460 likes | 761 Views
Noise Reduction of InSAR Time-series Analysis and Application to Kilauea Volcano. Yo Fukushima, DPRI, Kyoto University. …. InSAR Time-Series Analysis. Time. …. Time. Small BAseline Subset (SBAS) analysis (Berardino et al., 2002). Use SVD to obtain the min. norm for velocity. Time.
E N D
Noise Reduction of InSAR Time-series Analysis and Application to Kilauea Volcano • Yo Fukushima, DPRI, Kyoto University
… InSAR Time-Series Analysis Time … Time
Small BAseline Subset (SBAS) analysis (Berardino et al., 2002) Use SVD to obtain the min. norm for velocity Time
Typical (?) interferogram Atmospheric noise correlated with altitude Long-wavelength ramp due to orbital inaccuracy etc. + Offset + artifact due to DEM error Deformation We want to separate them.
Observation Equation UNWRAPPED phase at the k-th pixel of the i-th ifg: Phase (Obs.) Correlated w/altitude LOS displ. Offset Ramp (bilinear) DEM error contribution i k
Observation Equation UNWRAPPED phase at the k-th pixel of the i-th ifg: Phase (Obs.) Correlated w/altitude LOS displ. Offset Ramp (bilinear) DEM error contribution
Observation Equation UNWRAPPED phase at the k-th pixel of the i-th ifg: Phase (Obs.) Correlated w/altitude LOS displ. Offset Ramp (bilinear) DEM error contribution
Observation Equation UNWRAPPED phase at the k-th pixel of the i-th ifg: Phase (Obs.) Correlated w/altitude LOS displ. Offset Ramp (bilinear) DEM error contribution
Observation Equation Bperp UNWRAPPED phase at the k-th pixel of the i-th ifg: Phase (Obs.) Correlated w/altitude LOS displ. Offset Ramp (bilinear) DEM error contribution Range Inc. angle
Observation Equation UNWRAPPED phase at the k-th pixel of the i-th ifg: What to estimate: ・v:Velocity time-series at every px ・a, b, c, f:coefficients for every ifg ・δh:DEM error for every px
This can be solved in principle by a linear inversion Problem 1 The matrix is too large Problem 2 Rank deficiency Need constraints or SVD. How can we weight the model params. of different dimensions? -> Maximum likelihood Two-step approach: Inversion with down-sampled data. px-by-px inversion using corrected ifgs
Baseline M = 12 N = 21
Utrue (all color scale [-5 5] radian)
Dtrue Note: Range increase is positive (red) throughout my talk
Dtrue Dnoise Dinput=Dtrue+Dnoise D corrected
Dtrue Dnoise Dnoise estimated (except DEM correction) D=Dtrue+Dnoise
Utrue Uestimated DEM error DEM error estimated
Application to Kilauea volcano Mauna Loa Kilauea
ALOS/PALSAR, θ~10° 3 24 acquisitions 37 ifgs
Ascending Descending
June 2007 event Pre-event Co-event Post-event
Pre-event Uplift
Co-event Subsidence Dike intrusion
Post-event Subsidence
Summary • I proposed a noise reduction method in SBAS InSAR time-series approach • Applied to Kilauea data. • with respect to the 2007 father’s day event, we found: • - Pre-event: Uplift SW of the summit • - Co-event:Uplift due to dike intrusion and • subsidence at the summit • - Post-event:Subsidence at the summit & PuuOo • Acknowledgement: The used PALSAR data are shared by the PIXEL consortium and provided from JAXA. The GPS data The data are provided from JAXA under cooperative research contract with the GSI and with the Earthquake Research. I thank the Hawaiian Volcano Observatory for the GPS data processing and sharing.