript focuses on EF_d, given that it is activity-based and generally applicable (e.g., the ISO 19867–3:2018 rating framework for emissions is based on EF_d values). Summaries of mean overall energy emission factors grouped by stove and fuel are respectively summarized in Tables S5 and S6 of the SI; these groupings provide insight beyond that of individual stove/fuel combinations to explore overall effects of stove or fuel choice. Table S7 summarizes the percentage of test phases with emissions metrics either above DL, below DL, or zero. Finally, the Shapiro-Wilk test was employed to assess normality of data subset distributions (e.g., all PM_2.5 EF_d for MM-HW) and to inform subsequent statistical testing. Normality test results for individual stove/fuel combinations, and by stove- and fuel-groupings are summarized in Tables S8, S9, and S10, respectively. For all pollutant subsets (excluding CO₂), only 30% (28/93) were normally distributed and if grouped by stove and fuel types, only 4% (1/24) and 6% (2/32) were normally distributed. Thus, the use of non-parametric statistical approaches was considered appropriate; hence, pairwise Wilcoxon tests (vs. for example, Student’s t-tests) were applied. These results for stove- and fuel-groupings are summarized in Tables S11 and S12, respectively.

Figures 1a-d plot EF_d values for particle pollutants (PM_2.5, UFP, OC, and EC), respectively. The lowest overall PM_2.5 EF_d was for Phlp-PH (13.5 mg ± 1.3 MJ_d⁻¹), roughly four-fold the PM_2.5 emissions from LPG stoves (3.1 mg MJ_d⁻¹).³⁷ These results suggest that an ideally-matched pellet stove/fuel combination (as indicated by low emission factors) begins to approach PM_2.5 emissions performance of LPG, but likely cannot meet or exceed LPG performance. This is an important and likely inherent limitation of semi-gasifier cookstove technology, as PM_2.5 is considered to be one of the most important, if not the most important, component of air pollution contributing to health effects.⁴⁷ Conversely, ACE-WS (331 ± 60.1 mg MJ_d⁻¹) showed the highest overall PM_2.5 EF_d, similar to a natural-draft wood stove previously tested under identical conditions (mean overall PM_2.5 EF_d of 328 mg MJ_d⁻¹). Therefore, pellet stove PM_2.5 emissions varied by a factor of ~25 in this study. However, even the highest PM_2.5 emitting pellet stoves show a vast improvement over a traditional three stone fire (TSF) tested in the laboratory using the ISO protocol (943 mg MJ_d⁻¹).³⁷

Figure 1.

Box and whisker plots (with jittered points by ISO test phase) of emission factors based on energy delivered (EF_d) for particle pollutants (a) fine particulate matter (PM_2.5), (b) ultrafine particles (UFP), (c) organic carbon (OC), and (d) elemental carbon (EC).

Grouped by stove type, the mean overall PM_2.5 EF_d values in ascending order were (Table S5): MM ≈ Phlp < ACE. The ACE stove was also found to have statistically higher PM_2.5 EF_d values compared with both the MM and Phlp stoves (Table S11). It is unclear why the ACE resulted in the highest PM_2.5 emissions, and it is recommended that further research be conducted to examine differences in primary and secondary air flow rates between pellet cookstoves³⁵ (provided their generally similar design, use of materials, and build). When grouped by fuel type, mean overall PM_2.5 EF_d were (Table S6): RH ≈ PH < HW << WS; WS pellets had significantly higher PM_2.5 EF_d than all other pellet types (Table S12). For each stove type, the use of wheat straw pellets resulted in high and variable PM_2.5 EF_d values (ranging from 222–331 mg MJ_d⁻¹). No clear differences in fuel composition support this finding, however the pellet manufacture process itself (e.g., feedstock moisture, extrusion pressure) may impact the performance of the manufactured product. Regardless, use of wheat straw pellets in these stove types should be considered further when developing air quality policy.

