Lava Flow Forecasting and Remote Sensing During 2018 Kīlauea Eruption

Video Transcript
Download Video
Right-click and save to download

Detailed Description

Talk by Hannah Dietterich–Alaska Volcano Observatory geologist. Talk originally presented at the American Geophysical Union Fall Meeting 2020.

This talk includes information that was preliminarily shared with emergency responders during Kīlauea Volcano’s 2018 lower East Rift Zone eruption. The 2018 Kīlauea lower East Rift Zone eruption on the Island of Hawai‘i effused 0.9–1.5 cubic kilometers [0.2–0.4 cubic miles] DRE [dense rock equivalent] of lava, destroying infrastructure and homes across more than 30 square kilometers [11.5 square miles] of the lower Puna District. Over more than three months, lava erupted from numerous fissures and produced rapid and dramatic topographic changes; this had significant implications for tracking lava flow emplacement and assessing the ever-changing areas at risk from lava inundation as the eruption progressed. We integrated probabilistic lava flow modeling with remote sensing of topographic change and flow dynamics to provide timely data and forecasts to emergency managers. Flows were modeled from active vents and channel overflow locations over frequently updated topography using the DOWNFLOW model and similar codes based on steepest-descent paths, while approximating flow thickness and ponding. These tools produced flow routing forecasts, while flow advance forecasts were based on measured advance rates. Up-to-date elevation data was primarily derived through repeated surveys by small unoccupied aircraft systems, airborne syn-eruptive lidar and single-pass interferometry surveys, as well as daily mapping of flow extents and thicknesses using a variety of methods. Fast topographic data processing and rapid modeling allowed for flow forecasts to be issued promptly and with improved accuracy during eruption response. To retrospectively assess these efforts and the importance of updated topography, we compare real-time eruption forecasts with later flow mapping, as well as equivalent simulations over pre-eruptive topography. Our results demonstrate how the evolving lava flow field influenced later flow routing and highlight the value of repeat high-resolution topographic surveys for hazard response. Topographic monitoring of the eruption through time also captured the evolution in lava volume and morphology, providing a critical dataset for understanding lava flow dynamics.
 

Details

Date Taken:

Length: 00:13:30

Location Taken: HI, US

Transcript

Lava flow forecasting aided by remote sensing during the 2018 Kīlauea lower East Rift Zone eruption

Hannah Dietterich – USGS Alaska Volcano Observatory geologist

Talk originally presented at the American Geophysical Union Fall Meeting 2020

I’m Hannah Dietterich, and today I'll be talking about lava-flow monitoring and forecasting that was aided by remote sensing during the 2018 eruption of Kīlauea Volcano in Hawaii. This is work I've done with a number of USGS [U.S. Geological Survey] collaborators, but it's also greatly benefited from those who aided in eruption response.

Kīlauea Volcano is in East Hawaii Island [Island of Hawai‘i], and was until 2018 erupting for more than 35 years from the Pu‘u ‘Ō‘ō vent in the middle East Rift Zone of Kīlauea Volcano. More recently, there was an active summit lava lake for 10 years within the caldera of Kīlauea caldera. And then in 2018, magma propagated further down the rift and erupted in the lower East Rift Zone within a neighborhood called Leilani Estates.

The eruption, the effusive eruption, occurred in sort of three phases. The first phase, the first couple of weeks, were characterized by the opening of many, many small fissures. You can see those here in this thermal map overlaid on a satellite image of Leilani Estates. And these were all very, very small short-lived fissure eruptions producing very small pāhoehoe lava flows for the most part but starting in mid-May, the effusion rates increased quite dramatically. The flows were larger, they were longer-lived, and the magma that was erupting was hotter and lower viscosity. So these flows were able to travel much, much further and influence areas downhill from the neighborhood of Leilani Estates. And then in late May, activity really focused just on one fissure—fissure number eight—and produced a very long, long-lived channel system that persisted for more than two months until the end of the eruption in August.

So, what you can see from the chronology is some of the challenges in monitoring and forecasting this eruption. It was characterized by many, many vents; they were changing all the time. That eruption occurred over a seven kilometer [4 mile] stretch of the rift zone from the most up-rift to most down-rift fissures and fissures would erupt, stop erupting, and then reactivate. Also, overflows and levee failures within the flows produced flow branching so there were many sort of active sources to keep track of over the course of the eruption. As you can see, there were also evolving effusion rates, more than two orders of magnitude changes in effusion rates, as well as rheology that produced a really wide range of erupted behavior. And so to forecast flows—fundamentally lava flows are gravity currents, they flow downhill and if we have a good sense of the topography, we can sort of figure out where flows are most likely to go, the path they're most likely to take. But the complex dynamics of these flows, those changing effusion rates, the evolution from very small pāhoehoe lava flows to very large channelized ‘a‘ā flows, really limits our ability to actually forecast advance rates during this eruption.

So traditionally, lava-flow forecasting in Hawaii has been done using lines of steepest descent. These are the blue lines on this map, and they are just the centers of drainages, so equivalent to a watershed. These are the drainages that lava flows might follow when they follow the terrain. But of course, they don't incorporate any sort of uncertainty in the terrain, and they're somewhat simplified and don't start at an eruption source like an active vent, say.

