Introduction

Wildfires in the Northern Hemisphere have increased significantly in recent decades, mainly occurring in spring, summer, and autumn1,2,3. These wildfires pose severe threats to society, the economy, the energy sectors, human health, ecosystems and biodiversity4,5,6. Their primary impacts include elevated air pollution5,7,8 and carbon dioxide (CO2) emissions3, among other environmental and societal consequences.

Warm-season wildfires are driven by a complex interplay of weather and climate factors, influenced by both natural variability9,10,11 and anthropogenic climate change12,13,14,15. In addition to the availability of combustible vegetation, key contributors include lightning activity, windy conditions, surface warming, vapor pressure deficit (VPD), and atmospheric convection instability16,17,18. Any changes in these factors can influence wildfire patterns, including their ___location, frequency, extent, and intensity17,18. Research suggests that climate change has lengthened the wildfire season and increased both the frequency and total burned area19,20. The intensification of global warming has fueled a surge in summer wildfires worldwide2,21, mainly due to rising surface temperature, worsening terrestrial aridity and drought conditions22, and increased lightning ignitions in boreal forests and permafrost regions23. As a result, fire-prone area has increased by 29% globally, with boreal and temperate zones experiencing the most significant increases21.

Siberia24,25,26 and western United States27,28 have been identified as primary wildfire-prone regions of the Northern Hemisphere. The features of these wildfires are likely linked to changes in regional atmospheric circulation patterns15,29,30,31,32 and Arctic sea-ice loss18,33 during the summer. However, the summer of 2023 saw the most intense Canadian wildfire event since 200334,35, with fires spreading from western Canada to the northeastern regions of Canada and United States. This resulted in record levels of wildfire-induced air pollution and CO2 emissions in recent years36,37.

The extreme 2023 Canadian wildfires have been found to be tied to warm North Atlantic waters and reduced sea-ice concentration in Barents Sea38, likely due to increased atmospheric aridity and reduced precipitation39. However, the underlying causes of their prolonged duration, eastward expansion, widespread coverage, and exceptional intensity remain unclear, particularly in relation to variations in fire weather patterns and their interaction with large-scale climate variability, such as the Pacific Decadal Oscillation (PDO) and El Niño-Southern Oscillation (ENSO).

Previous studies have shown that anthropogenic warming and long-term climate change-related synoptic weather patterns in the Northern Hemisphere mid-to-high high latitudes drive the long-term variability and trend of the Northern Hemisphere wildfires13,14,15,19,33,40. However, they fail to explain the unprecedented severity of the 2023 Canadian wildfires. It is possible that anomalous atmospheric processes40,41, Pacific oceanic conditions1,13, and their interactions played a crucial role in shaping this extreme event, alongside the effect of climate warming22,34.

This study investigates the atmospheric and oceanic conditions that contributed to the intense and widespread Canadian wildfires of summer 2023, beyond the well-established drivers of rising air temperatures and decreasing relative humidity14,18. We find that the North American blocking events, characterized by increased frequency, persistence, expanded zonal extent and enhanced eastward movement, played a critical role in driving the severity of these wildfires. We further show that these blocking events were associated with positive Pacific North American (PNA) patterns, which were in turn tied to the Eastern Pacific El Niño and a particularly strong negative phase of the PDO.

Results

Linking Canadian wildfires of the 2023 summer to atmospheric blocking events

We define the periods of November-January (NDJ), February-April (FMA), and May-July (MJJ) as winter, spring, and summer, respectively. Figure 1 indicates that in 2023, the summer wildfire fraction (Fig. 1a) and CO2 emissions (Fig. 1b) in 2023 in Canada reached their highest levels since 2003, consistent with previous findings34,35,36,37. The extreme wildfires in Canada led to the evacuation of over 232,000 people, the loss of hundreds of homes, and the deaths of eight firefighters35. These fires also caused significant changes in forest landscape34, biomass and species composition37. In 2023, Canada experienced 834 large fires\(-\)2.5 times the 1986–2022 average of 320, whereas the total burned area reached approximately 15 Mha, seven times the annual mean of 2.1 Mha for 1986–202234. Very large fires (≥ 50,000 ha) mainly occurred over western Canada and persisted for extended periods. Wildfires burned for about five months in western Canada even when spreading eastward, continuing into 2024 with some zombie or overwintering fires smoldering in organic soils through the winter42. The summer wildfire fraction in 2023 was significantly higher in Canada (blue line in Fig. 1a) than in Siberia (red line in Fig.1a).

Fig. 1: Temporal variations of summer wildfire fraction and carbon dioxide (CO2) emissions and spatial patterns of summer mean wildfire weather parameters and wildfire fraction anomalies in 2023.
figure 1

a, b Normalized time series of summer (May to July, MJJ) mean (a) wildfire fraction anomalies (unit: non-dimensional) and (b) carbon dioxide (CO2) emissions (unit: kg m−2s−1), averaged over Siberia (60oE-120oE, 45oN-65oN) and Canada (140oW-60oW, 50oN-70oN) during 2003–2023. c–h 2023 anomalies in MJJ-mean (c) 500 hPa geopotential height (Z500; contour interval=10 gpm) and surface air temperature (SAT, color shading, unit: K), (d) vapor pressure deficit (VPD, unit: hPa), (e) precipitation (P, color shading, unit: mm), (f) soil moisture (SM, color shading, unit: m3 m−3), (g) snowmelt (color shading, unit: m, the negative value means the melt) and (h) wildfire fraction (WF, color shading, unit: non-dimensional). The anomaly of the wildfire fraction is defined as its derivation from the 2003–2023, whereas the anomalies of other variables are defined as their derivations from the 1979–2023 mean as shown below.

