Relationship between circum-Arctic atmospheric wave patterns and large-scale wildfires in boreal summer

Long-term assessment of severe wildfires and associated air pollution and related climate patterns in and around the Arctic is essential for assessing healthy human life status. To examine the relationships, we analyzed the National Aeronautics and Space Administration (NASA) modern-era retrospective analysis for research and applications, version 2 (MERRA-2). Our investigation based on this state-of-the-art atmospheric reanalysis data reveals that 13 out of the 20 months with the highest PM2.5 (corresponding to the highly elevated organic carbon in the particulate organic matter [POM] form) monthly mean mass concentration over the Arctic for 2003–2017 were all in summer (July and August), during which POM of ⩾ 0.5 μg m−3 and PM2.5 were positively correlated. This correlation suggests that high PM2.5 in the Arctic is linked to large wildfire contributions and characterized by significant anticyclonic anomalies (i.e. clockwise atmospheric circulation) with anomalous surface warmth and drier conditions over Siberia and subpolar North America, in addition to Europe. A similar climate pattern was also identified through an independent regression analysis for the July and August mean data between the same atmospheric variables and the sign-reversed Scandinavian pattern index. We named this pattern of recent atmospheric circulation anomalies the circum-Arctic wave (CAW) pattern as a manifestation of eastward group-velocity propagation of stationary Rossby waves (i.e. large-scale atmospheric waves). The CAW induces concomitant development of warm anticyclonic anomalies over Europe, Siberia, Alaska, and Canada, as observed in late June 2019. Surprisingly, the extended regression analysis of the 1980–2017 period revealed that the CAW pattern was not prominent before 2003. Understanding the CAW pattern under future climate change and global warming would lead to better prediction of co-occurrences of European heatwaves and large-scale wildfires with air pollution over Siberia, Alaska, and Canada in and around the Arctic in summer.


Introduction
marked co-occurrences of extremely high temperature during June and August 2019 (https://ds.dat a.jma.go.jp/tcc/tcc/products/climate/seasonal/season al_201906201908e.html) with heatwaves over Europe during June and July (https://ds.da ta.jma.go.jp/tcc/tcc/products/climate/annual/annual _2019e.html). Severe and extensive wildfires across Alaska in June and July 2019 (see the satellite image at NASA's worldview for 9 July 2019, at: https:// go.nasa.gov/3n6PlUq) and Siberia (www.nasa.gov/ image-feature/goddard/2019/siberian-smoke-headin g-towards-us-and-canada) also emitted air pollutants that were transported into distant areas. Characterized by atmospheric blacking, the Siberian wildfire in the summer of 2019 caused extensive air pollution, including PM 2.5, which exceeded the 2011-2019 average values (Bondur et al 2020). It has been reported that wildfires in and around Siberia have significant impacts not only on nearby areas but also on remote locations, such as Hokkaido in Japan, which has recorded considerable PM 2.5 increases (Ikeda andTanimoto 2015, Yasunari et al 2018), and the Arctic . The above studies suggest meaningful connections among wildfires, aerosols (air pollution), and climate patterns in and around the Arctic region in recent years.
Aerosols over the Arctic region contribute to the so-called Arctic haze (i.e. haze typically seen in spring in the Arctic; Shaw 1995, Quinn et al 2007 and affect cloud properties and radiative forcing (Zhao and Garrett 2015). Fractions of carbonaceous aerosols such as black carbon (BC) and organic carbon (OC) can increase owing to wildfires emissions (e.g. Bond et al 2004, Wang et al 2011, Noguchi et al 2015. Another consequence of wildfires is the snowdarkening effect (SDE), which is due to reductions in snow albedo caused by the depositions of lightabsorbing aerosols such as BC, OC, and dust (Warren and Wiscombe 1980, Flanner et al 2007, Yasunari et al 2010, Yasunari et al 2013, 2015, Qian et al 2015, Lau et al 2018, Hock et al 2019. Solar forcing by SDE due to post-wildfires has increased in the western US burned forests since 1999 (Gleason et al 2019). A recent study revealed that surface melting on the Greenland ice sheet is accelerated during warmer-than-normal summers because of high temperatures and increased BC deposition from wildfires (Keegan et al 2014). Therefore, in addition to the essential connections with climate patterns, wildfire-generated aerosols are also of significant concern for climate and snowdarkening interactions.
