Urban Air Quality Deterioration in Manaus During the 2023 Drought: Long-Range Wildfire Smoke Transport and Urban Sustainability

Abstract

Sustainable urban air quality in tropical cities is threatened by interactions between climate change, extreme drought, and long-range wildfire smoke transport. This study investigated the causes of PM_2.5 pollution in Manaus, Brazil, under El Niño conditions during the 2023 drought, focusing on long-range wildfire smoke transport. The links among hydroclimatic drying, wildfire activity, and urban air quality were examined using hourly PM_2.5 observations, meteorological data, long-term climate records, MODIS hotspot and fire radiative power (FRP) data, and air-mass trajectory analyses. Significant long-term warming, decreasing precipitation, and a declining standardized precipitation evapotranspiration index were observed around Manaus during 1981–2024, indicating persistent drying. In 2023, severe drought and increased wildfire activity caused an annual mean PM_2.5 concentration of 15.09 µg m⁻³. Directional analyses, upwind FRP, potential source contribution function, and backward trajectories consistently highlighted the eastern and southeastern source regions approximately 500–2200 km from Manaus. These results indicated that PM_2.5 levels were more sensitive to spatial alignment between upwind fires and prevailing winds than to total fire activity alone. In conclusion, the 2023 PM_2.5 surge was driven by long-range wildfire smoke transport under intensified drying and drought, with implications for urban sustainability, public health, and climate-resilient early warning systems.

1. Introduction

Wildfires, previously considered temporary disturbances confined to forest ecosystems, are increasingly recognized as large-scale environmental crises that directly affect air quality and public health. Wildfire smoke can travel hundreds of kilometers from the source area, sharply increasing fine particulate matter (PM_2.5) concentrations and health risks for locals [1,2]. Because health effects can occur even at relatively low concentrations, in its 2021 air quality guidelines, the World Health Organization (WHO) recommended annual mean guidelines for PM_2.5 at 5 µg m⁻³ and the 24 h guideline at 15 µg m⁻³ [1]. Exposure to wildfire-related PM_2.5 can be significantly associated with respiratory and cardiovascular diseases, and increased mortality risk [2]. This growing threat to urban air quality challenges sustainable urban development, particularly in tropical cities, where climate-driven wildfire smoke can impact public health and urban livability.

This issue has become increasingly important in the Amazon. Although the Amazon is a key region in the global carbon and hydrological cycles, its ecological stability has recently weakened with rising temperatures, drying, recurrent extreme droughts, land-use changes, and increasingly interacting fires [3]. In 2023, the Amazon simultaneously experienced record-breaking heat and drought, and the Rio Negro water level in Manaus decreased to its lowest level since 1902 [3]. The World Meteorological Organization classified the 2023/2024 El Niño event as one of the strongest on record, further amplifying the ongoing warming and hydrological instability [4]. Therefore, recent air quality problems in the Amazon should be considered as overlapping climate change and extreme climate events, instead of seasonal variability alone [3,4].

Manaus is a critical case study for understanding how these changes translate into urban air quality. Road corridors, land clearing, forest degradation, and fires have gradually expanded in Manaus, a large city located at the heart of the tropical rainforest and its surrounding areas [5,6]. The central Amazon previously had one of the cleanest background atmospheres in the world, with low PM_2.5 concentrations in the wet and dry seasons in preserved forest areas [7]. However, these findings mainly reflect the background atmospheric conditions in conserved forests, and it cannot be assumed that a large metropolitan city, such as Manaus, continues to maintain the same level of atmospheric cleanliness under conditions of high temperature, drought, and smoke transport [5,7]. Earlier assessments of the Amazon as having a clean atmosphere underscore the need to re-examine how recent climatic deterioration and expanding wildfires affect urban air quality [6,7].

During strong El Niño periods, the effects of wildfires on the air quality of the Manaus metropolitan area become evident. A study of the 2015 El Niño event reported that biomass-burning events increased in the Manaus metropolitan region, accompanied by significant increases in levels of wildfire-related indicators such as carbon monoxide and levoglucosan [8]. A study on PM in Manaus showed that urban air quality can be linked to surrounding fire activity [9]. Therefore, under recently observed conditions, in which background temperatures are increasing and drying is intensifying, wildfire smoke is likely to act as a factor causing temporary increases in PM_2.5 in Manaus and as a structural driver that heightens the risk of sustained exposure for locals [3,8,9].

This issue is directly related to Sustainable Development Goals (SDGs). SDG target 11.6 specifies that focus should be given to air quality in reducing the per capita environmental impact of cities, and SDG indicator 11.6.2 uses population-weighted urban PM_2.5 concentration as a key metric [10]. SDG target 3.9 aims to reduce illnesses and mortality caused by environmental pollution, including air pollution [11], and SDG 13 emphasizes the need for action to address climate change and its impacts [12]. Therefore, the growing likelihood that air quality in large tropical cities such as Manaus will deteriorate under the combined influence of climate change and increasing wildfires is a local environmental issue and is directly linked to the core SDG agendas of sustainable cities and healthy populations [10,11,12].