Full size image

We begin by examining the atmospheric conditions favoring the 2023 summer wildfires. From May to October, the mean air temperature over Canada was 2.2 °C warmer than the 1991–2020 average34, intensifying terrestrial aridity and drying surface fuels22,23. The vapor pressure deficit (VPD), reflecting the combined effect of air temperature and relative humidity (see “Definition and physical meaning of vapor pressure deficit” in Methods), is widely used as a key factor influencing forest dryness and aridity18. Elevated VPD conditions increase the likelihood of ignition and fire spread, leading to larger and severer wildfires31. As shown by the summer mean time series of surface air temperature (SAT), VPD, precipitation (P), soil moisture (SM) and snowmelt (Smlt) anomalies averaged over Canada (140oW-60oW, 50oN-70oN) (Supplementary Fig. 1), 2023 recorded the highest VPD and lowest SM values since 1979. Furthermore, SAT in 2003 reached its highest level since 2003, while P remained low and Smlt had negative anomaly. These results suggest that the extreme high VPD and SAT, along with exceptionally low SM, were primary drivers of the 2023 Canadian wildfires, while low P further contributed to increased wildfires1,15.

We observed that the 2023 summer wildfires affected the entire mid-high latitude region of North America, with the main impact centered in Canada (Fig.1h). These wildfires were associated with increased SAT (Fig. 1c), enhanced VPD (Fig. 1d), decreased precipitation (Fig.1e), elevated soil moisture deficit (Fig. 1f) and intensified snowmelt (Fig. 1g) across large areas of mid-high latitude North America. These changes coincided with a strong blocking anticyclonic anomaly over Canada (Fig. 1c). Blocking anticyclones are known to cause persistent warming and reduction in precipitation and soil moisture within the affected region33,40,41.

In 2023, the MJJ-mean blocking anticyclone over mid-high latitude North America induced an average SAT increase of 1.85 °C across Canada (120°W–80°W, 50°N–70°N), with localized increases reaching up to 5 °C (Fig. 1c). Note that the anomaly is calculated relative to the detrended 1979−2023 mean, and thus excludes the contribution from the long-term trend for a clarification. This blocking also induced a negative precipitation anomaly of −7.55 mm, a negative SM anomaly of −1.95 m3 m−3, a positive VPD anomaly of 0.79 hPa, and a snowmelt anomaly of −0.78 cm over Canada. These results suggest that the widespread and intense blocking anticyclone during summer 2023 was a key driver of the substantial increases in SAT and VPD, as well as the pronounced reductions in P and SM across North America’s mid-to-high latitudes. Previous studies43,44,45 suggest that the relationship between summer fire area and VPD and SAT in boreal forests is strongly nonlinear45. Specifically, elevated VPD or SAT can lead to an exponential increase in the wildfire burned area43,45, while decreased (increased) SM amplifies (inhibits) the effect of high VPD on the growing-season wildfires44.

Figure 2a, b show that during June 1 to 10, 2023, wildfires predominantly occurred in Canada (and parts of the United States) and Siberia, coinciding with the presence of dual blocking anticyclonic anomalies over these regions in Northern Hemisphere. This reinforces the strong link between mid-high latitude summer wildfires and blocking events. Notably, about 60% of wildfire ignitions resulting in burned areas larger than 20,000 ha in western North America are linked to the presence of blocking ridges40 that lead to increased SAT and VPD and decreased precipitation40,41.

Fig. 2: Relationship between wildfires and atmospheric blocking, and time-longitude evolutions of summer daily anomalies in 500 hPa geopotential height (Z500), surface air temperature and wildfires in given latitude regions in 2023.
figure 2

Time-mean fields averaged from 1 to 10 June, 2023 of (a) 500 hPa geopotential height (Z500, contour interval=30 gpm) and surface air temperature (SAT) (color shading, unit: K) and (b) wildfire fraction (color shading, unit: non-dimensional) anomalies, where the thick black line represents 5700 gpm. c–h Time-longitude evolutions of daily (c, f) 500 hPa geopotential height (Z500, color shading, unit: gpm), (d, g) SAT (color shading, unit: K) and (e, h) wildfire fraction (color shading, unit: non-dimensional) anomalies averaged over 45o-65oN for (c–e) Siberia and averaged over 50o-70oN for (f–h) Canada from May to July (MJJ) in 2023.

Full size image

Figure 2c shows that a blocking event occurred over the Ural-Siberia region from May 21 to June 10, while two cyclonic anomalies developed in the same area from May 11 to 20 and again from June 13 to July 3. This alternation of anticyclonic and cyclonic anomalies reduced the persistence of surface warming (Fig. 2d), thereby limiting the intensity and duration of summer wildfires. As a result, while the Ural-Siberian blocking event triggered intense wildfires during 1 to 10 June 2023 (Fig. 2b), wildfires across the region during the entire summer were not persistent and widespread (Figs. 1h and 2e). This was partly because of little positive temperature anomalies (Fig. 1c) and relatively weak negative anomalies in precipitation, soil moisture and snowmelt (Fig. 1e–g).