However, many studies have often focused on individual aerosol constituents that make up PM 2.5 , rather than assessments of comprehensive air pollution (e.g. Sharma et al 2004, Eleftheriadis et al 2009, Hegg et al 2009, Wang et al 2011, Yttri et al 2014 with a few PM 2.5 studies in and around the Arctic (Tran and Mölders 2011). Therefore, there is still a lack of statistical research on air pollution, specifically on PM 2.5 , and capturing long-term variability in PM 2.5 in the Arctic has thus far been difficult in the absence of continuous and extensive PM 2.5 measurements. Our understanding of long-term PM 2.5 variation, air quality in the Arctic and its surrounding areas, and their linkage to variations in climatic conditions, such as anomalous atmospheric circulation, is therefore still limited.
Recently, the National Aeronautics and Space Administration (NASA) produced MERRA-2 (Modern-Era Retrospective analysis for Research and Applications, Version 2), a long-term global reanalysis dataset for atmospheric and other conditions such as land and ocean (Bosilovich et al 2015). The best quality aerosol-related variables since 2003 in MERRA-2 are available because the most available satellite measurements were merged into the modeled data for this period (Randles et al 2017). We, therefore, utilize the MERRA-2 reanalysis data in this study.
Thus far, case studies for Siberian wildfires and associated air pollution in and around the Arctic have been reported (Ikeda and Tanimoto 2015. However, a number of questions remained unanswered. To address this knowledge gap, we focused on three key factors in this study: (a) statistical relationships among wildfires, PM 2.5 , and wildfire-related aerosols, especially under high air pollution (i.e. PM 2.5 ) conditions, (b) identifying the dominant climate pattern under high PM 2.5 conditions in the Arctic, and (c) identifying whether this pattern has occurred only in recent years. The state-of-the-art MERRA-2 reanalysis data allowed us to reveal the relationships among wildfires, aerosols (air pollution), and climate patterns in and around the Arctic. The outcome of this study provides a resource to aid in making better forecasts in future studies.

Data
To examine recent temporal variations of air pollutants in the Arctic, we used the monthly MERRA-2 reanalysis data from January 2003 to December 2017 (15 years) for the global atmosphere based on the GEOS-5 model (Rienecker et al 2008), including the data assimilation with observations since 1980 (Bosilovich et al 2015, Randles et al 2017 https://gmao.gsfc.nasa.gov/reanalysis/ MERRA-2/). The MERRA-2 atmospheric data from 1980 to 2017 was also used for some additional analysis. The advantage of using MERRA-2 is that observed aerosol optical depth (AOD) data from satellites and ground-based measurements (AER-ONET) are assimilated (Randles et al 2017). When observed AOD data are available, aerosol data simulated in the GEOS-5 model are modified to reflect the observations (Randles et al 2017). The best aerosol quality with lower bias was available in MERRA-2 from 2003 compared to that before 2003 (figure 5 of Randles et al 2017). MERRA-2 uses the monthlymean quick fire emissions dataset (QFED) emissions (Darmenov and Da Silva 2015) up to 2010 and daily QFED emissions afterward. While daily biomass burning (BB) emissions are critical for capturing day-to-day BB emission variability, the switch to daily BB emissions in 2010 does not introduce any long-term trends because the same monthly-mean emissions were used before and after 2010. Notice that despite the fact that monthly mean BB emissions were used prior to 2010, MERRA-2 can still capture daily variability in BB aerosols due to the three-hourly assimilation of aerosol data. Therefore, the switch to daily QFED emissions in 2010 should have minimal impacts on the analysis presented in this study. In the GEOS-5 model, the original air pollutants (i.e. aerosols) are simulated with the goddard chemistry aerosol radiation and transport (GOCART) module (Chin et al 2000, Ginoux et al 2001, Chin et al 2002. Besides, a case study for the Siberian wildfire in 2016 using MERRA-2, BC revealed long-range transport of air pollution to the Arctic , which further validated the usefulness of the MERRA-2 data in/around the Arctic region. We also used MODIS (moderate resolution imaging spectroradiometer) fire pixel counts (FPCs) retrieved by the Aqua satellite (available at: https:// feer.gsfc.nasa.gov/) to construct composites for 13 high PM 2.5 summer months. Monthly total FPCs aggregated at 1 • by 1 • both in latitude and longitude are used to discuss a potential source of POM.
