Forest fires in Australia are burning more land across more of the year than ever before, as climate-linked fire weather worsens.

The Black Summer forest fires of 2019–2020 burned more than 24 million hectares, directly causing 33 deaths and almost 450 more from smoke inhalation.

But were these fires unprecedented? You might remember sceptics questioning the idea that the Black Summer fires really were worse than conflagrations like the 1939 Black Friday fires in Victoria.

We can now confidently say that these fires were far from normal. Our new analysis of Australian forest fire trends just published in Nature Communications confirms for the first time the Black Summer fires are part of a clear trend of worsening fire weather and ever-larger forest areas burned by fires.

What did we find?

Our study found that the annual area burned by fire across Australia’s forests has been increasing by about 48,000 ha per year over the last three decades. After five years, that would be roughly the size of the entire Australian Capital Territory (235,000 hectares).

We found three out of four extreme forest fire years since states started keeping records 90 years ago have occurred since 2002.

And we found that the fire season is growing, moving out of spring and summer into autumn and winter.

These trends are almost entirely due to Australia’s increasingly severe fire weather and are consistent with predicted human-induced climate change.

Our study is based on satellite and ground-based estimates of burnt forest area, and trends of nine wildfire risk factors and indices that relate to characteristics of fuel loads, fire weather, extreme fire behaviour, and ignition.

We have focused here only on the most dangerous forest fires, not the fires affecting Australia’s savanna across the tropical north.

A graph showing burnt area of forest by year
Burnt area of forest by year. Data derived from satellite data (NOAA-AVHRR burned area) Author provided

Fire burns much more land than 25 years ago

Before the 1990s, Australia’s forest fires were infrequent, though damaging. A given area would burn at an interval between 20 to over 100 years.

The exception were rare summers which would see severe and extensive fires, such as 1939. Overall, only a small fraction of the total forest area burned in any year.

This pattern of fire behaviour no longer exists.

A graph showing years since the last forest fire from 1980-2019. The 2010-2019 time frame has less years between fires.
Years since the last forest fire (decadal mean). Data derived from satellite data (NASA-MODIS burned area) and ground/air-based data from states and territories. Author provided

Over the last 30 years, the areas affected by fire have grown enormously.

If we compare the satellite records from 1988–2001 to the period from 2002–2018, the annual average fire area has shot up by 350%.

If we include the 2019–20 Black Summer fires, that figure soars to 800% – an enormous leap.

We are seeing fires growing the most in areas once less likely to be affected by fire, such as cool wet Tasmanian forests unaccustomed to large fires as well as the warmest forests in Queensland previously kept safe from fire by rainfall and a humid microclimate. This includes ancient Gondwanan rainforests not adapted for fire.

More extreme fire years and longer fire seasons

Before 2002, there was just one megafire year in the 90 years Australian states have been keeping records – and that was 1939.

Since 2001, there have been three megafire years, defined as a year in which more than one million hectares burn.

Our fire seasons are also getting longer. Spring and summer used to be the time most forest fires would start. That’s no longer guaranteed.

Since 2001 winter fires have soared five-fold compared to 1988–2001 and autumn fires three-fold.

Overall, fires in the cooler months of March to August are growing exponentially at 14% a year.

Graph showing trends in autumn and winter burned areas increasing from 1990 to 2020.
Trends in autumn and winter burned areas over time. Data derived from satellite data (NOAA-AVHRR and NASA-MODIS burned area) Author provided

What’s driving these changes?

Imagine a forest fire starts from a lightning strike in remote bushland. What are the factors which would make it grow, spread and intensify?

A fire will get larger and more dangerous if it has access to more fuel (dry grass, fallen limbs and bark), and if the fire starts when the weather is hotter, drier and windier. Topography also plays a role, with fire able to move much faster uphill.

To get a sense of the overall risk of forest fire, temperature, humidity, windspeed and soil moisture are combined into a single figure, the Forest Fire Danger Index (FFDI).

As you might expect, this index has been steadily worsening over the past 40 years. The number of very high fire danger days in forest zones has been increasing by 1.6 days per decade.

So what does this mean for fire behaviour and spread?

In what we believe is a first, we have used 32 years of fire index data across Australia’s forest zones and compared the number of very high or severe fire danger days with areas subsequently burnt by fire.

We found a clear link, with a 300 to 500% increase in burnt area for every extra day of severe fire danger, and a 21% increase in burnt area for every extra day of very high fire danger.

Could fuel loads or prescribed burning be to blame? No. We looked for trends in these factors, and found nothing to explain the rise in burnt areas.

