NCCS-Enabled Satellite Data Assimilation Reveals
Vegetation Changes from Massive Australian Bushfires

The Orroral Valley Fire rages on January 28, 2020 near Tuggeranong, the southernmost town center of Australia’s capital city, Canberra. This bushfire consumed over 155,000 acres in a week. Photo by Nick-D; acquired from Wikimedia Commons.

In late 2019 to early 2020, Australia had a spring and summer that few who experienced it will soon forget. Massive bushfires burned 25 million acres of land, including 21% of the continent’s forests, with devastating impacts on life and property. To better understand how Australia’s unprecedented fire season affected vegetation, NASA Goddard Space Flight Center scientists and international collaborators mounted an experiment at the NASA Center for Climate Simulation (NCCS), assimilating NASA satellite data into a computer model and analyzing the output.

This map of Southeast Australia shows fires for November 1–December 5, 2019 detected by the Suomi NPP satellite's Visible Infrared Imaging Radiometer Suite (VIIRS). Figure by Joshua Stevens and Lauren Dauphin, NASA Earth Observatory.

The experiment focused on hard-hit Southeast Australia, which comprises southern Queensland, New South Wales, and part of Victoria (see map at right). The study area is about 1 1/2 times the size of Texas.

NASA satellites that observe Earth’s surface provide researchers with diverse data at different resolutions. For comparison, the researchers separately assimilated three datasets:

  • Soil moisture and vegetation optical depth (VOD) from NASA’s Soil Moisture Active Passive (SMAP) satellite at 9-kilometer (km) resolution.
  • Moderate Resolution Imaging Spectroradiometer (MODIS) leaf area index from NASA’s Terra and Aqua satellites at 500-meter resolution.
  • Climatological VOD averaged across 2015–2019 as a baseline.

These datasets were assimilated into NASA’s Land Information System (LIS) with the NOAH-Multiparameterization (Noah-MP) land surface model running at 1-km resolution over the period April 2015–February 2020.

Satellite images of Southeast Australia for January 22 over the years 2017 to 2020. Top: false color images from Suomi NPP VIIRS. Bottom: maps of negative anomalies from Soil Moisture Active Passive (SMAP) vegetation optical depth (VOD). As shown by dark red coloring in the bottom January 22, 2020 image, the SMAP satellite can capture the significant vegetation changes from the 2019–2020 drought and bushfires.​ Figure from Kumar et al. 2021.

Using 512 cores on the NCCS Discover supercomputer, the data assimilation runs took four weeks to finish. The experiments produced approximately 1.2 terabytes of data, which were stored on Discover and analyzed using Python and Matplotlib.

“The 20-member model ensemble at 1-km resolution experiment this large domain is a computationally intensive setup,” said Sujay Kumar, study leader and research physical scientist in NASA Goddard’s Hydrological Sciences Laboratory. “These runs would not have been possible without the NCCS resources.”

Impact: The study shows that vegetation alterations from the fires led to significant changes in the local water cycle. Post-fire, there was more runoff and less evaporation, which may have contributed to the significant flooding that occurred in New South Wales after the 2019–2020 fires.

The experiment reveals that the SMAP satellite can capture the vegetation disturbances from the bushfires, which Kumar said was “a bit surprising given that SMAP data is fairly coarse.” It also shows that the fires led to significant changes in Southeast Australia’s regional hydrology (see the Impact box above).

The study results compare favorably against five thermal remote sensing and in-situ datasets used for evaluation. The research team is currently running similar studies over other fire-prone regions of the world.

Graph (a) above shows changes in precipitation measured in millimeters (mm). It compares averaged precipitation for the 2015–2019 climatology period (red line) to averaged precipitation during the 2019–2020 bushfire period (black line). Note the significant precipitation increases in early February 2020. Figure from Kumar et al. 2021.

The four panels above show the percentage change between the SMAP VOD and climatological VOD data assimilation runs in (b) evapotranspiration, (c) runoff, and the evapotranspiration components (d) transpiration and (e) bare soil evaporation during February 7–14, 2020. These results confirm the role of decreased vegetation in the flooding that occurred in New South Wales after the early February precipitation events. Figure from Kumar et al. 2021.

Related Links

Jarrett Cohen, NASA Goddard Space Flight Center