For our regression analysis, a known climate index of the Scandinavian pattern (www.cpc.ncep.noaa.gov/data/teledoc/scand.shtml) was used; it was originally defined as the Eurasia type 1 pattern by Barnston and Livezey (1987).

Analytical techniques and statistics
In this study, PM 2.5 was calculated from the five MERRA-2 aerosol types in the PM 2.5 size-dust, sulfate, BC, OC (i.e. POM), and sea salt-by the method (equation (1)) of Buchard et al (2016), in which sulfate is assumed to be ammonium sulfate. . Thus, the remainder of the study focused on the summer months. These 13 summer months were then compared with the other 17 summer months (July and August) with the lower PM 2.5 levels through composite analysis shown later. Welch's t-test (Welch 1938) was used to assess the significant mean differences in the composites. Most composites adopted a twosided t-test for rejecting the null hypothesis, except for those presented in SI figure S2 and SI table S1 (see supplementary information: SI (available online at stacks.iop.org/ERL/16/064009/mmedia)). For these SI materials, we only would like to verify whether the PM 2.5 and aerosol mass concentrations over Siberia, Alaska, and Canada for the high PM 2.5 months in the Arctic are higher than those for the other months. Therefore, we only focused on the positive (i.e. increase) difference and thus used a one-sided t-test for them (i.e. under the alternative hypothesis).
The climate pattern's essential dynamic processes can be illustrated with a particular form of a Rossby wave-activity flux (WAF) (Takaya and Nakamura 2001), parallel to the group velocity of stationary Rossby waves (i.e. large-scale atmospheric waves). According to the linear theory, the flux magnitude is proportional to the squared wave amplitude and the eastward group velocity. The latter is proportional to the background westerly wind speed. The WAF of Takaya and Nakamura (2001) was calculated with the modified script based on the GrADS script for WAF (http://www.atmos.rcast.utokyo.ac.jp/nishii/programs/index.html).