The main driver for the growing areas burnt by fire is Australia’s increasingly severe fire weather, accounting for 75% of the variation observed in the total annual area of forest fires. This is consistent with predictions from climate change scenarios that severe fire weather conditions will intensify due to increasing greenhouse gas emissions.

Satellite image of New South Wales coast with red dots highlighting fires and the smoke from them.
Satellite imagery of the 2019 forest fires and smoke in New South Wales from November 9th. Joint Polar Satellite System/AAP

Other fire weather risks are also growing. We’re seeing more higher atmospheric conditions which can lead to the formation of fire-generated thunderstorms (known as pyrocumulonimbus clouds).

These thunderstorms emerging out of fire plumes can spread burning embers further and whip up more dangerous winds for unpredictable fire behaviour on the ground, as well as generate lightning in the fire plume that can ignite new fires far ahead of the fire front.

Dry lightning is the primary natural cause of fire ignitions. Here too, the trends are worsening in southeast Australia. We are now seeing 50% more dry lightning in forest areas in recent decades (2000–2016) compared to the previous two (1980–1999).

Under most climate change scenarios, fire weather is predicted to keep on worsening.

Can we predict our next megafire?

So could we have predicted how bad and how widespread the Black Summer fires would have been, if we had examined fire danger index forecasts in mid-2019?

In short, yes.

The huge amount of bush that burned is entirely consistent with the 34 days of very high forest fire danger across the forest zones that summer. That’s in line with the long-range bushfire weather forecasts provided to fire agencies earlier in 2019.

This means that in future years, we will be able to broadly predict the area likely to burn each fire season by examining fire index forecasts.

We can also safely – and sadly – predict that more and more of Australia will burn in years to come, with increasing numbers of megafire years.

While many factors contribute to catastrophic fire events, our Black Summer was not an aberration.

Rather, it was the continuation of fire trends beginning more than two decades ago. It is now clear that human-induced climate change is creating ever more dangerous conditions for fires in Australia.

We need to be ready for more Black Summers – and worse.

Garry Cook, Honorary Fellow, CSIRO; Andrew Dowdy, Principal Research Scientist, Australian Bureau of Meteorology; Juergen Knauer, Research fellow, CSIRO; Mick Meyer, Post Retirement Fellow, CSIRO; Pep Canadell, Chief research scientist, Climate Science Centre, CSIRO Oceans and Atmosphere; and Executive Director, Global Carbon Project, CSIRO, and Peter Briggs, Scientific Programmer and Data Analyst, CSIRO

This article is republished from The Conversation under a Creative Commons license. Read the original article.


  1. Hi Neville,

    Thanks for your comment. There is no available inventory of forest fuel loads over time at a national scale.

    Team CSIRO

  2. Fuel loads are the only variable we have any control over.
    Preventative burns cannot be dictated by a centralised control system with its lag periods of many months between lodging an application to burn, and receiving a permit (if the latter happe n s at all) The window to burn safely in an area may only be for a week or two eg, after rain, cool evenings, a slow breeze..

    Similarly with forestry and national park burns; by the time permission is received, the opportunity has usually passed, and if it hasn’t, then coordinating onerous requirements of having other brigades and police attend ensures ensures the final agreed date and time is not safe, and everyone goes home.

    It is the current system which reduces burning opportunities, and if the weather is getting hotter and drier, someone needs to come up with a way to manage the only tool available to us…. because all of the current net zero planning is still aiming at a 1.5 C temperature increase by the end of the century.

    It is all very interesting to blame it all on climate change, but what is the plan?

  3. Interesting to a point and then it just gets myopic.
    Can the next megafire be predicted.
    No there are far too many variable to predict such an event.
    But it is good to indicate that a mega firestorm can be predicted it frightens the children and scares the horses, or is it the other way round.

    I grew up walking the long paddocks with my Dad, the sheep and cattle cleared the sides of the roads.There were always at least a dozen goats who would eat the brambles that the sheep and cattle wouldn’t touch.
    Thus the roadsides were cleared of vegetation that was a carrier of flames during fire season.

    I grew up at the time when rural houses were not built in the middle of a scrub or forest, the land had to be cleared so that fires for the most part went around the houses, sheds would be lost but expensive machinery was normally shifted away from the fire heading.
    In the event of a fire farm gates were open so that livestock could move freely rather than have them perish in the cruelest of ways.

    I was also taught how to put out spot fires with a wet wheat bag around the house.

    Now houses are built in inaccessible landscapes, the gutters are overflowing with debris and trees overhang for aesthetic effect.
    The greenies won’t allow the long paddock to be cleared so the fire corridor is overgrown and filled with litter that just adds to the flammable materials.

