Cookstove emissions and performance evaluation using a new ISO protocol and comparison of results with previous test protocols

Abstract

In 2018, the International Organization for Standardization (ISO) 19867–1 “Harmonized laboratory test protocols” were released for establishing improved quality and comparability for data on cookstove air pollutant emissions, efficiency, safety, and durability. This is the first study that compares emissions [carbon dioxide, carbon monoxide, total hydrocarbons, methane, nitrogen oxides, fine particulate matter (PM_2.5), organic carbon, elemental carbon, and ultrafine particles] and efficiency data between the ISO protocol and the Water Boiling Test (WBT). The study examines six stove/fuel combinations [liquefied petroleum gas (LPG), pellet, wood fan, wood rocket, three stone fire, and charcoal] tested in the same US EPA laboratory. Evaluation of the ISO protocol shows improvements over previous test protocols and that results are relatively consistent with former WBT data in terms of tier ratings for emissions and efficiency, as defined by the ISO 19867–3 “Voluntary Performance Targets.” Most stove types remain similarly ranked using ISO and WBT protocols, except charcoal and LPG are in higher PM_2.5 tiers with the ISO protocol. Additionally, emissions data including polycyclic aromatic hydrocarbons are utilized to compare between the ISO and Firepower Sweep Test (FST) protocols. Compared to the FST, the ISO protocol results in generally higher PM_2.5 tier ratings.

Graphical Abstract

Introduction

Approximately 40% of the world’s population rely on solid-fuel cookstoves that emit pollutants affecting health and climate.¹ Cooking with solid-fuels contributes to 12% of the global ambient fine particulate matter (PM_2.5) burden,² and to 25% of ambient black carbon,³ an important short-term climate forcer. Household air pollution (HAP) from solid-fuel use caused between 1.4 and 1.9 million premature deaths in 2017,⁴ while exposures to ambient PM_2.5 from household solid-fuel use resulted in another 0.3–0.5 million annual premature deaths.^2,5,6 Cookstoves emit a wide range of pollutants, including particle- and gas-phase polycyclic aromatic hydrocarbons (PAHs),^7–10 many of which are known or probable human carcinogens.¹¹ The health of those in low- and middle-income countries is disproportionately affected by HAP.^12–14 Implementation of clean cooking technologies not only improves indoor and outdoor air quality, but provides potential advances in at least ten of the seventeen Sustainable Development Goals.^15,16

Laboratory testing of cookstoves provides high-quality, reproducible data useful for evaluating cookstove emissions and performance metrics. The Water Boiling Test (WBT) protocol,¹⁷ beginning in 1982,¹⁸ has been extensively employed for laboratory cookstove testing;^19–21 currently, the Clean Cooking Catalog²² provides >770 unique WBT results for stoves ranging from the traditional wood-fueled three stone fire (TSF) to clean-burning liquified petroleum gas (LPG). WBT data have been applied for technology development and used as guidance for stove evaluation prior to more expensive field trials.²³ If laboratory data are confirmed by field data, then emissions test results can be confidently used for estimation of health benefits²⁴ and in regional and global air pollution modeling.^25,26

There are limitations associated with laboratory cookstove testing, and with the WBT in particular, including: (1) focus on boiling and simmering of water (i.e., lack of applicability to all cooking activities), (2) ambiguity in testing procedures (e.g., ignition practice, quality assurance/control guidelines, fuel characteristics, determination of boiling point and definition of end time) that may result in inconsistent results between laboratories/testers, (3) focus on emissions measurements from the entire test phase vs. specific portions that can produce outsized emissions contributions (e.g., ignition), and (4) different and idealized operation of the stove in a laboratory vs. field setting. These limitations explain, in part, the tendency for laboratory testing to underpredict emissions from non-ideal, real-world practices observed in the field.^27–29 Therefore, testing protocols that include, for example, varying stove firepower,³⁰ varying pot sizes,³¹ or comparisons to field-based “burn cycles”³² were proposed. For example, the Firepower Sweep Test (FST)³⁰ for continuous-feed wood stoves employs periods of start-up, sweep-down (i.e., decreasing firepower), sweep-up, and shutdown to explore the range of stove operation in terms of firepower. The FST results in both higher and more variable emissions compared to the WBT,³⁰ and improved representation of field data. Best practices from the FST and other protocols were considered in the development of an ISO (International Organization for Standardization) test protocol.