In this context, this study aimed to interpret the increase in PM_2.5 observed in Manaus during the 2023 El Niño-induced drought in relation to the combined effects of long-term warming and drying, wildfire activity, and smoke transport. This study moved beyond the conventional view based on the region’s past relative atmospheric cleanliness to determine the extent to which urban air quality in Manaus has been affected by recent rising temperatures and intensified drought. Moreover, this study identified the importance of externally transported wildfire smoke in reframing air quality problems in Amazonian cities as a sustainability issue shaped by the overlapping influences of climate change, wildfires, and atmospheric transport.

2. Materials and Methods

2.1. Study Area

The study area, Manaus (3.1° S, 60.0° W), Amazonas, Brazil, is the largest city in the Amazon rainforest and is surrounded by extensive forested areas along its urban periphery, making it highly vulnerable to wildfire smoke during the dry season. The overall analysis period of this study spanned January 2021 to December 2024, whereas the long-range transport analysis focused on the dry season (July–November) in 2023.

The PM_2.5 observations used in this study were obtained from one monitoring site in Manaus. Therefore, the dataset was primarily used to examine the temporal variations in PM_2.5 and their relationship with regional wildfire smoke transport, rather than to characterize the full intra-urban spatial variability of air quality across Manaus.

2.2. PM_2.5 and Meteorological Data

The PurpleAir PA-II (PurpleAir, Inc., Draper, UT, USA) was equipped with dual Plantower PMS5003 light-scattering sensors to provide real-time particulate matter concentrations. The data were collected on an hourly basis in UTC using the PurpleAir API.

The PurpleAir data were corrected using the EPA national correction equation developed by Barkjohn et al. (2021) [13], expressed as pm25cal = 0.524 × PM2.5_cf_1 − 0.0862 × relative humidity (RH) + 5.75. Quality assurance and quality control (QA/QC) were applied following the EPA/AirNow PurpleAir cleaning framework, as far as possible with the available hourly dataset. Records were excluded when the difference between the two internal Plantower channels (A and B) exceeded both 5 µg m⁻³ and exceeded 70% of their mean. Timestamps were converted from UTC to Manaus local time (AMT, UTC−4) when necessary. The overall weighted mean QA/QC pass rate for 2021, 2022, 2023, and 2024 was 95.23%, with annual pass rates of 83.5%, 99.8%, 99.1%, and 97.9%, respectively.

To evaluate the applicability of the EPA correction equation under Amazonian humidity conditions, an RH-stratified sensitivity analysis was conducted for the 2023 dry season. The mean RH during this period was 64.9%, whereas observations with RH ≥ 95% accounted for only 1.2% of the dry-season hours. Mean PM_2.5 concentrations were lower during high-humidity periods (11.57 µg m⁻³ for RH ≥ 95% and 18.46 µg m⁻³ for RH ≥ 85%) than during lower-humidity periods (27.10 µg m⁻³ for RH < 95% and 28.08 µg m⁻³ for RH < 85%). In addition, the correction ratios (EPA-corrected PM_2.5/raw PM_2.5) were similar under RH ≤ 80% and RH > 80% conditions (0.572 and 0.560, respectively), indicating consistent correction performance across the observed humidity range.

Despite the above QA/QC procedure, the Plantower PMS5003 optical scattering sensors exhibit reduced accuracy under high PM_2.5 concentrations, primarily due to particle coincidence effects. Measurement uncertainty increases substantially above approximately 100–200 µg m⁻³ [14,15]. During the 2023 dry season, minimal hourly observations exceeded this range, including a recorded maximum of 436.2 µg m⁻³; therefore; these extreme values should be interpreted with caution. However, the primary conclusions of this study, including interannual comparisons, directional analyses, and the identification of transport-driven episodes, are based on temporal and directional patterns rather than absolute concentration values and are not expected to be materially affected by this limitation.

Meteorological data were obtained from hourly observations at the INMET Manaus Station (A101). Wind direction, wind speed, air temperature, relative humidity, and precipitation were used in the analysis. These meteorological data were merged with the PM_2.5 data to analyze pollution characteristics using wind direction and wildfire intensity in the upwind sector.

To assess the long-term climatic background, 3 × 3 grid-averaged data from NASA POWER (https://power.larc.nasa.gov/) were used to analyze the changes in temperature and precipitation from 1981 to 2024. Drought intensity was quantified using the standardized precipitation evapotranspiration index (SPEI) [16].

2.3. Wildfire Hotspot and Fire Radiative Power (FRP) Data

Wildfire data were analyzed using MODIS active fire hotspot data from the NASA FIRMS [17]. Only hotspots with a confidence level of ≥80% were included. Dry-season hotspot counts were calculated within radii of 100 km, 200 km, 300 km, 500 km, and 700 km, centered on Manaus. In addition, the wildfire intensity by direction was assessed using hotspot location and FRP. The upwind FRP was defined as the daily sum of the FRP from hotspots located in the upwind sector, based on the wind direction for each day.

