Temporal Characteristics and Sources of PM_2.5 in Porto Velho of Amazon Region in Brazil from 2020 to 2022

Abstract

Our study analyzed PM_2.5 concentrations in Porto Velho, Rondônia, during the April 2020 and October 2022 wildfire seasons. This study aimed to evaluate the temporal characteristics of PM_2.5 and the influence of long-distance pollution sources. Using PurpleAir data, we found that the average PM_2.5 concentration was 17.7 ± 24.0 µg m⁻³, with significant spikes in August. PM_2.5 concentrations decreased during the day but rose from nighttime to morning. The PM_2.5 concentration was observed to be distributed at a high level of 36.3 ± 31.1 µg m⁻³ in slow westerly winds. Moreover, even in the dominant northerly wind conditions, a similarly high concentration of PM_2.5 was detected, measuring at 33.2 ± 28.3 µg m⁻³. Air masses mainly originated from northeastern, southeastern, and southern regions, passing through Paraguay and Bolivia. Furthermore, PM_2.5 in Porto Velho was influenced by Brazil’s northern and Central-West areas. To meet the Sustainable Development Goal (SDG) Indicator 11.6.2 for clean air, it is recommended that wildfires in Porto Velho’s northern and western regions be reduced, and more robust deforestation policies are needed in areas with long-distance pollution sources.

1. Introduction

The Amazon rainforest plays a pivotal role as a carbon sink, absorbing greenhouse gases and thereby regulating climate change [1,2]. However, intentional wildfires are set in the Amazon during the dry season to clear land for pasture and agriculture [3,4]. These wildfires are significant carbon emission sources, producing carbon aerosols that amplify the impacts of climate change [5,6,7]. Exacerbated climate change conditions, characterized by more frequent droughts and longer dry seasons, further amplify wildfire occurrences [8].

Wildfires have the potential to adversely affect the health of residents both in the immediate vicinity of the fire outbreak and in downwind regions through long-range transport mechanisms [9,10,11]. Notably, PM_2.5 emissions from wildfires are known to be more harmful than those originating from fossil fuels, and the increase in wildfires correlates with a rise in atmospheric PM_2.5 concentrations [11,12,13]. Attention to PM_2.5 concentrations is crucial as they influence respiratory ailments, cardiovascular diseases, and cognitive function decline in individuals [14,15,16]. Thus, elevated levels of PM_2.5 run counter to Sustainable Development Goal (SDG) 11.6.2, which aims to reduce adverse health impacts from air pollutants, including fine particulate pollution [17,18]. Air quality enhancement through PM_2.5 reduction is pivotal to achieving sustainable development objectives.

An increase in the daily average concentration of wildfire-related PM_2.5 by 10 µg m⁻³ led to a 3.1% rise in mortality rate, whereas an increase in the annual or biennial average PM_2.5 concentration by 1 µg m⁻³ resulted in a 0.14% or 2% increase in mortality rate, respectively [10,15,19,20]. On the other hand, a decrease in PM_2.5 levels by 10 µg m⁻³, or meeting the WHO recommended level of 10 ug/m³, led to an increase in life expectancy [21,22]. While air quality improvement policies focus on reducing PM_2.5 and other pollutants in major urban areas, the effect of local and distant wildfires on urban PM_2.5 concentrations cannot be ignored [13].

Research related to PM_2.5 resulting from wildfires has also been conducted in the Amazon region [23,24]. Wildfire-induced PM_2.5 concentrations have been exceptionally high in the central-western parts of Brazil, affecting the health of children, women, and the elderly aged 65 and over [15,20,25,26]. Porto Velho, situated in the central-western Amazon of Brazil, is the capital of the Rondônia state and is home to approximately 500,000 residents. Rondônia is notably among the regions experiencing the most severe deforestation within Brazil. In the Amazon area, 80% of PM_2.5 emissions are attributed to deforestation [27]. Therefore, understanding the impact of wildfires stemming from deforestation on the sustainable well-being of the region’s inhabitants is of paramount importance.

