Realistic operation of two residential cordwood-fired outdoor hydronic heater appliances—Part 3: Optical properties of black and brown carbon emissions

ABSTRACT

Residential biomass combustion is a source of carbonaceous aerosol. Inefficient combustion, particularly of solid fuels produces large quantities of black and brown carbon (BC and BrC). These particle types are important as they have noted effects on climate forcing and human health. One method of measuring these quantities is by measurement of aerosol light-absorption and scattering, which can be performed using an aethalometer and nephelometer, respectively. These instruments are widely deployed in the study of ambient air and are frequently used in air quality modeling and source apportionment studies. In this study, we will describe (1) a method for measuring primary BC and BrC emissions from two residential log-fired wood hydronic heaters and (2) the BC and BrC emission from these devices over a wide range of operating conditions, such as cold-starts, warm-starts, four different levels of output ranging from 15% to 100% maximum rated output, and periods of repeated cycling. The range in flue-gas BC concentrations, measured using an aethalometer at the 880 nanometer (nm) wavelength, were between 5.09 × 10² and 2.24 × 10⁴ micrograms per cubic meter (µg/m³) while the scattering coefficient of the flue-gas, measured by a nephelometer at 880 nm, ranged between 2.20 × 10³ and 8.56 × 10⁵ inverse megameters (Mm⁻¹). The BrC concentrations, measured using the 370 nm wavelength of an aethalometer, were between 9.10 × 10¹ and 3.56 × 10⁴ µg/m³. The calculated Angstrom Absorption Exponent (AAE) of the flue-gas aerosol ranged between 1.54 and 3.63. Performing a comparison between the measured BC concentration and an external particulate matter (PM) concentration showed that overall BC makes up roughly a quarter of the PM emitted by either of the two appliances. Further for both appliances, the cold-start and the test phase immediately following it had the highest BC and BrC concentrations, the highest measured scattering coefficient, as well as a low AAE.

Implications: In this work we provide information on the black and brown carbon emissions from outdoor cordwood-fired hydronic heaters. Aethalometer based black carbon measurements are common in atmospheric science, but are uncommonly used in laboratory studies. This work helps to bridge that gap. This data helps to inform the work of modelers and policy makers interested in hydronic heaters and source apportioning biomass combustion emissions.

Introduction

Residential biomass combustion is a widely overlooked source of carbonaceous aerosol. Inefficient combustion, particularly of solid fuels, produces large quantities of black and brown carbon (BC and BrC). Both of these compounds are important as they have noted effects on climate forcing (Bond et al. 2013; Saleh et al. 2015) and human health (Grahame, Klemm, and Schlesinger 2014; WHO Regional Office for Europe 2012).

The health effects of exposure to BC are largely intertwined with the health effects of Particulate Matter (PM), which include excess mortality and cardiovascular and respiratory distress (Dockery et al. 1993; Schwartz and Dockery 1992). BC and BrC are principal components of combustion derived PM, which makes separation of these quantities from bulk PM challenging for epidemiological studies. Even so, within these studies, effect estimates for BC are higher than for PM_2.5 and PM₁₀ in general on a mass basis (units of microgram per cubic meter, µg/m³), particularly for short-term health effects (Janssen et al. 2011).

Residential biomass combustion has been shown to be responsible for elevating ambient BC, BrC, and PM in wintertime according to source apportionment studies (Allen and Rector 2020; Blanchard et al. 2021; Liu, Yan, and Zheng 2018; Mousavi et al. 2019; Rajesh and Ramachandran 2017). BC source apportionment can be performed in a variety of ways such as radiocarbon, macro-tracer, and aethalometer-based methods (Briggs and Long 2016; Favez et al. 2010; Peter et al. 2017; Sandradewi et al. 2008). Of these, the aethalometer-based method is especially useful as it requires fewer measurable quantities to come to a result than more sophisticated models.

At its core, aethalometer-based source apportionment uses reference values for Angstrom Absorption Exponent (AAE), which describes the wavelength dependence of an aerosols light absorption, to separate aerosols from traffic and biomass combustion (Prevot et al. 2008). This delineation is then used to quantify the amount of carbonaceous aerosol from each source in ambient air. Typically, an AAE value of 1 is used for traffic and values of ~2 are used for biomass combustion, specifically ranging from 0.9 to 3.5 (Brown et al. 2008; Helin et al. 2021; Kirchstetter, Novakov, and Hobbs 2004; Peter et al. 2017; Prevot et al. 2008; Saleh et al. 2013; Schnaiter et al. 2003). However, these choices must be made on a site-by-site basis as the composition of the aerosol from local sources, and the AAE of those aerosols can play a major role in the final apportionment (Helin et al. 2021). One reason for the variability in woodsmoke AAE, is that the composition and magnitude of PM emission from even a single source can vary based on combustion conditions.

Aethalometer data can also be used to estimate the fraction of BC in an aerosol. In most laboratory combustion emission studies the quantification of BC and BrC is performed on the basis of elemental and organic carbon (EC and OC), which is measured using a thermo-optical method (Karanasiou et al. 2015; WHO Regional Office for Europe 2012). However, this method requires collection of an integrated sample and as a result, the measurements cannot be parsed sufficiently to measure transient combustion conditions in many cases (Karanasiou et al. 2015). Recently, an aethalometer based method for near real-time EC and OC analysis has been proposed (Rigler et al. 2020). This method makes use of an aethalometer and a separate instrument to measure PM concentration. By comparing the two results, it is possible to derive an estimate of the BC fraction of the aerosol.PM emitted from biomass combustion appliances can vary due to factors such as appliance design, fuel type, size, and composition, and appliance operating conditions such as temperature, excess air, heat output condition, and burn rate (Johansson et al. 2004; Lillieblad et al. 2004; Obernberger, Brunner, and Barnthaler 2007; Md. Obaidullah, Verma, and De Ruyck 2012; Kinsey et al. 2012, Shen, Influence of fuel mass load, oxygen supply and burning rate on emission factor and size distribution of carbonaceous particulate matter from indoor corn straw burning 2013; Schmidl 2011; Vicente et al. 2015; Kortelainen et al. 2018) and it follows that the optical properties and BC emission from biomass combustion devices will also vary with these factors (Martinsson et al. 2015).