In development for more than four years and released in 2018, the ISO 19867–1 protocol was intended to unify best practices from existing protocols, and to provide a “standard test sequence to establish international comparability in the measurement of cookstove emissions and efficiency.” The ISO protocol was intended to improve harmonization in lab testing at multiple power levels, while reducing testing complexity. The associated ISO 19867–3 technical report employs a tier system³³ that provides a simplified way for non-experts (e.g., consumers, policy-makers) to rank stove performance. The ISO test protocol varies from the WBT and FST in some ways (as described below) and is intended to provide a practical and widely applicable laboratory test method. The ISO protocol is hypothesized to result in higher emissions of carbon monoxide (CO) and other pollutants compared to the WBT. To our knowledge, the present study is the first to examine this hypothesis by providing test results for the full ISO protocol and comparing it with WBT and FST protocols. More specifically, this study aims to: (1) compare gas- and particle-phase emissions and stove performance data for the ISO vs. WBT and FST protocols across a range of fuel and stove combinations, (2) highlight key differences between protocols that contribute to deviations in results, and (3) provide results in support of ongoing implementation and further development of cookstove standards in countries that may adopt or adapt the ISO protocol.

Methods

Cookstove/fuel combinations

Six stove/fuel combinations were tested as shown and described in Figure S1 and Section S1 of the Supporting Information (SI): (A) Mikachi stove burning LPG fuel, (B) Mimi Moto forced-draft semi-gasifier stove burning hardwood pellets, (C) Philips HD4012 forced-draft semi-gasifier stove burning cut red oak, (D) Envirofit G3300 natural-draft rocket stove burning cut red oak, (E) a “minimally tended” TSF burning cut red oak, and (F) Kenya Ceramic Jiko stove burning lump hardwood charcoal. These stoves are herein referred to respectively as: LPG, Pellet, Wood Fan, Wood Rocket, TSF, and Charcoal. The stoves were selected to: (1) span a range of current cookstove technologies, (2) test a representative model of each stove type, and (3) compare results of the same stove/fuel combinations previously tested using the WBT in the same facility,¹⁹ where applicable (as described below). Additionally, five of the six stove/fuel combinations (LPG, Pellet, Wood Rocket, TSF, and Charcoal) were compared against similar stove/fuel combinations previously tested using the FST.³⁴ A difference in the study comparison was that Bilsback and colleagues used the Philips HD4012 for the Pellet stove, whereas we used the Mimi Moto (both studies used hardwood pellets); the four other stove and fuel combinations matched.

Each cookstove was operated according to manufacturer instructions. Fuels are described in SI Section S2. Fuel moisture content was measured by oven drying using ASTM Method D4442–16. Fuel heat of combustion was measured by bomb calorimetry using ASTM Method D5865–13; per WBT and ISO specifications, the lower heating value was used in emission factor and efficiency calculations. Fuel analysis results are reported in Table S1. The cooking pots employed for testing are described in SI Section S3.

Test protocols

The ISO 19867–1 test protocol was employed in this study to determine cookstove power, energy efficiency, fuel use, and pollutant metrics.³⁵ This protocol employed test phases at three power levels (high, medium, and low), with each phase including a startup, an approximate 30-minute burning period, and a shutdown (Figure 1, upper panel). The three power levels were tested sequentially, and emissions sampled throughout the entirety of each phase included startup and shutdown.

Figure 1.

Schematic overview of ISO, WBT, and FST protocols. Schematics for the FST are for continuous-feed wood and batch-feed wood and charcoal stoves.

The WBT also consisted of three phases: high-power cold start, high-power hot start, and low-power simmer. Figure 1 illustrates key differences between the ISO and WBT, including (1) incorporation of a shutdown period in each phase for ISO, (2) incorporation of a startup period in the low-power phase (as opposed to continuation from high-power, hot start in the WBT), and (3) definition of test period durations based on time for all test phases for ISO vs. boiling point for WBT high-power phases.

The FST does not consist of distinct phases (although portions of the test may be defined for integrated sampling; e.g., filter sampling), but rather “sweeps” the firepower range of a stove to assess emissions and performance at both transition and pseudo-steady state operation. Figure 1 (bottom panel) shows idealized operation for a continuous-feed wood stove (solid line), and representative operation of batch-feed wood and charcoal stoves (dashed and dotted lines, respectively). The ISO and FST protocols were similar in that each incorporated startup and shutdown phases, were defined by approximate time intervals (as opposed to boiling point or water temperature), and spanned numerous stove power levels. However, the FST is effectively one continuous “sweep”, whereas the ISO protocol has distinct phases of steady operation at different power levels.

