In September, 2020 surpassed 2019 to become the most active fire year in the southern Amazon since 2012, the year that the VIIRS sensor was launched on the Suomi-NPP satellite. So far this year, the VIIRS instrument on Suomi-NPP has detected 574,000 active fires, compared with 509,000 for the same period last year. Importantly, 88% of those fire detections have occurred in just the past 60 days. The upper panel (Fig. 1a) of the graphic below compares VIIRS active fire detections, classified by fire type (colors), to the average fire detections in 2012-2019 during the same time period (black line). Deforestation fires account for >40% of all fire detections in the southern Amazon in 2020, with a growing contribution from understory forest fires in the Brazilian states of Mato Gross and Pará in the past month. Visit the new Amazon Dashboard for more information on this classification approach, or to track individual fire events in the Amazon and surrounding biomes this year.
The lower panel (Fig. 1b) of this graphic shows the cumulative fire activity over the same 60-day period for the entire VIIRS data record from 2012-2020. Whereas fire activity surged in August, 2019 from coordinated burning activity in the Brazilian Amazon, daily fire counts in 2020 were higher in the first weeks of September than in August. Large, multi-day fire events are one reason for this increase in daily fire detections as the dry season progresses in the southern Amazon. As of September 20th, thousands of large fires continued to burn in the southern Amazon, including large understory forest fires that have been burning for weeks in the Xingu River basin in Mato Grosso (Fig. 2).
With support from NASA, scientists have developed a new approach to cluster individual fire detections into fire events and classify each fire event across the southern Amazon and surrounding biomes. Of the three fire types occurring in forested landscapes, deforestation fires showed the largest contribution to VIIRS 375 m satellite fire detections, followed by small clearing and agricultural fires, and understory forest fires. Updated daily, the Amazon Dashboard provides quick access to new information on the location, fire type, start date, spread rate, and duration of individual fire events across the Amazon. For more information, or to explore and download the near-real time data, please see our Amazon dashboard.
As of May, 2020, sea surface temperatures in the tropical Pacific Ocean and North Atlantic Ocean were significantly warmer than the mean values during 2001-2015. Using the historic relationship between sea surface temperatures and regional fire activity, we forecast moderate to high fire season severity for the Southern Amazon in 2020. Climate-driven fire risk is highest for the Brazilian states of Acre, Mato Grosso, Rondônia, and Pará, and the departments of Pando, El Beni, and Santa Cruz in Bolivia.
Cumulative MODIS active fire detections for New South Wales, Victoria, and Queensland summed per ‘fire year’ running from July 1st to June 30th in the following year. This fire year, satellite fire detections in New South Wales are more than four times higher than the previous record year (the 2002-03 fire season). In Victoria, fires are above average at this point in 2019-2020 but have not yet surpassed previous extreme fire years in 2002-2003 and 2006-2007. In Queensland, MODIS fire counts were consistent with previous years, despite large wildfires in southern Queensland, since total fire activity in this state is dominated by savanna fires in northern Queensland, a natural part of these ecosystems.
Cumulative active fire detections from May 1st through October 1st from MODIS (Aqua + Terra) and VIIRS (SNPP) show that fire detections in 2019 have fallen below cumulative levels of fire activity detected in 2012 (the start of the VIIRS record) and 2017 across the Legal Amazon. Through the end of September, fires in 2019 were more intense than any year since 2012, measured in terms of fire radiative power, consistent with the observed increase in deforestation this year.
Cumulative active fire detections from May 1st through August 28th from MODIS (Aqua + Terra) and VIIRS (SNPP) confirm that the 2019 fire season has the highest fire count since 2012 (the start of the VIIRS record) across the Legal Amazon. In addition, fires in 2019 are more intense than previous years, measured in terms of fire radiative power, consistent with the observed increase in deforestation. The updated figures include cumulative fire counts, cumulative fire radiative power (FRP), and an estimate of the average FRP for active fire detections each day. Fire detections in August 2019 from both sensors have higher average FRP than other recent years (2012-2018) for the fire season across the southern Brazilian Amazon.
