School of Earth and Environment

Transcription

1 UNAVCO Short Coarse on TRAIN Toolbox for Reducing Atmospheric InSAR noise Troposphere correction methods for InSAR using TRAIN By David 2016 California Institute of Technology. Government sponsorship acknowledged.

2 Why a tropospheric correction for InSAR? [Bekaert et al., JGR 2015b] Tectonic Troposphere To extract smaller deformation signals Tropospheric delays can reach up to 15 cm With the tropospheric delay a superposition of - Short wavelength turbulent component - Topography correlated component - Long wavelength component cm Over 9 months 1 interferogram (ti tj)

3 Tropospheric correction methods in TRAIN Test areas for this demo Spectrometers: El Hierro MERIS (Envisat) MODIS (AQUA and TERRA) Italy Weather model ERA-I, MERRA, MERRA2 WRF Mexico Phase-based Uniform correction Non-uniform correction

4 Tropospheric delays from refractivity The troposphere is mainly described by isolines of - Temperature (T) - Pressure (P) - Water vapor (e) isolines And can be computed by integrating the refractivity N with height [e.g., Davies et al., 1985 Radio Science; Hanssen, 2001] constants

5 Tropospheric delays from weather models Use pressure, temperature and relative humidity Interpolate in time and compute the interferometric delay Available models e.g.: Used in this demo 70 km (6hrs) 32 km (3hrs) Locally run at 7km

6 Tropospheric delays from spectrometers Observations of Precipitable Water Vapor: Interpolate in time and compute the interferometric delay Available sources: MERIS and MODIS Used in this demo [e.g., Thayer, 1974 Radio Science; Li et al., 2003 JGR]

7 Tropospheric delays from a linear relationship Wet and hydrostatic delay combined Interferogram Linear est Assumes a uniform troposphere Tropo GPS InSAR and GPS data property of IGN isolines

8 A spatially varying troposphere [Bekaert et al., JGR 2015a] Wet and hydrostatic delay combined Topography est: Spectrometer & Linear However: Spatial variation of troposphere isolines

9 Tropospheric delays from a power-law relationship Allowing for spatial variation [Bekaert et al., JGR 2015a] Wet and hydrostatic delay combined h 0 With h 0 the lowest height at which the relative tropospheric delays ~ km from balloon sounding With α a power-law describing the decay of the tropospheric delay from balloon sounding data Estimate in a band-insensitive to deformation Sounding data provided by the University of Wyoming

10 α Δφ band K spatial ( h0 h ) Power-law example School of Earth and Environment rad 9.97 Interferogram (Δɸ) band [Y. Lin et al., 2010, G3] for a linear approach Band filtered: phase (Δɸband) & topography (h0-h)αband

11 α Δφ band K spatial ( h0 h ) Power-law example School of Earth and Environment band [Y. Lin et al., 2010, G3] for a linear approach Band filtered: phase (Δɸband) & topography (h0-h)αband

12 α Δφ band K spatial ( h0 h ) Power-law example School of Earth and Environment band [Y. Lin et al., 2010, G3] for a linear approach Band filtered: phase (Δɸband) & topography (h0-h)αband Anti-correlated! For each window: estimate Kspatial

13 α Δφ band K spatial ( h0 h ) Power-law example School of Earth and Environment band [Y. Lin et al., 2010, G3] for a linear approach Band filtered: phase (Δɸband) & topography (h0-h)αband For each window: estimate Kspatial

14 α Δφ band K spatial ( h0 h ) Power-law example School of Earth and Environment -1.1e-6 rad/mα 9.8e-5 Original phase (K (Δɸ) Tropo variability spatial) band [Bekaert et al., JGR 2015a] Band filtered: phase (Δɸband) & topography (h0-h)αband

15 α Δφ tropo = K spatial ( h0 h ) Power-law example School of Earth and Environment -1.1e-6 rad/mα 9.8e-5 Original phase (K (Δɸ) Tropo variability spatial) [Bekaert et al., JGR 2015a] 4.7e4 1/mα 2.4e rad 9.97 α α Topography (h0-h)(δɸ Band filtered: phase topography est (h-h(δɸ tropo) band) & Power-law 0) band

16 Power-law example School of Earth and Environment rad Allowing for spatial variation rad rad 9.97 Original phase (Δɸ) Power-law est (Δɸ tropo ) Spectrometer est (Δɸ tropo ) [Bekaert et al., JGR 2015a]

17 School Signal of Earth and Environment Error [Bekaert et al., RSE 2015] Mexico Signal Error of MERIS Accuracy [Li et al, 2009] =

18 Interferogram example Mexico [Bekaert et al., RSE 2015]

19 School of Earth and Environment Mexico techniques compared: profile AA MERIS Linear ERA-I Power-law

20 Mexico techniques compared: profile AA MERIS ERA-I Linear Power-law

21 Mexico techniques compared: profile AA MERIS ERA-I Linear Power-law

22 School Signal of Earth and Environment Error [Bekaert et al., RSE 2015] Italy Signal Error of MERIS Accuracy [Li et al, 2009] =

23 Interferogram example Italy [Bekaert et al., RSE 2015] Italy

24 School Signal of Earth and Environment Error [Bekaert et al., RSE 2015] El Hierro Signal Error of MERIS Accuracy [Li et al, 2009] =

25 Interferogram example El Hierro [Bekaert et al., RSE 2015] El Hierro

26 Validation ERA-I wet delay with MERIS BAD GOOD Calculate MERIS and ERA-I ZWD at ~1.8 km resolution ~100 MERIS/ERA-I comparisons for each box ( ) Calculate RMS misfit between the two estimates of ZWD [Walters et al., in prep.]

27 Major variations in ERA-I quality globally Strong latitude effect Mean RMS reduction is 70-90% but still several cm RMS misfit Misfit will be larger for InSAR ( 2 if σ slv = σ mst ) [Walters et al., in prep.]

28 Conclusions Different methods correct for different components of the troposphere All methods have there own limitations with varying accuracy Correction methods perform worse with increasing cloud cover No method is exclusive the best in reducing tropospheric delays Assuming e.g. not using MERIS for Envisat Rather than opting for a single method, jointly invert for tropospheric properties such different technique constrain each other This will require relative weighting, quality measure of methods Toolbox for Reducing Atmopsheric InSAR Noise = TRAIN Download from davidbekaert.com/#links 2016 California Institute of Technology. Government sponsorship acknowledged.