The Phlp-PH showed the highest overall UFP EF_d (1.92×10¹⁵ ± 1.80×10¹⁴ # MJ_d⁻¹), and ACE-HW was the lowest (1.05×10¹⁵ ± 6.44×10¹³ # MJ_d⁻¹). UFP emissions across all stove and fuel combinations varied by less than a factor of two (i.e., much less than bulk PM_2.5 emissions). UFP data (Figure 1b) generally align (albeit in the higher range) with previous laboratory-based test data for pellet stoves (1.3×10¹⁴ - 2.0×10¹⁵/MJ_d).^{18, 26, 48} The stove/fuel combinations tested here had higher UFP EF_d than natural-draft improved wood stoves (5.09×10¹⁴ # MJ_d⁻¹) and unimproved charcoal stoves (5.24×10¹⁴ # MJ_d⁻¹), but still provide emissions reductions over the traditional TSF (2.44×10¹⁵ # MJ_d⁻¹).³⁷ Notably, these pellet stoves had much higher UFP emissions than LPG stoves (3.90×10¹² # MJ_d⁻¹).³⁷ Importantly however, fresh UFP emissions from gas stove sources are typically sub-10 nm in aerodynamic diameter,^{49, 50} and therefore poorly characterized or missed entirely by the aerosol instrumentation used in many cookstoves tests (including the present study). Though these pellet stove/fuel combinations resulted in overall low PM_2.5 emission factors, the UFP emissions for some combinations (e.g., Phlp-PH) were only ~20% lower than comparable measurements from the TSF. This supports the understanding that improved combustion mechanisms driving reduced PM_2.5 mass emissions do not comport with comparable reductions in UFP number emissions. The relatively high combustion efficiency of pellet-fueled stoves (compared to their wood or charcoal counterparts) resulting in lower concentrations of volatile and semi-volatile organic compounds (S/VOCs) inhibit particle growth,⁵¹ resulting in reduced PM_2.5 emissions, yet relatively high UFP number emissions. Therefore, UFP emissions, becoming increasingly recognized for their ubiquitous role in damaging health,^{52, 53} remain a persistent concern for semi-gasifier stoves. It is unlikely that stove designers may reduce this phenomenon, owing to the fundamental processes occuring within and above the fuel bed/flame, and therefore a focus on informing and educating end-users to reduce their exposures through behavior change (e.g., avoid standing above the stove, cook upwind when possible) is encouraged.

The highly variable particle emissions from pellet stoves are further demonstrated by investigating the carbon composition of the emitted PM_2.5, informed by the well-established role of EC as a climate forcer⁵⁴ and OC as a significant contributor to health burdens caused by air pollution.⁵⁵ The ACE-HW showed highest overall OC EF_d (87.8 ± 77.2 mg MJ_d⁻¹), while the Phlp-RH showed the lowest (0.4 ± 0.2 mg MJ_d⁻¹). This finding may be due to improper mixing above the fuel bed, highlighting the importance and difficulty of proper fuel matching for this stove technology; the rate of fuel decomposition (and therefore formation of OC and PM_2.5 precursors) should be well-coupled to primary and secondary air flows. For EC, the highest-emitting combination was also ACE-HW (66.8 ± 24.1 mg MJ_d⁻¹), while the lowest-emitting was the MM-RH (0.7 ± 0.4 mg MJ_d¹). This respectively represents >200-fold and ~100-fold differences in OC and EC emissions among the stove/fuel combinations tested here. Grouped by stove (Table S5), mean overall OC EF_d were: Phlp < MM < ACE; mean overall EF_d for EC were: MM < Phlp < ACE. The ACE stove resulted in statistically higher OC and EC emissions than the MM and Phlp stoves (Table S11). When grouped by fuel (Table S6), mean overall OC EF_d were: WS < RH < PH < HW; mean overall EF_d for EC were: RH < WS < PH < HW. HW fuel had statistically significantly higher EC emissions than the other fuel types tested (Table S12).