The MODIS active fire product (Collection 6.1) provides thermal anomaly/fire detection at a spatial resolution of approximately 1 km and a temporal resolution of up to twice daily from the Terra and Aqua satellites. In this study, daily hotspot counts and summed FRP were used for the analysis. Because satellite fire detection can be affected by cloud cover, heavy smoke, and atmospheric interference, certain fires may be missed during certain periods, leading to conservative estimates of hotspot counts and FRP.

2.4. Directional and Statistical Analyses

For the long-term hydroclimatic series, trend slopes were estimated using the Theil–Sen estimator and statistical significance was assessed using the Kendall trend test. The confidence intervals for the Theil–Sen slopes were calculated at the 95% level. Distributions of wind and PM_2.5 by wind direction were evaluated using pollution rose analysis in the openair package [18]. Relationships between climatic variables and hotspot counts and between upwind FRP and PM_2.5, were evaluated using Spearman rank correlation analysis. Correlation coefficients and corresponding p-values were used to assess statistical significance. Spearman correlation was applied because the variables analyzed showed skewed and episodic distributions with occasional extreme values, making the nonparametric rank-based approach more appropriate for the study objectives. Hourly PM_2.5 concentrations were classified according to air quality index (AQI) categories, and the numbers of days exceeding the WHO daily guideline and the Brazilian 24-h standard were calculated.

2.5. Backward Trajectory Analysis

To trace the origin of the air masses, 72-h backward trajectories were calculated for the 2023 dry season (July–November) using the NOAA HYSPLIT model [19]. GDAS 1° × 1° meteorological fields were used as the input data. Trajectories were initiated every 6 h (00:00, 06:00, 12:00, and 18:00 UTC) from Manaus at a starting altitude of 500 m above ground level (AGL). The vertical motion method was used to input the model data (method 0). To identify the dominant inflow pathways, cluster analysis was performed using the total spatial variance in HYSPLIT. Cluster analysis was applied separately to the full dry season with days exceeding the WHO daily PM_2.5 guideline and the top 5% PM_2.5 episode days.

An arrival height of 500 m AGL was selected because this altitude is within the planetary boundary layer and is commonly used to represent regional wildfire smoke transport. To evaluate the sensitivity of the trajectory results to the selected arrival height, additional trajectory analyses were conducted using 100 m and 1000 m AGL, and the resulting transport pathways showed similar easterly source regions, indicating that the inferred transport patterns were not substantially affected by the selected arrival height.

3. Results

3.1. Long-Term Drying and the 2023 Drought Background

Although the hydroclimatic conditions around Manaus showed substantial interannual variability during 1981–2024, the overall long-term trend was toward drying (Figure 1). Mean annual temperature increased by +0.280 °C per decade (95% CI: 0.140 to 0.436; Kendall trend test p = 0.00012), whereas annual total precipitation decreased by −153.5 mm per decade (95% CI: −236.9 to −63.8; p = 0.00073). Annual SPEI decreased by −0.408 per decade (95% CI: −0.607 to −0.184; p = 0.00023), indicating a long-term reduction in water availability. The time series in Figure 1 shows that not all years are dry in the same manner; however, despite repeated interannual fluctuations between wet and dry years, the long-term trend lines clearly indicate rising temperatures, decreasing precipitation, and a declining SPEI. In other words, the recent drought and air quality deterioration cannot be interpreted as problems confined to a single anomalous year but rather as events that emerged against a long-term dry climatic background that has accumulated over time.

Figure 1c shows that relatively wet years occur frequently from the mid-1980s to the mid-1990s, whereas negative SPEI values become more frequent after the 2000s, with severely dry conditions in the mid-2010s and 2024. This indicates that long-term drying did not proceed linearly but rather reflected a gradual shift toward drier conditions superimposed on strong interannual variability. Therefore, the recent climate change around Manaus is better understood as a process in which the intensity and frequency of dry years have gradually increased during recurring wet–dry oscillations.

Against this long-term background, the drought conditions in 2023 and 2024 indicated an abrupt recent deterioration (Figure 2a). The mean SPEI for the dry season remained within the normal to wet range in 2021 (+0.456) and 2022 (+0.897), but decreased sharply to −1.267 in 2023, entering the moderate drought category, and declined to −1.732 in 2024, approaching severe drought conditions. In particular, the dry season mean temperature in 2023 was 1.35 °C higher than that in 2022, and although precipitation was greater, the SPEI was lower. This suggests that the 2023 drought cannot be explained simply by a lack of precipitation, but rather by the combined effects of increased potential evapotranspiration and worsening water balance at higher temperatures [3,16]. In other words, the drying observed in 2023 should be understood not as a simple precipitation deficit resulting from increased atmospheric moisture demand and soil moisture loss intensified by high temperatures.