In this study, we (1) show a method for measuring real-time primary BC and BrC emissions from the flue stack of two residential cordwood-fired hydronic heaters (HHs) and (2) describe the BC, BrC, scattering coefficient, and aerosol AAE over a wide range of operating conditions, including cold-starts, warm-starts, four different levels of heat output ranging from 15% to 100% maximum rated output, as well as during periods of repeated cycling. Particulate emissions in terms of number concentration and particle size from the same set of experiments are discussed in (Lindberg et al. 2022) and gaseous pollutant and particulate mass emissions and emission rates are discussed in Trojanowski et al. (2022). The results we show represent a direct measure of BC emission from biomass hydronic heating appliances at different operating conditions and as an indicator of the possible range of emission and AAE values reflecting changes in duty-cycle.g

Materials & methods

In this study, we investigated two wood-fired hydronic heaters operating with combustion conditions typical in the field. The heating appliances are representative of modern biomass hydronic heaters used in the northern United States (U.S.). The fuel used in each test was red oak cord wood prepared by the State University of New York (SUNY), College of Environmental Science and Forestry (ESF) using a partial kiln drying procedure (Smith 2014). The specific test method which the appliances operated under is outlined in “A Test Method for Certification of Cord Wood-Fired Hydronic Heating Appliances Based on a Load Profile: Measurement of Particulate Matter (PM) and Carbon Monoxide (CO) Emissions and Heating Efficiency of Wood-Fired Hydronic Heating Appliances” (Northeast States for Coordinated Air Use Management. 2021). Optical measurements of Particle Mass Concentration (PMC), and particulate BC and BrC concentrations were made using a custom-built dilution sampling system. A Thermo Scientific personal Data Ram Model 1500 (pDR) and Magee Scientific Aethalometer Model AE33 (AE33) were used to measure optical PMC and BC and BrC, respectively. The data from these instruments were applied further, to determine the aerosol AAE, and to estimate the relative BC and BrC contribution to an external gravimetric measure of PM concentration.

Appliances, fuel, and test method

The wood-fired hydronic heaters investigated in this study are residential scale biomass units representative of devices certified by EPA for sale in the U.S. Overall, both appliances had reported mass emission rates of 30–32 mg/MJ with similar maximum rated heat outputs, between 30 and 35 kW. The reported thermal efficiency of the appliances was 67–68%. Both appliances featured two-stage combustion, with the first stage being a gasification step. Gasification combustion is a two-stage process; wherein wood gas produced in the primary firebox by inadequately oxygenating the coal bed followed by a secondary combustion process using wood gas as fuel (van Loo and Koppejan 2008). The second combustion stage of each appliance was different. The second stage for hydronic heater A (HH A) was catalytic combustion, while hydronic heater B (HH B) relied on a standard combustion techniques. The fuel used in each experiment was red oak cordwood. Red oak was chosen, due to its high heating value, low volatiles content, prevalence in the northeast (Fine et al. 2001). The moisture content and weight of each piece of fuel was measured prior to testing to ensure the proper amount of fuel was used in each test section and that the average moisture content of each fuel charge was between 19% and 25% on a dry basis. The test protocol includes ten distinct phases in which the appliance is fueled three times, in order to test multiple permutations of fuel loading density and heat output. A graphical representation of the heat output and firebox loading density is shown in Figure 1 and descriptions of each burn phase are given in Table 1, adapted from Trojanowski et al. (2022).

Table 1. Multi-phase testing protocol.

Phase	Description	Milestone to Signify End of Phase
Cold Start Fuel charge -kindling + starter fuel
1	Cold Start: Includes kindling (10% of fuel charge weight) made up of 50% small pieces and 50% larger kindling pieces	Once 80% of kindling fuel has been consumed
First fuel charge
2	Ramp-up to full load	Once 20% of first fuel charge has been consumed during this phase
3	Maximum burn rate	Once 20% of first fuel charge has been consumed during this phase
4	Steady state – less than 25% load	1.25 hr
5	Steady state – less than 50% load	Once 20% of first fuel charge has been consumed
6	End phase – burn out	Once 85% of the first fuel charge has been consumed, establishing a coal bed of 15% of the first fuel charge
Second fuel charge
7	Reload – equivalent of a hot start	Only 10% of second fuel charge has been consumed during this phase
8	Steady state – less than 15% load	Greater of 2.0 hr or time required for 1 reheat firing cycles
9	Steady state 100% load	Only 20% of second fuel charge has been consumed during this phase
10	Cyclic operation (10 min off, 5 min on)	One the HH temperature has returned back to its starting point of temperature set-point reached
End of sampling		Let remainder of fuel burn out at maximum load – no sampling

Figure 1. Graphical representation of appliance output and fuel load, fuel mass is shown as a red-dashed line, and appliance output is shown as a black-solid line.

This testing protocol was designed to test a wide range of operating conditions which are prevalent in use by a consumer. The test protocol is explained in detail in the test method, which is available online (Northeast States for Coordinated Air Use Management. 2021). Information specific to the application of the method to the two appliances for this study is given in Trojanowski et al. (2022). However, we will present a summary of the method here in addition.