Overall EC to TC (i.e., total carbon, the sum of OC and EC) ratios ranged widely from 0.06 (MM-PH) to 0.87 (Phlp-HW). Generally, higher EC:TC ratios represent stronger climate forcing of the PM_2.5 in the atmosphere,⁵⁶ and recent work has called for improved understanding of the impacts of cookstove PM_2.5 composition on human health effects.⁵⁷ The Phlp stove generally had higher EC:TC ratios (overall mean: 0.77) compared to the ACE (0.40) and MM (0.19) stoves. Stove operation can strongly influence PM composition from pellet stoves. For example, Shen et al.²² observed that stoves operating without secondary air have much higher EC:TC ratios compared with those operating with secondary air (0.75–0.76 vs. 0.18–0.39, respectively). The three pellet-fueled cookstoves evaluated here had primary air injected at the bottom and secondary air injected at the top of the combustion chamber and had widely ranging EC:TC values. Air flow rates and distribution patterns as well as primary/secondary air ratios may have affected emissions performance, especially for EC.

Figures 2a-d respectively plot EF_d values for CO, THC, CH₄, and NO_x. For CO, the highest overall EF_d was for ACE-HW (4.5 ± 1.3 g MJ_d⁻¹) while the lowest was for Phlp-PH (0.3 ± 0.1 g MJ_d⁻¹), comparable to an LPG stove (0.2 g MJ_d⁻¹),^{37, 58} suggesting that CO emissions performance of some pellet stove-fuel combinations can indeed approach gas stove-like performance under ideal conditions (i.e., favorable fuel/stove matching and careful operation). The highest-emitting combination in terms of CO (ACE-HW) was comparable to a natural-draft wood stove (4.5 g MJ_d⁻¹) but much cleaner than traditional TSF and unimproved charcoal stoves (15.5 and 19.3 g MJ_d⁻¹, respectively). Still, acute and chronic CO exposures are a persistent public health concern⁵⁹ and must be a consideration when adopting any cooking technology. Grouped by stove type (Table S5), mean overall CO EF_d were: Phlp < MM < ACE; the Phlp had statistically significantly lower CO emissions compared to the ACE and MM stoves (Table S11). When grouped by fuel type, mean overall CO EF_d were: PH ≈ RH < HW < WS. The WS pellets had significantly higher CO emissions compared with the other fuels tested (Table S12), similar to the PM_2.5 emissions performance, underscoring the relatively poor performance observed here for this pellet type.

Figure 2.

Box and whisker plots (with jittered points by ISO test phase) of emission factors based on energy delivered (EF_d) for gaseous pollutants (a) carbon monoxide (CO), (b) total hydrocarbons (THC), (c) methane (CH₄), and (d) nitrogen oxides (NO_x).

Finally, THC and CH₄ emissions were generally low for all stove/fuel combinations, ranging from 0.0–1.0 and 0.0–0.3 g MJ_d⁻¹, respectively. THC can contain health-damaging compounds,⁶⁰ while CH₄ is an important climate forcer.⁶¹ The cleanest stove/fuel combination (Phlp-PH) (here, under idealized operation) was comparable with LPG stove performance (0.06 g MJ_d⁻¹ of THC), and all combinations were cleaner than traditional wood and charcoal stoves (2.0 and 1.3 g MJ_d⁻¹, respectively).³⁷ NO_x EF_d values ranged between 0.1 ± 0.0 g MJ_d⁻¹ (ACE-HW) and 0.9 ± 0.2 g MJ_d⁻¹ (Phlp-PH), consistently higher than LPG stoves tested using the ISO protocol which narrowly ranged between 0.05–0.06 g MJ_d^-1.³⁷ Compared with particle emissions, gaseous emissions were generally less variable and thus less influenced by stove and fuel matching.