So in this eruption, we wanted to have a tool to forecast from specific active new vents, active overflows or flow fronts, but still incorporating the terrain so we employed the downflow model from Favalli et al. at INGV Pisa [The National Institute of Geophysics and Volcanology in Pisa, Italy], which allows you to simulate many, many steepest-descent lines from a given location with stochastic perturbation of the topography that allows the flows to spread a little bit to overcome small topographic obstacles and therefore behave a little bit more like a lava flow. And we can display the results probabilistically, in terms of how frequently a given area forecasts lava flow paths going through it. So, areas that more of the flow paths traverse are more likely to be the route the flow takes. And so, the results look like this, where the main drainage is of course highlighted in both the downflow model and the steepest descent lines shown in blue. And if we compare where the flow went to where the flow was forecast to go, we can see that in general flows went where we thought they might. In this example, which is the initial forecast of the fissure 8 reactivated flow, we sort of had routes to the north and to the east and sort of two out of three of these were forecast. The one to the north, actually there were large ground cracks in this area that may have caused the flow to stall in that direction very locally.

We can see that the downflow model has some advantages over the steepest descent model, not just in that it’s originating at the lava source but also the flow paths can traverse areas between the steepest-descent lines. So, allowing the flow to overcome some topography allows it to enter multiple drainages as lava flows tend to do. And they can also show the degree to which the topography is confined or not, areas where the flow is more likely to spread or narrow so showing you a little bit about what flow width might look like as well. But we can see that there is also a flow path potential here to the south that goes straight over the flows that were already in place which are shown in dark gray.

And so that really highlights one of the biggest challenges we had during this eruption, which was as the eruption went on topography was changing more and more from what our pre-eruptive surface looked like. But that of course greatly influences where the flows are likely to go. So, if we had run the simulation over just the pre-eruptive topography, on the left, you can see a number of flow paths going to the south straight over where lava flows have already been emplaced. But we were updating flow extent maps as frequently as we could during the eruption at high resolution, sometimes twice a day. So, what we would often do during the eruption is take that most up-to-date flow extent map, add artificial thickness to the terrain and use that to inform our forecasts to make them more accurate, but then we were also trying to collect as much syn-eruptive topographic data as we could. So, this example on the far right uses a DEM [Digital Elevation Model] from single-pass InSAR [Interferometric Synthetic Aperture Radar] to actually quantify the new terrain where flows have been emplaced and run simulations over up-to-date topographic data. And you can see the effect that this has.

In order to collect up-to-date topographic data, we relied on a number of remote sensing techniques. The one we used very frequently was aerial surveys with unoccupied aircraft systems. These are small UAS, including hexacopters and fixed-wing aircraft. And these would fly photo survey flights that we could use structural motion processing to reconstruct the terrain, or things like orthophotos mapping the flow field. These have somewhat limited extent but could be flown quite frequently. So, getting updated topography as often as 45 minutes. We also were able to conduct two syn-eruptive LiDAR [Light Detection and Ranging] surveys, in particular one in June focused on the heavily densely-vegetated rain forest within Leilani Estates and really captured well a graben feature where the terrain changed by meters right above where the dike was emplaced and these new features were erupting and we were worried that this might impact where flows were going to be going within Leilani Estates.

We also used a single-pass InSAR flown on a jet. This is the NASA GLISTIN [Glacier and Ice Surface Topography Interferometer] instrument and work that Paul Lundgren lead. And this allowed collection of terrain, both from the summit of the volcano in the caldera, as well as throughout the whole lava flow field, capturing things like this series of flow thickness maps, seen on the right. So, a nice example of how updated topography was really critical in the eruption. This is an example from May 20—I've covered the date with my face—but the lava flows so far have been emplaced in the pink area and everything's gone to the south so far in the eruption in this area. All the flows have gone to the south, but with a new flow reported on the north side of this area. Now, a lot of vent and lava material has constructed and changed the typography here. The question was whether flows were going to eventually start going to the north instead. And so we wanted to collect updated terrain data in this area, this thickness map and hillshade from a digital surface model from a drone flight. Just of this region around the vent area, you can see that a lot of topography has been constructed in this area that might affect where the flow will go. So if we run simulations on the initial terrain, again the flow paths largely go to the south. But we know that there's now already lava there. And if we incorporate some of that, just this little area outlined in red, where a new DEM has been swapped in, we can see that there's now enough terrain that some of the flow paths are expected to go to the north instead and that that would be something to look for.

We can use this sequence of terrain data as well to characterize eruptive volume and the evolution of the effusion rate over the course of the eruption, which is also important for keeping track of evolving hazards and whether the eruption is waxing or waning. So we used again all these datasets, including post-eruptive multi-beam sonar, and are able to construct eruptive volume over time. You can see volume data for the entire flow field is somewhat sparse because the flow field was very large and beyond the extent of just drone mapping. And so, to supplement this we also used drones to record video of the channel near the vent over time, to record regular observations of velocities in the channel to track flow rates and thus effusion rates. So, the blue dots here show the DEM time-average discharge rates of the eruption, showing the initial low effusion rates that popped up to much higher effusion rates later in the eruption, and the drone data is shown in black. That shows again the sort of maximum peak flow of the eruption in mid-June, and then continued high rates until right before the eruption ended in August, including little details like post-caldera-collapse events causing surges in lava effusion rate further down the channel that were important for hazard assessments at the time as well. These were associated with many overflows. So, in conclusion, we can see that simple, on-demand, very simple lava-flow forecasting just based on terrain was able to offer up-to-date and accurate information to emergency managers. This was really enabled by a collection of repeat high-resolution topographic mapping. Without new topographic data, our forecasts would not have been as accurate. And what we can also see is that this data informs hazard response, but it also informs our understanding of lava-flow emplacement. The syn-eruptive remote sensing data offers a lot of insights into eruption dynamics. So we hope to learn much more about lava flow emplacement and improve our lava-flow forecast models using eruption data from this eruption. So, thank you.