The test protocol has 10 distinct burn phases. Fuel is added to the appliance during Phases 1, 2, and 7 the amount of fuel is different for each fuel charge. Phase 1 can be thought of as a cold-start condition, during Phase 1 a small fuel charge is added to an empty firebox and the fuel is ignited. In Phase 2, a large fuel charge, sized nominally to fill the combustion chamber, is added on top of the fire started in Phase 1 and the appliance is allowed to come up to typical operational temperature. Phases 3, 4, and 5 are steady-state output conditions where a heat demand of 100%, 25%, and 50% of the HH maximum rated output are applied, respectively. Phase 6 can be referred to as a “burnout”; in this phase, the appliance is directed to fire at maximum output to develop a coal bed onto which the next fuel charge can be added. In Phase 7 a large fuel charge is added on top of the existing coalbed, this fuel charge is sized to overfill the firebox. In Phases 8 and 9 a heat demand of 15% and 100% of the maximum rated output of the HH are continuously applied. Phase 10 includes repetitions of 10-min off, where no heat demand is applied, and 5-min on, where a demand of the maximum rated output is applied. This section tests how the appliance handles rapid changes in heat demand. Over the course of an experiment, the appliance’s control system runs continuously to modulate the firing rate to meet the imposed heat demand. As such the emissions are dependent on the control systems response to the test method. This experimental procedure was performed in triplicate on both appliances. Six tests were performed in all. For more detailed information on the appliances, fuel specifications, and test method please refer to Trojanowski et al. (2022).

Instrumentation

The experiments discussed in this study featured simultaneous measurements of pollutant gas, PM mass, PM number, and optical measurements of BC and BrC. The particle number concentration emission metrics and optical carbon compositional measurements were made using a dilution sampling system in tandem with the actual measurement instrumentation. A schematic of the experimental setup is shown in Figure 2.

Figure 2. Schematic of combustion test setup featuring locations of instrumentation.

Measurements of the gaseous pollutant and PM mass emissions were made directly in the flue stack, and in the dilution tunnel, respectively. The measurements of these quantities are discussed in greater length in Trojanowski et al. (2022). Similarly, the particulate matter number concentration data is also discussed separately in Lindberg et al. (2022). Following this delineation, there are three components of interest in Figure 2. The dilution system, the pDR and the AE33.

The dilution system consisted of a sampling probe, a two-stage ejector dilutor, and ancillary dilution air generation and conditioning equipment. In all tests, a 1 liter-per-minute (LPM) critical flow orifice was installed in the dilution system to limit sample flow rate and to provide a point to monitor the inlet flow via pressure drop for occlusion. The dilution ratio of the system was set to 108:1 using the primary and secondary dilution air controls. Variations in dilution factor occurred as a result of occlusion within the sample flow orifice. In order to minimize this effect, the dilution factor was measured throughout testing by periodic measurements of sample and dilution air gas flow rates. When the sample flow rate had strayed noticeably from the initial value, the critical flow orifice was cleaned and replaced. The average dilution ratio for each of the six experiments performed during the sampling campaign were as follows: 110 ± 22, 121 ± 30, and 109 ± 19 for the three tests of HH A and 210 ± 109, 133 ± 16, and 146 ± 60 for the three tests of HH B. During data processing the average dilution ratio for each respective experiment was applied to the measured particle counts across the test. The addition of the clean, dry, dilution air was sufficient to reduce relative humidity of the sample aerosol below 25% and temperature to below 35°C during all experiments. The aerosol measurement instrumentation used in this study was connected to the outlet of the dilution system. Excess diluted sample gas was directed from the overflow outlet of the sampling system outside the building.

The AE33 is a multiwavelength aethalometer, which measures light absorption at seven wavelengths: 370, 470, 520, 590, 660, 880, and 950 nm. The AE33 was operated with a sampling flow rate of 2 LPM with data averaged on a 1-min time basis. The spot change attenuation was set to 120. These settings were chosen to extend the life of instrument tape and minimize spot changes during an experiment. These factors were prioritized due to the long duration and broad time variation expected during the experiments. The AE33 data was treated using the standard techniques for the instrument. Namely, the BC concentrations given are those output by the instrument, and are therefore corrected using the “dual-spot” method (Drinovec et al. 2015). The single scattering correction factor used was 2.14, which is also standard for this instrument and well supported in literature (Weingartener et al. 2003). Data from the 370 and 880 nm wavelengths are discussed in this work and both absorption and concentration units are used. When transforming between absorption (units: inverse megameters, Mm⁻¹) and concentration (units: micrograms per cubic meter, µg/m³) the manufacturer recommended Mass Absorption Cross Section (MAC) was used; these MAC values are given in the AE33 user manual as 18.47 m squared per gram (m²/g) at the 370 nm wavelength and 7.77 m²/g at the 880 nm wavelength. Various studies have found that the uncertainty in aethalometer absorption measurements are between 0% and ±50% of the measured value (Di Biagio et al. 2017; Healy et al. 2017; Zanatta et al. 2017), based on these findings we estimate an uncertainty of ±25%. Including the uncertainty of the dilution ratio for each appliance gives an appliance average mass concentration uncertainty value of ±33% for HH A and ±43% for HH B. These values are less than the statistical uncertainties associated with the phase mean concentrations in most cases, due to the wide variations in concentration within a given phase; therefore, statistical uncertainties rather than experimental uncertainties are used in interpreting the results. The aethalometer data were also used to calculate the AAE of the absorbing aerosol, the equation used for this purpose is given as supplemental information. No attempt was made to separate the pure absorbing component from the aerosol as this approach would require additional assumptions about the composition and optical properties of the aerosol. Propagating the uncertainty of the absorption measurements through the equation for AAE gives a value of ±35%.

The pDR is a nephelometer instrument, which measures the light scattering coefficient of the sample aerosol with units of inverse megameters (Mm⁻¹). The pDR uses light with a wavelength of 880 nm. The pDR was operated with the total aerosol inlet at a flow rate of 2.3 LPM. The internal relative humidity correction was not applied as relative humidity in the sample was low due to the dilution process. Data were collected by the instrument on a 1-min time basis. The accuracy of the instrument operated with these settings is ±5%; combined with the dilution ratio, the uncertainty associated with light scattering measurements of HH A is ±21% and for HH B is ±36%.

Data analysis

Our measurements of optical BC and BrC, AAE, scattering coefficient, and BC concentration mass fraction show that there are differences in these quantities between different appliance operating conditions. These differences occur in response to heat demand and to fuel additions. The response to either stimulus is appliance dependent.