Figure 2 shows that the drought background temporally coincides with changes in air quality. During the second half of 2023, the SPEI declined continuously and fell below the moderate drought threshold during the late dry season, whereas the daily mean PM_2.5 increased markedly during the same period and frequently exceeded the WHO daily guideline of 15 µg m⁻³. In contrast, during the dry seasons, when the SPEI remained within the normal or wet range, an increase in PM_2.5 was relatively limited. Thus, the increase in PM_2.5 in 2023 can be interpreted as the outcome of extreme drought, formed against a background of long-term drying, that contributed to worsening air quality in the late dry season [3].

The severe drought conditions observed during 2023–2024 are associated with a strong El Niño event. El Niño suppresses rainfall and enhances drought conditions across large portions of the Amazon Basin through changes in atmospheric circulation and moisture transport. Such drought conditions increase vegetation dryness and wildfire susceptibility, thereby creating favorable conditions for enhanced smoke emissions and subsequent deterioration in air quality. Large-scale ocean-atmosphere variability, including tropical Atlantic sea-surface temperature anomalies, may further modulate these hydroclimatic conditions.

3.2. Wildfire Activity and Its Climatic Drivers Around Manaus

During the 2021–2024 dry season, the wildfire activity around Manaus showed clear interannual differences depending on the distance range and timing, with the strongest increase occurring in 2023 within the near- and mid-range areas surrounding the city (

Table 1. Annual wildfire hotspot counts around Manaus by cumulative radius.

Radius	2021	2022	2023	2024	2023/2021
≤100 km	641	860	2140	876	3.3×
≤200 km	1376	2232	5020	2628	3.6×
≤300 km	2034	3000	7433	4153	3.7×
≤500 km	10,445	16,406	20,632	24,734	2.0×

and Figure 3a). The number of dry-season hotspots within a 100 km radius increased from 641 in 2021 and 860 in 2022 to 2140 in 2023, reaching 3.3 times that in 2021. Over the same period, the number of hotspots within the 200 km and 300 km radii increased by factors of 3.6 and 3.7, respectively, whereas the 2023 hotspot count within the 500-km radius was more than twice that in 2021. This pattern shows that the expansion of wildfire activity in 2023 was not confined to areas immediately adjacent to Manaus, but intensified simultaneously across the broader near- to mid-range zone surrounding the city. In contrast, in 2024, hotspot counts within the 100–300 km range decreased relative to those in 2023, whereas higher levels were maintained at larger radii, suggesting that the total amount of wildfire activity and its spatial distribution experienced annual variations.

The monthly variation highlighted the distinctiveness of 2023 (Figure 3b). In 2021 and 2022, hotspots within 100 km showed relatively short peaks, mainly concentrated between August and October, followed by a rapid decline. However, by 2023, hotspot numbers increased in July, peaked in September and October, and remained high throughout November. This indicates that in 2023, the wildfire season began earlier and lasted longer, creating conditions in which smoke generation and accumulation could persist in the late dry season. In other words, wildfire activity in 2023 was characterized by a greater number of hotspots and intensified timing and persistence.

These spatiotemporal patterns are consistent with the long-term drying trend and 2023 drought background described above. Climatic conditions and wildfire activity were compared, and mean annual temperature and the number of dry season hotspots within a 100 km radius were significantly positively correlated (r = 0.591, p = 0.013), whereas annual precipitation and hotspot counts were significantly negatively correlated (r = −0.679, p = 0.003). In 2023, high temperatures, relatively low precipitation, and numerous hotspots occurred under both conditions. These results indicate that the expansion of wildfire activity during the 2023 dry season was closely associated with drier climatic conditions [3,17] and that long-term warming and a worsening water balance contributed to the intensification of regional wildfire activity.

Figure 3 shows that in 2023, the scale, onset, duration, and spatial concentration of wildfire activity around Manaus all intensify simultaneously. In particular, the rapid increase in hotspot counts across the near- to mid-range areas surrounding the city and the persistence of these elevated levels in the late dry season provide an important context for interpreting the subsequent analyses of PM_2.5 increases and smoke transport. Therefore, the deterioration of air quality in Manaus in 2023 should be understood as the result of changes in emissions from within the city, and as a phenomenon linked to the intensification of regional wildfire activity that began early and persisted for an extended period under dry climatic conditions.

3.3. PM_2.5 Increase and Air Quality Episodes in Manaus

PM_2.5 concentrations in Manaus showed clear interannual differences during 2021–2024, with the highest levels observed in 2023 (

Table 2. Annual distribution of AQI categories from 2021 to 2024 (%).