Previous studies have used satellite data and modeling techniques, presenting challenges in assessing specific areas in the Amazon [20,25,28,29]. Recent observational data for PM_2.5 in the region are outdated, and the sampling period for PM_2.5 exceeded two days, limiting the understanding of the diurnal variation of PM_2.5 from wildfires [25,27].

This research utilizes hourly average PM_2.5 data from PurpleAir. It aims to evaluate the temporal characteristics of PM_2.5 in conjunction with meteorological data during wildfire periods and the influence of long-distance pollution sources. Through this, we expect to contribute to the sustainable goal of clean atmospheric quality (SDG Indicator11.6.2) for the area’s residents.

2. Methods

2.1. PM_2.5 and Meteorological Data

In the Amazon region, inclusive of the Porto Velho area, there are no regulatory-grade monitors. Instead, hourly measurement data regarding particulate matter are acquired using PurpleAir’s Low-Cost Sensor (PALCS). This study utilized PM_2.5 data provided by PurpleAir (https://www2.purpleair.com/ (accessed on 30 July 2023)) from April 2020 to October 2022. The data can be downloaded using the PurpleAir Data Download Tool. The PALCS system comprises two channels, labeled as A and B. Utilizing these two channels, it simultaneously measures the mass concentrations of PM₁, PM_2.5, and PM₁₀, along with particle counts for sizes of 0.3, 0.5, 1.0, 2.5, 5.0, and 10.0 μm. Additionally, the system records meteorological variables, including temperature, relative humidity, and atmospheric pressure. Data from both channels are utilized for the purpose of data verification. It is notably employed in areas experiencing extensive wildfires [30,31,32]. Ardon-Dryer et al. [23] found a good agreement (75% of the comparison had a R² > 0.8) between regulatory-grade monitors and low-cost sensors in wildfire areas. To conduct Quality Control (QC) in our research, we first calculated the differences and relative percentiles of the raw data, precisely the PA_cf1 measurements, obtained from channels A and B. Adhering to the QC criteria proposed by the U.S. EPA, only data exhibiting a difference of 5 µg m⁻³ or less, or a relative percentile of 70% or less, were utilized in our analysis [33]. Consequently, 1.5% of the data were excluded from this study. In studies on wildfire-affected areas, there have been cases where data removal rates through QC for Channel A and Channel B of PALCS were 2.6% and 4%, respectively [33,34]. The correlation of PM_2.5 between Channels A and B showed a good fit (R² = 0.99). Additionally, the Root Mean Square Error (RMSE) showed a 2.8% improvement following QC.

Secondly, an issue was identified with the PALCS’s final data, denoted as PA_atm, which appeared to overestimate measurements compared to regulatory-grade monitors [35]. The U.S. EPA has proposed Equation (1) to adjust the PA_atm data, leveraging the results from regulatory-grade monitors and incorporating PM_2.5 concentration and relative humidity [33].

PM_2.5 = PA_cf1 × 0.524 − 0.0862 × RH + 5.75

(1)

where PM_2.5 is the corrected PurpleAir sensor mass concentration, PA_cf1 is the raw data of PM_2.5 concentration measured with PALCS, and RH is the relative humidity.

In this research, we also adopted this method, rectifying the excessive measurements of PALCS (Equation (2))

PM_2.5 = 0.762 × PA_atm − 1.704

(2)

Meteorological data (wind speed, wind direction, temperature, relative humidity, visibility) were sourced “from timeanddate website (https://www.timeanddate.com (accessed on 30 July 2023))”. The mixing height data were determined using the Vmixing program of the NOAA’s Lagrangian integrated trajectory (Hysplit4 model, https://www.ready.noaa.gov/HYSPLIT.php (accessed on 30 July 2023)). The data utilized in the Vmixing program were based on the GDAS1 dataset (https://www.ready.noaa.gov/data/archives/gdas1/ (accessed on 30 July 2023)), which provides reanalysis data at a 1° by 1° resolution.

