Fires are a distinct natural disturbance in combustible vegetation areas and release carbons into the atmosphere on a short time scale1,2. The emission from fires originating from mid and high latitudes, in particular, is comparable to the fuel emission3,4. Hence, to reduce negative social and economic impacts, it is a matter of great importance to investigate dominant factors for generating severe and frequent fires. Fire weather index (FWI)5 has been defined as an indicator of fire potential considering that fire ignition, spread, and frequency can be influenced by specific weather conditions such as temperature, relative humidity, precipitation, and wind speed. It has been suggested that those favorable fire weather conditions can be modulated by human-induced global warming and natural climate variabilities on interannual to interdecadal timescales6-11. The Arctic oscillation and associated jet streams play a role in a fire in the high latitude12-13, while ENSO affects tropical and mid-latitude fires11,14. In a warmer climate, fires are expected to be intensified and occur more frequently with hotter and drier conditions and more lightning occurrences6-8.
The Madden–Julian oscillation (MJO) is a subseasonal variability with 20–100-day periods in the tropics. The MJO is represented by the eastward-propagating deep convective system in the Indian Ocean and western Pacific. The MJO teleconnections have a profound impact on rainfall15,16,20, drought19, hurricane17,18, and regional monsoons21,22. The convective heating causes divergent flow in the upper troposphere, generating Rossby wave trains23-26. The MJO teleconnection could be a trigger for extreme weather events by generating a heatwave and drying in high-population mid-latitudes, particularly, in East Asia, eastern Europe, and United State23-27. However, impacts of MJO-related weather conditions on fire occurrences during boreal spring and summer on a global scale are not quantified and remain unclear. Here we present the relationship between station-based daily fire emission and fire weather conditions regulated by MJO convective anomalies over the eastern Indian Ocean and western Pacific. We bring forward that MJO could be a trigger and booster for boreal spring and summer fire activities in many tropical and mid-latitude regions by generating severe weather conditions favorable for fires.
Sub-seasonal variability of fire
Climatologically major fire emission regions are California, Alaska, southern US, Syberia, South America, Central Africa, east and south Asia, eastern Europe, and northern and eastern Australia (Supplementary Figure S1a). In addition to that, daily fire emission data exhibit significant subseasonal variability across the globe (Fig.1). Regions with considerable 20-100-day variance include the southern U.S., East Asia, central Africa, eastern Europe, eastern Australia, and South America (Fig. 1), indicating that climatologically abundant fire emission areas are consistent with those with significant subseasonal variance. Note that the filtered variance of fire emission is relatively larger in all seasons except boreal wind winter boreal (Supplementary Fig. S2). In most tropical land regions, the contribution of the 20-100-day variance to total variance is 20-40%, particularly, in South Africa, northern Australia, South Asia, and South America. In the midlatitude, the filtered variance explains 8-20% of the total variance: eastern Europe and the Siberia region show relatively higher variance by 15-20%, suggesting that the subseasonal variability of fire emission may explain a large portion of fire emission variability (Supplementary Fig. S1b). Note that portion of subseason variability in the Syberia, Alaska, and North California increases in the boreal summer and autumn (Supplementary Fig. S2). The power spectrum of the daily emission data selected major land regions highlights significant power at 20-100 days periods (Fig.1b-g). Fire emission in the Southern US, central Africa, and eastern Europe shows relatively longer periodicities (60-80 days), while those in east Asia, southern US, and South America represent relatively shorter ones. Analysis of Fire Radiative Power (FRP)28, energy from fire based on satellite observation, further supports the large portion of subseasonal variance to the total variance in most tropics and mid-latitudes (Supplementary Fig. S3).
FWI during active MJO and global fire
We hypothesize that the strong subseasonal variability of fire emissions across the globe is modulated by the MJO taking into account the dominant timescales shown in Fig.1. To prove it, a composite analysis is applied to the emission data, collecting the daily emission anomaly for each season when significant MJO events (standard deviation is greater than 1, see Methods) occur in the Indian ocean (phase 2 and 3) and western Pacific (phases 5 and 6). It is to note that the MJO teleconnection originated from convective activities over the Indian Ocean and western Pacific is closely associated with fire events in both tropics and mid-latitudes, particularly southeast and west coast of the United States, southeast of South America, northeast and southeast Asia, northern and eastern Australia, eastern Europe, and Central Africa (Fig. 2 and Supplementary Figure S4-S6). The emission anomalies are dependent on the MJO phases; strong negative anomalies are observed during MJO phases 2 and 3 in Africa, East Asia, Eastern Europe, and Argentina, while positive anomalies during MJO phases 5 and 6. Additionally, the amplitude and sign of emission anomalies associated with MJO teleconnections depend on seasons, particularly high latitudes of the Northern Hemisphere (e.g. Syberia and Alaska).
