Monitoring the Kilauea Eruption Through Cloud Cover with Radar and AI
If you look for optical satellite images of the volcanic eruption in Kilauea over the last month, you will have difficulty finding a clear shot due to a combination of clouds, gases and dust obscuring the satellite’s view. Optical images might have been somewhat useful on the day of the initial eruption, but as time went on, the visibility decreased as clouds moved in and as the volcano spewed more gases and dust.
PlanetWatchers utilizes both Synthetic Aperture Radar (SAR) and optical imagery for its multi-source analytics platform. Radar satellites can collect data through cloudy or hazy conditions and at night, making them the ideal data source to analyze a volcanic eruption over time.
Using this advanced technology, PlanetWatchers was able to begin monitoring this event right away and on a reliable schedule through the clouds and smoke. Figure 1 shows our analysis of areas newly covered by lava flows over the last several weeks.
We achieved this analysis by using our in-house algorithms, which we normally use for applications in forestry, agriculture and other natural resource domains, such as monitoring illegal logging activities or tracking the growth of agricultural fields and forests.Figures 2, 3 and 4 use Sentinel-2 optical imagery to show a few examples of the massive cloud and smoke cover that is prominent in this area.
The U.S. Geological Survey (USGS) is monitoring this eruption and the resulting fissures and lava flows to ensure the public is safe and up to date. Since May 18th, the USGS has been releasing updated maps of the affected areas roughly every 24 hours. They create these detailed maps via thermal imaging from a helicopter. This is an effective method of remote sensing for this event due to the clouds and smoke and because the eruption area is quite limited in size and is located very close to residential communities. The USGS cannot utilize Landsat because the satellite revisit time is too long and does not have all-weather capabilities. For large scale remote areas, PlanetWatchers takes advantage of the all-weather capabilities of radar to monitor frequently and reliably to generate insights using our advanced analytics.
Figure 5 shows a comparison of USGS results derived from helicopter thermal imaging and PlanetWatchers analysis using Sentinel-1 radar satellite imagery.
PlanetWatchers analyzed Sentinel-1 SAR imagery of the Kilauea eruption on a 12-day repeat cycle as follows:
- May 2nd (the day before the initial earthquake and eruption)
- May 14th (four days before the initial release of the USGS daily maps)
- May 26th
PlanetWatchers found that during the period of May 2nd to May 14th, there were 42 acres of Kilauea that were covered by newly formed lava flows, which damaged or destroyed around nine homes and structures in Leilani Estates. During the period of May 14th to May 26th, there were 3,354 acres of Kilauea covered by lava flows, which damaged or destroyed around 162 homes and structures in Leilani Estates and the surrounding areas. This 3,354 acre area included 540 acres of farmland and 555 acres of developed land (e.g. roads and housing). This analysis shows that the second period of observation was very active and supports the current observations that the event is not over.
Our analytics platform estimates that between May 14th and May 26th, there were 2,259 acres of forested areas being covered by newly formed lava flows destroying roughly 250,000 trees. The native trees in this area are mostly banana, coconut gum and koa trees. It will take many years for these areas to start recovering and some may never be the same.
We will continue to monitor this area and share our findings. We hope this information provides valuable insights and that everyone in the area remains safe.
You can contact PlanetWatchers at email@example.com or follow us on Twitter at @planetwatchers for more information on changes in the Kilauea eruption. We will continue monitoring this event and would be happy to assist you with your monitoring needs.