Black and brown carbon

Using the AE33 the BC and BrC concentrations of the flue gas aerosol were measured. In this analysis BC concentration specifically refers to the AE33 measurement at the 880 nm wavelength converted to mass units with the standard MAC value of 7.77 m²/g. BrC concentration refers to the light absorption measurement at 370 nm wavelength minus the light absorption measurement at 880 nm projected to 370 nm assuming an AAE of 1, this result is then converted to mass units with the standard MAC value of 18.47 m²/g. The mean BC and BrC concentration during each phase, for each test, of both appliances is given in Tables 2 and 3, and shown in Figures 3 and 4, respectively. Occasional negative values occurred during the sampling campaign, which is common with this instrument as the reported concentrations are based on a difference of discrete measurements. Negative test section mean values for BC are reported in Table 2, for HH B Test 1 Phase 6 and in Table 3 for HH B Test 1 Phase 6 and Test 3 Phase 5.

Table 2. Mean black carbon concentration, in units of micrograms per cubic meter (µg/m³), measured using the Magee Scientific Aethalometer Model 33, during each phase, of each test, for two wood-fired hydronic heating appliances.

Hydronic Heater	Test	Phase 1	Phase 2	Phase 3	Phase 4	Phase 5	Phase 6	Phase 7	Phase 8	Phase 9	Phase 10	Entire Test	Weighted
A	1	9.8E+02	1.3E+04	3.6E+03	5.1E+03	9.5E+02	1.1E+03	1.4E+04	3.9E+03	2.4E+03	4.5E+02	4.6E+03	5.3E+03
	2	5.6E+04	2.4E+04	6.6E+03	5.7E+03	5.4E+02	2.5E+03	1.5E+03	5.9E+03	5.5E+01	3.1E+03	1.1E+04	1.1E+04
	3	1.1E+04	2.2E+03	7.2E+02	8.6E+02	3.4E+01	1.1E+03	1.2E+03	9.2E+02	6.0E+02	7.1E+02	1.9E+03	2.0E+03
	Average	2.2E+04	1.3E+04	3.6E+03	3.9E+03	5.1E+02	1.5E+03	5.6E+03	3.6E+03	1.0E+03	1.4E+03	5.7E+03	6.2E+03
B
	1	1.5E+04	1.8E+03	5.6E+02	3.4E+02	1.3E+02	−1.1E+01	1.1E+03	6.1E+02	3.0E+03	6.9E+02	2.3E03	2.2E+03
	2	3.2E+03	1.7E+03	3.4E+03	3.2E+03	1.1E+03	1.8E+02	1.9E+03	8.6E+02	4.1E+03	3.5E+03	2.3E+03	2.1E+03
	3	3.8E+04	1.1E+03	2.1E+03	1.9E+02	6.7E+00	1.8E+03	7.8E+03	1.3E+03	2.4E+03	4.4E+03	5.9E+03	5.5E+03
	Average	1.9E+04	1.5E+03	2.0E+03	1.2E+03	4.1E+02	6.7E+02	3.6E+03	9.3E+02	3.2E+03	2.9E+03	3.5E+03	3.3E+03

Table 3. Mean brown carbon concentration, in units of micrograms per cubic meter (µg/m³), measured using the Magee Scientific Aethalometer Model 33, during each phase, of each test, for two wood-fired hydronic heating appliances.

Hydronic Heater	Test	Phase 1	Phase 2	Phase 3	Phase 4	Phase 5	Phase 6	Phase 7	Phase 8	Phase 9	Phase 10	Entire Test	Weighted
A	1	1.37E+03	3.56E+04	4.94E+03	6.82E+02	4.51E+02	7.28E+02	5.10E+03	4.61E+03	1.03E+03	−5.21E+01	5.45E+03	5.64E+03
	2	1.67E+03	8.77E+03	7.64E+02	−6.80E+02	3.87E+01	4.85E+01	1.38E+03	8.88E+03	3.43E+02	3.08E+03	2.43E+03	3.41E+03
	3	1.82E+03	1.85E+03	4.24E+02	2.71E+02	6.70E+01	5.60E+02	8.96E+02	8.80E+03	1.12E+02	5.22E+02	7.41E+02	8.08E+02
	Average	1.62E+03	1.54E+04	2.04E+03	9.10E+01	1.86E+02	4.46E+02	2.46E+03	4.79E+03	4.96E+02	1.18E+03	2.87E+03	3.29E+03
B
	1	1.58E+04	2.42E+03	1.11E+03	1.88E+02	4.81E+02	−4.49E+01	1.43E+03	4.99E+02	1.14E+03	2.13E+03	2.52E+03	2.33E+03
	2	7.84E+03	2.85E+03	1.36E+04	1.32E+03	3.21E+03	2.74E+03	6.69E+03	2.20E+03	1.29E+04	1.96E+03	5.53E+03	3.66E+03
	3	8.32E+04	1.72E+03	8.61E+03	1.01E+02	−3.36E+01	2.08E+04	1.52E+04	3.11E+03	3.01E+03	9.55E+03	1.45E+04	1.25E+04
	Average	3.56E+04	2.33E+03	7.76E+03	5.35E+02	1.22E+03	7.82E+03	7.78E+03	1.94E+03	5.69E+03	4.55E+03	7.52E+03	6.16E+03

Figure 3. Boxplot of minute-by-minute black carbon concentration measurements made using the Magee Scientific Aethalometer Model 33 during each phase for both appliances. Red with single hatch is hydronic heater A, blue with double hatch is hydronic heater B. The median is shown as an orange bar and the mean as a green triangle. The notches extend to 95% confidence interval around the median. Boxes extend to 25th and 75th percentile. Whiskers extend to the 5th and 95th percentiles. Outliers are suppressed.

Figure 4. Boxplot of minute-by-minute brown carbon concentration measurements during each operating condition for both appliances. Red with single hatch is hydronic heater A, blue with double hatch is hydronic heater B. The median is shown as an orange bar and the mean as a green triangle. The notches extend to 95% confidence interval around the median. Boxes extend to 25th and 75th percentile. Whiskers extend to the 5th and 95th percentiles. Outliers are suppressed.