Although a detailed analysis of climate impacts/benefits (i.e., global warming commitment or GWC) brought about by interventions employing pellet-fueled stoves was outside the scope of the present work (owing to high variability in the many assumptions needed for such analysis, e.g., non-renewable biomass fractions and stove use patterns globally), it is important to discuss potential environmental/climate impacts from pellet stove use. Carbon dioxide equivalents (CO₂-eq), which account for climate warming/cooling behavior of combustion emissions, are a common means of estimating such impacts. Previous work has identified pellet cookstoves as offering CO₂-eq on-par with LPG stoves (i.e., offering substantial climate benefits).²¹ However, a recent study in Ghana²⁵ comparing climate benefits (but ignoring OC climate cooling impacts) from pellet stove use vs. traditional and improved wood and charcoal stoves determined pellet stoves to rank between charcoal and improved wood stoves. Therefore, climate benefits of pellet stoves are not as clearly defined as expected health benefits, and should be more well characterized for this technology.

To explore relationships between pollutant emissions, Figure S2 plots EF_d values for all stove and fuel combinations and for all pollutants against CO EF_d data. CO was chosen as the basis here, as it is the key gaseous pollutant reported for cookstove emissions, and often used for comparison against other pollutants (typically, PM_2.5).^{36, 62} Use of CO as a proxy for PM_2.5 emissions has sometimes been suggested for indoor concentrations in homes using wood and gas cookstoves,⁶³ though this approach is generally not recommended.^{64, 65} Additionally, CO emission factors are, by definition, inversely related to modified combustion efficiency (MCE, defined as the molar ratio of CO₂ to the sum of CO and CO₂). Table S13 summarizes statistical test results of these linear fits. All pollutants were significantly correlated to CO EF_d, with PM_2.5, OC, EC, THC, and CH₄ having positive relationships, and UFP, NO_x, and CO₂ negative. This result is unsurprising, as aerosol and organic pollutants are expected to increase as combustion efficiency decreases, while UFP and CO₂ emissions decrease owing to greater particle growth (i.e., conversion of UFP to PM_2.5) and more fuel carbon reaching complete thermal decomposition.

3.2. Relationships between firepower and emissions metrics

The ability to evaluate stoves at variable power levels (specifically an additional medium-power level) is an advantage of the ISO protocol compared to the WBT. Using the EF_d data, power level-means (i.e., mean of all stove/fuel combinations at a given power level as reported in Table 1) were greatest in the high-power phase for PM_2.5, and lowest in the low-power phase. These results were similar for UFP, and, as expected (but not tabulated here), for CO₂ as well. This trend was reversed (i.e., lower-power operation resulted in higher EF_d values) for OC, CO, THC, and CH₄. Only EC had the highest emissions during the medium-power phase. These trends highlight how operationally dependent (here, in terms of fuel loading and fan setting) the emissions performance of pellet stoves can be.

Table 1.

Summary of power level-mean (i.e., mean of all stove/fuel combinations at a given power level) and standard deviations (SD, here defined as standard deviations of all stove/fuel combination means at a given power level) of emission factors based on energy delivered (EF_d) for all pollutants excluding carbon dioxide (CO₂).

	PM2.5 (mg MJd−1)		UFP (# MJd−1)		OC (mg MJd−1)		EC (mg MJd−1)		CO (g MJd−1)		THC (g MJd−1)		CH4 (g MJd−1)		NOx (g MJd−1)
	Mean	SD	Mean	SD	Mean	SD	Mean	SD	Mean	SD	Mean	SD	Mean	SD	Mean	SD
High	156	220	1.91×1015	2.05×1014	1.4	1.0	1.6	1.7	1.7	1.5	0.1	0.0	0.0	0.0	0.6	0.3
Medium	89.6	86.5	1.35×1015	2.71×1014	15.7	30.8	16.5	29.4	1.9	1.4	0.3	0.3	0.1	0.1	0.4	0.2
Low	80.7	64.5	1.18×1015	3.14×1014	31.3	44.9	15.4	28.2	3.0	2.1	0.5	0.5	0.2	0.1	0.4	0.2

It is well established that higher firepower generally increases emissions from many cookstove types,^{34, 37} but these relationships have not been extensively explored for pellet-fueled cookstoves. Table S14 summarizes p-values and slope values from linear fits of EF_d values (by pollutant) against firepower for all stove/fuel combinations. Figure S3 plots the significant (p<0.05) correlations between pollutant EF_d values and firepower. PM_2.5 and UFP EF_d values (as well as EC:TC ratio) were significantly positively correlated with firepower, while CO, THC, and CH₄ were anticorrelated. Findings from this study therefore suggest that operation of pellet-fueled cookstoves at higher firepower generally results in higher particle emissions, while operation at lower firepower results in higher gaseous pollutant emissions (apart from CO₂ emissions, which increased with firepower).

