Health impacts of PM_2.5 originating from residential wood combustion in four nordic cities

This study focuses on three Nordic capital regions: Helsinki Metropolitan Area, Oslo, and Copenhagen, and in Sweden on the smaller city of Umeå (where detailed exposure data were currently available), and their neighboring areas. We will focus on the larger urban areas instead of solely focusing on the areas of the cities themselves. For instance, we examine the four separate cities, which together comprise the Helsinki Metropolitan Area, as well as Umeå and its neighboring villages. However, we will subsequently refer to the regions simply as Helsinki, Oslo, Copenhagen, and Umeå. The locations of the selected cities and their domains are thoroughly presented in an earlier analysis by Kukkonen et al. [31].

The population exposure to RWC induced air pollution from local sources is based on both the modelled annual average PM_2.5 concentration and the population density in different parts of the urban areas. PM_2.5 has been chosen as a comprehensive indicator of several pollutants originating from RWC [31, 32]. For the air pollution modelling, the following information was collected from three sources: (i) wood usage for combustion and combustion appliances sources (details are presented in the earlier analysis by Kukkonen et al. [31]), (ii) emissions factors, and (iii) temporal allocation of emissions. The assessment of emissions from RWC for these four cities is also described by Kukkonen et al. [31]. The RWC induced PM_2.5 concentrations were calculated by using air pollution modelling.

The spatial resolution (size of grid-square cells) was 250 m × 250 m for Umeå and Helsinki, and 1 km × 1 km for Oslo and Copenhagen. The results represent the year 2011 for Umeå, 2013 for Helsinki and Oslo, and 2014 for Copenhagen. All four cities used different dispersion models: DISPERSION for Umeå (a multiple-source Gaussian model), UDM-FMI and SILAM for Helsinki (a multiple-source Gaussian model, and a global and regional scale chemical transport model), EPISODE for Oslo (a 3D Eulerian and Lagrangian model), and DEHM/UBM for Copenhagen (a 3D Eulerian and Gaussian models). Details of the models are presented in the earlier analysis by Kukkonen et al. [31].

For Helsinki, the gridded population data were obtained from the population grid of the Helsinki Metropolitan Area [33]. The population grid consists of grid-based statistics of this region that include information of the total population, the population density, and the age distribution of the population. In this study, the population data of 2018 were used. For Oslo, the 2013 gridded population data from Statistics Norway were used. The dataset was based on information from the national building and population registries. In addition, this dataset was combined with 2013 population data by age and gender in Oslo from Statistics Norway. For Copenhagen, the population data were obtained from the central personal register (CPR) for the year 2014, and aggregated to the 1 km x 1 km resolution. For Umeå, gridded population data for different age groups with a resolution of 100 m x 100 m were provided by Statistics Sweden for the year 2011.

Based on these data (Fig. 1), the population-weighted concentration (C_pop) of RWC in each city was calculated. For this, the calculated concentration in each grid-square cell (Ci) was multiplied with the number of people in the corresponding grid-square cell (Pi), and then all products were summed. The result of this sum was finally divided by the total population (Eq. 1).

Using this principle, the grids with a higher population density give a larger contribution to the population-weighted mean concentration.

The numbers of premature deaths were calculated based on health impact assessment (HIA) principles. HIA combines the results of RWC air pollution exposure, baseline mortality in the population, and concentration-response functions (CRF) from earlier epidemiological studies. The baseline mortality data in the population were retrieved from Statistics Sweden [34], Statistics Finland [35], Statistics Norway [36], and Statistics Denmark [37]. We applied the most recent total mortality data (2019) in the age group 30+. Considering that the health impact calculations apply to Nordic cities, we employed CRF from Hvidtfeldt et al.‘s [38] study from Denmark. This resulted in a Hazard Ratio (HR) of 1.26 (95% CI: 1.10–1.42) per 10 µg m^− 3 increment in PM_2.5 for all-cause mortality. Moreover, this HR is in line with Turner et al. [39] who found a HR of 1.26 (95% CI: 1.19–1.34) per 10 µg m^− 3 increase in near-source PM_2.5 in a large study using cohort data from the American Cancer Society Cancer Prevention Study II (ACS CPS II).