In contrast, three blocking events took place in Canada during the summer of 2023 (Fig. 2f). These events were also stronger and exhibited a pronounced eastward movement (Fig. 2f), resulting in sustained surface warming (Fig. 2g), persistent decreases in precipitation40 and soil moisture (Fig. 1e, f), persistent increases in VPD40 (Fig. 1d), and intensified zonal winds along the northern flank of the blocking anticyclone41. These conditions supported extensive and prolonged wildfire activity across Canada throughout the summer (Fig. 2h). As a result, the MJJ-mean 500 hPa geopotential height (Z500) anomaly associated with blocking anticyclones over Canada in the summer of 2023 (Fig. 1c) leads to more persistent, widespread, and intense surface warming, snowmelt, soil moisture and precipitation deficits across the North American mid-high latitudes (Fig. 1c–g). Under these conditions, large wildfires that ignited in western Canada were able to spread eastward into northeastern Canada and the northeastern United States (Fig. 1h), following the eastward migration of the blocking system (Fig. 2f).

Moreover, the stronger MMJ-mean blocking anticyclones over Canada than over Ural-Siberia was associated with much more intense zonal winds in the higher latitudes of the blocking anticyclones, which further facilitated the spread of wildfires41. Taken together, the extreme Canadian forest fires in 2023 were likely due to frequent, long-lasting, and eastward-moving blocking events over mid-high latitude North America, which induced extensive warming, snowmelt, soil moisture and precipitation deficits, and strong high-latitude zonal winds.

Precursors leading to summer Canadian wildfires in 2023

Since antecedent climate conditions in winter and spring can significantly influence summer wildfire activity46,47,48, it is also important to examine the pre-summer conditions of the 2023 Canadian wildfires to fully understand the factors behind their extreme severity.

Wildfires in Canada began as early as May 2023 (Fig. 2h). This early onset of the wildfires is likely linked to decreased precipitation and soil moisture during the 2022 winter (November 2022 to January 2023) and the 2023 spring, mainly in western and central Canada north of 50oN and west of 80oW (Fig. 3a, d). Our analyses reveal that P, SM and Smlt anomalies averaged over the region (140oW-80oW, 50oN-70oN) are −4.88 mm, −0.80 m3m−3 and −0.11 cm in winter, and −2.44 mm, −0.69 m3 m-3 and −0.32 cm in spring, respectively.

Fig. 3: Spatial patterns of winter and spring mean precipitation, soil moisture and snowmelt anomalies in 2022/2023.
figure 3

a, c, e Winter (November 2022 to January 2023, NDJ) and (b, d, f) spring (February to April 2023, FMA) mean (a, b) precipitation (unit: mm), (c, d) soil moisture (unit: m3 m−3) and (e, f) snowmelt (unit: m) anomalies in 2022/2023.

Full size image

In high-latitude regions, reduced winter and spring precipitations often correspond to negative snowfall anomalies or diminished snowpacks in mountainous aeras. While no strong blocking anticyclones were observed over Canada in the 2022 winter and the 2023 spring (Supplementary Fig. 2), a winter blocking anticyclone was present over the North Pacific (Supplementary Fig. 2a). This suggests that the large reductions in winter precipitation and soil moisture anomalies over Western Canada were not produced by North American blocking anticyclones but were instead influenced by a North Pacific blocking anticyclone, which extended northeastward toward western Canada (Supplementary Fig. 2a). Furthermore, the strong reductions in winter and spring P and SM across western and central Canada in 2022/2023 might also be influenced by climate change1,15 and Arctic warming12, warranting further investigation.

Reductions in winter and spring snowpacks can accelerate spring snowmelt (Fig. 3f), enhancing SAT and evaporation through the snow-albedo feedback. This process depletes soil and vegetation moistures (Fig. 1f), ultimately contributing to longer and drier summers48. Consequently, the dry summers over western Canada are often associated with large negative snowfall anomalies during the preceding winter and spring, coupled with strong snowmelt49. In addition, reduced soil moisture during winter and spring can further contribute to a warm, dry summer.

As a result, the extreme dry summer (Fig. 1e, f), partly driven by large decreases in precipitation, snowfall, and soil moisture across western and central Canada during the winter and spring (Fig. 3), led to widespread and severe 2023 summer wildfires. In contrast, the less intense summer wildfires in the Ural-Siberia region (Fig. 1h) are likely in part due to small decreases in precipitation and soil moisture during the preceding winter and spring (Fig. 3).

We further observed that increased snowmelt primarily occurred in Alaska and western Canada during spring (Fig. 3f), but shifted eastward to the central and eastern parts of the North American mid-high latitudes in the summer (Fig. 1g). This spatial transition from enhanced spring snowmelt in the west to increased summer snowmelt in the central and eastern regions also likely contributed to both severity and the eastward expansion of the 2023 wildfires.