Characteristics of PM 2.5 and aerosols, and their relationships with wildfire
Here, we investigate monthly PM 2.5 variations during 2003-2017 by selecting 20 months with the worst air quality in the Arctic region. From the areamean data over the Arctic for 2003-2017 (figure 1), the fractions of dust, ammonium sulfate, sea salt, BC, and OC (POM; hereafter called POM; proportional to OC in MERRA-2: see section 2) are about 2.7%-42.5%, 9.5%-37.5%, 16.7%-73.1%, 0.5%-2.8%, and 1.5%-58.0% of the monthly mean Arctic PM 2.5 , respectively. Generally, the dust contribution tends to be smaller in summer and fall but more prominent in spring. Sulfate aerosols are also smaller in summer. Meanwhile, BC and POM peaked during summer, with the summer fractions reaching 1.0%-2.8% and 13.8%-58.0%, respectively. These  findings imply a common emission source in summer. Sea salt markedly increases in fall but decreases to a minimum, mainly in spring; its fraction never goes below approximately 16.7%, and the minimum contribution is higher than the other aerosol constituents. Of the 20 months of the worst air quality based on PM 2.5 , 13 months were in summer (figure 1). These 13 significantly high PM 2.5 months in summer also correspond to higher POM months for 2003-2017 (figure 2). When the Arctic-averaged monthly-mean POM is not less than 0.5 µg m −3 , POM and PM 2.5 were well explained by its strong linear relationship (R 2 = 0.77). Those 13 high PM 2.5 months in the Arctic summer were also characterized by regionally high POM ( figure 3(a)). The POM mass concentrations were greater over Siberia, Alaska, and Canada (figure 3(e)), where wildfires were very active (figure 4). The POM increased significantly in the 13 summer months (July and August) in Siberia and Canada compared with the other July and August months, where increased BC was also prominent (SI figures S1 and S2; SI table S1). The POM increase in Alaska was significant only in July (SI table S1). Those results indicate that the 13 highest PM 2.5 months in summer (July and August) in the Arctic in 2003-2017 were also higher POM months due to active wildfires over Siberia and subpolar North America in and around the Arctic. Also, the POM increase tended to be accompanied by BC increases from the same areas ( figure 3(d)).

Summer climate pattern under high PM 2.5 conditions in the Arctic
During the 13 highest PM 2.5 months in summer (i.e. July and August), significant anticyclonic anomalies (i.e. clockwise atmospheric circulation) from the lower (850 hPa) to mid-troposphere (500 hPa) were identified (figures 5(b) and (c)) over central Siberia and western Canada. The local PM 2.5 and POM mass concentrations over the regions were significantly high (figures 3(a), (e); SI figures S1 and S2; SI table S1) owing to active wildfires (figure 4). These anticyclonic anomalies tend to accompany anomalous high surface air temperatures (figure 5(a)) that are significant over Siberia and Canada in July and August, as well as significantly drier conditions over Siberia, Alaska, and Canada (SI figure S3). The anomaly pattern in figure 5 also resembles the warm anticyclonic anomalies observed over western Europe, central Siberia, Alaska, and Canada in late June 2019 (SI figure S4). Figure 6 shows WAF in the atmosphere evaluated from the composite anomaly fields for the 13 summer months and their differences in geostrophic stream functions (from geopotential height) with the composite of horizontal winds for the 17 other summer months at 500 hPa. The warm anticyclonic anomalies are accompanied by two eastward-developing Rossby wave trains along the westerly jet stream, one from Europe to Alaska via Siberia and the other from the North Pacific to Alaska (see the vectors in figure 6).

Regression analysis with the scandinavian pattern index
We further performed an independent regression analysis with the same three meteorological variables used in figure 5. Here we use the Scandinavian pattern index; a known climate pattern originally referred to as the Eurasia-1 pattern (Barnston and Livezey 1987). The index is valid for assessing unusual climate such as blocking, temperature anomalies, etc, over Eurasia from Scandinavia to East Eurasia (www.cpc.ncep.noaa.gov/data/teledoc/ scand.shtml). Regression maps between the signreversed July and August mean Scandinavian pattern index and the three meteorological variables ( figure 7) show a very similar climate pattern in the atmosphere obtained from the composite differences (figure 5). Positive anticyclonic anomalies at 850 and 500 hPa over Europe, Siberia, and subpolar North America (Alaska and Canada) were evident, corresponding to the near-surface warm anomalies (figures 5 and 7). These relationships imply that the climate pattern that characterizes the 13 highest PM 2.5 summer months in the Arctic by the composite analysis (figure 5) is well explained by the inter-annual variations of the sign-reversed July and August mean Scandinavian pattern index for the 2003-2017 period. To emphasize the characteristic of pressure anomalies aligned zonally around the Arctic    from Europe to Siberia, Alaska, Canada, and back to Europe, we refer to this summertime climate pattern as the circum-Arctic wave (CAW) pattern.