2.2. Wildfire Information

Visible Infrared Imaging Radiometer Suite (VIIRS) was developed to enhance the capabilities of the Advanced Very-High Resolution Radiometer (AVHRR) [36]. Currently, VIIRS’s active fire detection products are available in both 750 m and 375 m spatial resolutions [37]. The 375 m resolution product from VIIRS offers improved sensitivity to smaller fires and boasts a higher rate of detection [36]. The VIIRS data were downloaded from the BDQEUMADAS website, which is operated by the Brazilian National Institute for Space Research under the Ministry of Science, Technology, and Innovation (http://terrabrasilis.dgi.inpe.br/queimadas/bdqueimadas (accessed on 30 July 2023)). The data are based on the 375 m resolution Visible Infrared Imaging Radiometer Suite (VIIRS) dataset.

2.3. Back-Trajectory Analysis

To investigate the long-range transport of PM_2.5, the Hysplit4 model was employed to conduct a back trajectory analysis of the air inflow at a height of 500 m above the measurement site over 72 h (https://www.ready.noaa.gov/hypub-bin/trajtype.pl?runtype=archive (accessed on 30 July 2023)).

The meteorological data used for the back trajectory transport were from the GFS (Global Forecast Model) at a resolution of 0.25° by 0.25°. Additionally, to determine the long-range wildfire regions contributing to the PM_2.5 concentrations in the Porto Velho area, we applied the potential source contribution function (PSCF) and concentration-weighted trajectory (CWT) models, leveraging trajectory analysis data. The Openeair package in the R program was used to analyze PSCF, CWT, and cluster analysis in order to evaluate long-range emission sources [27].

2.4. PSCF Model

The PSCF model calculates a conditional probability, assessing the chance that a trajectory moving through a particular geographic grid cell will lead to a concentration surpassing a defined threshold (in this study, 24 h PM_2.5 standard 50 µg m⁻³) when it reaches the receptor site (Equation (3)).

PSCF_ij = m_ij/n_ij

(3)

where m_ij denotes the count of trajectory points that traverse the ij grid cell when the concentration at the site of Porto Velho exceeds a specific criterion, and n_ij represents the total number of air masses that enter the ith and jth cells over the study period.

The domain was defined with a latitude range from 26 S to 1 N and a longitude range from 45 S to 73 S, with a grid size set at 0.5° by 0.5°. While grids with high probability values (PSCF_ij) can be identified as emission sources, those with limited trajectories can exhibit significant uncertainties. To address this limitation, a weighting factor (W_ij) can be employed. In this study, the Openair package in R, which we utilized, incorporates a weighting factor as described in Equation (4) [38]. A weighting factor was applied when the trajectory points were less than twice this average, as suggested by Zeng and Hopke [39].

$${\mathrm{W}}_{\mathrm{i}\mathrm{j}}=\left\{\begin{array}{ll}\hfill 1.0,& {\mathrm{n}}_{\mathrm{i}\mathrm{j}}>2·{\mathrm{n}}_{\mathrm{a}\mathrm{v}\mathrm{e}}\\ \hfill 0.75,& {\mathrm{n}}_{\mathrm{a}\mathrm{v}\mathrm{e}}<{\mathrm{n}}_{\mathrm{i}\mathrm{j}}\le 2·{\mathrm{n}}_{\mathrm{a}\mathrm{v}\mathrm{e}}\\ \hfill 0.5,& {0.5·\mathrm{n}}_{\mathrm{a}\mathrm{v}\mathrm{e}}<{\mathrm{n}}_{\mathrm{i}\mathrm{j}}\le {\mathrm{n}}_{\mathrm{a}\mathrm{v}\mathrm{e}}\\ \hfill 0.15,& {\mathrm{n}}_{\mathrm{i}\mathrm{j}}<0.5·{\mathrm{n}}_{\mathrm{a}\mathrm{v}\mathrm{e}}\end{array}\right.$$

(4)