Pacific sea surface temperature anomalies associated with 2023 summer Canadian wildfires

While previous studies have largely attributed the escalating forest wildfires in the western US to anthropogenic warming over the North American continent1,2,9,10,43, interannual to decadal fluctuations in wildfire activity are primarily driven by Pacific sea surface temperature (SST) anomalies through their influence on drought conditions in the western US47,50. However, the influence of Pacific SST anomalies has generally been limited to the western US1,7. This suggests that earlier findings cannot explain the severity and geographic extent of the 2023 summer wildfires, particularly their widespread occurrence across Canada. Therefore, it is crucial to investigate the specific oceanic SST conditions that may have served as background conditions contributing to the exceptional wildfires observed across broader regions of Canada in the summer of 2023.

As shown in Fig. 1c, North American blocking during the summer is part of the positive PNA pattern. We further observe that the time-mean Z500 anomaly, averaged from June 1 to 10, 2023, resembles a positive PNA pattern (Fig. 2a), coinciding with the North American wildfire event during this period (Fig. 2b). This suggests that summer Pacific SST anomalies can influence North American blocking by altering the PNA patterns and westerly winds over the North Pacific51. In particular, the eastward movement of atmospheric blocking system is largely governed by the intensity of upstream westerly winds52. Hence, the observed eastward movement of the North American blocking in early summer 2023 is likely tied to intensified mean zonal winds in the North Pacific mid-high latitudes.

In the summer of 2023, the Pacific was dominated by a negative PDO (PDO−) with a value of −1.68 and an EP El Niño. The Pacific SST anomaly is classified as the PDO− and EP or CP El Niño combination when the 9-yr low-pass PDO index is negative, and the 9-yr high-pass EP or CP ENSO index is at or above 0.5 standard deviations (Fig. 4a, see “Definition of Pacific Decadal Oscillation and Eastern Pacific ENSO events” in Methods). The MJJ-mean Pacific SST anomaly in 2023 (Fig. 4c) closely resembled the composite pattern of the PDO− and EP El Niño combination (PDO− & EP El Niño hereafter) (Fig. 4b). However, the midlatitude warm anomaly associated with the PDO− in the 2023 summer was especially strong and extended farther east, near the west coast of US (Fig. 4c), compared to the composite midlatitude warm anomaly in the PDO− & EP El Niño summers (Fig. 4b). This suggests that the PDO− in the summer 2023 represented an exceptionally strong negative PDO.

Fig. 4: Temporal variations of summer Pacific Decadal Oscillation (PDO), central Pacific (CP) and eastern Pacific (EP) ENSO indices and their corresponding soil, atmospheric and oceanic anomaly fields during 1979–2023.
figure 4

a Time series of normalized summer (May to July, MJJ) mean 9-yr low pass PDO, 9-yr high pass central Pacific (CP) and eastern Pacific (EP) ENSO (Niño3.4) indices during 1979–2023 and (b) composite MJJ-mean SST (color shading, unit: K), (c) 2023 MJJ-mean SST anomaly, (d) composite MJJ-mean Z500 (contours; unit: gpm) & SAT (color shading; unit: K), (e) precipitation (P, unit: mm), (f) vapor pressure deficit (VPD, unit: hPa), (g) soil moisture (SM, unit: m3 m−3), (h) snowmelt (unit: m) and (i) fire weather index (FWI, unit: non-dimensional) anomalies for both the negative PDO (PDO-) and EP El Niño. The dot denotes the region being statistically significant at the 95% confidence level.

Full size image

Figure 4d further shows that, corresponding to PDO- & EP El Niño composite, a strong and expansive blocking anticyclonic anomaly developed over the North American high latitudes (north of 50oN) during the summer. This atmospheric pattern led to widespread surface warming, decreased precipitation, enhanced VPD, declined soil moisture and increased snowmelt, and intensified fire weather index across a vast region of North American mid-high latitudes, from the west to the southeast (Fig. 4d–i). The composite MJJ-mean FWI anomaly distribution (Fig. 4i) in PDO− & EP El Niño summer closely mirrors the wildfire fraction pattern seen over Canada in 2023 (Fig. 1h). In contrast, for the PDO− and CP El Niño combination (PDO- & CP El Niño hereafter), no strong VPD anomalies or widespread wildfires were observed in the North American high latitudes. This is likely because the MJJ-mean blocking anticyclone is weak or nearly absent in the North American mid-high latitudes (Supplementary Fig. 3). We thus conclude that the PDO- and EP El Niño combination likely represents the key oceanic condition that contributed to the intense and widespread wildfire events over Canada in 2023.

Oceanic modulation on North American blocking events and wildfire events

We further discuss the joint role of Pacific decadal and interannual variability in the 2023 wildfires, focusing on their modulation on North American blocking events. Based on our findings, different combinations of PDO− with the EP and CP types of El Niño can influence wildfire activity across the North American mid-high latitudes, likely by altering the frequency and persistence of blocking events. Our analysis reveals that during 1979–2023, there were 8 PDO− & CP El Niño summers and 5 PDO− & EP El Niño summers. Using a blocking index (see “Blocking index and the selection of North American blocking events” in Methods), we identified a total of 15 blocking events over North America during the PDO− & CP El Niño summers and 8 events during the PDO− & EP El Niño summers, averaging 1.88 and 1.60 blocking events per summer, respectively. While the difference in the mean number of blocking events between these two combinations is modest, the PDO- & EP El Niño has a slightly lower frequency than the mean of 1979-2023 (72 events).