However, if we performed the same regression analysis extended for a long-term period (i.e. including 1980-2002), the climate pattern only shows a NW-SE oriented (line-shaped) atmospheric teleconnection pattern from Scandinavia to East Asia ( figure 8(b)) rather than the CAW. This pattern resembles a typical Scandinavian pattern, as reported in Bueh and Nakamura (2007), although they excluded the summer season from their analysis. This climate pattern exclusively identified over Eurasia is also dominant in the regression for the whole period (1980-2017; figure 8(a)). This characteristic indicates that the typical Scandinavian pattern was dominant over Eurasia before 2003, and afterward, the climate pattern has somehow been modified into forming the CAW. The CAW-associated positive anticyclonic (geopotential height) anomalies at 500 hPa over Europe and Siberia are also stronger than their counterpart associated with the Scandinavian pattern before 2003. This change would imply a possible shift of the summertime climate pattern (teleconnection pattern; i.e. climate links between remote regions) before and after 2003 and forming the co-occurrence of warm surface anomalies over Europe, Siberia, Alaska, and Canada in CAW (figures 5 and 7).

Discussion and conclusions
We have found high PM 2.5 mass concentrations over the summertime Arctic in 2003-2017 (figure 1). These high PM 2.5 months also correspond to higher POM (figure 2) and its significant increase over Siberia and Canada (SI figures S1 and S2; POM in Alaska only significant in July: SI table S1), where wildfires were very active (figure 4). OC (also BC) is a good indicator of wildfire activity (BB; e.g. Bond et al 2004, Spracklen et al 2007, Noguchi et al 2015, Shikwambana 2019. Therefore, we argue that the significantly increasing contributions of organic aerosols to Arctic PM 2.5 are mainly due to wildfire smoke in and around the Arctic. This characteristic is consistent with results modeled from October 1999 to May 2005 in a previous study (Stohl 2006). In conclusion, for 2003-2017, the 13 higher PM 2.5 months out of the 20 worst air quality months in the Arctic are explained by the greatly increased OC (POM) with some BC increases from active wildfires in Siberia, Alaska, and Canada.
We have also examined the causes for summers when the Arctic air quality worsens with dominant climate patterns. Investigating those climatic patterns involved in forming warm anticyclonic anomalies over Siberia, Alaska, and Canada under the high PM 2.5 conditions (i.e. worse air quality) in the Arctic is essential for better prediction of large-scale wildfire occurrences in the future. For this purpose, composite and regression analyses for 2003-2017 were performed (figures 5 and 7). The composite analysis extracted anticyclonic anomalies from the lower to mid-troposphere over Europe, Siberia, and Canada (figures 5(b) and (c)), which locally caused anomalous near-surface warmth and dryness from near the surface to the lower troposphere (figure 5(a); SI figure S3). A recent study with the MODIS fire and MERRA reanalysis (the former product of MERRA-2) data concluded that the primary months of wildfire activity in the circumpolar Arctic tundra region were July and August, and those occurred under warm and dry conditions (Masrur et al 2018). Their results strongly support our results. The extreme anomalous warmth is also likely to induce heatwaves (Chase et al 2006, Feudale andShukla 2011); the associated abnormal dryness favors active wildfires (Bondur 2011). Our composited climate pattern in summer (figure 5) is likely to set a favorable condition for heatwaves over Europe and active wildfires (figure 4) over Siberia, Alaska, and Canada, causing high PM 2.5 and POM and BC emissions in and around the Arctic.