3.3. Influence of fuel composition

Owing to high variability in biomass/agricultural feedstocks across different regions of the world,⁶⁶ here we conduct a comparison of fuel parameters to stove emissions and performance. The objective of this analysis is to inform potential stove use outcomes based on general pellet characteristics to be applied to prospective pellet production based on locally-available feedstock for a given area. Figure S4 plots thermal efficiency (i.e., the percentage of fuel energy that contributed to heating the water, as defined in the ISO 19867–3:2018 protocol) against the fuel characteristics of volatile matter (VM), ash, fuel O:C ratio, and fuel H:C ratio (as summarized in Table S15). A similar approach was taken for CO and PM_2.5 EF_d values and MCE, but no clear trends were observed. Importantly, thermal efficiency and MCE are not inherently linked, as stoves with high thermal efficiency can exhibit poor combustion efficiency. The VM content of pellets is known to strongly influence combustion in residential heating stoves, as the devolatilization and subsequent gas-phase combustion occur much more rapidly than the char oxidation (of fixed carbon and formed char).⁶⁷ Additionally, low O:C and H:C ratios in processed fuels (e.g., briquettes, pellets) tend to reduce water vapor (and particle emissions) when burned, contributing to (and owing to) higher combustion efficiencies.^{68, 69} In the present study, increasing VM content and fuel O:C ratios resulted in generally higher thermal efficiencies, while increasing ash content resulted in lower thermal efficiencies.

3.4. Ranking of stove/fuel combinations using ISO tier rating system

Figure 3 plots EF_d values for PM_2.5 and CO using the tier rating system defined by ISO 19867–3:2018;²⁴ tier ratings are defined by the upper bounds of emissions estimates (i.e., 90% confidence intervals) and therefore account for variabilities in emissions estimates vs. use of solely central tendencies. Included in Figure 3 are mean PM_2.5 and CO EF_d values for an array of literature pellet cookstove and fuel combinations, from lab, test kitchen, and field studies (shown in the figure as “L”, “K”, and “F”, respectively), as summarized in Tables S16 and S17 of the SI. Data from those studies (using reported mean values as opposed to the ISO protocol-specified upper-bounds of the 90% confidence intervals) span from Tiers 1–4 for PM_2.5 and Tiers 0–5 for CO. As a result, the EF_d values reported in the present study appear generally less variable and in the lower end of the range of values from the literature. This outcome is unsurprising given the controlled testing and adherence to the ISO protocol here.

Figure 3.

Emission factors (based on energy delivered) of fine particulate matter (PM_2.5) and carbon monoxide (CO) plotted against ISO Tiers for distinct test phases (i.e., power levels) and overall (i.e., mean of phase-averaged) results for all three stoves (a-c). Error bars represent the 90% confidence intervals about the mean. Mean values are also plotted from the literature (d) for lab (L), test kitchen (K), and field (F) studies, as summarized in Tables S16 and S17 of the Supplemental Information (SI).^{13, 18, 20–22, 26, 70}

