Approaches to mitigate atmosphere artifacts in SAR interferograms:

GPS vs. WRF modelP. Mateus, G. Nico, R. Tomé, J. Catalão, P. Miranda

Lisbon University Instituto Dom Luiz (IDL)

Lisbon, Portugal

Outline

GPS data analysis

Weather Research & Forecasting (WRF) Model

InSAR data

Results (GPS vs WRF, GPS and WRF vs InSAR)

Conclusions

GPS network over Lisbon region

Total Zenith Delay (GPS)

The GPS network consists in 12 stations. All stations were installed between 1997 and the early 2009.

The Total Zenith Delay (TZD) was estimated by means of GAMIT software (v10.34).

The wet zenith delay for each station are modeled by a piecewise-linearfunction over the span of the observations.

The IGS precise orbits were used in solution and the cut-off angle chosen for the GPS data was 20°.

For each site the tropospheric TZD was determined at each GPS site every 15 min.

Weather Research & Forecasting

1×1 kmIV

3×3 kmIII

9×9 kmII50 layers

27×27 kmI

VerticalHorizontalGrid

Resolution

1×1 kmIV

3×3 kmIII

9×9 kmII50 layers

27×27 kmI

VerticalHorizontalGrid

Resolution

Weather Research and Forecasting Model (WRF-ARW), version 3

Initial and time-dependent boundary conditions derived from the European Centre for the Medium range Weather Forecasting (ECMWF) high resolutionanalysis.

Setup:

1. Four two-way nested domain at 27x27, 9x9, 3x3 and 1x1 km horizontal gridresolutions.

2. Top atmosphere located at 10 hPa level

3. 50 vertical levels

4. Lowest model layer 40m thick.

Simulations referred to the 12 hours preceding each SAR image acquisition.

Atmospheric refractivityThe atmospheric refractivity N is computed as

⎟⎠⎞

⎜⎝⎛ +⋅+⋅=+=

Tkk

Te

TPkNNN wetdry

3'21

1253

1'2

11

hPaK1075.3hPaK3.23hPaK6.77

−−

−

−

⋅⋅=⋅=⋅=

kkk

The following values were used for the constants

The atmospheric delay at the time acquisition of the SAR data is

( )∫ ⋅+⋅=Δ+Δ=Δ− h

hydwethydwetatm dhNNRRR0

6

cos10

ϑ

GPS vs WRF - Time (1)

GPS (blue) TZD vs WRF (red) ZTD at April 12, 2009.

GPS vs WRF – Space (1)

Station Hydrostatic Wet

Mean RMS Max Mean RMS Max.

MAFR 2.2 2.4 3.7 1.8 6.4 12.1

GRID 0.0 1.2 1.8 6.5 13.3 25.4

ALCO 2.1 2.3 3.8 4.3 9.8 26.3

FCUL -0.5 1.0 1.3 -0.6 5.9 9.5

PARC 13.8 13.8 15.1 -2.7 6.7 13.9

IGP0 -3.0 3.2 -1.2 -1.9 6.4 7.8

VNOV -0.9 1.4 1.4 9.1 12.3 25.8

PALM -10.0 10.1 -8.1 1.8 5.4 9.8

ARRA -34.9 34.9 -32.7 -2.0 4.6 5.4

CRAI 7.4 7.5 9.4 0.9 5.9 16.4

SMAR 4.1 4.3 6.6 10.9 16.4 34.1

CASC 3.6 3.8 5.0 -4.1 8.5 5.4

Zenith delay differences (mm) Dry and wet components (April 12, 2009)

GPS vs WRF – Space Variation (1)

Difference of TZD (May 17, 2009 – April 12, 2009)DRY WET

WRF

GPSm

m

mm

mm

mm

GPS vs WRF – Space Variation (2)

Station Hydrostatic Wet

Variation Variation

WRF GPS Diff. WRF GPS Diff.

MAFR -3.1 -5.3 2.2 49.2 40.5 8.7

GRID -3.2 -2.6 -0.6 68.7 51.8 16.9

ALCO 3.5 -3.6 0.1 69.1 8.8 60.3

FCUL -3.6 -3.1 -0.5 51.4 41.0 10.4

PARC -3.6 -3.1 0.5 48.3 51.5 -3.2

IGP0 -3.4 -3.5 -0.1 21.4 22.7 28.7

VNOV -2.5 -1.9 -0.6 75.4 70.3 5.1

PALM -3.1 -2.7 0.4 64.0 63.0 1.0

ARRA -2.5 -3.3 0.8 67.5 41.6 25.9

CRAI -4.0 -1.6 -2.4 45.0 55.3 -10.3

SMAR -2.5 -2.4 -0.1 67.9 57.4 10.6

CASC -3.4 -2.5 -0.9 48.4 65.9 17.5

Zenith delay differences (mm) Dry and wet components (May 17, 2009 - April 12, 2009)

SAR Interferometry

A total amount of 25 SAR interferograms with a 35-day temporal baseline were processed using the DORIS (Delft object-oriented radar interferometric software) software.

The SAR data was acquired by ENVISAT/ASAR over Lisbon region during the period from 2003 to 2005 and from 2008 to 2009 along and ascending orbit.

Terrain deformations related to known geological phenomena in Lisbon area are negligible at this time scale of 35 days.

These interferograms give an estimate of the variation of the total atmospheric delay occurred between the acquisition DATM of the two SAR images

ATMATM Dλπϕ 4

=

GPS and WRF vs InSAR (1)

ENVISAT/ASAR interferogram (May 17, 2009 – April 12, 2009)Spatially averaged to 160x160m to reduce phase noise

rad

GPS and WRF vs InSAR (2)

Synthetic “interferogram” from WRF TZD (May 17, 2009 – April 12, 2009)

rad

ATMATM Dλπϕ 4

=

GPS and WRF vs InSAR (3)

Synthetic “interferogram” from GPS TZD (May 17, 2009 – April 12, 2009)

rad

ATMATM Dλπϕ 4

=

Conclusions

GPS measurements and WRF forecasts of atmospheric phase delay are in agreement with phase patterns in SAR interferograms with 35-day baseline.

When available, GPS data are more effective to capture local variations of atmospheric delay

In regions without GPS data, WRF information is a promising tool to mitigate atmospheric delay effectsin SAR interferograms