AQI Category	2021	2022	2023	2024
Good	94.3	79.8	66.4	78.4
Moderate	5.5	18.1	24	17.1
Unhealthy for Sensitive Groups	0.1	1.9	5	2.7
Unhealthy	0	0.3	4.1	1.9
Very Unhealthy	0	0	0.3	0
Hazardous	0	0	0.3	0

and Figure 4). The annual mean PM_2.5 concentration increased from 5.08 ± 4.17 µg m⁻³ in 2021 and 8.38 ± 8.71 µg m⁻³ in 2022 to 15.09 ± 25.13 µg m⁻³ in 2023, and then decreased to 9.93 ± 12.86 µg m⁻³ in 2024. The dry-season mean showed the same pattern, with values of 7.32 µg m⁻³, 12.12 µg m⁻³, 23.90 µg m⁻³, and 17.78 µg m⁻³ in 2021, 2022, 2023, and 2024, respectively. The median ratio of the dry to wet season increased from 1.77 in 2021 to 2.74 in 2022 and 4.19 in 2023, before declining to 2.94 in 2024, indicating a strong seasonal contrast in PM_2.5 concentration in 2023. This pattern indicated that 2023 recorded the highest annual mean PM_2.5 level and that pollution was more strongly concentrated during the dry season.

The monthly variation highlighted the distinctiveness of 2023 (Figure 4). In 2021, PM_2.5 remained generally low across seasons, whereas in 2022, a temporary increase was observed mainly in August and September; however, its persistence was limited. In contrast, in 2023, the concentrations increased in July and remained high throughout September to November. Monthly mean PM_2.5 concentrations in October and November 2023 were 40.00 µg m⁻³ and 36.02 µg m⁻³, which were 2.8 and 3.8 times higher, respectively, than those in the same months of 2022. In 2024, although the annual mean and dry-season mean were lower than in 2023, the monthly mean in August reached 28.8 µg m⁻³, exceeding the August 2023 level and indicating that the risk of high-concentration episodes remained during specific periods. Overall, 2023 was characterized by an earlier onset of concentration increase and a longer duration of high-PM_2.5 conditions.

The frequency of guideline exceedance and AQI distribution was most unfavorable in 2023 (Table 2 and Figure 4). The number of days exceeding the WHO daily guideline of 15 µg m⁻³ was 5 in 2021, 43 in 2022, 85 in 2023, and 75 in 2024, whereas the number of days exceeding the Brazilian 24-h standard of 50 µg m⁻³ increased from 0 in 2021 to 1 in 2022 and 16 in 2023. In the AQI distribution, the Unhealthy, Very Unhealthy, and Hazardous categories appeared simultaneously for the first time in 2023, and the maximum hourly PM_2.5 concentration peaked at 436.2 µg m⁻³. The mean concentration of the top 5% dry-season episodes was 25.9 µg m⁻³ in 2021, 44.9 µg m⁻³ in 2022, 131.3 µg m⁻³ in 2023, and 77.8 µg m⁻³ in 2024, showing that the intensity of high-pollution episodes in 2023 was markedly greater than in other years. These results indicate that the deterioration of air quality in 2023 was not limited to an increase in mean concentrations, but was the most severe year in terms of the frequency of exceedances and the intensity of extreme values.

The high-concentration period in 2023 can be divided into two major waves in the time series: 28 September to 15 October and 31 October to 6 November. These two waves indicate that the high-PM_2.5 episodes in 2023 were not a single isolated event, but part of a series of recurrent pollution periods during the late dry season. Because these episodes were mainly concentrated in October and November, the PM_2.5 accumulation was strongest in the late dry season, when drought and wildfire activity had already accumulated.

The diurnal variation showed a distinctive pattern in 2023 (Figure 5). In all years, PM_2.5 was consistently higher from nighttime to early morning and lower in the afternoon; however, this contrast was greatest in 2023. The nighttime peak concentration during the dry season was 10.1 µg m⁻³ in 2021, 16.2 µg m⁻³ in 2022, 34.7 µg m⁻³ in 2023, and 23.3 µg m⁻³ in 2024, whereas the ratio of the nighttime maximum to the afternoon minimum was highest in 2023 at 1.74. The annulus plot showed higher concentrations during the nighttime–early morning period in 2023 than in the other years, confirming that dry-season PM_2.5 in that year differed in its mean level and time-of-day accumulation pattern. This indicates that the deterioration of air quality in 2023 was not evenly distributed throughout the day but was accompanied by an exposure pattern that was more strongly concentrated during the night and early morning [9,20]. This pronounced nocturnal peak is consistent with stronger nighttime accumulation under dry-season conditions, possibly reflecting reduced nocturnal mixing and a shallower boundary layer during drought periods. During the nighttime and early morning, reduced solar heating weakens turbulent mixing and lowers the planetary boundary layer height, favoring pollutant accumulation near the surface. In contrast, daytime convective mixing promotes the vertical dispersion and dilution of PM_2.5.