To elucidate the relationship between the climate patterns of recent years (2013-2017) and the longterm, we focused on the Scandinavian pattern as a well-known climate pattern. A recent paper also reported that the Scandinavian pattern is relevant to the summer heatwave activity after the mid-1990s (Choi et al 2020), substantiating this index's usefulness for our analysis. As illustrated in the regression patterns in figure 7, the negative phase (i.e. signreversed index) of the Scandinavian pattern is characterized by mid-tropospheric anticyclonic anomalies in Europe, central Siberia, Alaska, and Canada. Unlike its counterpart in the cold season (Bueh and Nakamura 2007), the recent summertime Scandinavian pattern or CAW (figure 7(c)) accompanies anticyclonic anomalies from Europe to Alaska along the subpolar westerly jet stream that forms in July and August along the Siberian coast Fukamachi 2004, Tachibana et al 2004). Notably, a similar anomaly pattern, which slightly shifted westward from the corresponding regression patterns (figure 7), is also apparent in the pressure anomalies composited for the 13 summer months with the high PM 2.5 (figure 5). As demonstrated by the WAF in the atmosphere (figure 6; Takaya and Nakamura 2001), the CAW is likely a manifestation of eastward groupvelocity propagation of a stationary Rossby wave train from Europe to Alaska along the upper-troposphere westerly jet stream. The CAW is also associated with near-surface warm and tropospheric dry anomalies, as shown in figure 7(a) and SI figure S3. Our analysis substantiates the previously discussed importance of mid-tropospheric anticyclonic circulations in wildfire occurrences in Siberia and Alaska (Hayasaka et al 2016, Bondur et al 2020. Note that the similarity between figures 5 and 7 was unexpected.
A similar eastward-extending Rossby wavy pattern along the subpolar jet stream was observed in summer 2010, which induced extreme warmness, especially in eastern Europe, Russia, and Japan (Otomi et al 2013). The particular wave train and that associated the CAW both form along the subpolar jet stream over northern Eurasia in association with the enhanced poleward temperature gradient in summer between the warmer Eurasian continent and the cooler Arctic (Serreze et al 2001, Nakamura and Fukamachi 2004, Tachibana et al 2010. Interestingly, the CAW and associated anticyclonic anomalies over Europe, Siberia, and subpolar North America became prominent only from 2003 (figure 8). Amplified anomalies in recent years over Alaska and western Canada can be explained by converging WAF. Those were associated with the wave trains (the one from Siberia via the Arctic Ocean; the other from the North Pacific) if the composite difference pattern in figure 5 and the 'negative' Scandinavian pattern in figure 7 have a similar mechanism (figure 6). Elucidating the CAW formation is beyond the scope of this work but should focus on future studies. A recent numerical simulation showed that a response to a heat source prescribed in western Russia could be similar to the Scandinavian pattern over Siberia (Choi et al 2020). Their results may explain the summer CAW observed in the recent years (figure 7(c)), resulting in co-occurrences of heatwaves over Europe and large-scale wildfires in Siberia and subpolar North America, notably Alaska and Canada (figures 4, 5(a) and 7(a)).
Therefore, it is not surprising to observe a quasistationary wave train in late June 2019, similar to the CAW as a recent example (SI figure S4). This situation Figure 9. Schematic summary of the results obtained in this study. Anomalous anticyclones developed concomitantly in summer (July and August) over the circumpolar regions: Europe, Siberia, and subpolar North America (i.e. Alaska and Canada). Those anticyclones induce warm and dry forcings from the surface to the mid-troposphere. This study names this climate pattern the CAW pattern. Those anomalous anticyclones induce heatwaves over Europe and active wildfires over Siberia and subpolar North America. The wildfire smoke emitted OC and BC aerosols into the atmosphere, and those aerosols could reach the Arctic region to increase PM2.5 there. The interactions between the CAW pattern and the atmospheric aerosols must be investigated in future studies. Reproduced from Hayanon Science Manga Studio. CC BY 3.0.
likely induced the concomitant occurrence of a heatwave in Europe, and the Siberian and Alaskan wildfires, as mentioned in section 1. The climatic impacts of the Scandinavian pattern during fall, winter, and spring have been well explored (Blackburn and Hoskins 2001, Bueh and Nakamura 2007, Pall et al 2011. However, a recent study (Choi et al 2020) and this study have only emphasized the importance of the corresponding summertime pattern on heatwaves and related large-scale wildfires.
