Monitoring Cascade Volcanoes using Multi-temporal InSAR
Parker, Amy1; Biggs, Juliet1; Lu, Zhong2; Wright, Tim3
1University of Bristol, UNITED KINGDOM; 2USGS Cascades Volcano Observatory, UNITED STATES; 3University of Leeds, UNITED KINGDOM
Measurements of ground deformation have proven to be a key component of successful volcano monitoring networks. Interferometric synthetic aperture radar (InSAR) produces ground deformation measurements using successive satellite images and is therefore becoming increasingly important in regions where other monitoring equipment is not available or is difficult to deploy. A large InSAR dataset has been acquired for the Cascade volcanoes in the western USA, but its application has been limited by snow cover, vegetation, steep topography and atmospheric artefacts, which cause incoherence and compromise the accuracy of measurements. We are working to overcome these limitations by using multi-temporal InSAR techniques to investigate recent ground deformation at Cascade volcanoes. Here we focus on Medicine Lake Volcano (MLV) located at the southern end of the Cascade chain in northern California.
Leveling and GPS surveys throughout the 20th century have revealed 8-11 mm/yr of caldera-wide subsidence at MLV. However, the low spatial and temporal resolution of measurements has limited interpretations of the deformation signal. InSAR studies have added little to the analysis of ground deformation due to poor coherence across the caldera. To demonstrate how InSAR may be used more successfully in this region, we apply three analysis techniques: stacking, π-RATE, and persistent scatterer InSAR (PSInSAR) to datasets from the ERS, ENVISAT and ALOS satellites.
We investigate the noise contributions to each dataset and design a stacking strategy that minimises the effects of atmospheric noise whilst maximising the number of interferograms included in the stack. We then apply π-RATE, an advanced stacking technique that uses a pixel-wise least squares inversion to calculate the deformation rate at each pixel. π-RATE has not previously been applied to volcanic data, but this pixel-wise approach is advantageous as it allows us to calculate deformation rates at pixels that are coherent in different numbers of interferograms. Finally, we use the StaMPS algorithm to produce rate maps of PS pixels across the caldera. We appraise each of the methods used and find that π-RATE is the most successful: it offers a greater improvement in coherence than PSInSAR and reduces the effects of atmospheric and orbital errors that remain in traditional stacking.
Results from all datasets are indicative of slow, steady subsidence of the edifice. We compare these results to those from past geodetic surveys and find a decrease in the rate of subsidence at MLV. Building upon the models of past studies, we use our results as inputs to inverse models of volume loss at depth. By coupling these analytical solutions to a thermal model, we investigate the possibility that present day subsidence at MLV is due to cooling and crystallisation of a sill.
Results from ENVISAT showing subsidence across Medicine Lake Volcano. Left to right are results using: stacking, π-rate and StaMPS. The caldera is shown by the black ellipse. The extent of Medicine Lake lavas and major surficial flows are also outlined in black.