3.4. Evidence of Long-Range Wildfire Smoke Transport to Manaus

The wind direction and directional distribution of PM_2.5 during the dry season were compared, showing that the background wind field in Manaus was dominated by the NE–E sector from 2021 to 2024 (Figure 6). The frequency of easterly winds (45–135°) was 44.6% in 2021, 51.8% in 2022, 59.5% in 2023, and 58.3% in 2024, accounting for approximately half of all dry-season hours. Consistent with this pattern, the pollution rose showed a low and relatively homogeneous PM_2.5 distribution across all directions in 2021, whereas relatively higher concentrations began to appear in the NE–E sector in 2022, and the contribution of high-concentration events from the eastern (E–NE) sector became most pronounced in 2023. The mean pollution rose concentration for the entire 2023 dry season was 29.39 µg m⁻³, and under the top 5% PM_2.5 conditions, the eastern sector was the most frequent, accounting for 53%, with a mean concentration reaching 158 µg m⁻³. This directional pattern indicated that high PM_2.5 concentrations during the 2023 dry season occurred together with the predominance of easterly winds.

When the mean concentrations were compared according to wind direction, PM_2.5 during the 2023 dry season was highest under southerly winds at 34.1 µg m⁻³, followed by westerly, easterly, and northerly winds at 28.5 µg m⁻³, 22.3 µg m⁻³, and 20.5 µg m⁻³, respectively. However, southerly and easterly winds accounted for 8.9% and 59.5%, respectively, of the dry-season hours. This indicates that individual high-concentration events should be distinguished from high background concentrations that persist throughout the dry season. Southerly winds were relatively infrequent but were associated with intense short-term episodes, whereas easterly winds appeared more frequently and contributed to persistent PM_2.5 background throughout the dry season. Therefore, PM_2.5 in Manaus during 2023 is better interpreted as being more closely related to broad smoke inflow from the upwind sector than to the isolated influence of a specific local emission source.

The multi-year analysis yielded results that supported this interpretation. In the Spearman analysis based on 524 dry-season days during 2021–2024, daily mean upwind FRP was significantly positively correlated with PM_2.5 (r = 0.449, p < 0.001). In the directional FRP distribution, the FRP of dry-season hotspots within a 500-km radius was generally concentrated in the southern sector; however, the proportion of FRP from the eastern sector increased from 14.5% in 2021 and 14.7% in 2022 to 27.6% in 2023, before decreasing to 9.9% in 2024. Additionally, during the high-pollution days of the 2023 dry season (the daily mean PM_2.5 ≥ 29.0 µg m⁻³), the mean upwind FRP was 1571 MW, 5.4 times higher than the 290 MW observed on low-pollution days (≤8.8 µg m⁻³). In 2023, the eastern sector at 300–500 km showed the largest FRP, at 117,175 MW, and monthly peaks in upwind FRP occurred during the same period as the PM_2.5 increase in September–October. To distinguish the effects of fire distance, fire activity was stratified into <300 km, 300–1000 km, and 1000–2200 km. Significant positive correlations between the daily PM_2.5 and FRP were observed for all classes (

Table 3. Distance-stratified Spearman correlations between daily PM_2.5 concentrations and wildfire fire radiative power (FRP) during the dry seasons (July–November) of 2021–2024. Correlations are shown for all fire directions and for the easterly transport sector (45–135°).

Distance Class	All Directions r	Easterly Sector r
<300 km	0.428	0.379
300–1000 km	0.464	0.515
1000–2200 km	0.44	0.459

). Under easterly wind conditions (45–135°), the strongest association was found for the 300–1000 km class, followed by the 1000–2200 km and <300 km classes. This pattern suggests that PM_2.5 variability in Manaus was influenced by wildfire activity across distance ranges, with the strongest association occurring for fires located 300–1000 km from the city.

These results indicate that high PM_2.5 concentrations were more closely linked to changes in wildfire intensity in the upwind sector than to simple accumulation within the city. In particular, the higher total FRP in 2024 and markedly lower proportion of FRP from the eastern sector suggest that the distribution of fires in the upwind sector may be more important for PM_2.5 levels in Manaus than the total amount of fire activity alone. However, these results do not exclude potential local contributions; rather, they indicate the temporal and directional pattern of PM_2.5 in 2023 was more consistent with regional wildfire smoke transport than with local emissions alone.

The map showing the potential source contribution function (PSCF) and FRP provides spatial support for these directional results under the 2023 dry-season conditions (Figure 7). Under the WHO daily guideline exceedance and top 5% of PM_2.5 conditions, high PSCF values were concentrated over broad areas to the east and southeast of Manaus, and high FRP was distributed in the same regions. In particular, in the top 5% PM_2.5 condition, the potential source contribution area appeared stronger and more concentrated, indicating that the most severe pollution episodes were more closely linked to fire activity in the eastern upwind region. This spatial overlap increases the likelihood of the major contributors to PM_2.5 in Manaus were not point sources within the city, but rather large-scale wildfire smoke distributed across the eastern and southeastern upwind regions [6].

