InSAR water vapour correction models: GPS, MODIS, MERIS ...earth.esa.int/fringe2005/proceedings/presentations/109_li.pdf · InSAR water vapour correction models: GPS, MODIS, MERIS

InSAR water vapour correction models: GPS, MODIS, MERIS and InSAR integration

Zhenhong Li(1), Eric J. Fielding(2), Paul Cross(1), and Jan-Peter Muller(1)

(1): Department of Geomatic Engineering, University College London, UK

(2): JPL, California Institute of Technology, Pasadena, California, USA

Sponsored by:

ContentsIntroduction to water vapour products

Water vapour effects on InSAR measurements

InSAR water vapour correction models

GPS Topography-dependent Turbulence Model (GTTM)

MODIS-based water vapour correction model

MERIS water vapour correction model

MERIS+MODIS water vapour correction model

Discussion and Conclusions

Water vapour products used in this study

GPS((GGlobal lobal PPositioning ositioning SSystem)ystem)

MODIS ((ModModerate Resolution erate Resolution IImaging maging SSpectroradiometerpectroradiometer))

MERIS ((MEMEdiumdium RResolution esolution IImaging maging SSpectrometer)pectrometer)

PWV retrievals rely on channel ratio techniques

MODIS/MERIS Channel Positions Related to PWV

MODIS: 2 non-absorbing 3 absorbing

MERIS:1 non-absorbing 1 absorbing

(Figure adapted from Gaoand Kaufman [1998])

Comparisons of GPS, MODIS and MERIS PWV

GPS, MODIS and MERIS PWV products are complementary!GPS, MODIS and MERIS PWV products are complementary!

Water vapour effects on InSAR measurements--Example 1: Southern California (29 Jan 2005 – 09 Apr 2005)

• 1 mm of PWV => ~6.2 mm of ZPD• ZPDDM: Zenith Path Delay (ZPD) Difference Map

ZPDDM• Max:

7.8 cm• Min:

-5.2 cm• Mean:

-0.9 cm• StdDev:

1.0 cm

Water vapour effects on InSAR measurements--Example 2: Southern California (07 Aug 2004 – 09 Apr 2005)

ZPDDM: Max: 2.7 cm; Min: -12.8cm; Mean: -2.9 cm; StdDev: 2.9 cm

Water vapour effects on InSAR measurements--Example 3: Southern California (11 Sep 2004 – 16 Oct 2004)

ZPDDM: Max: 5.8 cm ; Min: -16.5cm; Mean: -9.0 cm; StdDev: 3.5 cm

GPS Topography-dependent Turbulence Model (GTTM)

Underlying assumptions of GTTM:

Water vapour variations conform temporally and spatially to a statistical turbulent model;

Water vapour distributions are correlated with topography.

Key Finding: Interpolation methods should be applied to Zenith Path Delay differences (NOT ZPD themselves) (Li et al., 2005, JGR, in press)

Case study 1: GTTM and IDW correction

Original interferogram: 1996/01/10-1996/01/11

After GTTM Correction

After IDW Correction

(Li et al., 2005, JGR, in press)Unwrapped interferogram

2D Spatial Structure Function

Case study 2: GTTM correction(11 Sep 2004 – 16 Oct 2004)

(Li et al., 2005, ION)

Left: Before correction1.94 rad

Right: After Correction1.32 rad

Comparison of range changes:GPS vs. InSAR0.8cm ⇒ 0.6cm

Case study 2: Water vapour variation + Clouds (11 Sep 2004 – 16 Oct 2004)

• Strong water vapour variation (11/09/2004)• Presence of clouds (16/10/2004)

• This highlights the major advantage of the GTTM (over MODIS and MERIS): It can reduce water vapour effects even under cloudy conditions!