AirQ + was used for the calculation of the following four variables: (i) the number of premature deaths, (ii) the years of life lost (YLL), (iii) the average YLL per premature death case, and (iv) the decrease in life expectancy. AirQ+, developed by the WHO, is a software tool for determining health risk assessment of air pollution exposure [40]. It has previously been used for the Nordic area, and it has also been compared and validated against other similar tools [41]. It enables users to apply (i) concentration-response functions for selected air pollutants, (ii) data regarding air quality, and (iii) data regarding the population in terms of population size, age distribution, and baseline mortality. With these data entered into AirQ+, it is possible to calculate (i) the number of premature deaths, (ii) YLL, and (iii) decrease in life expectancy (i.e. the health effects from exposure to air pollutants) using different cut-offs and CRFs.

In this study, the health effects were calculated for the population aged 30 years and older, as no effects on premature mortality have been noticed among younger ages [42]. No cut-off concentrations were used as combustion particles are expected to be equally toxic at low concentrations. This principle has been applied in recent similar studies [43].

The current study shows that RWC has a relatively high health burden in terms of premature deaths, years of life lost, and decrease in life expectancy. However, the health impacts vary in the four Nordic city areas. The average annual estimates of the number of premature deaths range from 19 (Umeå) up to 232 (Oslo), and these estimates are reflected by differences in (i) the amount of exposure to PM_2.5 from RWC, (ii) the population size, and (iii) the age distribution (e.g. proportion of elderly). The average decrease in life expectancy also varies, but significantly less than the number of premature deaths. This is because life expectancy is largely determined by the amount of exposure to PM_2.5 from RWC, and to some extent also by the age distribution. However, life expectancy is not affected by the total number of inhabitants.

In Finland, the health effects associated with exposure to PM_2.5 from RWC in 2015 were calculated in an earlier study conducted by Savolahti et al. [44]. The concentrations of PM_2.5 from RWC were in the range of 0.5 − 2 µg m^− 3 in the proximity of most urban areas, resulting in approximately 31 premature deaths in the cities with > 200,000 inhabitants in 2015. This estimation is significantly lower than the results in this study with 53 premature deaths in Helsinki in 2013. The Helsinki region has approximately one million inhabitants which constitutes around three-quarters of such urban agglomerations in Finland that have more than 200,000 inhabitants. The main differences in these two assessments were the different emission and dispersion computations, and, most significantly, the fact that different RRs have been used. Savolahti et al. [44] applied a RR of 1.062 (95% CI: 1.040–1.083) per 10 µg m^− 3 increase in PM_2.5, whereas more recent HIAs, including our study (e.g. [43]), have applied significantly larger RRs for combustion particles. The trends in the concentrations of PM_2.5 originating from residential wood combustion in Helsinki region have been analyzed by Kukkonen et al. [45] during a multidecadal period with a slightly increasing trend from the 1980’s to the mid-2010s, caused by the more widespread use of residential wood combustion in the area.

The choice of concentration-response function has a crucial importance for the health impacts calculations. In this current study, the health impact calculations are based on a HR of 1.26 (95% CI: 1.10–1.42) per 10 µg m^− 3 increase in PM_2.5 for all-cause mortality based on a 15-year exposure to PM_2.5 at the residential addresses as according to Hvidtfeldt et al. [38]. Hvidtfeldt et al.‘s study collected detailed data on air pollution exposure, lifestyle factors, and socio-demography on a group of participants who lived in the areas of Copenhagen and Aarhus from 1997 to 2015. This hazard ratio is based on exposure to PM_2.5 in general, and not exclusively on RWC induced PM_2.5. As the aim of this study is to calculate the health effects exclusively caused by exposure to RWC induced PM_2.5, the application of a general hazard ratio based on PM_2.5 exposure is expected to create some uncertainty.