In boreal spring (February-April), the negative emission anomalies during MJO phases 2 and 3 are dominant in most tropics and mid-latitude regions, particularly, central Africa, East Asia, Eastern Europe, southern US, and East of Australia, while positive anomalies during the MJO phase 5 and 6, indicating that MJO teleconnection contributes to increasing fire emission during MJO phase 5 and 6 (Fig. 2a and c). Noted that the emission changes by MJO phases are associated with the change of geopotential height anomalies in the upper troposphere by eastward movement of MJO convection in the tropics. From May to July, the negative impact of MJO teleconnection increases in Siberia, South America, Central Africa, and the west of North America during MJO phases 2 and 3 but positive impact during MJO phases 5 and 6 in the same regions. In August-October, compared to boreal spring, positive emission anomalies are dominant in eastern Europe, western Australia, and Alaska. However, during MJO phases 5 and 6, emission anomalies show opposite signs. In boreal winter, the emission anomalies are reduced in most mid-latitude regions of the Northern Hemisphere, suggesting that the impact of MJO teleconnection is relatively weakened during November-January. In summary, the MJO-related emission anomalies are significant for boreal spring to autumn in the mid-latitude regions but the amplitude and sign strongly depend on seasons and MJO phases.
The relationship between MJO-related FWI and fire
MJO deep convection and associated strong divergence can affect weather conditions in the mid-latitudes by exciting traveling atmospheric waves. Warming and drying local weather conditions induced by The MJO would be favorable to fire occurrences. The FWI represents favorable weather conditions for fire activities considering temperature, rainfall, humidity, and wind. Thus, it is of particular importance to investigate whether the local FWI anomalies driven by MJO are significantly linked to changes in fire emission or not. For this, we calculated composites of FWI anomalies for MJO phases 2 and 3 or MJO phases 5 and 6 during each season (Fig. 2b and 2d). The results show that the emission anomalies are closely linked to the FWI anomalies related to the MJO phases. During MJO phases 2 and 3, the pattern of emission anomalies is consistent with those in the FWI anomalies in boreal spring; the horizontal map of emission anomalies in central Africa, South America, Australia, East Asia, southern US, and eastern Europe is similar to those in the FWI. Note that the sensitivities of emission anomalies to FWI changes seem to be different from each region. In boreal summer, the pattern of emission anomalies is consistent with those of the corresponding FWI except in Australia, where the signs of the anomalies are opposite. In boreal autumn, the spatial patterns of emission anomalies in central Africa and the western US resemble those in the FWI but do not match in western Australia and eastern Europe. In boreal winter, the magnitude of both emission and FWI anomalies are relatively weak in most mid and high latitudes, particularly, in the Syberia and Alaska.
When MJO convection occurs in the western Pacific (MJO phases 5 and 6), the pattern of FWI anomalies is capable of capturing those in emission anomalies in the US, east Asia, eastern Europe, northern Australia, and central Africa during boreal spring. Compared to the MJO phases 2 and 3, the consistency between the FWI and emission anomalies is intensified, particularly in the southern US (February-October), East Asia (all season), and eastern Europe (February-October, August-October). It may be attributed to stronger MJO convective anomalies and longer duration during MJO phases 5 and 629.
This study brings forward that the FWI changes induced by MJO convection should be a key factor in determining the changes in the fire emission, particularly in the mid-latitudes. To further validate this, we compare the relationships between daily FWI and emission anomalies during the active MJO period to those during total periods. The results show that the linear relationship between FWI and emission is robust during active MJO periods but very weak during total periods in boreal spring. The relationships between the FWI and emission during total periods are weak with low correlations of 0.17 to 0.34, while they are very strong during active MJO periods with high correlations of 0.58-0.68. The most strong relationship is seen in eastern Europe but weak in eastern Australia. The strength of the impact of MJO teleconnection (that is, the changes in fire emission by unit change of FWI) depends on the local region and MJO phases. Compared to those during MJO 2 and 3, the strength during MJO 5 and 6 is higher in the southern US, while lower in Eastern Australia. On the other hand, there are no significant changes in strength in eastern Asia and eastern Europe.
Contribution of seasonal-mean FWI on subseasonal fire
A seasonal mean-state weather condition may be a critical factor for interannual variability of fire emission rather than weather change by MJO teleconnection although it may be a trigger for generating or intensifying fire emission. We examined the relationship between interannual variability of MJO-related fire emission (or FWI) anomalies and seasonal mean-state FWI in high-population of midlatitudes, particularly the US, East Asia, Eastern Europe, and Australia. The results showed that during MJO phases 2 and 3 of boreal spring, the MJO-related emission change at the interannual time scale is closely linked to changes of corresponding FWI anomalies with a correlation of 0.55-0.86. For large changes in FWI anomalies, the fire response is more concrete. On the other hand, when we compared emission anomalies (blue line in Supplementary Fig. S8) to present seasonal mean FWI (gray bar in Supplementary Fig. S8), the emission change is less linked to the change of mean-state FWI with a low correlation of 0.01-0.25. Also, a last winter mean FWI may affect the emission anomalies due to lagged effect through overwintering drying. We compared the emission anomalies in boreal spring to seasonal mean FWI in the previous winter. The results showed that the impact of atmospheric conditions in the previous winter on emission in the next spring is not significant. Note that the interannual relationships between emission and FWI are also significant during MJO 5 and 6 (Supplementary Fig. S9 and S10).