3.1 Biomass samples
A total of 30 biomass pellet sample bags, each weighing 15 kg, were purchased from retail outlets in the Upper Silesia region of south-central Poland. Among the acquired fuels, 26 were wood pellets and 4 were of non-wood origin (agro pellets), including sunflower husk, dry bean pods, and bran. Almost half of the samples were ENplus and/or DINplus certified (Table 1; Drobniak et al., 2024).
All samples underwent physicochemical analysis according to the DINplus certification, ensuring a thorough examination of key physical and chemical attributes. These parameters included pellet diameter and length, fines content, mechanical durability, and bulk density, as well as moisture, ash, sulfur, and nitrogen contents, ash melting temperature, and net calorific value (DINplus 2023).
Additionally, samples were submitted for petrographic analysis to evaluate their composition and detect possible contaminants. The process involved the preparation of microscopic plugs, a 1,000-point petrographic analysis using reflected light, and the identification and classification of pellet components, following the methodology outlined by Drobniak et al. (2022, 2023b). The identified components included woody and non-woody biomass, bark, charcoal, fossil and processed organic matter, inorganic matter, petroleum products, binders, and other additives. Based on the data, the total amount of impurities in each sample was calculated.
3.2 Description of combustion experiments
To measure the type and amount of released emissions gases and particulate matter, aliquots of pellet samples were combusted in an Invicta Bassano 5 pellet stove located on the ground floor of a one-story building and operating at the maximum power of 5 kW. The stove was cleared of ash after each sample was burned. The next sample was then incinerated for 30 minutes before emission measurements were initiated. Combustion data were collected at the chimney's exit on the building’s roof (Figure 1).
Table 1. Information about the pellet fuel provided by producers. NP – information not provided.
Pellet type
|
Source of biomass
|
Pellet
certification
|
Wood pellets
|
NP
|
ENplus A1
|
NP
|
ENplus A1
|
NP
|
ENplus A1
|
NP
|
ENplus A1
|
NP
|
none
|
NP
|
ENplus A1
|
NP
|
none
|
NP
|
ENplus A1 and DINplus A1
|
coniferous wood
|
none
|
NP
|
none
|
NP
|
ENplus A1
|
NP
|
ENplus A1
|
spruce and pine
|
none
|
NP
|
none
|
NP
|
ENplus A1
|
coniferous wood
|
DINplus
|
NP
|
ENplus A1 and DINplus
|
NP
|
ENplus A1 and DINplus
|
coniferous wood
|
ENplus A1
|
NP
|
none
|
NP
|
DINplus
|
NP
|
none
|
NP
|
none
|
coniferous wood
|
none
|
pine and oak
|
none
|
pine and hornbeam
|
none
|
Non-wood pellets
|
sunflower husks
|
none
|
sunflower husks
|
none
|
sunflower husks and bean pods
|
none
|
bran
|
none
|
Wind speed and direction were also recorded by anemometer, and based on this information, the analyzers were positioned upwind before each measurement.
Emission data were recorded every second for 15 minutes using a Testo 440 and two Nanosens AtmonFL S.M.O.K pollution analyzers, which measured concentrations of CO, CO2, NO2, NH3, Cl2, PM2.5, PM10, H2S, SO2, HCHO, and respiratory tract irritants (NO2 + O3 + Cl2 + HC), as well as outside temperature, pressure, and humidity. Information on the instruments’ precision for particulate matter and gases measurements is available on the manufacturers’ websites (AeroMind 2024; Testo 2024). Furthermore, the air quality surrounding the building underwent daily analysis before initiating combustions. The emission data presented in this paper have had background values subtracted, focusing solely on emission levels resulting from fuel combustion.
3.3 Statistical analysis
Statistica 13 software by StatSoft (Poland) was used to explore relationships between physical properties, chemical composition, petrographic components of pellets, and their emission parameters. Each combustion variable in the database underwent testing for its probability distribution. Due to a significant number of peak values, the variables generally did not follow a normal distribution (Shapiro-Wilk test, p < 0.05; Schinazi, 2022). Therefore, the median value for analyzed emissions was used, along with the 25th and 75th percentiles of data distribution, as well as maximum and minimum values. Furthermore, to prevent an ill-conditioned matrix, parameters exhibiting a Pearson correlation coefficient (r²) above 0.9999 were removed. Given the close correlation between biomass and impurities contents derived from petrographic analysis (their combined sum equals 100 vol. %) and hence, their redundancy, only the impurities values were integrated into the model.