However, the summer of 2023 stands out as an outlier, with three blocking events over North America far exceeding the average observed for both PDO- & CP El Niño and PDO- & CP El Niño summers. Furthermore, our results indicate that the North American blocking events, defined by a time-mean Z500 anomaly threshold exceeding 20 gpm, tend to be longer-lived (Fig. 5l) under the PDO− & EP El Niño condition, whereas they are generally shorter-lived (Fig. 5k) under PDO− & CP El Niño combination. Specifically, under the PDO- & EP El Niño condition, blocking anticyclones exhibit a large zonal scale of about 90 longitudes ( ~ 4920 km) and a long lifetime of ~10 days in North America (50o-70oN). In contrast, under the PDO− & CP El Niño condition, the blocking anticyclones are smaller and shorter-lived, with a small zonal scale of about 40 longitudes ( ~ 2187 km) and a short lifetime of ~6 days. The unusually wide and persistent blocking events in 2023 may be partly attributed to the elevated index of the EP El Niño in the presence of strong negative PDO.

Fig. 5: Time-mean composite daily atmospheric fields and wildfire weather parameter anomalies averaged during the blocking lifecycle of North American blocking events in summer under different combinations of negative PDO (PDO−) with the central Pacific (CP) and eastern Pacific (EP) types of El Niño during 1979–2023.
figure 5

a–j Time-mean anomaly fields averaged from lag −10 to 10 days of composite daily (a, b) 500-hPa geopotential height (Z500, contour interval=10 gpm) and surface air temperature (SAT, color shading, unit: K), (c, d) vapor pressure deficit (VPD, color shading, unit: hPa), (e, f) precipitation (P, color shading, unit: mm/day), (g, h) soil moisture (SM, color shading, unit: m3 m−3) and (i, j) fire weather index (FWI, color shading, unit: non-dimensional) of summer North American blocking events for the combinations of the negative PDO (PDO−) with the (a, c, e, g, i) CP and (b, d, f, h, j) EP types of El Niño during 1979–2023, where lag 0 denotes the peak day of each blocking event. The color shading denotes the region being statistically significant at the 95% confidence level. k, l Time-longitude evolutions of composite daily Z500 anomalies of summer North American blocking events averaged over the high latitudes (50o-70oN) during the blocking process for the (k) PDO− & CP El Niño and (l) PDO− & EP El Niño during 1979–2023.

Full size image

Our composite analysis shows that the time-mean Z500 anomaly, averaged from 10 days before to 10 days after the peak (lag 0) of each blocking event, reveals a more intense and widespread anticyclonic anomaly over North America under the PDO− & EP El Niño condition (Fig. 5b) than under the PDO− & CP El Niño condition (Fig. 5a). This enhanced blocking anticyclone is accompanied by higher VPD anomalies, greater reduction in P and SM, and larger FWI (Fig. 5d, f, h, j versus Fig. 5c, e, g, i), which is also concurrent with the stronger snowmelt (Supplementary Fig. 4). The VPD, SAT and FWI over Canada increase significantly with the growth of North American blocking under the PDO− & EP El Niño condition (Supplementary Fig. 5). When the North American blocking is strongest, the VPD, SAT and FWI reach their maximum intensities and areas, while the decrease in SM slightly lags the blocking intensification (Supplementary Fig. 5). A strong VPD (Fig. 5d) and a sharp decline in SM (Fig. 5h) in a large region of Canada can be caused by persistent, strong and widespread surface warming (Supplementary Fig. 5a) and a large reduction of precipitation (Fig. 5f) associated with long-lasting blocking (Fig. 5l) under the PDO− & EP El Niño condition. The long-lasting blocking can increase the burned area of wildfires due to a persistent increase in VPD or SAT and a persistent decrease in SM or P53,54. We, thus, infer that the presence of long-lived North American blocking can contribute to intense, widespread and long-lasting wildfires in Canada by producing strong positive SAT and VPD anomalies as well as large negative SM, Smlt and precipitation anomalies.

Further quantification shows that the time-mean blocking anticyclonic anomaly, averaged over the blocking period from lag -10 to 10 days, induced an area-averaged surface warming of ~0.50 K over Canada (140o-60oW, 50o-70oN) under the PDO− & CP El Niño condition (Fig. 5a), but a substantially greater warming of ~1.62 K under the PDO− & EP El Niño condition (Fig. 5b). Concurrently, the PDO- & CP El Niño blocking events produced a mean positive VPD anomaly of ~0.14 hPa, accompanied by P and SM anomalies of −0.07 mm/day and −0.25 m3m−3, respectively. In contrast, under the PDO− & EP El Niño condition, the corresponding VPD anomaly rose to ~0.34 hPa, with larger negative P (−0.1 mm/day) and SM (−0.53 m3m−3) anomalies.

Moreover, North American blocking during PDO- & EP El Niño summer induced more intense and widespread snowmelt over Canada (Supplementary Fig. 4b) than under the PDO−& CP El Niño condition (Supplementary Fig. 4a). Specifically, the area- and time-mean of composite daily Smlt anomaly over Canada during blocking episodes was −0.36 \(\times\) 10−5 m for the PDO− & EP El Niño case and 0.33 \(\times\) 10−5m for the PDO− & CP El Niño case. Thus, it is suggested that blocking-induced surface warming or heat dome over North America produces more intense, prolonged, and widespread wildfires under the PDO− & EP El Niño combination (Fig. 5j) compared to the PDO− & CP El Niño condition (Fig. 5i). However, during the summer of 2023, the concurrent presence of both PDO− & EP El Niño and PDO− & CP El Niño conditions likely acted in concert, reinforcing each other’s impacts and amplifying the severity of wildfire activity, while the PDO− & CP El Niño condition plays a minor role.