As summarized in Table 2, PM_2.5 tier ratings for the stove/fuel combinations were determined for overall performance (50% Tier 4, 25% Tier 3, and 25% Tier 2), high-power (75% Tier 4, 8% Tier 3, and 17% Tier 2), medium-power (50% Tier 4, 25% Tier 3, and 25% Tier 2), and low-power (42% Tier 4, 42% Tier 3, and 17% Tier 2; sum ≠ 100% due to rounding). During high-power operation and for all fuels apart from wheat straw pellets, the combinations tested achieved Tier 4 for PM_2.5; this may be considered excellent performance for a solid fuel stove in general. CO tier ratings for the stove/fuel combinations were determined for overall (67% Tier 5, 25% Tier 4, and 8% Tier 3), high-power (75% Tier 5, 8% Tier 4, and 17% Tier 3), medium-power (67% Tier 5, 25% Tier 4, and 8% Tier 3), and low-power (33% Tier 5, 25% Tier 4, 25% Tier 3, and 17% Tier 2). Higher CO tier ratings were observed at higher power levels, consistent with the firepower analyses discussed in Section 3.2. The Philips (Phlp) stove was rated Tier 5 for CO for all fuel types except WS, suggesting that it was the most effective stove in terms of reducing CO emissions, independent of fuel type. Importantly, climate impacts of pellet cookstoves are not included in the ISO-based tiering framework. The higher combustion efficiencies observed here contributed to increased CO₂ emissions and generally elevated EC:TC ratios, both of which contribute to increased GWC.⁷¹ As such, it is recommended that climate impacts are included in cost-benefit analyses⁷² for prospective interventions seeking to further distribute pellet-fueled cookstoves.

Table 2.

Summary of ISO tier ratings for all stove/fuel combinations tested and for overall, high-, medium-, and low-power phases based on upper-bound of 90% confidence interval for fine particulate matter (PM_2.5) and carbon monoxide (CO) emission factors based on energy delivered (EF_d), as specified in ISO 19867–3:2018. Higher values indicate better performance.

	PM2.5 Tier				CO Tier
Stove/Fuel	Overall	High	Medium	Low	Overall	High	Medium	Low
ACE-HW	3	4	2	2	3	5	3	2
ACE-PH	3	4	3	3	5	5	4	3
ACE-RH	4	4	3	4	5	5	5	4
ACE-WS	2	1	2	3	4	3	4	4
MM-HW	4	4	4	4	5	5	5	4
MM-PH	3	4	4	2	4	5	5	2
MM-RH	4	4	4	3	5	5	5	3
MM-WS	2	2	3	3	4	4	4	3
Phlp-HW	4	4	4	4	5	5	5	5
Phlp-PH	4	4	4	4	5	5	5	5
Phlp-RH	4	4	4	4	5	5	5	5
Phlp-WS	2	1	2	3	5	3	5	5

3.5. Study limitations

A small set of stoves and fuels was tested and only a single laboratory test protocol was applied for this investigation. However, the stoves selected are representative of those used in a variety of global markets, and the ISO protocol (intended to replace the former WBT) generally provides emissions results between the WBT and the more field-representative FST.³⁷ Testing approaches may play a significant role in the emissions data collected for pellet-fueled cookstoves. For example, Hays et al.⁷³ observed that in a pellet-fueled biomass boiler, the effect of test cycle (i.e., protocol) had a greater impact on PM emissions, in terms of both mass emitted and composition, than the pellet type itself. Lastly, emissions measured in laboratory settings have repeatedly been shown to underestimate those observed in field settings,^{36, 74} therefore future research should compare ISO protocol to field performance results.

Importantly, the applicability of our results is limited by the relative lack of literature/data identifying pellet types commonly used in different regions and/or countries for cooking. However, assuming that agricultural residue may represent that feedstock available for cooking pellet manufacture,⁷⁵ we observe that across 23 low- and middle-income economy nations (where pellet use for cooking is most relevant) the prevalence of rice husk and wheat straw comprising the entire crop residue makeup vary from 0–11% and 0–59%, respectively.⁶⁶ Therefore a subset of the fuels (RH and WS) studied here are highly relevant in some locales, yet non-applicable to others. The HW fuel studied here is likely applicable to regions utilizing hardwood waste (e.g., sawdust) for pellet manufacture. Additionally, our findings may not be applied to all real-world conditions. For example, ignition and burnout periods of pellet stove use contribute outsized proportions of black carbon and CO, respectfully,²¹ and laboratory-based cookstove testing in general often results in lower emissions when compared to real-world, field-based testing.³⁶