The 72-h HYSPLIT backward trajectory analysis independently supported this process (Figure 8). In the cluster analysis for the 2023 dry season, all clusters showed origins east of Manaus, and some clusters had 72-h origins located at distances > 1000 km and altitudes > 1000 m. For days exceeding the WHO daily guidelines, all clusters originated from the east, and some extended to distances of >2000 km. The same eastern origin was maintained for the top 5% PM_2.5 episode days, and compared with the full dry season, their source regions were located at greater distances and higher altitudes. This indicated that during the full dry season, exceedance days, and extreme episodes, the main origin of the air masses entering Manaus was consistently located to the east. Particularly, more distant eastern origins were associated with higher PM_2.5 conditions, suggesting that the 2023 PM_2.5 episodes were not limited to near-range emissions alone.

The 2024 data provided a comparative test for this interpretation. Although the drought conditions were more severe and the total FRP was higher in 2024, the annual mean PM_2.5 concentration was lower than in 2023. During the same period, the proportion of FRP from the eastern sector decreased from 27.6% in 2023 to 9.9% in 2024, and the pollution rose showed that the dominant direction of the top 5% of episodes shifted from east in 2023 to west–southwest in 2024. This contrast indicated that PM_2.5 levels in Manaus were more closely related to the spatial alignment between hotspot locations and the prevailing wind direction than to the total amount of fire activity itself. In other words, the urban air quality responded more sensitively to the distribution of fires in the upwind sector and transport conditions than to the absolute magnitude of the regional fire burden. This pattern cannot be explained using local urban emissions alone, which are not expected to vary in conjunction with the directional distribution of regional fires. Therefore, the 2023–2024 contrast provides an informative comparative case showing that the spatial alignment between upwind fire activity and prevailing winds was a dominant control on PM_2.5 variability in Manaus. These results suggest that air quality management in Amazonian metropolitan areas such as Manaus should consider reductions in emissions from within the city, upwind fire monitoring, and early warning systems [6,10,11,12].

4. Discussion

This study shows that the deterioration of PM_2.5 in Manaus in 2023 should be interpreted as the result of the combined effects of extreme drought, wildfire expansion, and long-range smoke transport against a background of long-term drying. Rising temperatures, decreasing precipitation, and a declining SPEI were observed around Manaus during 1981–2024, and the SPEI decreased markedly during the dry seasons of 2023 and 2024. In particular, although dry-season precipitation was higher in 2023 than in 2022, the SPEI was lower, suggesting that the intensification of drought was closely related to the deterioration of the water balance at higher temperatures [3,16]. This indicates that the air quality deterioration in 2023 was not a temporary anomaly, but rather a case in which long-term climatic drying translated into an urban air quality crisis [3].

The expansion of wildfire activity during the 2023 dry season was directly linked to this climatic background. Hotspot counts around Manaus increased substantially in 2023 compared to those in 2021–2022, particularly within the 100–300 km range. In addition, the hotspot increases began in July and remained high throughout November in 2023, indicating that the wildfire season began earlier and lasted longer. This shows that long-term drying and extreme droughts are coupled with the intensification of regional wildfire activity [3,17].

Interannual and seasonal variations in PM_2.5 were consistent with this interpretation. By 2023, annual mean PM_2.5, dry-season mean PM_2.5, the number of days exceeding the WHO daily guideline and the Brazilian 24-h standard, and the degree of AQI deterioration all showed the most unfavorable values. Monthly mean PM_2.5 increased sharply in October and November, and the high-concentration period appeared repeatedly in two major waves. This indicates that the air quality deterioration in 2023 was not the result of a gradual increase in the mean concentration, but was formed through the accumulation of recurrent high-concentration smoke episodes during the late dry season. The diurnal pattern, in which concentrations became even higher during the night and early morning, showed that this pollution accumulated more strongly at specific times of the day, a feature similar to that reported for other cities in the Amazon [9,20].

The central finding of this study was that the high PM_2.5 concentrations observed in Manaus in 2023 are better explained by large-scale wildfire smoke transport than by local urban emissions. During the dry season, the prevailing winds belonged to the NE–E sector, and the frequency of easterly winds was highest in 2023. The pollution rose showed a clear strengthening of high-concentration contributions from the eastern sector, and upwind FRP and PM_2.5 were significantly positively correlated. The PSCF and FRP maps indicated potential contributions from eastern and southeastern upwind regions under WHO exceedance days and top 5% PM_2.5 conditions, whereas the HYSPLIT backward trajectories showed that the main origin of air masses was consistently located to the east during the full dry season, WHO exceedance days, and top 5% episodes. In particular, higher-concentration conditions were associated with more distant eastern origins, suggesting that the most severe PM_2.5 episodes were difficult to explain by near-range emissions alone and were strongly influenced by large-scale smoke transport. This interpretation is consistent with studies reporting enhanced biomass-burning effects in the Manaus metropolitan area during the 2015 El Niño, and with recent research on the 2023 Manaus smoke crisis [6,8].

The year 2024 provides an important comparative case study to test this interpretation. Although the drought was more severe and the total FRP was higher in 2024, the annual mean PM_2.5 concentration was lower than in 2023. During the same period, the proportion of FRP from the eastern sector decreased, and the dominant direction of the top episodes in the pollution rose shifted from east to west–southwest. PM_2.5 in Manaus responded more sensitively to how strongly fires were concentrated in the upwind sector of the prevailing winds than to the total amount of fire activity itself.