The fine fuel moisture code (FFMC), which integrates air temperature, relative humidity, wind and rain, has been used as a key component of the FWI system55. This parameter measures the flammability of fine fuels and predicts the potential for the ignition and spread of surface fires. A higher FFMC indicates a greater risk of rapid fire spread. In fact, blocking ridges can enhance FFMC, Duff moisture code (DMC), drought code (DC) and initial spread index (ISI)40 through temperature increase and precipitation decrease55. Here, we use the daily FFMC and soil moisture data to examine how the FFMC and SM associated with North American blocking events depend on the type of El Niño under PDO−. Our result shows that the FFMC and the SM deficit are more significantly intensified in a large region from Canada to the east part of the United States during the blocking episode due to the blocking effect under the PDO− & EP El Niño condition than under the PDO− & CP El Niño condition (Supplementary Fig. 6). Hence, North American blocking events under the PDO− & EP El Niño condition are more favorable for large wildfires in Canada. Some studies56,57 have revealed that the atmospheric response to the El Niño resembles a positive PNA pattern with larger negative anomalies over the Northern Pacific, and a more intense Canadian positive height anomaly compared to that of the La Niña. The North American blocking associated with the positive PNA pattern is stronger for the PDO− & EP El Niño (Fig. 5b) than for the PDO− & CP El Niño (Fig. 5a) partly because the PNA pattern associated with the PDO− & CP El Niño condition is weaker (Fig. 5a). It seems that positive PNA patterns and long-lived North American blocking events, and their influences are closely tied to EP El Niño and PDO−. Thus, the wildfires in Canada rely on the type of El Niño under PDO− due to changes in North American blocking anticyclones associated with positive PNA patterns.

Furthermore, the westerly winds over the North Pacific mid-high latitudes in the summer of 2023 were stronger than the composite zonal winds observed during the PDO− & EP El Niño summers (Fig. 6c). This led to a stronger eastward movement of North American blocking events in 2023 (Fig. 2f) compared to the PDO− & EP El Niño (Fig. 5l). This eastward progression facilitated the spread of wildfires from western Canada to eastern US or Canada. In addition, the blocking frequency over North America in 2023 is ~2 times of that in the PDO− & EP El Niño composite, which increases the persistence of wildfires in Canada. These frequent and long-lasting North American blocking events with enhanced eastward movement, which were associated with the particularly strong PDO− and EP El Niño combination, were thus the primary drivers of the massive wildfires in Canada in 2023.

Fig. 6: Background conditions of summer North American blocking events in 2023 and the PDO- & EP El Niño.
figure 6

a–c Summer (May to July, MJJ) mean 500 hPa zonal wind (U500, unit: m s−1) and (d–f) non-dimensional meridional potential vorticity gradient (PVy, unit: non-dimensional) anomalies for North American blocking events excluded (the days of blocking events from lag −10 to 10 days are removed, where lag 0 denotes the strongest day of the blocking event) for (a, d) 2023 and (b, e) PDO− & EP El Niño and (c, f) their difference (a minus b and d minus e). The color shading represents the area being statistically significant at the 95% confidence level.

Full size image

A reduced meridional background potential vorticity gradient (PVy) over North American mid-high latitudes can increase the frequency, persistence and zonal scale of North American blocking events, as suggested by atmospheric blocking theory based on a nonlinear multiscale interaction (NMI) model58,59,60 (see “Nonlinear multi-scale interaction model of atmospheric blocking” in Methods). A smaller PVy in the mid-high latitude North America can be produced by oceanic condition via changes in background westerly winds over North American mid-high latitudes (Fig. 6). In 2023, PVy was smaller over the North American mid-high latitudes (40o-60oN) compared to the typical PDO− & EP El Niño condition (Fig. 6f). This likely occurred due to more pronounced intensification (weakening) of westerly winds over high (middle) latitudes in North America in 2023 than in the PDO− & EP El Niño composite (Fig. 6a, c), leading to a smaller PVy in the mid-high latitudes (40o-60oN) in 2023 (Fig. 6d, f).

In 2023, the strong PDO− contributed to strengthening westerly winds over the North Pacific mid-high latitudes through the intensified Aleutian low61, which extended toward North American high latitudes and created a smaller PVy over North American mid-high latitudes (Fig. 6d). This was achieved by enhancing the meridional SST gradient in North Pacific mid-high latitudes, suppressing the retrogression of North American blocking. Thus, a weaker summer PVy in the North American mid-high latitudes and stronger North Pacific mid-high latitude westerly winds in 2023, associated with the strong PDO− and EP El Niño combination, favored the occurrence, prolonged persistence, reduced retrogression, and larger zonal scale of North American blocking events58,59,60.