Recently, Vohra et al. [43] applied a low concentration RR of 1.129 (95% CI: 1.109–1.150) for all-cause mortality in all ages associated with a 10 µg m^− 3 increase in PM_2.5 exposure in order to calculate the global mortality from outdoor fine particle pollution generated by fossil fuel combustion. Another recent European multi-cohort study suggested a similar RR [46]. According to Vodonos et al. [29], a RR at low concentrations could be even higher (1.24, 95% CI: 1.08–1.40 per 10 µg m^− 3 increase in a meta-regression restricted to studies with mean concentrations of PM_2.5 below 10 µg m^− 3). A higher meta-estimate for the RR at low concentrations is also reported by Chen and Hoek [47]. In addition, Turner et al. [39] found a similar increased risk (RR = 1.26, 95% CI: 1.19–1.34 per 10 µg m^− 3 increase in PM_2.5) in a large study using cohort data from the American Cancer Society Cancer Prevention Study II (ACS CPS II) which is comparable to the RR in Hvidtfeldt et al. [38] that we have applied in this current HIA.

In an earlier Swedish report from 2018, where the population’s exposure to air pollutants was calculated as annual average concentrations, the number of premature deaths in Sweden associated with exposure to RWC induced PM_2.5 was estimated to be 935 (95% CI: 292 − 1577) in the age group 30 + during the year 2015 [48]. In this study, and based on earlier data from a subset of subjects from the American Cancer Society Cohort in Los Angeles County [49], a smaller concentration-response function (RR = 1.17, 95% CI: 1.05–1.30 per 10 µg m^− 3 increase in PM_2.5) was applied. Thus, our estimation is almost two times higher due to applying a higher RR.

Another question we address is whether wood-smoke particles pose different levels of risk compared to other ambient particles of similar size. In general, it is difficult to determine the difference between long-term exposure to RWC induced PM_2.5 and long-term exposure to PM_2.5 in original epidemiological studies. This is due to the fact that people are exposed to a mixture of fine particles from different sources. The difference in short-term health effects associated with exposure to RWC induced PM₁₀ as well as to PM₁₀ from point sources and mobile sources has been analyzed in Chile in the cities of Temuco and Pudahuel [50]. In Temuco, the main source of pollution was RWC, while in Pudahuel, the main sources of pollution were point sources and mobile sources. The findings of this study showed that the RRs for cardiovascular and respiratory mortality were slightly higher in Temuco as compared to Pudahuel.

In contrast, a study in the Estonian city of Tartu [51], where traffic and RWC induced particles were modelled separately, no association was found between RWC induced particles and health symptoms, although traffic-induced particles increased the odds of cardiac disease. As this study used self-reported health data, cross-sectional design, and modelled exposure data, conclusions should be taken with reservation and more epidemiological studies focusing specifically on RWC are needed. Furthermore, some experimental studies have found different effects. For example, Riddervold et al. [52] reported that wood smoke at a concentration normally found in a residential area can cause a mild inflammatory response. In contrast, Forchhammer et al. [53] did not find any effects either on markers of oxidative stress, DNA damage, cell adhesion, cytokines, or microvascular function in the same 20 atopic subjects.

Finally, in a review study with 22 identified publications based on the results from twelve studies on controlled human exposures to wood smoke, a range of different combustion conditions, exposure concentrations, and durations were applied. Different effects on the airways and the cardiovascular system as well as systemic endpoints were assessed. Large variations regarding study design in the analyzed studies make it difficult to draw any general conclusions. However, the findings were broadly consistent with respect to the effects on the airways, but there were no clear patterns regarding the effects on oxidative stress, systemic inflammation, and cardiovascular physiology [25].

In the current analysis, we have used modelled RWC induced PM_2.5 concentrations from local sources and applied annual average concentrations at home addresses. On the one hand, this approach is similar to the original epidemiological studies from which the concentration-response functions have been obtained. On the other hand, the real-life situations are much more complex, and there could be several uncertainties in those exposure estimations. Firstly, the models often apply relatively large grid-square cells. Secondly, these grid-square cells may vary in size in the different city areas. In Oslo and Copenhagen, a cruder modelling domain was used as compared to Helsinki and Umeå. If a more finely spaced receptor grid had been used for Oslo and Copenhagen, the predicted exposure and health effect values would have been somewhat higher. This effect is caused by both the positive spatial correlation of the emissions of RWC and the locations of the population in these cities.