2.5. CWT Model

The PSCF model determines backward trajectories that display concentrations exceeding specific thresholds (in this study, the 24 h environmental standard of 50 µg m⁻³). However, this approach has an inherent drawback: it might underestimate trajectories with even greater concentrations, making it challenging to differentiate and assess them individually. The CWT model combines the back trajectory with the concentration in the measurement area to express the degree of impact of the back trajectory through each grid cell on the measurement area as a concentration (Equation (5)) [40]. The weighting factor is applied identically, as in the case of PSCF.

$${\mathrm{CWT}}_{\mathrm{ij}}=\left(\frac{1}{{\sum }_{\mathrm{T}=1}^{\mathrm{M}}{\tau }_{\mathrm{i}\mathrm{j}}\mathrm{T}}\right)\times {\sum }_{\mathrm{T}=1}^{\mathrm{M}}\mathrm{C}{\tau }_{\mathrm{i}\mathrm{j}}\mathrm{T}$$

(5)

where CWT_ij represents the 1 h PM_2.5 concentration within grid cell ij at the time the backward trajectory arrives, C is the concentration observed on arrival of trajectory T, τ_ijT is the number of trajectory segment endpoints for an individual backward trajectory (T) as it passes through a 0.5° × 0.5° grid cell ij, i is the index of the trajectory, and M is the total number of back trajectories.

3. Results

3.1. Temporal Variation of PM_2.5

During the analysis period in Porto Velho, the average PM_2.5 concentration was 17.7 ± 24.0 µg m⁻³, equivalent to Brazil’s annual standard of 17 µg m⁻³. Wildfires contribute more than 80% of PM_2.5 during the dry season in the western Amazon region [41]. During the wildfire season, July through October, the average concentration was 34.2 µg m⁻³, seven times higher than the average for the wet season (December–May), with more than 25% of high concentrations above 100 µg m⁻³ occurring in August when the average concentration was 54.5 µg m⁻³ (Figure 1).

In the Amazon basin, including Porto Velho, observational studies on PM_2.5 are scarce, with the majority of the research being dated (

Table 1. Comparison of PM_2.5 average concentrations with those of previous studies in the Amazon rainforest, Brazil.

Location (State)	Year	PM2.5 (µg m−3)	Reference
Manaus (Amazonas)	2008–2012	2.4 (January–June 1.3, July–December 3.4)	Artaxo et al. [42]
Porto Velho (Rondonia)	2009–2012	21.6 (January–June 10.2, July–December 33.0)	Artaxo et al. [42]
Iranduba (Amazonas)	2016	March–April 4.7 August–September 6.7	Fernandes et al. [43]
Rio Blanco (Acre state)	2018–2019	10.1	Coker et al. [31]
Porto Velho (Rondonia)	2020–2022	17.7 (January–June 6.2 July–December 24.6)	This study

). The most commonly employed method is filter-based, and for low-cost sensors (LCS), a laser light source is utilized to measure PM_2.5. In Mauna, located in the Amazonas state, and in Porto Velho of the Rondônia state, samples were collected using filters over durations of two to five days. In Iranduba, also in the Amazonas state, sampling on the filters lasted for 10 h. Conversely, in the Rio Blanco area of the Acre state, real-time measurements were conducted utilizing low-cost sensors with a laser light source, and the analyzed data were based on daily averages. However, these studies yielded low-resolution data, presenting challenges in accurately capturing daily variations in ambient conditions.

In the state of Amazonas, both the Iranduba and Mauna regions maintained notably clean air quality in terms of PM_2.5 concentrations. Emissions in the Iranduba area originated from fugitive dust, open burning, and vehicles, while the Mauna region was influenced by long-range pollutants from forest fires. In Porto Velho, the average PM_2.5 concentration measured from 2008 to 2012 was 28.6% higher than in this study due to a drought in 2010 that increased wildfires.