Discussion and summary

Canada experienced its worst wildfires on record during the summer of 2023, marked by the largest burned area, unprecedented smoke pollution36, and the highest wildfire-induced CO2 emissions37. In this study, we examined the underlying causes of this extreme wildfire season and identified the key atmospheric and oceanic drivers. Our analysis reveals that unusually frequent, persistent, and eastward-moving blocking events with a large zonal scale over North America, alongside widespread and long-lasting heat domes, were the primary meteorological factors fueling the intense and widespread wildfires of the summer 2023. Also crucially, precipitation and soil moisture were well below normal during the preceding winter (November 2022 to January 2023) and spring (February to April 2023) as well as strong spring snowmelt, serving as important precursors that increase fire risk across Canada. Thus, persistent, eastward-moving and frequent North American blocking events were a key atmospheric driver of the extreme Canadian wildfire in summer 2023.

We found that the North Pacific SST anomaly in 2023 resembled a PDO−, coinciding with both an EP El Niño and a CP El Niño. The PDO- in the summer of 2023 was exceptionally strong, with its midlatitude warm SST anomaly extending farther east than in a typical PDO−. This eastward extension strengthened mid-high latitude westerly winds over the North Pacific, enhancing the eastward movement of North American blocking and thereby facilitating the eastward spread of the 2023 wildfires.

Our analysis further shows that the combination of the PDO− with EP El Niño promotes the development of North American blocking events that are more persistent, less retrogressive, broader in zonal extent, and stronger in amplitude, largely through the amplification of positive PNA patterns56,57. These blocking characteristics led to persistent and widespread surface warming and elevated VPD, alongside significant precipitation and soil moisture deficits as well as strong snowmelt across North American mid-high latitudes during summer. In addition, the combination of a PDO− and CP El Niño also induces warm and dry conditions over the North America while its impacts are weaker compared to those associated with the PDO− and EP El Niño pairing. Together, the Pacific SST anomalies related to the strong PDO− and EP El Niño represent key oceanic conditions driving the intense and extensive wildfires over Canada in 2023.

In summary, our results indicate that the 2023 summer extreme wildfires in Canada can be attributed to the following main factors: (1) unusually frequent, long-lasting, and eastward-moving blocking events with large zonal extent across North America, (2) substantial decreases in precipitation (or snowfall) and soil moisture during preceding winter and spring, along with strong spring snowmelt, and (3) the presence of an exceptionally strong PDO−, in conjunction with mainly an EP El Niño.

Although our study reveals the critical role of oceanic and atmospheric processes in driving the 2023 wildfires in Canada, land surface processes, such as fuel types, ecosystem conditions, land management actions and ignition sources, are also expected to have an effect15,62,63,64,65,66,67. Wildfires in the western North America were also influenced by anthropogenic warming9,10 via strengthening summer heatwaves and prolonged drought68,69,70, and the Atlantic Multi-decadal Oscillation (AMO)71 in addition to PDO and ENSO70. Thus, it is of importance to further consider the roles of anthropogenic warming, ground conditions, multi-scale atmospheric and oceanic processes and their interactions in the 2023 Canadian wildfires. Assessing the relative contributions of these factors to the 2023 wildfires is an important issue that warrants further investigation. Furthermore, the PVy changes are also associated with the Arctic Amplification and influence North American blocking events from the NMI model view. Additionally, the occurrence of summer wildfires in Canada is an integrative result of numerous weather and climate factors. Therefore, the generation and variability of the 2023 wildfires may involve a more complex mechanism, which requires further study.

While climate warming contributes to the long-term increase in summer wildfire activity in boreal regions1,2, the occurrence, persistence and extent of wildfires also depend on weather patterns associated with natural climate variability. For example, the milder Canadian wildfire season in summer 2024 coincided with fewer blocking events, partly due to the influence of a Pacific La Niña in 2024, which, as shown in our study, tends to suppress blocking activity. Further investigation is needed to disentangle the relative contributions of naturally variability and anthropogenic forcing to wildfires.

It is important to note that wildfires in North America respond to a wide range of interactive factors across multiple spatial and temporal scales. Our study has focused on the broader atmospheric and oceanic contexts and large-scale drivers, revealing how certain conditions set the stage for wildfire-prone scenarios and shape the evolution of fire events. Moreover, the pronounced precipitation and soil moisture deficits observed over western Canda during the preceding winter and spring remain insufficiently understood and merit further investigations.

Methods

Data treatment

We used monthly and daily atmospheric data that include surface air temperature (SAT) at 2 m, dewpoint temperature at 2 m and 500 hPa geopotential height (Z500) from the fifth generation European Centre for Medium-Range Weather Forecasts (ECMWF) reanalysis for the global climate and weather (ERA5 with a 1° grid point)72 in winter (November to January, NDJ), spring (February to April, FMA) and summer (May to July, MJJ) from January 1979/February 1979 to June 2023/August 2023 (1979–2023). The daily precipitation data used in this study is taken from ERA5. We used the summer daily and monthly Fire weather index (FWI) and Fine fuel moisture code (FFMC), which is taken from the Copernicus Emergency Management Service for the European Forest Fire Information System (EFFIS) with horizontal resolution of 0.5o × 0.5o grid points from 1979 to 2023. The FWI indicates the fire intensity by combining the rate of fire spread with the amount of fuel being consumed and is used for general public information about fire danger condition. FFMC is representative of the top litter layer less than 1–2 cm deep and values change rapidly because of a high surface area to volume ratio, and direct exposure to changing environmental conditions. The summer monthly sea surface temperature (SST) is taken from Hadley Centre (HadISST with a 1° grid point)73, whereas the monthly precipitation we used is taken from the global station data from Global Precipitation Climatology Centre (GPCC). The daily or monthly anomaly fields were obtained by removing the long-term (1979–2023) mean and linear trend of all the daily or monthly data for each calendar day or month. We used the daily and monthly wildfire fraction and atmospheric carbon dioxide (CO2), which was derived from the Global Fire Assimilation System (GFAS) with horizontal resolution of 0.1o × 0.1o grid points from 2003 to 2023 based on the satellite observation. The time series with a 9-year low pass smoothing is used to represent the decadal and interdecadal variations based on the 9-year low pass filtering of a time series via the Lanczos filter method74.