Combined Terra + Aqua MODIS fire counts for the Legal Amazon since 2003
The combined MODIS data record from the Terra and Aqua satellites begins in 2003. Interpreting the longer MODIS record requires careful attention to economic and climatic variability over the past two decades. The MODIS time series of fire activity in the southern Amazon includes drought years (2005, 2007, 2010, and 2015-2016) and periods of higher deforestation activity (2003-2008). 2019 is not an extreme drought, despite the potential for lingering impacts from a weak El Niño that developed in late 2018 and warm sea surface temperatures in the tropical north Atlantic (see seasonal forecast information, below). The reported increase in deforestation in 2019 is also not at the level of clearing mapped by INPE during 2003-2008. Thus, MODIS fire counts to date in 2019 are a remarkable departure from recent years, but not a record for fire activity during the MODIS era. The figure below shows cumulative fire detections, cumulative fire radiative power (FRP), and mean FRP for Terra and Aqua MODIS fire detections from May 1st to August 28th for 2003-2019. Mean FRP per fire detection in 2019 is consistent with observations from years with more deforestation fires (2003-2007).
Cumulative active fire detections of the fire season from May 1st through August 22nd, 2019 from MODIS (Aqua + Terra) and VIIRS (SNPP) confirm that the 2019 fire season has the highest fire count since 2012 (the start of the VIIRS record) across the Legal Amazon. In addition, fires in 2019 are more intense than previous years, measured in terms of fire radiative power, consistent with the observed increase in deforestation.
Heavy rains have been reported in Kalimantan starting October 26 and since then the fire season has been coming to an end. There are still fires burning but nothing compared to those that burned earlier in the season.
This figure shows how the current fire season progressed in relation to previous ones (2003-2014). On October 21 this year passed 2006, which was the highest fire year in the MODIS satellite era, 2000 onwards. The main fire season is August through October when the southern part of Indonesia experiences its dry season. However, in some years including 2014 the fire season in the northern part of Sumatra is prominent as well, burning in February and March.
We expect that the GFED estimate for the 2015 fires will be about 1.75 billion metric ton of CO2 equivalents, with substantial uncertainty.
Above the greenhouse gas emissions from Indonesian fires are plotted
according to GFED for 1997-2014 with estimates for 2015 based on active
fires. These are converted to emissions based on a relation between the
two, established using data from previous years, see the figure and text
below for more information. The numbers on the right indicate fossil
fuel CO2 emissions for various countries for 2013 derived from the EDGAR database.
In general, fire CO2 emissions are compensated for by regrowing vegetation after a fire and should not be compared to fossil fuel emissions, but that is not the case when forests are burned to make way for other land uses or when peat is burned. That is exactly what happens with the vast majority of the fires in Indonesia and these fires are thus a net source of CO2 as well as other greenhouse gases.
Conversion of active fires to emissions
This graph shows how we derive the 2015 estimates. The grey dots
indicate the total annual active fire observations in Indonesia on the
horizontal axis and the corresponding GFED estimates are on the vertical
axis with the years 2006 and 2014 labeled. Each grey dot represents one
year between 2003 and 2014. The relation is not perfect and adds some
uncertainty to those that are in these estimates already. The
non-linearity is probably related to smoke obscuration of active fires
in high fire years.
Note: we have adjusted the trendline describing the relation between active fire detections and emissions on October 20 to better represent the high fire years, especially 2006. This led to a small increase in emissions compared to the relation used before October 20.
Using the conversion from active fires to emissions we can calculate
daily emissions which is shown above. This has generated a lot of media
interest after WRI showed that on many days the rate exeeds that of fossil fuel emissions in the US
(roughly 15 million ton CO2 per day). Keep in mind though that these
fires do not burn continuously at this rate: on a global annual scale
they are far less important for climate change than other sources of
A few things to consider:
These estimates contain a substantial amount of uncertainty related
to the complex fire situation in Indonesia. We would very much welcome
opportunities to lower uncertainties which is poorly supported by the
regular national agencies.
GFED starts in 1997, there is information on earlier years thanks to work of, amongst others, Wooster et al. (2012) and Field et al. (2009).
That work supports the tight link between (El Nino related) drought and
fires. However, El Nino is not the root cause: the work of Field shows
that without humans these extreme fire seasons would not exist.
In addition to greenhouse gas emissions from fires, drainage of
peatlands also leads to CO2 emissions from decomposition. The emissions
are not as visible as those from fires but may be of similar magnitude
according to the work of Hooijer et al. (2010), although not as variable from year to year.