In general, BC concentration maintained a constant value throughout an experiment, punctuated by spikes, which occur intermittently throughout each test. The nature of these variations causes poor agreement between the median and mean values for each phase, as can be seen in Figure 3. The highest mean BC concentrations occur during sections with fuel loading for both HH A and HH B. The lowest BC concentrations are found during sections when the appliance output was increasing for HH A, but to a lesser extent for HH B. The overall average BC concentration of both appliances are of similar magnitude, with BC concentration plus or minus twice the standard mean error values of 5.7 ± 2.4 × 10³ and 3.5 ± 1.8 × 10³ µg/m³ for HH A and B, respectively.

For HH A the first, second, and third highest mean BC concentrations are measured during periods with fuel loading. The highest mean BC concentration, 2.2 ± 1.8 × 10⁴ µg/m³, occurred during Phase 1; The second highest concentration, 1.3 ± 0.4 × 10⁴µg/m³ occurred during Phase 2; The third highest occurred during Phase 7 with a value of 5.6 ± 5.8.6 × 10³ µg/m³. The lowest BC concentrations are measured during test sections where heat output is increased from the previous test section. The lowest BC values occur during Phases 5 and 9 with values of 5.1 ± 2.4 × 10² µg/m³ and 1.0 ± 1.2 × 10³ µg/m³, respectively.

For HH B the highest average BC concentrations were measured during Phases 1 and 7, with values of 1.9 ± 1.1 × 10⁴ and 3.6 ± 1.3 × 10³ µg/m³ respectively. During Phase 2, the BC concentration was not significantly elevated. The lowest BC for HH B was found during Phase 5, with a value of 4.1 ± 1.8 × 10² µg/m³. Phase 9 mean value was not significantly depressed.

Overall, we found relative peaks in BC concentration during fuel loading events. These periods are commonly punctuated by inefficient combustion, characterized by high PM mass and number concentration in the flue gas, due to factors such as lower temperature, and increased fuel volatiles and moisture content (Johansson et al. 2004; Shen 2013a) and it is likely that these same factors result in peak BC emission. Conversely, decreased BC concentrations were found in test sections following an increase in heat output: Phase 5, 6, and 9. One possible explanation is as follows: in a low output condition, fuel may catch fire, but not have sufficient oxygen to support flaming combustion or generate a high enough temperature to combust/vaporize all the volatiles present in the fuel. Increasing heat output signals the appliance to increase oxygen flow, which allows the fuel which had partially decomposed during the low output phase to combust more fully in the subsequent high output test section, resulting in lower BC production.

The BrC measurements show the same trend with fuel loading seen in the BC results. The overall experimental averages for BrC were 2.9 ± 1.7 × 10³ and 0.8 ± 5.0 × 10⁴ µg/m³ for HH A and B, respectively.

For HH A, the highest average BrC concentration was measured during Phase 2 with a value of 1.5 ± 0.6 × 10⁴ µg/m³. The BrC concentration is also high during Phase 8 with a value of 4.8 ± 1.6 × 10³ µg/m³. The lowest BrC measurements for HH A occur during Phases 4 and 5, with values of 0.9 ± 2.0 × 10² and 1.9 ± 0.8 × 10² µg/m³ respectively.

For HH B, the highest averages are measured during Phase 1, with a value of 3.6 ± 3.8 × 10⁴ µg/m³. The lowest BrC concentrations are measured during the lowest output test sections: Phases 4, 5, and 8. During Phase 4, the BrC concentration is 5.3 ± 2.4 × 10² µg/m³. During Phase 5, the BrC concentration is 1.2 ± 0.4 × 10³ µg/m³. During Phase 8, the measured BrC concentration is 1.9 ± 08 × 10³ µg/m³.

The consistent highs between the BC and BrC results indicate that combustion quality is poor during fuel loading events. This lends credence to the need for more measurements of biomass combustion appliance performance with a focus on transient operating conditions and particle composition. The relatively high BrC measurement during Phase 8 for HH A was not seen in the BC results. Although this finding it is not unexpected; during Phase 8 section, the appliances are tested at a very low heat output state, which fosters poor combustion quality due to factors such as low temperatures and low air flow, and often result in high emissions (Shen 2013, 2013aa, van Loo and Koppejan 2008; Alves 2018). In contrast, for HH B, lower output produces lower BrC results than high output periods. Based on the overall BC and BrC concentration data set, it seems that HH A shows a clear decreasing trend between fuel loading events. Similarly, HH B shows a decreasing trend between the first and final fuel addition, but also shows low BC and BrC concentrations during low heat output periods. It follows that HH B is better optimized for low output conditions. Some possible explanations include that the updraft airflow pattern employed in HH A does not mix the air and gasified fuel prior to secondary combustion, or that the catalyst employed by HH A may have been overwhelmed by the amount of gasified fuel created during the low output periods, due to the low catalyst temperature and low air-to-fuel ratio required during this period, both of which inhibit catalytic combustion.

Angstrom absorption exponent analysis

The calculated mean AAE for each appliance during each test section is given in Table 4and agraphical comparison of the AAE is given in Figure 5. The overall experimental average AAEs were 1.94 ± 0.32 and 2.64 ± 0.26 for HH A and B, respectively. The AAE for HH B was significantly larger than HH A on average, however the AAEs for both appliances fall well within the range reported in literature of values around 2 (Brown et al. 2008; Helin et al. 2021; Kirchstetter, Novakov, and Hobbs 2004; Peter et al. 2017; Prevot et al. 2008; Saleh et al. 2013). Analysis of the test section mean AAEs indicated an increasing trend between AAE and time from fuel loading events, which is opposite to the trend found between BC and BrC concentrations and fuel loading events.

Table 4. Calculated angstrom absorption exponent for each appliance during each test phase for hydronic heater A (upper) and hydronic heater B (lower).