The computations were done on a spatial resolution of 250 × 250 m for Umeå and Helsinki, and 1 × 1 km for Oslo and Copenhagen. We have selected the finest possible spatial resolutions for the computations for all four cities. However, the spatial resolution influences the predicted exposure and health values. Such impacts have been examined previously by Karvosenoja et al. [54] and Korhonen et al. [55]. Both studies showed that the predicted exposure values were lower for computations with a coarser spatial resolution. It is therefore essential to use a sufficiently fine model resolution in view of the assessment of health impacts. This is especially important for primary particles from emission sources at low emission heights. Moreover, for Copenhagen and Umea region a larger modelling domain is shown in order to describe the contribution from emissions outside the city in detail. Consequently, in the present study, we estimate that the use of different spatial resolutions (250 × 250 m or 1 × 1 km) is expected to result in a difference of less than 10% in the health impact estimates, based on the results by Korhonen et al. [54].

Another important aspect is to find the best possible proxy of human exposure. We have used air pollution concentrations at home addresses as a proxy. However, people are mobile and, thus, they are exposed during the day to air pollution concentrations at different locations (e.g. at home, at work, during shopping, whilst commuting, etc.). A higher spatial resolution of exposure data will not necessarily give a better estimation of the personal exposures, and using very high-resolution exposure data requires a spatio-temporal personal exposure model for estimating the exposure in different environments during the day, which is not available at the moment. Coarser resolution exposure data gives an average of the exposure during a day in the area at home and in the nearest surroundings.

The highest concentrations of RWC particles from local sources were found in Oslo. This concentration was, on average, three times larger than the corresponding values in Umeå and Copenhagen, and five times larger than in Helsinki. Clearly, all modelling results are dependent on the accuracy of the information of wood usage for combustion and the adopted emission coefficients for combustion appliances. The most common sources of information regarding firewood consumption are usage statistics and different questionnaires [56]. For instance, in Denmark, the wood usage has been estimated through questionnaires for approximately 6 000 households with wood stoves. In the current study, different years with different meteorology have been modelled where outdoor temperature and windiness might have affected the energy demand for heating, and, consequently, also the emissions and dispersions. However, Kukkonen et al. [31] have thoroughly described and evaluated the applied RWC emission inventories, and these contain the best available emission data in each of the target cities. The emission inventories and meteorological data in the current analysis partly correspond to different years in the target cities, selected based on the availability of relevant data. According to Kukkonen et al. [31], none of the considered years was rare in any of these cities in terms of the ambient temperatures. Although comparing the results from different years includes an uncertainty, the differences in relevant meteorological conditions were not substantial for the selected years. In addition, only small variations in the populations of the target cities occurred during the period from 2011 to 2014, and this is not considered to have any noticeable impact on the health impact estimates.

One limitation of this study is that we only focused on local emissions and local impacts. PM_2.5 originating from wood burning is also long-range transported in the atmosphere where it can spread up to thousands of kilometres away from the emission source. Therefore, the emissions from these four cities also influence health outside the city areas, and wood burning outside the cities contributes to the health impacts within these cities. Another limitation of this study is that we addressed only ambient RWC concentrations, as we were not able to estimate the contribution of RWC to indoor air pollution. Previous studies have shown that in some regions, the particle concentrations can be more than two times higher in homes with residential stoves [57, 58]. According to Vicente et al. [59], this increase is especially high during open fireplace operation where PM₁₀ concentrations can rise up to twelve times as compared to background concentrations. Moreover, candles are very often used during wintertime in Denmark [60], and a high concentration of candle induced PM_2.5 has shown a mild inflammatory response among young asthmatics as a result of five hours of exposure [61]. Nevertheless, it has been discussed that addressing ambient and indoor RWC exposure as separate risk factors can lead to double counting due to their interrelated nature [44]. Therefore, in order to avoid double counting, the results of this study should not be combined with the burden of disease estimates of indoor RWC exposure. Infiltration of outdoor particles indoors can be significant even in well-insulated buildings due to the operation of windows and doors, and cracks in the building envelope and windows, and door frames [62]. Population exposure can therefore be significantly different, depending on the structure and ventilation of buildings. The infiltration factors of PM_2.5 have been estimated to range from 0.47 to 0.59 in the Helsinki area [63].