In the region of Porto Velho, diurnal variations of PM_2.5 exhibited distinct characteristics between the rainy (December–May) season and periods of wildfire (Figure 2). During the rainy season, the peak concentration occurred at 20:00 Local Time (LT), while the minimum concentration was observed at 6:00 LT. In contrast, during wildfire periods, concentrations were generally lower during the day, but from dawn until 7:00 LT, the PM_2.5 concentration surged to its daily maximum. The diurnal variation of PM_2.5 during wildfire periods correlated with the diurnal attributes of wind speed and mixing height. As the wind speed increased and the mixing height rose during the day, PM_2.5 concentrations decreased due to enhanced dispersion and dilution. However, starting from 18:00 LT, when the mixing height reduced and the wind speed weakened, there was a rising trend in PM_2.5 concentrations. By 20:00 LT, with the mixing height dropping to 70 m, PM_2.5 concentrations intensified, continuing this upward trend until 7:00 LT the next morning. This can be attributed to the weaker horizontal dispersion and the challenges in dilution at night when the mixing height is exceptionally low, especially during wildfires. When high concentrations of PM_2.5 are introduced indoors during the nighttime, the duration of their presence indoors increases, leading to elevated levels of PM_2.5 concentration within the indoor environment [43]. During the day, if smoke plumes from wildfires enter the atmosphere and stagnate by evening, the resulting decrease in mixing height and concentration can adversely impact the air quality in Porto Velho. The average PM_2.5 concentrations during the rainy season and wildfire periods were 4.8 ± 5.8 and 34.2 ± 30.1 µg m⁻³, respectively. This underscores the fact that although rising mixing heights can lead to decreased PM_2.5 concentrations, increased wildfires can boost both the peak and minimum concentrations in the diurnal pattern. In August, the average wind speed reduced by 67% compared to that in the rainy season, while the maximum and minimum wind speeds decreased by 63% and 48.5%, respectively. This reduction in wind speed, leading to minimal horizontal dispersion, contributed to the elevated PM_2.5 concentrations. Additionally, the relative humidity in Porto Velho was, on average, 27.2% lower at 16:00 LT during wildfire periods compared to during the rainy season, providing more favorable conditions for wildfire occurrences.

3.2. PM_2.5 Concentrations Based on Wildfire Occurrence and Wind Conditions

Figure 3a shows the average annual number of wildfires within 30 km of the study area and the number of wildfires during the fire season from July to October. From 2012 to 2022, there was an average of 1121 wildfires per year, and trends in wildfire occurrence were not statistically significant (Theil-Sen, p > 0.05). Most wildfires occurred during the July to October fire season, accounting for 92.9% of annual wildfires. In particular, an average of 15.7 wildfires per day occurred in August, when 42% of the annual wildfires occurred.

Figure 3b shows the wildfires that occurred within 30 km of the measurement site during the analysis period in August when PM_2.5 concentrations were the highest. Figure 3c shows the wildfires that occurred during the analysis period in August in Brazil. During a wildfire season, plumes from wildfires in upwind areas can also contribute to PM_2.5 enhancement and those near the measurement site [44]. Higher mixing heights can increase the vertical rise of wildfire plumes, which can increase PM_2.5 concentrations in the atmosphere in the downwind region [45].

Wildfires that frequently occur within a 30 km radius of the measurement station can impact the PM_2.5 concentration levels at the station, contingent on the wind direction and speed. During the period of wildfire occurrences, the wind direction in the Porto Velho region was most frequently northerly and southerly, accounting for 28.4% and 16.4% of the time, respectively (Figure 4). In westerly, southwesterly, and northwesterly winds, light winds of 0–2 ms⁻¹ occurred at an average rate of 95.5%, while relatively strong winds exceeding 4 ms⁻¹ were infrequent, averaging just 0.2%. In contrast, easterly, southeasterly, and northeasterly winds had a comparatively higher percentage of winds exceeding 4 ms⁻¹, averaging 9.2%, while light winds made up an average of 67.4%. Northerly and southerly winds predominantly had light winds of 0–2 ms⁻¹, occurring 77.5% and 78.7% of the time, respectively, with relatively strong winds of 4 m/s or more occurring 2.5% and 6.1% of the time.

