(1) ISEIS, Chinese University of Hong Kong, NT, Shatin, Hong Kong

1

(1) (1) ISEIS, Chinese University of Hong Kong, NT, Shatin, Hong KongISEIS, Chinese University of Hong Kong, NT, Shatin, Hong Kong(2) (2) DEI, Politecnico di Milano, Milan, ItalyDEI, Politecnico di Milano, Milan, Italy(3) (3) DIEI, Università la Sapienza, Rome, ItalyDIEI, Università la Sapienza, Rome, Italy(4) (4) CETEMPS, University of L’Aquila, ItalyCETEMPS, University of L’Aquila, Italy(5) (5) DIIAR, Politecnico di Milano, Milan, ItalyDIIAR, Politecnico di Milano, Milan, Italy(6) (6) ESA,ESA, ESTEC ESTEC Noordwijk, The NetherlandsNoordwijk, The Netherlands

Vancouver, 28Vancouver, 28thth July 2011 July 2011

Mitigation of atmospheric delay in InSAR: the Mitigation of atmospheric delay in InSAR: the ESA METAWAVE projectESA METAWAVE project

Daniele Perissin Daniele Perissin (1)(1),,Fabio RoccaFabio Rocca （（ 22 ）） , Mauro Pierdicca, Mauro Pierdicca （（ 33 ）） , Emanuela , Emanuela

PichelliPichelli （（ 44 ）） , Domenico Cimini, Domenico Cimini （（ 44 ）） , Giovanna Venuti, Giovanna Venuti （（ 55 ）） , , Bjorn RommenBjorn Rommen （（ 66 ））

2

Mitigation of atmospheric delay in InSAR: Mitigation of atmospheric delay in InSAR: the ESA METAWAVE projectthe ESA METAWAVE project

Table of Contents

1. decomposition of atmospheric signal

2. connection between APS and IWV

3. experiments and performances (GPS, Meris, NWP)

4. PS precision assessment

5. Conclusions

3

Decomposition of atmospheric signal Decomposition of atmospheric signal (from InSAR point of view)(from InSAR point of view)

Atmospheric components…

- stratification (correlated with topography)- turbulence- spatially linear component

- stationary part- variational part

…which can be divided into

Points to keep in mind

- The APS contains only the variational part of the atmosphere

- The APS gathers spatially correlated noise components (also orbital artifacts) spatially linear trends must be removed!

4

Connection between APS and IWVConnection between APS and IWV

- in APS domain: differential way (multi-master)

2 different strategies for comparison/correction of APS

- in IWV domain: stationary term + spatial linear terms must be provided by external data

To be able to extract Water Vapor from the APS

- in APS domain: pseudo-absolute way (single Master) the Master delay is estimated and removed, so the atmospheric delay can be compared day by day

5

Experiments and performancesExperiments and performances

Table of Contents

NWP data in Rome vs APS

Meris data in Rome vs APS

GPS data in Como vs APS

6

InSAR vs NWP

T351, desce, 10UTC

30 images

10 std IWV maps

T172, asce, 21UTC

41 images

20 std IWV maps

Rome Envisat datasets

7

InSAR vs NWP NWP domain and topography

8

InSAR vs NWP “differential” comparison

IWV APS APS-IWV

Scatter plotAPS vs IWV

APS-IWV APS-IWVvs SRTM

9

InSAR vs NWP “differential” comparison

IWV std APS std APS-IWV std

IWV stratif. APS stratif. APS-IWV stratif.

10

Estimated Master APSAverage IWV

qk iii ii kkk 0

InSAR vs NWP “pseudo-absolute” comparison

11

IWV APS APS-IWV

Scatter plotAPS vs IWV

APS-IWV APS-IWVvs SRTM

InSAR vs NWP “pseudo-absolute” comparison

12

Scatter plot disp: 0.7 mm/km

vs InSAR disp:1.3 mm/km

InSAR vs NWP Variational stratification

MM5 can help reducing thestratification component

13

IWV APS APS-IWV

Scatter plotAPS vs IWV

APS-IWV APS-IWVvs SRTM

InSAR vs NWP Comparison of turbulent terms

14

19940131 19940203 19940212 19940215 19940227 19940305 19940308

APS 0.51 0.64 0.33 0.26 0.45 0.85 0.33APS-IWV 0.62 0.58 0.5 0.42 0.64 0.8 0.51

Spatial cross-correlation

Standard deviations [mm]

InSAR vs NWP Comparison of turbulent terms

15

19940131 19940203 19940212 19940215 19940227 19940305 19940308

KS 0.12 0.24 0.26 0.23 0.11 0.27 0.1

Cumulative distribution functions

Test statistics

Kolmogorov-Smirnov test

InSAR vs NWP Comparison of turbulent terms

MM5 turbulent termhas very low correlation

with the APS one

16

InSAR vs NWP IWV evolution in time

17

Standard deviation vs delay

InSAR vs NWP NWP-APS synchronization

NO significant improvement

18The average map has been subtracted

00

06

11

InSAR vs NWP Impact of starting time

06

11

3 October 08, residuals after subtraction of stationary term

Strong random componentin MM5 simulations!!

19

InSAR vs MERIS

T351, desce, 10UTC

30 images

26 Meris image

T172, asce, 21UTC

41 images

The Rome dataset

No use for night passes

Meris can be used onlyfor day time passes

20

Examples in the Rome dataset

Rome T351, morning passes

40% of loss

InSAR vs MERIS

Meris needs clear sky conditions

21

InSAR vs MERIS Spectral analysis

Meris has spectral contentcloser to the APS one

22

Scatter plots and correlation

Meris IWV [mm]

InSAR IWV [mm]

InSAR vs MERIS

In our experiment theMeris success rate is quite low

23

InSAR vs GPS The Como test-site

480 descending, 10am (28 images) 487 ascending, 9pm (38 images)

5 GPS stations,5 overlapping days

24

descending track

Different ways for estimating the GPS stationary term

InSAR vs GPS

ascending track

25

Ascending Descending

GPS std

InSAR std

diff std

corr coeff

GPS std

InSAR std

diff std

corr coeff

3.69 3.373.43

0.53 3.43 1.722.85

0.56  

2.34 4.274.27

0.28 2.73 5.263.52

0.79  

3.61 5.001.98

0.95 0.63 3.353.03

0.59  

2.94 3.361.06

0.95 5.36 6.072.79

0.89  

2.18 1.151.27

0.89 1.67 1.832.05

0.32  

Correlation and deviation reductionInSAR vs GPS

GPS has a 50% success rate

26

D =10004.6mm*sqrt(0.036)=4.6*0.19=0.9mm.

1000036.0

Dq

InSAR alone

1mm path delay error if we interpolate PS’s

distributed along a circle with 10km diameter

APS interpolation in presence of PS’s

By F. Rocca

27

Experiments and performancesExperiments and performances

Conclusions at this time

Meris: spectral content closer to APShowever usable only in clear skystratification not very robust

NWP: powerful tools in space and timestrong random componentuseful for long spatial wavelengths and stratification

GPS: highest accuracyreliability depends on density of ground stations

PS: where PS’s are present, no better way to estimate the APS