Controlled experiments with different types of wood stoves have resulted in significantly different emissions. Fuel moisture, charge size, feeding rate, and air ventilation are also important factors in terms of emissions of particles [64]; e.g. high moisture content fuels will result in increased PM_2.5 emissions [62].

During a laboratory study, where different kinds of wood species were burned in a wood stove during different burning conditions, it was evidenced that the user of the stove could largely influence the emissions both by fuel selection and through the choice of burning conditions. There were three main factors that were crucial when it came to creating a complete combustion: (i) a proper amount of well-mixed oxygen present in relation to the fuel, (ii) a suitable temperature, and (iii) an optimal residence time of the fuel/oxygen mixture. There were, however, no significant differences regarding the emissions from different types of wood [65].

With respect to the amount of emissions, the age of the burning device was according to laboratory tests an important factor in which there were large variations in the emissions from different wood combustion appliances. In general, more modern devices with newer technology gave rise to lower PM emissions compared to older devices. However, there are large differences in the emissions that occur during laboratory conditions compared to the emissions that occur in real-life conditions. There are only a limited number of tests that have been conducted under real-life conditions. Further studies and a better understanding of the prevailing real-life conditions is, therefore, needed since the user’s behavior is an important factor regarding the amount of RWC emissions [66].

In order to reduce the health effects caused by wood combustion, several policy measures should be applied. These could include stricter guidelines for air quality, emission reduction measures, and improvements of pre-processing, storage, and combustion practices to lessen the associated health impacts. However, there has been a historical misconception that wood smoke is something natural that does not cause any serious health effects [67]. Furthermore, even though wood has been considered earlier as a renewable fuel with climate benefits, the validity of this statement depends on forest management policies and several other factors [68].

Over the past years, biomass combustion for residential heating has been increasing, and globally, it has been projected to become the major source of primary particle emissions over the next 5 − 15 years [69]. It is also important to bear in mind that wood burning emits black carbon [70] which is a short-lived climate forcer (SLCF) with a warming effect [71]. Thus, urgent actions are needed.

One of the measures regulating air quality has been the Ambient Air Quality Directive (2008/50/EC), established by the European Union that entered into force in 2015. According to this Directive, the limit value for PM_2.5 aims for a maximum concentration of 25 µg m^− 3 as a yearly average in many parts of the EU. However, and directly based on scientific evidence on the health impacts of fine particulate matter, the health-based guideline issued by the WHO for the annually averaged PM_2.5 concentration is 10 µg m^− 3. These limit and guideline values are substantially higher than the concentrations originating from RWC that were found in this study (max 7.22 µg m^− 3, but mostly < 1 µg m^− 3). In addition, benzo(a)pyrene, which is emitted during small-scale wood burning [72], is restricted as a target value in the EU (the annual mean value may not exceed 1 ng m^− 3). Those target values are exceeded in several countries in the EU [73].

Another type of regulation is emission control. The European legislation for solid fuel boilers (2015/1189/EU) and local space heaters (2015/1185/EU) include emission limit values that have to be met. Unfortunately, these requirements entered into force in 2020 for boilers and will enter into force in 2022 for local space heaters. Moreover, these regulations regulate new boilers, but not the existing ones. For the already existing boilers and wood stoves, filters can be installed into the chimneys by using different technologies (e.g. electrostatic (precipitator) filters, cyclones, etc.) [74].

Thirdly, several Nordic countries have implemented scrapping payments and replacement subsidies [75]. Those programs have provided grants to individual property owners for the replacement of boilers with new low-emissions Eco-labelled products. Some studies have also addressed the effectiveness of stove exchange programs in order to reduce the PM emissions and ambient concentrations with different outcomes depending on factors such as the level of implementation [76‐79].

However, emissions from firewood stoves considerably depend on the user´s behavior and habits, and the quality of the firewood [65]. In fact, all four Nordic countries have been instructed/notified by information campaigns in terms of the use of wood stoves and the proper storage of fuels [31]. Therefore, information campaigns should also promote cleaner domestic burning practices [31] and certifications of the quality of firewood.