Both ISO 19867–1 and WBT v4 protocols were employed in this study and test-phase-specific emissions data were reported. Fuel burning rates (and subsequently firepower) for ISO testing were selected to represent reasonable upper and lower operating limits of the stove, and to align with those of the respective WBT phase for comparison (e.g., ISO high-power vs. WBT high-power, cold start).¹⁹ Pollutant emissions were measured and reported for each of the three ISO phases. To augment the literature WBT dataset used for Wood Fan, Wood Rocket, TSF, and Charcoal stoves,¹⁹ WBTs were conducted for LPG and Pellet stoves. Results for each phase are reported as averages with sample standard deviations, with test replicates of at least five as specified by the ISO protocol and at least three as specified by the WBT (see Table S2 for experimental matrix). The FST was not employed in the present study, but literature data for the overall FST³⁴ were used for comparison against ISO data generated here; discretized power level data for the FST were reported previously,³⁰ but for differing stove/fuel combinations, and thus not used in comparisons here. Additional descriptions of the test protocols are provided in SI Section S4.

Cookstove Test Facility

Testing was conducted at the US EPA’s Household Energy Laboratory in Research Triangle Park, NC, USA. A schematic of the emissions testing system and description of the facility are provided in the SI of our laboratory’s previous work.¹⁹ Briefly, stove emissions were collected into a stainless-steel hood connected to a dilution tunnel (~4.2 m³ min⁻¹) from which emissions 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.

Emissions characterization

CO, carbon dioxide (CO₂), total hydrocarbons (THCs), and methane (CH₄) were continuously monitored with nondispersive infrared (NDIR) and flame ionization detector (FID) analyzers (Models 600 NDIR/HFID, California Analytical, Orange, CA); nitrogen oxides (NO_x) were monitored with a chemiluminescence analyzer (Model 200EH, Teledyne, San Diego, CA). Continuous measurements were averaged and recorded every 5 s. PM_2.5 was sampled isokinetically and collected 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 the gas-phase artefact.³⁶ PM_2.5 mass was measured gravimetrically with a microbalance with a resolution of 1 μg (Model MC5, Sartorius, Göttingen, Germany or Model XP2U, Mettler Toledo, Columbus, OH). 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 analyzer (Sunset Laboratory, Tigard, OR) with a modified NIOSH protocol.³⁷ Annular glass sorbent tubes (6 mm OD × 4 mm ID × 178 mm, Sigma-Aldrich, St. Louis, MO) filled with 40 mm of Carbotrap-F, 20 mm of Carbotrap-C, and upstream QFF punches, were used to sample emissions for triplicate ISO high- and low-power phases for all stove/fuel combinations tested. Sorbent tubes containing the QFF punches were analyzed using a thermal desorption gas chromatography mass spectrometer (TD/GC/MS) (Gerstel TDS2, MD, USA) that reports mass of gas- and particle-phase (i.e., total) PAHs sampled. The GC/MS analysis was conducted in single ion monitoring (SIM) mode, and results are reported for 26 compounds. This approach reduced the total number of samples requiring analysis and simplified data work-up for the semi-volatile organic compounds (SVOCs).

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). A tapered element oscillating microbalance (TEOM; Model 1405; Thermo Fisher; Waltham, MA) quantified real-time PM_2.5 mass emissions for the ISO tests. Emissions of all pollutants were quantified using the total-capture mass-flow method with continuous monitoring of the dilution tunnel air flow, temperature, pressure, and relative humidity over the test period. The carbon-balance method³⁸ was utilized to compare carbon measured in fuel and emissions for internal quality assurance.