Definition and physical meaning of vapor pressure deficit

The vapor pressure deficit (VPD) is a combination of air temperature and relative humidity. When the VPD is high, there are high air temperature and low relative humidity, it can cause the most fire-prone conditions and increases fire ignition, growth and burned area though the dying of fuel75. Meanwhile, the high VPD is also related to the drought, low precipitation and soil moisture deficit76, and associated with high pressure systems13.

The VPD is defined as the difference between the saturation vapor pressure (es) and actual vapor pressure (ea), which is written as75

$${{{\rm{VPD}}}}={e}_{{{{\rm{s}}}}}-{e}_{{{{\rm{a}}}}}$$
(1)

where \({e}_{s}=6.109\times {e}^{\left(\frac{17.625\times T}{T+243.04}\right)}\) and \({e}_{a}=6.109\times {e}^{\left(\frac{17.625\times {T}_{d}}{{T}_{d}+243.04}\right)}\), where \(T\) is the near surface (2 m) air temperature (°C) and \({T}_{d}\) is the dew point temperature (°C).

Blocking index and the selection of North American blocking events

To identify North American blocking events, we pick blocking events confined in the longitudinal sector from 120oW-80oW based on the one-dimensional blocking index of Tibaldi and Molteni (TM hereafter)77. The TM index is defined in terms of the reversal of the meridional gradient of Z500:

$${GHGN}=[Z500({\phi }_{N})-Z500({\phi }_{o})]/({\phi }_{N}-{\phi }_{o})$$
(2)
$${GHGS}=[Z500({\phi }_{o})-Z500({\phi }_{s})]/({\phi }_{o}-{\phi }_{s})$$
(3)

at the three given reference latitudes: \({\phi }_{N}={80}^{o}N+\,\Delta\), \({\phi }_{o}={60}^{o}N+\,\Delta\), \({\phi }_{s}={40}^{o}N+\,\Delta\) and \(\Delta \,=\,{-4}^{o},{0}^{o},{4}^{o}\) in a fixed longitude. A blocking event is defined to have taken place if the conditions GHGS > 0 and GHGN < −10 gpm (deg lat)−1 hold for at least three consecutive days and are satisfied for at least one choice of \(\Delta\) in a given zonal region covering at least 15o of longitudes.

Definition of Pacific Decadal Oscillation and Eastern Pacific ENSO events

We define the value of 9-yr low-pass MJJ-mean Pacific Decadal Oscillation (PDO) index greater than (less than) 0 standard deviations (STDs) as the positive (negative) phase of the summer PDO or PDO+ (PDO−). Similarly, the value of 9-yr high pass MJJ-mean EP or CP ENSO index with ≥0.5 STDs is defined as the EP or CP El Niño in summer78.

Nonlinear multi-scale interaction model of atmospheric blocking

The background meridional potential vorticity gradient (PVy) is a key factor influencing the spatiotemporal evolution of atmospheric blocking based on the nonlinear multiscale interaction (NMI) model of atmospheric blocking58,59,60. When PVy is smaller, the atmospheric blocking is favored, which has longer lifetime, weaker eastward movement, larger zonal scale and stronger intensity due to stronger nonlinearity and weaker non-dispersion59. The non-dimensional three-dimensional baroclinic PVy scaled by the characteristic horizontal velocity \(\widetilde{U}\) ( ~ 10 m/s), length \(\widetilde{L}\) ( ~ 106 m) and vertical height \(\widetilde{H}\) ( ~ 104 m) can be expressed as

$${{{{\rm{PV}}}}}_{{{{\rm{y}}}}}=\beta -\frac{{\partial }^{2}U}{{\partial y}^{2}}-\frac{1}{{\rho }_{0}}\frac{\partial }{\partial z}({\rho }_{0}{F}_{r}\frac{\partial U}{\partial z})$$
(4)

where \({\rho }_{0}={\rho }_{s}{e}^{-z}\) with \({\rho }_{s}\) being the density of atmospheric reference state at the Earth’s surface (z = 0), \({F}_{r}\) = \(\frac{{\widetilde{L}^{2}}{{f}}_{o}^{2}}{{N}^{2}{\widetilde{H}^{2}}}\) \(\approx 0.7259\) and \(\beta ={\beta }_{0}\widetilde{L}\)2 /\(\widetilde{U}\), \({N}^{2}\) is the Brunt-Vaisala frequency, U(y, z) is the non-dimensional basic zonal wind varied in the meridional and vertical directions. \({f}_{0}\) and \({\beta }_{0}\) are the Coriolis parameter and its meridional gradient at a given reference latitude \({\varphi }_{0}\).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.