MODIS-based (or GPS/MODIS integrated) water vapour correction model

Basic principles:

There is a scale uncertainty in MODIS near-IR water vapour products (Li et al., 2003, JGR)

Only one continuous GPS station is required to calibrate MODIS scale uncertainty within a 2,030 km × 1,354 km MODIS scene

GPS and MODIS data can be integrated to provide regional water vapour fields with a spatial resolution of 1 km × 1 km

~60 minute time difference between ENVISAT and MODIS

Li, Z, J-P Muller, P Cross and E Fielding, InSAR atmospheric correction: GPS, MODIS and InSAR integration. JGR - Solid Earth, 110, B03410, 2005

MODIS case study: 2000/09/02 – 2000/12/16

Left: Before correction2.48 rad

Right: After Correction1.47 rad

Comparison of range changes:GPS vs. InSAR1.1cm ⇒ 0.5cm

(a) Black rectangles: PWV effects; Black oval: uplift(b) Dashed rectangle and oval: uncertainties due to clouds

(Li et al., 2005, JGR)

MERIS water vapour correction model

MERIS data can be acquired at the same time as ASAR data (time differences between MODIS and SAR data: ~1 hour)

MERIS has better spatial resolution, up to 300 m against 1km for MODIS

MERIS near IR water vapour product agrees more closely with GPS than MODIS (Li et al., 2005, IJRS)

A paper on MERIS water vapour correction model is under review for Geophysical Research Letters

MERIS case study: 2004/08/07 – 2005/01/29

Left: Before correction2.38 rad

Right: After Correction1.49 rad

Comparison of range changes:GPS vs. InSAR0.9cm ⇒ 0.5cm

(1) black solid triangles represent GPS stations w/o changes after correction

(2) white squares with black borders imply improvement after correction

(3) red solid circles indicate deterioration after correction

(Li et al., 2005, under review for GRL)

Comparison between MERIS and GTTM(2004/08/07 – 2005/01/29)

0.9 cmOriginal

0.7 cm0.5 cmInSAR vs. GPSGTTMMERIS

Comparison between MERIS and GTTM(2004/08/07 – 2005/04/09)

0.6 cmMERIS

0.7 cm0.8 cmOriginal

0.9cmInSAR vs. GPSGTTM BottomGTTM All

NB: The density of CGPS stations may be crucial under certain conditions!

Comparison between MERIS and MERIS-MODIS(2004/08/07 – 2005/04/09)

0.8 cmOriginal

0.6 cm0.6 cmInSAR vs. GPSMERIS-MODIS*MERIS

*: 2004/08/07(MERIS)- 2005/04/09(MODIS): Time diff: 65 min

Impact of time differences on MODIS-based water vapour correction models

Terra MODIS descending node: 10:30 am (local time)

ERS-2 descending node: 10:30 am (local time)

ENVISAT descending node: 10:00 am (local time)

MODIS-ERS time difference: <60 min (usually)

MODIS-ASAR time difference: ~60 min (normally)

Impact of time differences on MODIS-based correction model (2003/09/27)

• Time difference across scene: 65 min• Little water vapour variation

Impact of time differences on MODIS-based correction model (2004/08/07)

• Time difference across scene: 50 min• Strong local water vapour variation

NB: Stripes are due to radiometric calibration errors in MODIS sensors.

Impact of time differences on MODIS-based correction model (2005/01/29)

• Time difference across scene: 5 min• Little water vapour variation

Impact of time differences on MODIS-based correction model (2005/04/09)

• Time difference across scene: 65 min• Little water vapour variation

Global average frequency: ~25%

Good News from MODIS: Tibet: 38%SCIGN: 48% Iran: Even higher?!

Seasonal Frequencies of cloud free conditions

← Area: Eastern Tibet(Sep 2001 - Aug 2004)

← Highest: WinterLowest: Summer(Li et al., 2005, IJRS)

ConclusionsTHREE water vapour correction models (GTTM, MODIS & MERIS) have been developed & validated

It shows that the order of water vapour effects on interferograms can be reduced from ~10 mm to ~5 mm using water vapour correction model(s)

The impacts of time differences on MODIS-based models depend on water vapour distribution

Cloud free frequency is a major limitation for MODIS/MERIS models. BUT……

Current & future work

Water vapour field stacking + correction (Encouraging results achieved)

Water vapour correction + phase stacking(Ongoing)

Water vapour + Time series analysis (Ongoing)

Other water vapour products, e.g. infrared

Questions?