To explore real-time trends (and define which periods contributed to differences between protocols), instantaneous emission rates (IER_P = QΔ[P]) were calculated and analyzed; IER_P (μg s⁻¹) is the instantaneous emission rate for pollutant P, Q is the duct flow rate (m³ s⁻¹), and Δ[P] is the background-corrected mass concentration of pollutant P (μg m⁻³). IERs were normalized with respect to both test duration and mass of pollutant emitted for CO, THC, PM_2.5, and UFP (ultrafine particle, number) using an approach similar to prior laboratory and field studies.^39,40 In brief, IERs were normalized to total pollutant emitted ($[IE{R}_{norm,P,t}=IE{R}_{P,t}/{\sum }_{t_0}^{t_f}\left(IE{R}_{P,t}\right)]$, where IER_P is as defined above, P is the pollutant, t is a specific point during the test duration, t₀ is test start, and t_f is test finish) and then integrated to develop cumulative distribution functions (CDFs). These CDFs were not employed for the FST comparison, as these real-time data were not available.

Detection limits

Detection limits (DLs) were calculated using an approach described previously.⁴¹ Briefly, DLs were defined as three times the standard deviation of the noise in the background concentrations measured for at least 10 mins before and after tests for gaseous pollutants. For particle mass measurements, a similar method was employed using daily background filter sample masses. Results averaged from replicate tests were calculated as follows: (1) if a background-corrected emission metric was ≤zero, it was included in the average as a zero, (2) if a metric was greater than zero and ≤DL, it was averaged in conservatively as the DL itself, and (3) if a metric was above the DL, it was averaged in as its calculated value. This approach differs from that of Bilsback et al.,³⁴ where the imputed value of DL/√2 was used for those values meeting the second criterion above. Imputation, or non-imputation, of non-detects is effectively arbitrary,⁴² and our approach allows reporting of statistics for stove/fuel combinations with some replicate values below the DL (e.g., LPG and Pellet, as shown in Tables S3 and S4) and provides a conservative estimate of emissions metrics for these cleaner-burning stoves. DLs are plotted along with metrics, as described in Section S5, in Figures S2–S7.

Results and Discussion

Stove, fuel, and test protocol comparisons using the voluntary performance target system

As discussed, non-experts need a method to compare stove performance. Thus, Figure 2 introduces the ISO 19867–3 ‘Voluntary Performance Targets’.³³ The lowest tier, Tier 0, represents ‘performance typical of open fires and the simplest types of solid-fuel cookstoves’, while the highest tier, Tier 5, represents best performance (i.e., higher tier = lower emissions). For example, Tier 5 for PM_2.5 represents emissions modeled to result in ≥90% of homes (using the default assumptions in the ISO indoor air quality model) meeting the WHO annual mean air quality guideline (10 μg m⁻³), a level associated with minimal adverse health risk; Tiers 1–4 represent incremental improvements over the baseline case of Tier 0. Placement of emission factors (EFs) within the tier system is defined by the upper bounds of 90% mean confidence intervals; the FST Pellet CO EF is a single value however, as replicate data were unavailable. It is important to note that while the former ISO International Workshop Agreement (IWA) performance tiers specified that a stove must meet thresholds for both CO and PM_2.5 to be classified (e.g., as a “Tier 4 stove for emissions”), the ISO 19867–3 tier system specifies separate tiers for each pollutant. Also, the ISO tier system is intended for use with averaged values obtained from testing at all power levels, as in Figure 2b, but values for high-power only are shown in Figure 2a for the purpose of comparing against WBT results. Lastly, the ISO tier system is intended for use with ISO test data (and not other protocols), though results obtained with other protocols are shown here for comparison.

Figure 2.

CO and PM_2.5 EF_d values for (a) ISO (high-power) and WBT (mean of high-power CS and HS), and (b) ISO (overall) and FST (overall) rated with the ISO 19867–3 ‘Voluntary Tiers of Performance’. Confidence intervals (90%), as specified in ISO 19867–1, using pooled standard deviations are represented by the error bars.

Figure 2a plots CO and PM_2.5 EFs based on useful energy delivered to the cooking pot (EF_d) for the ISO high-power phase and the high-power WBT phases (mean of cold start (CS) and hot start (HS) phase test data, as defined by the IWA tier system for WBT data); these data are summarized in Tables S5 and S6. CO emissions produced using both protocols show the LPG, Pellet, and Wood Fan stoves in Tier 5, and the Wood Rocket, TSF, and Charcoal stoves in Tiers 3, 1, and 0, respectively. PM_2.5 emissions using both protocols show Pellet in Tier 4, Wood Fan in Tier 3, Wood Rocket in Tier 1, and TSF in Tier 0; although, the ISO protocol (vs. WBT) did increase the tier rankings for the LPG (Tier 5 vs. 4) and Charcoal stoves (Tier 3 vs. 1). Therefore, tier levels vary by protocol for the most advanced stove type (LPG) and one of the traditional stove types (Charcoal): the LPG and Charcoal stoves show lower average PM_2.5 emissions with the new ISO protocol. This observed difference is likely due to the typically higher start-up emissions averaged over a longer test phase duration with the ISO protocol.