A schematic in figure 9 summarizes all the possible connections from the unique circumpolar atmospheric circulation pattern in summer, CAW, which has become predominant since 2003, to highly increased aerosols in the Arctic. The CAW pattern can induce the anomalous warmth (figure 7(a)) and dryness near the surface and in the lower troposphere (figure S3) under anomalous high-pressure systems (figures 7(b) and (c)). The anomalous anticyclones are likely to cause heatwave over Europe and wildfire occurrences in Siberia and subpolar North America (Alaska and Canada) (figure 4) with massive aerosol injections such as POM and BC into the atmosphere (figure 3; SI figures S1 and S2; SI table S1). The CAW-associated atmospheric circulation can eventually transport the wildfire-induced aerosols into the Arctic, acting to lower air quality (figure 2). The interaction between the aerosols and CAW may be possible through the impact of carbonaceous aerosols on climate via radiative effects (e.g. Liu et al 2014, Zhang et al 2017. However, we have presented no solid evidence to substantiate this interaction. Therefore, further investigation is needed in the future. Our study reveals the climatic significance of the recent July and August mean Scandinavian pattern, particularly its 'negative phase' (i.e. CAW pattern). As a stationary atmospheric Rossby wave train with fast eastward group velocity (figure 6), the CAW can act as a useful dynamic indicator in summer of the simultaneous occurrences of heatwaves in Europe, Siberia, Alaska, and Canada, and also works to increase the probability of large-scale wildfires in Siberia, Alaska, and Canada, and thereby higher PM 2.5 and OC (POM) concentrations in the Arctic. A recent study reported the relationships among wildfire activity in the Arctic tundra and meteorological variables (Masrur et al 2018). In conjunction with that study, this study further provides perspectives on their connections to wildfire-induced air pollution and spatiotemporal climate pattern. As increasing wildfires projected under global warming have been discussed (Veira et al 2016), further investigations are required to deepen our understanding of the summertime CAW pattern. The trigger mechanism and its persistence under such climate conditions as seasurface temperature anomalies are yet to be explored. It is also essential to assess whether the CAW will likely emerge throughout the extended summer season under the warming climate. A better projection of this will help take effective measures of large-scale summer heatwaves and severe wildfires that lead to massive air pollutants in and around the Arctic.

Data availability
The data used in this study are available from the corresponding author or co-authors upon reasonable request. MERRA-2 and fire data from MODIS are also publicly available at: https://gmao.gsfc.nasa.gov/ reanalysis/MERRA-2/; https://feer.gsfc.nasa.gov/ index.php.
The data that support the findings of this study are available upon reasonable request from the authors.

Acknowledgments
We want to thank two anonymous reviewers for their helpful comments in revising the paper. This study is supported in part by the Japanese Ministry of Education, Culture, Sports, Science and Technology through the Arctic Challenge for Sustainability (ArCS; JPMXD1300000000) project and its successor project (ArCS II; JPMXD1420318865), and by the Japan Society for the Promotion of Science through the Grants-in-Aid for Scientific Research (JSPS KAKENHI 17H02958, 17KT0066, 18H01278, 19H01976, 19H05668, 19H05698, 19H05702, 20H01970, and 20K12197) and by the Japanese Ministry of Environment through Environment Research and Technology Development Fund JPMEERF20192004. NASA's Global Modelling and Assimilation Office (GMAO) produced the MERRA-2 reanalysis data, and we also used NCCS for data analyses. We also thank NASA's MODIS team for making the fire pixel count data available. Shunsuke Tei (Forestry and Forest Products Research Institute) provided a helpful discussion of the calculations of PM 2.5 and aerosol composite differences (statistics). Kazuaki Nishii (Mie University) helped with plotting the WAF with the GrADS script. We want to thank Editage (www.editage.com) for English language editing before the initial submission. We appreciate Science Manga Studio (Hayanon, noguchi.m) to make the schematic figure for summarizing our results.