Hydronic Heater	Test	Phase 1	Phase 2	Phase 3	Phase 4	Phase 5	Phase 6	Phase 7	Phase 8	Phase 9	Phase 10	Entire Test	Weighted
A	1	1.92	2.58	2.52	2.02	2.39	2.40	1.86	2.38	2.33	1.48	2.19	2.17
	2	2.02	1.21	1.85	0.94	1.77	1.65	1.80	2.11	3.29	2.36	1.90	1.72
	3	1.34	2.03	1.55	1.67	2.46	1.56	1.81	2.09	2.31	1.70	1.85	1.86
	Average	1.76	1.94	1.97	1.54	2.21	1.87	1.82	2.19	2.64	1.85	1.98	1.91
B
	1	1.65	2.08	2.21	3.36	2.87	2.79	2.07	2.79	1.74	2.68	2.42	2.62
	2	2.35	2.41	2.64	2.45	3.06	4.22	2.72	2.96	2.76	2.17	2.77	2.71
	3	2.24	2.21	2.66	2.78	2.49	3.88	2.57	2.59	2.97	2.66	2.70	2.64
	Average	2.08	2.23	2.50	2.86	2.81	3.63	2.45	2.78	2.49	2.50	2.63	2.66

Figure 5. Boxplot of minute-by-minute AAE during each operating condition for both appliances. Red with single hatch is hydronic heater A, blue with double hatch is hydronic heater B. The median is shown as an orange bar and the mean as a green triangle. The notches extend to 95% confidence interval around the median. Boxes extend to 25th and 75th percentile. Whiskers extend to the 5th and 95th percentiles. Outliers are suppressed.

Analysis of the AAE time series shows similar behavior across tests, generally described as noisy data around a mean value. This behavior may be attributable to changes in aerosol particle composition from inhomogeneous mixing in the firebox inherent in solid fuel combustion. Notably, during Phase 4 and Phase 10 for HH A, negative AAE values were found at the 1-min time scale indicating that the absorption coefficient for shorter wavelengths was higher than for longer wavelengths. The data for HH B had no negative AAEs. This break from expected form is largely a product of the high variance in emission during those periods. Rapid changes in BC and BrC emission during these periods result in coefficients of variation greater than 1 and it is possible that that at the extremes of the measured BrC concentration, the instrument struggles to give an accurate measure, resulting in negative AAEs. In this analysis, negative AAE values were removed from the dataset.

The overall average AAEs were similar for both appliances and the test section average results showed similar trends. Most notably, the mean AAE was lowest during test sections where fuel was added, such as during Phases 1, 2, and 7.

The highest average AAEs for HH A were measured during Phases 9, 5, and 8 with values of 2.64 ± 0.22, 2.21 ± 0.26, and 2.19 ± 0.10, respectively. The lowest average AAEs were measured during Phases 4, 1, and 7 with values of 1.54 ± 0.18, 1.76 ± 0.54, and 1.82 ± 0.24, respectively. In general, the AAE increases between Phases 1 and 6, with the exception of Phase 3 section. Similarly, the AAE during Phase 7 is a local minimum and AAE increases during the subsequent phases.

For HH B the highest average AAEs were measured during Phases 6, 4 and 5 with values of 3.63 ± 0.24, 2.87 ± 0.32, and 2.80 ± 0.34, respectively. The lowest average AAEs were measured during Phases 1 and 2 with values of 2.08 ± 0.46 and 2.24 ± 0.20, respectively. The AAE consistently increases between Phase 1 and 6 for HH B.

These results show that the magnitude of AAE and the trends in BC and BrC concentration are inversely related. The AAE for both appliances trends with time from fuel load, where AAE increases between fuel additions, BC and BrC decrease between fuel additions. However, some notable exceptions to this rule occur: Phase 4 for HH A, and Phase 10 for both appliances. In the latter case, the cause is likely related to fluctuating combustion conditions in response to the heat demand. Overall, the cause of these opposing trends is mathematical. For both appliances, we found high BC and BrC concentration during periods with fuel additions. However, during these periods, the difference between the values was not large. Therefore, AAE which is based on the logarithmic difference of the two, is low. This trend agrees with the findings of Kortelainen et al. (2018) who found high refractory and elemental carbon emission during start-ups, followed by a short period of increased carbon production, and low carbon emission during the char combustion for wood fired masonry heaters (Kortelainen et al. 2018). In contrast, Martinsson et al. (2015) found decreasing trend between AAE and time between fuel loading events for wood burning stoves. The differences between these two studies are most likely due to the different appliance types and combustion technology tested in each study.

Light scattering analysis

Typically, BC is thought of as an absorbing aerosol, but when dealing with many particles, particularly small particles, some fraction of the light is scattered. BrC aerosol can both scatter and absorb light. The pDR instrument measures total light scattering, representing a composite of the BC and BrC results. It follows the pDR data show common features from both BC and BrC measurement analysis. The measured mean light scattering coefficient for each appliance during each test section is given in Table 5.

Table 5. Mean aerosol optical depth (units: inverse megameters, Mm⁻¹) measurements made using the Thermo Scientific personal DataRam Model 1500 for each test, separated by phase, and appliance.