4. Implications

This study determined the emissions performance of pellet-fueled semi-gasifier cookstoves burning a range of pellet types. The combinations tested generally met Tiers 3–4 and 4–5 for PM_2.5 and CO, respectively, which suggest substantial health benefits over the use of traditional stove types. Climate benefits were not included in the analyses here, but previous work suggests pellet stoves to offer GWC between LPG and wood rocket stoves.^{21, 25} That is, pellet stoves offer clear climate benefits over charcoal stoves, but further research should be conducted to determine climate benefits on global or regional scales. Laboratory results suggest that pellet stoves may be considered a transitional technology that “bridges the gap” between readily available traditional stoves that exhibit poor emissions performance (Tiers 0~2) and clean-burning gas stoves that perform well (Tier 5) but are not yet accessible for some populations currently reliant on solid fuels. The combination of Phlp-PH showed the lowest overall PM_2.5, CO, and THC EF_d values of all stove/fuel combinations tested. However, this combination also had the highest overall UFP EF_d value, highlighting an important characteristic of many pellet cookstoves: a trade-off between bulk PM_2.5 emissions and UFP number emissions. We encourage interpretation of these results to account for these nuanced results: pellet-fueled cookstoves offer emissions reductions of many health-damaging pollutants, but the relatively high emissions of UFPs warrant discussion with policy-makers and end-users. PM_2.5 emissions were highest from all stove types burning the WS pellets, and therefore the use of this pellet type is discouraged unless new technology can be developed to provide cleaner combustion. This finding highlights the need for future work to further characterize the most promising combinations of pellet stoves and fuels and underscores the importance of proper stove and fuel matching (as informed by laboratory- or field-based emissions testing). Finally, across stove/fuel combinations, we observed particle pollutants to generally increase with stove firepower, and gaseous pollutants (apart from CO₂) to decrease.

In addition to fuel efficiency and emissions performance, the design of an appropriate pellet stove must account for multiple practicalities including household size, cooking frequency and duration, cooking methods and foods cooked, and characteristics of the locally available fuel.¹³ As such, there can be no “one size fits all approach.” Field experience has highlighted the importance of collaboration between practitioners (e.g., on-the-ground entities doing stove implementation) and stove designers, as evidenced in Rwanda.¹⁶ However, the implementation of pellet-fueled cookstoves is inherently difficult in practice. Past initiatives have attempted to make pellet stoves cost-competitive with the local dominant fuel (e.g., charcoal),^{12, 16} but achieving in-country pellet manufacture at-scale can be challenging. Primary operating costs in pellet production are for feedstock, bagging, and manufacture,⁷⁶ and hopefully future operational advances may continue to drive pellet fuel costs down thereby promoting transitions away from dirtier solid fuel energy sources. Many users currently reliant on solid fuels have access to free/harvested biomass and therefore are less incentivized to use pellets, which are a commercial fuel subject to affordability and availability constraints. For example, a recent case study in Kenya observed substantial barriers to adoption of pellet-fueled cookstoves owing to lack of affordability.⁷⁷ Lastly, peer influence and perception (in terms of, e.g., reduced expenses, convenience, increased cleanliness, and improved health and safety) will continue to drive factors related to preferred cooking technology.^{12, 14}

Approaches to clean cooking require a community-scale lens with which to assess issues, as the focus on single-technology interventions will likely continue to be unsatisfactory in terms of reducing public health burdens.⁷⁸ We know that stove and fuel “stacking” is prevalent.^79–81 As such, pellet-fueled semi-gasifier cookstoves offer the potential to promote a “cleaner stack”, especially in more rural areas where transitions up the energy ladder/gradient are slower (e.g., decades vs. years). While urban areas have greater opportunity to develop markets for cleaner fuels, and peri-urban areas have both greater access to and awareness of modern cooking techniques,¹⁴ rural populations may benefit most from pellet-fueled cookstove technology during long-term energy transitions.

Synopsis:

Pellet cookstoves may “bridge the gap” between polluting traditional and clean electric/gas types, especially for rural populations. Here, we assess emissions and performance of twelve pellet stove/fuel combinations using the latest international test protocol.