Figure 2b tiers CO and PM_2.5 EF_d values for the overall (i.e., mean of all power levels) ISO and FST test data (Tables S5 and S6). For CO, the LPG and Pellet stoves are Tier 5 using both protocols, while TSF and Charcoal stoves are Tiers 1 and 0, respectively. Only the CO ranking for the Wood Rocket stove varies by test protocol; it is Tier 3 using ISO and Tier 0 using FST. Note that FST data are not available for the Wood Fan stove, which is Tier 4 using the ISO protocol. For PM_2.5, only one stove (TSF) is the same tier using both protocols (Tier 0). Figure 2b shows that ISO protocol results are a single higher tier for PM_2.5 emissions for the four other stove/fuel combinations tested with comparable ISO and FST data. Therefore, the ISO protocol results in generally lower PM_2.5 emissions results compared to the FST. These observed differences between protocols are likely due to the operation of cookstoves at more nearly steady-state conditions per the ISO protocol compared to more transitional conditions per the FST. It is important to note however that FST sampling (for PM and PAHs) excluded start-up and shutdown periods, likely reducing overall emissions metrics for these pollutants.

Comparison of ISO and WBT results for stove/fuel combinations

While tiering is useful, it is limited (e.g., focus on target pollutants only, small number of tiers), and further investigation into differences in data produced using different protocols is warranted. Inclusion of shutdown periods for the ISO protocol was expected to increase CO emissions compared to the WBT, while the startup period in the ISO low-power phase was expected to increase gas- and particle-phase emissions. The relatively longer test phase durations, especially for higher power stoves where the WBT high-power phases would end sooner, were expected to result in high-emitting events being averaged over longer time periods, and hence lower average emissions. Our study was designed to test these assumptions.

We compare performance and emissions metrics between the ISO and WBT protocols using two-way Wilcoxon rank sum tests within stove and fuel combinations (Figures S8–S20); this statistical test was chosen over the Student’s t-test due to the relatively limited number of test phase replicates per stove/fuel combination. Statistical significance is defined here as p ≤ 0.05; however, note that many significant but relatively small differences were observed. For ease of discussion across protocols, the first phases of both protocols are referred to as “high-power”, and the third phases as “low-power”; the second phases (ISO medium-power and WBT high-power, hot start) are not compared given their inherent firepower differences.

Figure S8 and Table SS1 (of the SI summary spreadsheet) show box-and-whisker plots and a summary of fuel burning rates for ISO and WBT testing. Fuel burning rate is a metric required by ISO 19867–1 to document test conditions. Only the Pellet stove showed statistically significant (but small) differences in average burning rates between ISO and WBT testing during high- and low-power phases. Pellet stove burning rate and firepower are largely controlled by stove fan speed and combustion chamber diameter as opposed to fuel loading given the nature of fixed-bed combustion.^43,44 TSF and wood rocket stove fuel burning rates are largely controlled by fuel feeding by the operator. Average fuel burning rates for all stoves may be affected by differences between protocols in the duration of test phases.

Figure S9 and Table SS2 show modified combustion efficiencies (MCEs) [defined as net CO₂/(CO+CO₂) on a molar basis]. ISO 19867–1 requires measurement of CO and CO₂ for emissions testing, and although MCE is not a required metric, it is recommended as a useful proxy indicator of actual combustion efficiency for evaluating stove performance. During the high-power phase, high CO emissions measured during shutdown with the ISO protocol decrease test-mean MCE. In the low-power phase, though char typically remains from the WBT high-power hot-start phase for natural-draft stoves, it is consumed through flaming combustion and it is again inclusion of the shutdown that likely contributes to lower observed MCEs using the ISO protocol. Median MCEs were lower for all stoves using ISO vs. WBT in both high- and low-power phases, and MCEs were significantly lower with the ISO protocol for all stoves except LPG.