Hydronic Heater	Test	Phase 1	Phase 2	Phase 3	Phase 4	Phase 5	Phase 6	Phase 7	Phase 8	Phase 9	Phase 10	Entire Test	Weighted
A	1	8.1E+02	1.4E+06	9.5E+04	2.8E+04	5.4E+03	4.1E+03	2.2E+04	2.2E+04	1.1E+04	2.8E+03	1.6E+05	1.6E+05
	2	2.5E+06	4.1E+05	2.7E+04	4.0E+04	1.0E+04	9.4E+03	2.1E+03	1.8E+04	5.1E+03	4.3E+04	3.3E+05	3.5E+05
	3	3.3E+04	1.1E+04	4.7E+03	4.8E+03	5.5E+03	2.8E+03	7.2E+03	5.5E+03	3.4E+03	6.5E+03	8.0E+03	8.4E+05
	Average	8.6E+05	6.1E+05	4.2E+04	2.4E+04	5.3E+04	5.4E+03	1.0E+04	6.8E+03	6.6E+03	1.7E+04	1.6E+05	1.7E+05
B
	1	1.5E+05	1.1E+04	4.6E+03	5.4E+03	1.8E+03	3.3E+02	5.0E+03	2.3E+03	6.9E+03	6.4E+03	1.9E+04	1.9E+04
	2	1.8E+04	1.2E+04	3.4E+04	1.7E+04	4.4E+03	2.7E+03	1.3E+04	1.0E+04	1.4E+04	1.4E+04	1.4E+04	1.3E+04
	3	0.0E+00	9.1E+03	2.2E+04	1.3E+03	3.7E+02	3.4E+04	4.7E+04	1.4E+04	7.2E+04	5.6E+04	1.9E+04	1.4E+04
	Average	5.5E+04	1.1E+04	2.0E+04	8.0E+03	2.2E+03	1.2E+04	2.2E+04	9.0E+03	1.0E+04	2.5E+04	1.8E+04	1.6E+04

An analysis of the pDR data time series shows similar structure to the BC and BrC results. However, the pDR data have considerably less noise, likely because the pDR data are continuous measurement, while the AE33 data are based on the difference between consecutive discrete measurements. Even so, the time variation in the data within a single test section, was still high. This may be due in part to variations in combustion quality within a test section, which was not evident in the AE33 results due to the semicontinuous nature of the filter tape readings. The high degree of time variation in the data is apparent in Figure 6 as a large inter-quartile range. The best examples of which are for HH A: Phases 1 and 2, and for HH B: Phase 6.

Figure 6. Boxplot of minute-by-minute aerosol optical depth measurements during each test section for both appliances. Red with single hatch is hydronic heater A, blue with double hatch is hydronic heater B. The median is shown as an orange bar and the mean as a green triangle. The notches extend to 95% confidence interval around the median. Boxes extend to 25th and 75th percentile. Whiskers extend to the 5th and 95th percentiles. Outliers are suppressed.

Overall, the trend in light scattering coefficient means for each test section were similar across appliances, but with a higher coefficient for HH A relative to HH B. The overall appliance average light scattering coefficient was 1.6 ± 0.9 × 10⁵ and 1.7 ± 0.9 × 10⁴ Mm⁻¹ for HH A and B respectively.

For HH A, the highest mean coefficients occurred during Phases 1 and 2 with values of 8.6 ± 7.0 × 10⁵ and 6.1 ± 1.2 × 10⁵ Mm⁻¹, respectively. The lowest mean coefficients for HH A were measured during Phases 5 and 6 with 5.3 ± 2.9 × 10³ and 5.4 ± 3.5 × 10³ Mm⁻¹, respectively. The large coefficient measurement during Phases 1 and 2 aligns well with the BC and BrC results. Similarly, the low measurement in Phase 5 also agrees with the previous findings.

The highest mean coefficient measurement for HH B was found during Phase 1 as well, with a value of 5.5 ± 4.5 × 10⁴ Mm⁻¹. The lowest mean light scattering coefficient measurements were found during Phases 5, 4, and 8 sections with values of 2.2 ± 3.6 × 10³, 8.0 ± 5.1 × 10³, and 9.0 ± 3.2 × 10³ Mm⁻¹, respectively. Again, the high coefficient during Phase 1 and low coefficient during Phase 5 agree with previous BC and BrC findings.

The in-stack light scattering coefficient measurements generally agree with the BC and BrC analysis. Specifically, the data for HH A show a dependence on fuel loading at least during Phase 1, and for HH B, both Phases 1 and 2 test sections have high coefficients. The HH B results also indicate a dependence on heat output, where Phases 8, 4, and 5, the low output phases generate substantially lower light scattering coefficients than the high output test sections, Phases 3 and 9. It follows that the coefficient measurements are fairly consistent with previous analyses of BC and BrC and are therefore responsive to the test conditions. However, the standard mean errors of the measurements are often high, which could indicate further time variation within operating conditions.

Black carbon fraction

In most laboratory combustion emission studies, the quantification of BC and BrC is performed on the basis of EC and OC, which is measured using a thermo-optical method. However, the thermo-optical method does not have sufficient time-resolution for our purposes. Here, we present an estimate of EC and OC using the aethalometer method, by comparing the BC data from our AE33 in comparison with a gravimetric filter-based PM mass concentration measurement taken using an EPA Method 5 sampling train taken from Trojanowski et al. (2022). The two mass measures, BC mass from the AE33 and PM mass as measured using the gravimetric method, which we are using as our total carbon measurement, are compared in Figure 7. Overall, this analysis showed that BC makes up approximately a quarter of the PM emitted from the HHs we tested.

Figure 7. Bar chart comparison plots of gravimetric PM_2.5 and AE33 BC measurements for two biomass hydronic heating appliances. Data from hydronic heater A is shown in the topmost chart (a), and data from hydronic heater B is shown in the bottommost chart (b).

The overall appliance average BC fraction results were very reasonable 0.24 ± 0.15 and 0.30 ± 0.12 for HH A and B, respectively. Further, the PM and BC concentration measurements trend well together. However, during some test periods the BC measurement exceeds the PM measurement resulting in a fraction greater than 1. In the majority of cases, the phase-means between the two measurements overlap within two standard mean errors, leaving a single aberrant case during Phase 4 for HH A.

The BC fractions for HH A are reasonable, except during Phase 4 section where the BC measurement is clearly larger than the PM measurement. Additionally, during Phases 6 and 7 the mean BC fraction is greater than one, but well within two standard mean errors of the PM measurement. For HH A, the BC fraction of the PM measurement is between 0.06 and 3.88 with an overall average fraction of 0.24 ± 0.15. The lowest BC fraction for HH A was measured during Phase 1, with a value of 0.06 ± 0.09, despite this section having the highest measured BC concentration during the entire test.