As for the PM_2.5 concentration according to wind direction, westerly, northwesterly, northerly, and southwesterly winds exhibited concentrations of 36.8 ± 30.2 µg m⁻³, 35.6 ± 25.2 µg m⁻³, 32.8 ± 27.8 µg m⁻³, and 32.7 ± 28.3 µg m⁻³, respectively. Conversely, the concentrations of PM_2.5 in southeasterly, southerly, easterly, and northeasterly winds were distributed at 26.4 ± 25.6 µg m⁻³, 27.5 ± 26.1 µg m⁻³, 27.5 ± 25.3 µg m⁻³, and 30.1 ± 25.2 µg m⁻³, respectively.

3.3. SDG Indicator 11.6.2 and Air Quality Index (AQI)

Good air quality is vital for the sustainable well-being of citizens. The Sustainable Development Goal (SDG) Indicator 11.6.2 quantifies the “annual mean levels of fine particulate matter (e.g., PM_2.5) in cities (population-weighted)”. In Brazil, the annual standard for PM_2.5 is set at 17 µg m⁻³. For 2021, the Porto Velho region recorded an annual average PM_2.5 concentration of 16.9 ± 25.3 µg m⁻³, meeting the annual environmental standard. Nevertheless, this value is three times higher than the World Health Organization (WHO)’s annual guideline of 5 µg m⁻³.

During the rainy season, the PM_2.5 concentration fully met Brazil’s standard and achieved 96.7% of the WHO’s standard (24 h environmental standard at 15 µg m⁻³). However, during the wildfire season, these concentrations met 78.3% of Brazil’s standard and only 23% of the WHO’s standard. Specifically, in August, the concentrations complied with 48.9% of Brazil’s standard and a mere 8.7% of the WHO’s criteria.

The Brazilian government has established a five-tier air quality index to help citizens understand and respond to health risks associated with air quality conditions. The AQI for PM_2.5 consistently indicated a “Good” index during the rainy season. During the wildfire period, there was a 21.2% occurrence above the “Bad” index, and this proportion increased to 51.1% in August (

Table 2. Wildfire frequency, visibility, and wind speed characteristics by AQI, Brazil.

AQI	Range of PM2.5 Concentration (24 h Average, µg m−3)	July–October			August
		Frequency (%)	Visibility (km)	Wind Speed (ms−1)	Frequency (%)	Visibility (km)	Wind Speed (ms−1)
Good	0–25	49.7	22.7	1.5	16.3	22.6	0.7
Moderate	>25–50	29.1	21.1	1.5	32.6	20.3	0.6
Bad	>50–75	13.3	14.2	1.4	31.5	12.8	0.5
Very Bad	>75–125	7.0	8.9	1.1	17.4	8.8	0.5
Hazardous	>125	0.9	5.2	1.1	2.2	5.6	0.4

). When the AQI deteriorated from a “Good” index to a “Hazardous” index, the wind speed decreased by 0.4 ms⁻¹ and 0.3 ms⁻¹ during the wildfire period and in August, respectively. The average wind speed of 0.7 ms⁻¹ in August corresponds to the daytime (from 12:00 to 18:00 LT) average seen in Figure 2b, while 0.4 ms⁻¹ is on par with the average wind speed at night (from 17:00 to 06:00), illustrating the difference in wind speeds between daytime and nighttime. The lower average wind speeds in wildfire period and August when the AQI index is categorized as “Very Bad” and “Hazardous” suggest a low mixing height and stagnant air conditions, predominantly during nighttime. Visibility, which provides a direct visual indication of air quality status for citizens, deteriorated by 77.1% during the wildfire period and by 75% in August.