Figures S10–S11 and Tables SS3–SS4 summarize thermal efficiencies of the stove/fuel combinations. Interestingly, the forced-draft (Pellet and Wood Fan) and Charcoal stoves had slightly higher thermal efficiencies (including energy credit for char) with statistically significant differences with the ISO protocol during the high-power phase compared to WBT, potentially due to additional heat capture by the cooking pot during shutdown. Thermal efficiency is not measured in the WBT simmer phase. Emissions and performance data are analyzed further below, and direct comparisons are made between high- and low-power phases of each protocol.

Figures 3 and 4 plot EFs based on dry fuel mass (EF_m) for the ISO and WBT high- and low-power phases; note that the high energy density of LPG results in some EF_m values in the same range as those of solid fuel stoves. Previous WBT test results¹⁹ utilized particle size distributions (from 14.6–661 nm) from a scanning mobility particle sizer (SMPS; Models 3080 and 3010; TSI; Shoreview, MN). Given the differences in aerosol size ranges and bin widths for the SMPS and the EEPS used in this study, UFP emissions (on a particle number basis) over the ranges of 14.6–100 nm and 14.3–100 nm (herein referred to as 14–100 nm) are reported for the SMPS and EEPS, respectively.

Figure 3.

Summary of emissions factors for ISO and WBT protocols for pollutants measured in both the present study and by Jetter et al., ¹⁹ excluding CO (see Figure 4). Values of test phase replicates are plotted for each test protocol and power level.

Figure 4.

Summary of CO emission factors for ISO and WBT protocols using data from both the present study and Jetter et al.¹⁹ Values of test phase replicates are plotted for each test protocol and power level.

Data generally align between protocols, with differences for each stove/fuel combination discussed in SI Section S8. The influence of protocol on emissions and performance data is expected to be highly dependent upon the type of combustion that occurs with different fuels and stove types (e.g., forced vs. natural draft, batch- vs. continuous-feed). Statistically significant differences between protocols in terms of individual emission metrics (e.g., CO EF_m) are also discussed in SI Section S8 and tabulated in Table S7. Figures S12–S20 and Tables SS5–SS31 show emissions metrics [emission rate (ER), EF_d, and EF_m] for CO₂, CO, THC, CH₄, NO_x, PM_2.5, OC, EC, and UFPs. Note that comparisons for EF_d are not available between low-power phases, because useful energy is not accurately measured during the WBT simmer phase.¹⁹ Emissions of OC, EC, and NO_x were not previously measured for the Wood Fan, Wood Rocket, TSF, and Charcoal stoves.¹⁹

Emission mass ratios, CO:CO₂ and PM_2.5:CO₂, versus firepower are plotted in Figure S27. Linear relationships are shown for stove/fuel combinations using both ISO and WBT data. CO:CO₂ mass ratio vs. firepower relationships were significant (p ≤ 0.05) and positive for LPG and Charcoal stoves, and significant and negative for Pellet and Wood Fan stoves. PM_2.5:CO₂ mass ratio vs. firepower relationships were significant and positive for Pellet, Wood Rocket, Wood Traditional, and Charcoal stoves. It is interesting to note the different relationships for some stoves/fuels, such as CO:CO₂ increases with firepower for LPG and decreases with firepower for Pellet. This phenomenon is likely due to variation in combustion efficiency over the power range of cookstoves.

Figure S28 plots trends in ERs for UFP vs. PM_2.5; only the Wood Fan stove had a significant (and positive) relationship during high-power operation, while both traditional stove types (TSF and Charcoal) had significant (and positive) relationships during low-power operation. Figure S29 plots PM_2.5 determined by the QFF (defined as EC plus estimated organic matter (OM) using OM:OC ratios of 1.2 for LPG and charcoal, and 1.5 for wood and pellet stoves)³⁴ vs. PM_2.5 as determined by PTFE filters; data between filter sampling methods are in relative agreement (m = 1.08 and 0.85, and R² = 0.98 and 0.99 for high- and low-power phases, respectively).

Comparison of ISO and FST results

Figure 5 plots pollutant EF_d values measured in the present study using the ISO protocol, and EF_d values from Bilsback et al.³⁴ using the FST protocol. The FST, like the ISO, probes stove performance and emissions at multiple power levels, aims to provide improved alignment of laboratory and field test data, and has been the focus of recent testing.^30,34 Both datasets (ISO and FST) include results for total (i.e., gas- and particle-phase) PAHs. Exposures to ambient and household PAHs have been extensively linked to health effects,⁴⁵ and PAHs in cookstove particle emissions have been correlated to mutagenicity potential.^46–48 In addition to the 11 PAHs analyzed in both studies, EF_d data for 15 additional PAHs are reported in SI Section S11.