    Fire breaks around houses are a thing of the past.
    Houses no longer have rain water tanks.

    Once a fire starts to run then it will continue to run until it either runs out of fuel or it rains enough to total extinguish the flames. All the firefighters can is slow and if fortunate divert the fire front.

    Early warning work to a point for tsunamis and earthquakes and at a stretch volcanoes.
    But every year we have tsunamis earthquakes and erupting volcanoes.

    Predictions for firestorms, sorry I am just not going to buy that one.

  4. There is a question in the above article “Could fuel loads or prescribed burning be to blame? The authors comment notes “No. We looked for trends in these factors, and found nothing to explain the rise in burnt areas”.
    In relation to this comment in regards to prescribed burning and fuel loads, I provide quick dot points:
    • In the paper that the authors released they use an average of 1 % prescribed burning across Australia. They note “only 1% of forests are subject to fuel reduction burns every year, it is very likely that fuel management had no effect on the observed multi-decadal increasing trend in the burned area of forest fires” The authors note elsewhere in an article that “this is a really small amount”, and I totally agree it is, but at unsatisfactory levels. This is the issue, the prescribed burning in eastern states has been way too low over a very long period, and fuel loads and 3D fuels continue to increase as they have over the last 30 years, safe levels of fuels haven’t been achieved and are getting worse.
    • The incidence of large wildfires in Western Australian forests over the last 67 years data unequivocally show that when the area of prescribed burning trends down, the area of uncontrolled bushfires (wildfires) trends up. The ideal area of forest burnt annually appears to be about 8 %.
    • The WA prescribed burning data has been combined with eastern data, as noted elsewhere this is like comparing apples with oranges. A key weakness is the bundling together of quite different forest types and management practices between west and east. In WA, the prescribed burned area is nearly half the national total, but it has considerably less forest area than the big eastern states. By bundling WA into the analysis, the authors arrive at conclusions that do not appear to be valid: For example: They say: “However, given the lack of trend and the fact that on average, only 1% of forests are subject to fuel reduction burns every year, it is very likely that fuel management had no effect on the observed multi-decadal increasing trend in the burned area of forest fires”. This includes much undetectable controlled burning in other states with wetter forests, but this does not apply nearly so strongly in WA (with some 1-2 million ha/year of controlled burning in mainly in more open jarrah forest).
    • On data I have, areas of annual prescribed burning in NSW have reduced since 2000. The 1998 NSW Auditor General performance audit report refers to 600,000 hectares per year in NSW around the year 2000 in NSW. The prescribed burning of the 7 million hectares of NPWS land in 2020/ 21 was 53,145 ha which is 0.76 % of NPWS lands.
    • It is important to note that prescribed burning is not expected to stop fires, it reduces intensity and damage.
    • I suggest that there is an inbuilt bias in this data, Modis/ Landgate would increase fire/ bushfire areas in more recent years, as well as potentially pick up mild burning, but they have tried to explain this. As well, in earlier years, much of the private prescribed burning would not have been recorded.
    • Reduced grazing in previous State Forests and transfer of lands to parks is another issue increasing fuel loads and increasing bushfire risks. Another potential influence in the data is the increasing population over time, increasing bushfire risks over time. I am not sure if arson has increased.
    • Safe and healthy landscapes and chronic eucalypt decline has been missed as an issue. Mild burning is needed to keep forests healthy, something they aren’t in many cases.
    • There are many many authors and practitioners who have highlighted the importance and success of mild burning across landscapes.
    • Fuel loads in the paper are simulated. An open discussion on actual (not simulated) fuel loads in forests across landscapes in NSW and other eastern states is warranted, to tease out fact from fiction.
    • It is important to highlight the fact that human fires have had a pivotal role in Australia’s ecological history. A useful reference Vic Jurskis, Roger Underwood, Neil Burrows. How Australian Aborigines Shaped and Maintained Fire Regimes and the Biota. Ecology and Evolutionary Biology. As noted recently, the 1824 and 1851 bushfires in Victoria related to increased fuel loads and scrub thickening after Aboriginal dispossession and disease which decimated the Aboriginal people. Key information such as this, human fires and Aboriginal cultural burning evidence, provides land managers with information to undertake adaptive land management practices.

    1. Hi John,

      Thanks for your comment. At a national scale, interannual variation in burnt area in forest over the satellite era is best explained by changes in FFDI. A completely independent analysis ( came to the same conclusion.

      Team CSIRO

  5. I would like to see the data for forest production industries (timber harvesting and other activities) included in the analysis. Fuel reduction programs are only part of the picture in WA. peter

What do you think?

We love hearing from you, but we have a few guidelines.