Improving volcano monitoring through rapid, automatic InSAR processing

Abstract:

Over the last two decades, InSAR has become a proven method for quantifying surface deformation resulting from magma movements. Historically, data availability was rather limited, which combined with long processing times meant that results often were not available until weeks to months after the event. The successful launch of the Sentinel-1a and ALOS-2 satellites will increase the over the last years already much improved availability of SAR images over volcanically active areas. This higher data volume potentially allows us to monitor volcanoes in near-realtime, but only if we can reduce the long processing time of current time series techniques. Here we present results from a rapid algorithm, developed and tested as part of the FUTUREVOLC project in Iceland, able to automatically produce high signal-to-noise results within hours.

Our algorithm identifies for each pixel neighboring pixels that behave statistically similar through the time series, known as Statistically Homogeneous Pixels (SHP). We use these SHP to calculate the complex coherence of the pixel in question in every interferogram. This is different from other time series techniques, which typically select a single set of pixels in all interferograms. We work under the assumption that the SHP do not change rapidly over time, which allows us to estimate them on an a priori set of interferograms, and store them for later use on new acquisitions. This greatly reduces the processing time for new images.

The algorithm allows us to produce interferograms with respect to a new image within 10 minutes. Using 8 processor cores in parallel, we can then estimate the coherence for each pixel in the 15 closest pairs in approximately 45 minutes, assuming a 5000 by 5000 scene. Phase unwrapping and estimating the deformation since the last acquisition is achieved within a further 45 minutes. Our system will be integrated with existing near-realtime monitoring systems, like GNSS, already in place in Iceland. The integrated system will improve our ability to classify movements as resulting from magma movements or not, and allow models to better constrain depth and volume during ongoing events. Providing this kind of information in a timely manner to civil authorities, whom we are working with as part of FUTUREVOLC, is of great help in steering their response in times of crisis.