Figure 5.

Summary of emissions factors for ISO and FST protocols for pollutants measured in both the present study (circle markers) and Bilsback et al.³⁴ (square markers). Values of test phase replicates are plotted for each study. Note that zero values are not shown due to use of log scales.

The data generally indicate consistency between the two protocols, with the following notable observations: (1) pollutant emissions are generally lower for LPG stoves using the ISO protocol (aligning with observed difference in tier levels; Figure 2b), and (2) PAH emissions are generally lower for Pellet and Charcoal stoves using the ISO. Differences in PAH (and PM) data may be due to sampling and analysis method variables. To our knowledge, this study is the first to report PAH emissions using the filter-in-tube method. Bilsback et al.³⁴ sampled gas- and particle-phase PAHs using polyurethane foam plugs and QFFs, respectively, and reported the sum as total PAHs. Additionally, integrated samples for the FST data were collected during the pseudo-steady state periods and excluded startup and shutdown periods, whereas samples in the present study were collected for entire ISO high- and low-power phases. Exclusion of these transitional periods in the FST data likely resulted in lower emissions estimates.

The Wood Rocket stove had the highest mean test phase-specific PAH EFs of the stove/fuel combinations tested (Figure S30; Tables S8–S9) when reported on an energy delivered basis, while the TSF had the highest reported on a mass fuel consumed basis. The higher flame temperatures of the Wood Rocket stove compared to the TSF may explain the increased PAH emissions, as observed previously.³⁴ A previous laboratory study of cookstoves¹⁰ found total PAH emissions to range between 24–114 mg kg⁻¹ for biomass fuels (fuel wood and rice husk briquettes), effectively overlapping EF_m values observed here for the Pellet, Wood Fan, and TSF stoves. A field study of rural wood and coal burning stoves⁷ determined the mean (± standard deviation) sum of total PAHs (22 compounds) to be 74.2 (± 50.1) mg kg⁻¹, in close agreement with our results for the TSF (81.7 mg kg⁻¹).

Napthalene was the predominant PAH measured for Wood Fan, Wood Rocket, TSF, and Charcoal stoves, while phenanthrene and fluoranthene were predominant for LPG, and phenanthrene and napthalene for the Pellet stove. The importance of measuring total gas- and particle-phase PAHs is apparent, as in our previous study⁸ measuring only particle phase PAHs, pyrene was found to be predominant for forced- and natural-draft stoves. Previous laboratory studies have determined gas-phase fractions of all PAHs measured to exceed 86%,¹⁰ and that for napthalene in particular to exceed ~95%;³⁴ therefore, it is recommended that studies sample for both gas- and particle-phase PAHs. Additionally, gas-phase PAH emissions are important owing to their potential to form secondary organic aerosol (SOA).⁴⁹

Lower volatility compounds dominate contributions to the PAH matrix for the LPG (ISO high- and low-power) and Pellet (ISO high-power only) stoves (Figure S31). Higher volatility compounds that likely exist in a gas-phase at ambient conditions (e.g., naphthalene) are more prevalent for the biomass stoves (including the Pellet stove at ISO low-power operation). Therefore, the cleaner-burning stoves (LPG at both power levels and Pellet at high-power) emit lower quantities of SOA precursors, in agreement with previous work.^49,50 Regarding health effects, the 16 EPA priority PAHs (designated as such for their human toxicological effects) accounted for the majority of total PAH emissions from all stove/fuel combinations. Percent contributions of priority PAHs to total PAHs at high-power operation were highest for the Pellet stove (95%) and lowest for LPG (65%); at low-power operation, the forced-draft stoves (Pellet, Wood Fan) and Wood Rocket had the highest percent composition of priority PAHs (89%), while Charcoal had the lowest (72%). It is important to note however, that health effects will vary based on exposure route (e.g., gas- vs. particle-phase).