The results for HH B were similar to those for HH A; PM_2.5 and BC concentration show the same trends, but with fractions greater than 1 found during Phases 8 and 4. In both cases, the BC measurement is within two standard mean errors of the PM measurement. The BC fraction range for HH B was between 0.03 and 2.94, with an overall average value of 0.30 ± 0.12. The lowest measured BC fraction for HH B was found during Phase 6 with 0.03 ± 0.05.

Overall, this analysis would indicate that BC represents a sizable fraction of PM_2.5; approximately 0.24 ± 0.15 in the case of HH A, and 0.30 ± 0.12 in the case of HH B which is well within the range found in literature (Chow et al. 2011). While the overall correlation between the aethalometer BC and PM_2.5 was fairly good, the large range of BC fraction and particularly the fractions greater than one present concerns. Given these issues, more research comparing the real-time EC/OC technique to standard methods should be performed. Further, the phases where these inversions occur seem to be correlated with low PM mass concentrations. Gravimetric PM analysis suffers from low accuracy due to low PM mass catch during these conditions, which suggests a real-time PM concentration measurement may be better suited for this analysis.

Conclusion

In summary, our measurements show the range of BC and BrC concentrations in the flue gas of two modern wood-fired HHs. For HH A, BC concentration ranged between 3.42 × 10¹ and 5.57 × 10⁴ µg/m³ and for HH B −1.14 × 10¹ and 3.78 × 10⁴ µg/m³. BrC concentrations for HH A ranged between −6.80 × 10² and 3.56 × 10⁴ µg/m^3, and for HH B between −4.49 × 10¹ and 8.32 × 10⁴ µg/m³. Negative values were present, typically found during burnout periods where literature suggest the aerosol is mostly comprised of small salt particles (Kortelainen et al. 2018). The AAE results ranged between 1.52 and 2.65 for HH A, and for HH B between 2.08 and 3.63. AOD measurements ranged between 5.33 × 10³ and 8.56 × 10⁵ Mm⁻¹ for HH A, and between 2.20 × 10³ and 5.50 × 10⁴ Mm⁻¹ for HH B. The BC fraction estimates were 0.24 ± 0.15 and 0.30 ± 0.12 for HH A and B, respectively.

Further analysis of the data shows that trends in emission response to heat output and fueling condition are appliance dependent, likely due to their designs, employed combustion technology, and control systems. In general, for both appliances Phases 1 and 2 had the highest BC and BrC concentrations and light scattering coefficients, as well as a low AAE.

More specifically, for HH A, a trend between emissions and fuel condition was found; emissions decreased with time from introduction of the fuel load. In the case of HH B, a trend between heat output and emissions was identified; high heat output test conditions produced higher emissions than low heat output conditions, with a step function increase between heating categories. The presence of a heat output correlation for HH B, which was not identified for HH A was likely due to the design of the HH; where HH A relies on a catalyst for emissions reduction across operating conditions, HH B utilizes gasification which requires more tuning of the airflow systems at different output conditions.

These findings highlight the importance of characterizing transient operations in biomass heating appliances. Different HH models react differently to heat output and fuel load conditions and transient periods such as Phases 1 and 2 can cause higher polluting conditions than seen in a steady state operation. The characteristics of the carbonaceous component of PM depend upon appliance operation patterns. These alterations are not captured by current laboratory tests relied on by the EPA and others to evaluate emissions impacts. To properly gauge the climate, environmental and health impacts from residential biomass combustion for heating, more robust test protocols are needed, including transient emission test cases reflecting more realistic operating conditions. This information will be necessary to improve source apportionment models, pollutant emissions inventories and modeling of radiative forcing potential.

Data availability

The data that support the findings of this study are available from the corresponding author, JL, upon reasonable request.

Acknowledgment

The authors would like to acknowledge that the New York State Department of Health and in particular Patricia Fritz and Nicole Vitillo, everyone at the Energy Conversion Group at Brookhaven National Laboratory (BNL) for organizing the experiments and operating the appliances, and Northeast States for Coordinated Air Use Management (NESCAUM) for developing the test method used in this study. Financial support from New York State Energy Research Development Authority (NYSERDA) has made this work possible. This research was funded through NYSERDA Agreement 63033.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the New York State Energy Research and Development Authority [63033].

Notes on contributors

Jake Lindberg

Jake Lindberg is a recent doctoral graduate at Stony Brook University in the Chemical and Molecular Engineering Department in Stony Brook, New York, USA and works closely with the Energy Conversion Group, at Brookhaven National Laboratory in Upton, New York, USA.

Marilyn Wurth

Marilyn Wurth is a research scientist at the New York State Department of Environmental Conservation, in the Emissions Measurement Research Group within the Division of Air Resources, Bureau of Mobile Sources & Technology Development in Albany, New York, USA.

Brian P. Frank

Brian P. Frank is the section chief of the Emissions Measurement Research Group at the New York State Department of Environmental Conservation, Division of Air Resources, Bureau of Mobile Sources & Technology Development in Albany, New York, USA.

Shida Tang

Shida Tang is a research scientist at the New York State Department of Environmental Conservation, in the Emissions Measurement Research Group within the Division of Air Resources, Bureau of Mobile Sources & Technology Development in Albany, New York, USA.

Gil LaDuke

Gil LaDuke is a research scientist at the New York State Department of Environmental Conservation, in the Emissions Measurement Research Group within the Division of Air Resources, Bureau of Mobile Sources & Technology Development in Albany, New York, USA.

Rebecca Trojanowski

Rebecca Trojanowski is a research scientist at Brookhaven National Laboratory, in the Interdisciplinary Science Department, Energy Conversion Group, in Upton, New York, USA, and is a doctoral candidate at Columbia University in the Department of Earth and Environmental Engineering, in New York, New York, USA.

Thomas Butcher

Thomas Butcher is a research scientist at Brookhaven National Laboratory and is the leader of the Energy Conversion Group within the Interdisciplinary Science Department, in Upton, New York, USA.

Devinder Mahajan

Devinder Mahajan is a research professor in the Department of Materials Science and Chemical Engineering at State University of New York at Stony Brook in Stony Brook, New York, USA.