Particulate matters emitted by an firewood cook stove

4. Results and Discussion

The PM(s) concentration is presented in Fig. 3 with marks for the cooking period and heavy rain in late April 30. with a The data were averaged into a 20-minute interval. For a higher-resolution data, Fig. 1 shows the 1-minute interval concentration of PM(s).

20-min resolution
Fig. 3: PM2.5 and PM10 of a 3-day sampling campaign with 20-min resolution with marking for cook time and heavy rain
1-min resolution
Fig. 4: PM2.5 and PM10 of a 3-day sampling campaign with 1-min resolution

Fig. 5 presents the main outcomes of this study. Each box shows the median, 25th and 75 percentiles with the box. The ends of error bars show the minimum and the maxium values. The dots outside the error bars are treated as outliers by seaborn plotbox.

As expected, the peaks of higher PM(s) concentration are occurred on cooking time and supporting the presumption that a higher PM(s) concentration emitted by active cook stoves. The outliers located mostly during the cooking periods, which could interpret as PM(s) concentration when the users blew air to stoke the fire.

day1 day2 day3
Fig. 5: The average of PM2.5 and PM10 from the immediate background concentration (two hours before and after cooking), the daily background concentration (daily average excluding the cooking period), and the cooking period when fire is active.

Table 1 lists the concentration of PM(s) while cooking and the ratios to IBC and ABC. The average concentration included all the points in Fig. 1 and not excluding the outliers as in the boxplot. The ratios of cooking to ABC are consistent during the 3-day run. PM(s) concentration emitted by this cook stove was 64-95% more than the ABC. The ratios with IBC in a wider range, from 6% in the day 2 to 125% more in day 3. The wide range to IBC reflects a difficult nature to measure IBC. A detailed description of measuring IBC is useful to process data and post analysis.

Table 1: Average values and ratios of PM(s) while cooking to the average of before and after cooking
episode day 1 day 2 day 3
Target PM2.5 PM10 PM2.5 PM10 PM2.5 PM10
Cooking (µg/m3) 102 118 72 84 74 90
Cooking/IBC 1.38 1.28 1.16 1.06 2.24 2.25
Cooking/ABC 1.76 1.64 1.71 1.65 1.95 1.91

A significant increment of PM(s) while cooking points to a higher risk to a long term exposure for those who cook or tend the fire. The height of the sampling kit is similar to the base of the stove. This setup potentially reduced the actually PM(s) concentration that human exposed because of the height of the human's head when hovering over the stove to check the foods. In addition, firewood stoves are used to often cook other foods not only boiling water in this case. Thus, the risk of PM(s) from frying foods, for instance, induces even greater risk to those who are close to the stoves.

The background concentration (BC) variation is less expected. To the author's surprise, the BC varied wildly from around 80 µg/m3 in the first day down to 40-70 µg/m3, and around 10-40 µg/m3 in the third day. The assumption that a stable BC of PM(s) in a rural area is not supported by the obtained data in this campaign. This observation reflects more complex nature of PM(s) and the author is skeptical to draw any meaningful and reliable conclusion from a relatively short sampling. A longer sampling period with other sensors could provide more insight into the absolute and variation of the PM(s) value.

The PMS7003 is a low-cost sensor using the laser-scattering method. This ~$15 sensor is convenient to a quick sampling and yields a second resolution. However, the sensor has inherent drawbacks, most importantly, a lower accuracy than PM(s) monitors using BAM or gravimetric methods. The PMS7003 tends to overestimate PM(s) concentration about two times than a BAM monitor as observed by the author. A recent study published on May 22, 2019, concurred the author's observation that the PMS7003 over-estimated the research-grade equipment TEOM 1400a by a factor of 2.2. The readers should be cautious when interpreting the absolute concentration of PM2.5 and PM10 by PMS7003 to the health standard and the national standards.

Temperature and relative humidity presented in Fig. 6 is not useful to delf into the PM(s) variation and changing BC. The environment data, however, reflects a heavy rain in the late afternoon of April 30 by a dropping 8 °C in temperature and increasing RH by 15%.

Fig. 6: Temperature and relative humidity of sampling location. The x-axis in this graph covers less time than the Figs.1-2 above because of additional time for installing the environment sensor

Besides sampling the rural home as the primary location, the author did a trial run in Hong Linh town. The PM(s) is shown in Fig. 7. Of the total PM10, the PM2.5 consisted of 0.86 ± 0.05. The profile in the rural home is 0.83 ± 0.07. These data are in aggreement with EPA report, Fig. 2.4 that the PM2.5 is the dominant group in PM10 when the emission is from combustion or from secondary pathways. The secondary pathways are the combination of precursors in the atmosphere to form PM2.5.

Hong Linh
Fig. 7: Pilot sampling in Hong Linh town (Ha Tinh), no cooking nearby. The sampling is about 100-m from the national highway (A1)

The author did two rounds of search on Google Scholar to find a similar study but did not find the research that compares the PM(s) emission by the stove fire. Most research mentioned the absolute concentration of PM2.5 with solid fuels from 154-6901 µg/m3[1]. Pokhrel el at. (2015) evaluated different types of cooking fuels and found that using biomass emitted the most PM(s) as 656 µg/m3 and the least is with electricity as 80 µg/m3. Another research in China by Hu el at. (2014), Table 2 listed 6 types of fuels with PM2.5 concentration with 5 types of stoves. With vented stoves, the PM2.5 is cut by 35-55% with the fuel as wood. This finding suggested well ventilation provides immediate mitigation to personal exposure to PM2.5 emitted from fire pits.