All stove/fuel combinations, apart from the Pellet stove, had lower total PAH EF_d values during the low- vs. high-power phase (Figure S30; Tables S8–S9); there were 3.8-, 1.1-, 3.5-, 3.0-, and 45-fold decreases during low- vs. high-power operation for the LPG, Wood Fan, Wood Rocket, TSF, and Charcoal stoves, respectively. Therefore, the Charcoal stove was most influenced by operation in terms of total PAH emissions. Notably, the Pellet stove had 6.1-fold higher PAH EF_d values during low- vs. high-power operation, despite a corresponding 1.9-fold decrease in PM_2.5 EF_d values. Further research would be needed to better understand PAH formation mechanisms, but during low-power operation, the Pellet stove’s lower fan speed may result in less mixing of combustion gases and less burn-off of gas-phase PAHs in secondary combustion. Interestingly, naphthalene (expected to exist in the gas-phase at ambient and sampling conditions) accounted for 51% of all PAHs by mass for low-power vs. 12% for high-power. Compared with high-power operation, low-power operation of the Pellet stove resulted in both higher overall PAH emissions, as well as a higher proportion of likely SOA precursors.⁵¹ Nevertheless, Pellet PAH emissions at both power levels were much lower than Wood Rocket and TSF emissions, as shown in Figure S30.

Study limitations and future work

This study was intended to examine potential implications of the ISO protocol, rather than extensively characterize each stove/fuel combination. Therefore, a limitation of this work is that only one stove model per stove type/family was tested. Future work is needed to explore performance of a range of stove models using the ISO protocol to expand the limited ISO emissions database. An additional limitation of this study was the matching of fuel burning rates to previous WBT data, as arguments have been made for the need to test stoves at “high…even ‘overfueled’ conditions”.³⁰ Given the relationship between emissions and firepower and the tendency for in-field stove users to operate stoves at higher-than-intended power levels, future work is needed to evaluate these stove types at additional power levels.

Study Implications

Although performance in terms of tier ratings was generally similar between ISO and WBT protocols (apart from the LPG and Charcoal stoves for PM_2.5), there were statistically significant differences in emissions metrics for most of the stove/fuel combinations tested. In particular, the advanced wood and charcoal combinations were most influenced, due likely to inclusion of the shutdown phase in the ISO protocol; gasifier stoves are sensitive to operating practice and can emit markedly higher CO during shutdown.^40,44,52 When compared with the FST, the ISO protocol generally produced lower PM_2.5 emissions with tiers typically one level higher (towards better performance) for all stoves tested except for the TSF; tiers for CO are generally unaffected, apart from the Wood Rocket stove. FST PM_2.5 emissions results are both higher and more variable than the ISO results, in agreement with previous observations of the FST vs. WBT.³⁰ Implications from this finding are that the ISO protocol can provide more controlled testing conditions for repeatability and comparability of results, while the FST can provide less controlled conditions for better simulating field performance.

Wood Rocket stoves had the highest total PAH EF_d values of the stove/fuel combinations tested at both ISO high- and low-power levels. Regarding future development and implementation of advanced biomass stoves, trade-offs among PM mass, UFP number, and total PAH emissions should be considered in terms of their potential health effects. Compared with larger sized particles, UFPs are considered highly reactive and promote increased cellular activity (e.g., inflammation, translocation).^53–55 Additionally, many PAH are carcinogenic.¹¹

In addition to differences in cooking activity (i.e., boiling water vs. cooking food), key discrepancies in emissions results from controlled laboratory testing can be due to the burn cycle.⁵⁶ Real-time emissions data show distinct combustion events producing outsized pollutant contributions, especially for advanced biomass stoves. For example, higher CO IERs were observed for pellet stoves for both startup and shutdown phases during field testing in Rwanda.⁴⁰ Bilsback et al.³⁰ also identified the importance of individual ‘transitional’ combustion events that occur during cooking (e.g., refueling, shutdown). Thus, combining the startup, steady-state operation, and shutdown phases for integrated filter sampling during each test phase is an ISO protocol limitation. However, the ISO protocol is intended for practical widespread use and testing is facilitated by this approach.

Finally, discrete power level emissions metrics may be employed to develop country- or even community-scale EFs based on power level weightings specific to cooking practices (e.g., tendencies to cook at lower vs. higher power). The phases of each protocol to some extent simulate the wide variety of cookstove uses (e.g., high-power represents boiling water or frying, low-power represents simmering). However, significant efforts would be needed to prepare an accurate “regional cooking practices database.” Findings from this study support ongoing implementation and further development of international standards for cookstoves in many countries.

Synopsis.

Results of this study inform further development and implementation of international standards for cookstoves to reduce air pollution and fuel use.