3.4. Estimation of PM_2.5 Sources

3.4.1. Conditional Probability Function (CPF) and Polar Annulus

CPF analysis serves as a valuable method for assessing the sources of specific PM_2.5 concentrations, taking into account wind speed and direction. The AQI uses a 24 h average, which exhibited a maximum wind speed difference of 0.4 m/s during wildfire periods. However, CPF analysis employs hourly data, thus revealing a significant fluctuation in wind speeds, ranging from a minimum of 0.1 ms⁻¹ to a maximum of 13 ms⁻¹ (Figure 5). When the hourly PM_2.5 concentration was measured at 14 µg m⁻³ or lower, the predominant winds were either southeasterly, ranging from 8 to 14 ms⁻¹, or easterly winds of 4 ms⁻¹ or higher (Figure 5a). For PM_2.5 concentrations between 15 and 25 µg m⁻³, northwesterly winds exceeding 4 ms⁻¹ were the main contributing factor (Figure 5b).

At moderate air quality levels, where PM_2.5 concentrations were between 25 and 50 µg m⁻³, weaker winds of 2 ms⁻¹ or less were influential (Figure 5c). However, at poor air quality levels with PM_2.5 concentrations ranging from 50 to 75 µg m⁻³, strong northeasterly winds between 12 and 14 ms⁻¹ were prevalent (Figure 5d). Moreover, in instances where PM_2.5 concentrations exceeded 75 µg m⁻³ (Figure 5e), the distribution occurred under conditions of low wind speed and stagnant air.

The CPF analysis demonstrated that the distribution of PM_2.5 concentrations varies based on the characteristics of wind direction and speed. The polar annulus approach is used to analyze how PM_2.5 concentrations change over time based on the wind direction (Figure 6). In the Porto Velho region, higher concentrations of PM_2.5 were observed from nighttime to morning in the northerly wind sector and from dawn to daytime over an extended period in the westerly wind sector (Figure 6a). In the easterly wind, high concentrations of PM_2.5 were distributed from late dawn to morning. In the Porto Velho region, daytime PM_2.5 concentrations are comparatively lower, largely driven by the strong easterly wind influx. Conversely, the elevated PM_2.5 concentrations observed in the westerly wind patterns can be attributed to the prevailing weaker wind velocities (Figure 6b). Visibility is a good indicator of air quality, including pollutants such as PM_2.5 [46]. In this region, the visibility was excellent when the southeasterly wind system, which has high wind speed and low PM_2.5 concentration, was introduced (Figure 6c). Furthermore, atmospheric pressure was observed to be relatively higher in the southerly wind and lower with northerly and westerly winds (Figure 6d). Such observations imply that wildfires may occur more frequently in the north and west than in the southern regions.

3.4.2. Back Trajectory Analysis

In the Amazon rainforest, PM_2.5 from forest fires can affect hundreds of kilometers of horizontal distances [11]. Moreover, plumes of wildfire smoke can rise as high as 2500 m and, depending on weather conditions, influence PM_2.5 enhancement in downwind areas [47]. Back trajectory analysis can assess the contribution of PM_2.5 from wildfires to the target area through horizontal and vertical transport [44,48].

Figure 7a displays the probability of trajectories with PM_2.5 concentrations exceeding 50 µg m⁻³ crossing each grid. This suggests pollution sources in the North Region (comprising the states of Amazonas and Pará). The CWT analysis further indicated that areas contributing to the PM_2.5 in Porto Velho include parts of the North and Central-West regions (the states of Mato Grosso, Figure 7b). From 2020 to 2022, in Mato Grosso, Para, and Amazonas, the areas damaged by wildfires constituted 42.9% (471,721 ha), 0.7%, and 0.09%, respectively, of the total wildfire-affected areas in Brazil [49].