A progressive stepwise regression analysis was utilized to assess whether the combustion behavior of pellets can be predicted using various predictor variables. This method iteratively selects independent variables for the model based on their statistical significance, intending to enhance predictive accuracy while minimizing complexity. The progressive stepwise regression aimed to construct equations that predict the expected value of the dependent variable based on the values of several independent variables. The general form of the progressive stepwise regression equation can be expressed as:
Y = β0 + β1X1 + β2X2+... + βpXp + ε
where Y represents the predicted or expected value of the dependent variable, while X₁ through Xₚ denote p distinct independent or predictor variables, ε is a random error, and the coefficients β0 through βₚ represent the estimated regression coefficients.
Subsequently, F-tests were conducted to evaluate the overall significance of the regression models, and each coefficient was assessed using statistical tests, such as t-tests or p-values, to determine individual contributions of predictor variables. This process ensured that only statistically significant variables were retained in the final regression models, thus enhancing the validity and interpretability of results. Furthermore, coefficients of determination (R²) were calculated to evaluate the extent to which independent parameters included in the regression models account for the variance in the dependent variable. This analysis offered insights into the overall "Goodness-of-Fit" of the regression models, enabling the assessment of the effectiveness of the selected predictor variables in explaining the variability in pellet combustion behavior.
THEORY
Although the combustion of biomass leads to emissions of gaseous pollutants, volatile organic compounds (VOC) including polycyclic aromatic hydrocarbons (PAHs), and particulate matter, the human health risks associated with the emissions in residential settings remain inadequately investigated. Though there has been significant improvement in our comprehension of the dynamic interplay between fuel properties, combustion conditions, emissions, and their impact on human health in recent years, further research is needed to fully grasp the complexities of these relationships (Badyda et al., 2020; Chandrasekaran et al., 2013, 2016; Chen et al., 2017, 2023; Drobniak et al., 2023a, 2024; Freiberg et al., 2018; Gordon et al., 2014; Harun and Afzal, 2016; Jelonek et al., 2020, 2021; Jenkins et al., 1998; Karanasiou et al., 2021; Kasurinen et al., 2016; Kurmi et al., 2012; Naeher et al., 2007; Najser et al., 2019; Orasche et al., 2012; Ozgen et al., 2021; Perez-Jimenez, 2015; Petrocelli and Lezzi, 2014; Price-Allison et al., 2023; Rabaçal, 2015; Rabaçal et al., 2013; Rabbat et al., 2023; Shojaeiarani et al., 2019; Sigsgaard et al., 2015; Sterman et al., 2022; Su et al., 2023; Tran et al., 2023; Venturini et al., 2018; Vicente et al., 2018; Williams et al., 2012; Yang et al., 2019; Zeng et al., 2016; Zhang et al., 2020).
Research into a prediction of combustion emissions and their effects on human health is, however, a difficult task considering the intricate interaction of various factors like the variability of the biomass feedstock, complexity of the combustion process, atmospheric chemistry, meteorological factors, and variability of human exposure. Additionally, the challenge involves setting up a combustion experiment, an analytical approach, and subsequently appropriate statistical methods to draw conclusions.
Each type of biomass carries unique compositional characteristics and combustion behavior. Emissions from combustion can vary depending on the technical specifications of stoves, boilers, or grills, and on the chosen temperature and ventilation settings. Once released from chimneys, the emitted pollutants can either disperse via wind or pose a health hazard when concentrated in smoggy conditions. Equally difficult is estimating the human health impact of biomass utilization, particularly concerning long-term and cumulative effects. Previous studies, often combined with air quality monitoring, have been carried out to assess exposure levels and associated health risks (such as respiratory and cardiovascular diseases), and toxicological evaluations provided insight into the mechanisms by which biomass emissions affect human health and impact climate (Tomlin 2021; Vicente et al. 2021; Lazaro and Baba 2023; Jiang et al. 2024). Nevertheless, integrating data from various sources and incorporating additional factors like proximity to emission sources, weather conditions, seasons, human outdoor time-activity patterns, and individual susceptibility is challenging.
However, the results of such research are invaluable, as they shed light on challenges associated with biomass utilization and provide data allowing the enhancement of sustainable practices throughout the entire biomass energy production and utilization cycle. This, in turn, allows us not only to capitalize on the advantages of biomass as a renewable energy source but also to effectively address associated environmental and health challenges in pursuit of sustainable energy solutions.