These findings have practical implications for health protection and urban sustainability in Amazonian metropolitan areas that face increasing climate change and wildfire risks. From the perspective of the 2021 WHO guidelines SDGs 11.6.2, 3.9, and 13, the Manaus case shows that climate change and wildfires can be directly linked to health and the urban environment through urban PM_2.5 exposure [1,10,11,12]. This study combined corrected low-cost sensor data with trajectory and potential source contribution analyses to demonstrate the link between PM_2.5 deterioration in Manaus in 2023 and large-scale wildfire smoke transport.

This study had several limitations. First, the analysis was based on a single ground-based monitoring site, which limits the ability to characterize the intra-urban spatial variability in PM_2.5 across Manaus. Second, combined use of directional PM_2.5 analysis, upwind FRP, PSCF, and HYSPLIT trajectories provides consistent transport-based evidence of regional wildfire smoke influence; however, the absence of biomass-burning chemical tracers, such as levoglucosan or water-soluble potassium, prevents direct chemical confirmation and formal quantitative source apportionment of local emission contributions (e.g., traffic and industry). Third, uncertainties remain in corrected low-cost sensor data and satellite fire detection, particularly under extreme smoke, humidity, cloud cover, or heavy aerosol conditions. However, the convergence of multiple independent indicators reduces the likelihood of the main interpretation being an artifact of a single dataset or method. Future studies should extend this framework to multiple Amazonian cities and multiple drought years by incorporating chemical tracers, vertical smoke profiles, boundary-layer dynamics, and formal source apportionment techniques where available. Taken together, although extensive evidence still supports the interpretation that the 2023 Manaus PM_2.5 episode was closely linked to drought-enhanced wildfire activity and large-scale smoke transport, these limitations should be considered when interpreting the results [3,6].

5. Conclusions

This study integrated PM_2.5 observations for 2021–2024 with long-term climate data, MODIS hotspot and FRP data, directional analyses, PSCF, and HYSPLIT backward trajectories to examine how severe air pollution developed in Manaus during 2023. Manaus experienced long-term drying during 1981–2024, characterized by rising temperatures, decreasing precipitation, and a declining SPEI. Against this climatic background, drought conditions intensified markedly in 2023 and 2024. In 2023, wildfire activity increased substantially within the near- and mid-range surroundings of Manaus, whereas PM_2.5 concentrations peaked in terms of annual mean, dry-season mean, the number of WHO exceedance days, and the intensity of high-pollution episodes.

Directional and transport analyses revealed that severe PM_2.5 episodes observed in 2023 cannot be explained sufficiently by local urban emissions alone. During the dry season, northeasterly to easterly winds prevailed over Manaus, and the pollution increase analysis showed a clear strengthening of high-concentration contributions from the eastern sector. Upwind FRP was positively associated with PM_2.5, whereas PSCF and backward trajectory analyses consistently linked WHO exceedance days and top 5% pollution episodes to eastern and southeastern upwind regions. These findings indicate that the 2023 PM_2.5 surge in Manaus was collectively driven by intensified wildfire activity in upwind regions and long-range smoke transport.

A comparison with data from 2024 strengthens this interpretation. Although drought conditions were more severe and the total FRP was higher in 2024, the PM_2.5 concentrations remained lower than those in 2023. This suggests that the air quality in Manaus is more sensitive to the spatial alignment between upwind fire activity and prevailing winds than to total fire activity alone. In other words, the severe PM_2.5 pollution in Manaus depended strongly on whether major fire activity was located in sectors that efficiently transported smoke toward the city.

These results show that the 2023 Manaus case cannot be understood solely in terms of local emissions but should be interpreted within a broader climate–fire–transport system operating over long distances. As the long-term drying and recurrent extreme droughts intensify, Manaus and other Amazonian cities may become increasingly vulnerable to wildfire smoke intrusions. These findings have direct implications on public health, urban living conditions, and sustainability. Therefore, air quality management in Manaus and other Amazonian cities should go beyond conventional local emission controls and incorporate upwind fire monitoring, early warning systems, and climate adaptation strategies. Thus, this study contributes to a better understanding of the 2023 Manaus PM_2.5 crisis as an urban environmental event shaped by the interaction of long-term drying, wildfire expansion, and long-range smoke transport. Because this study focused on Manaus during the 2023 El Niño drought, the findings should be interpreted as a case-specific assessment rather than direct evidence for all Amazonian cities. Future multi-city and multi-year studies are required to evaluate whether similar climate–fire–transport interactions affect other Amazonian urban areas under increasing drought and wildfire conditions.

Overall, this study underscores that sustainable air quality management in Amazonian cities requires strategies linking regional fire management, smoke-transport monitoring, climate adaptation, and public-health protection.