Figure 7c shows the five clusters of winds entering the station, with the influence of northeasterly winds (Clusters 1 and 2), southeasterly winds (Clusters 3 and 4), and southerly winds (Cluster 5). PM_2.5 concentrations from Clusters 2, 3, and 4 were 38.6, 34.6, and 37.6 µg m⁻³. Cluster 1 was 28.4 µg m⁻³, and Cluster 5 was 24.3 µg m⁻³. Cluster 5 reflected the longest pathway from Paraguay and Bolivia. Clusters 2 and 4 had relatively fast inflows from Para and Mato Gosso. Clusters 1 and 3 had shorter travel paths, reflecting the impact of wildfires in areas close to Porto Velho. Deforestation is a major cause of wildfires in the Brazilian Amazon region, and deforestation occurs along roads. Cluster 1 and Cluster 2 were found to be affected by wildfires occurring in the area around highways 319 and 230 [50,51,52]; therefore, policies to reduce wildfires are needed to reduce exposure to high concentrations of PM_2.5 for citizens along roads and in downwind areas [48,53]. Cluster 2, Cluster 4, and Cluster 5 traveled relatively long distances through wildfire areas, contributing to the deterioration of air quality in the region. This is because the more wildfires that occur, the more likely air traveling through the wildfire zone will contain high concentrations of PM_2.5 [54].

4. Discussion

The Porto Velho region is known for its recurrent forest fires due to deforestation. Nevertheless, with a population of approximately 500,000, research on their PM_2.5 exposure has been notably limited. In particular, there has been no detailed study on the hourly distribution of PM_2.5 and the areas most affected by forest fires regarding air quality.

The PM_2.5 concentration in the Porto Velho region was consistent with Brazil’s annual environmental standards. However, from July to October, the period recognized as the wildfire period, there was a notable increase in PM_2.5 concentrations attributed to the rise in wildfires, wind velocities, and mixing height.

During wildfire periods, PM_2.5 concentrations increased from 20:00 to 07:00 LT, corresponding to mixing height and wind speed changes. While decreased mixing height intensified PM_2.5 concentrations, the continued rise until 07:00 LT indicates wildfires as a significant PM_2.5 source during the night and early morning. Wildfires during these hours further enhance PM_2.5 concentrations due to limited vertical dispersion [55,56].

During the day, increased wildfire activity can elevate PM_2.5 concentrations [57,58]. This can intensify PM_2.5 concentrations in downwind areas, especially in the afternoon when wind speed drops and mixing height decreases [58,59]. These patterns are pronounced in diurnal variations, particularly during wildfire seasons and in August. Despite the increased mixing height in August, the PM_2.5 concentrations peaked due to frequent wildfires and a 67% decrease in wind speed compared to those of the rainy season, hindering PM_2.5 dispersion.

In Porto Velho, PM_2.5 concentrations during wildfire seasons need reduction to meet SDG Indicator 11.6.2. The air quality index indicated “Bad” or worse levels in 21.2% of the period, with August reaching a concerning 51.1%. This poses significant health risks to its inhabitants.

In the Porto Velho region, northerly and westerly winds correlated with increased PM_2.5 concentrations due to frequent wildfires. In contrast, stronger southerly or easterly winds led to decreased PM_2.5 concentrations and clearer visibility, with the highest levels observed in the early to mid-morning. The main long-distance pollution contributors were wildfires from Amazonas, Para, and Mato Grosso. Cluster analysis also highlighted air influx from northeastern highways, central areas, and regions in Paraguay and Bolivia.

In the Amazon rainforest region, including the Porto Velho area, the absence of regulatory-grade monitoring stations precludes direct comparison with PurpleAir’s measurement data. However, the PM_2.5 emitted from the wildfires in the Amazon can settle in areas where residents of downwind regions live [47]. Low-cost sensors allow for real-time data collection from multiple monitoring points, enabling the construction of diverse datasets. Through this, the accuracy of the analysis can be enhanced [33]. Consequently, evaluating emission source characteristics through even low-cost sensors is an essential preliminary study to formulate policies to reduce PM_2.5.

Local wildfires from the north and west, along with long-range transport from wildfires from Mato Grosso, Amazonas, and Para, impact PM_2.5 concentrations in the region. To achieve clean air goals (SDG Indicator 11.6.2), mitigating these wildfires and reducing deforestation in Brazil’s northern and central areas is crucial. Given the immediate and long-term impacts of elevated PM_2.5 levels, the results of this study underscore the urgent need for comprehensive policies aimed at improving air quality.