Continental mass change from GRACE over 2002-2011 and its ... · CSR, release 04 (temporal...

24
GGHS 2012 Venice | Italy | 9–12 October 2012 Continental mass change from GRACE over 2002-2011 and its impact on sea level Oliver Baur 1 , Michel Kuhn 2 , Will Featherstone 2 1 Space Research Institute, Austrian Academy of Sciences, Graz 2 Western Australian Centre for Geodesy, Curtin University of Technology, Perth

Transcript of Continental mass change from GRACE over 2002-2011 and its ... · CSR, release 04 (temporal...

Page 1: Continental mass change from GRACE over 2002-2011 and its ... · CSR, release 04 (temporal resolution ≈1 month) SLR-derived geocenter coordinates (taken from CSR) replaced by results

GGHS 2012Venice | Italy | 9–12 October 2012

Continental mass changefrom GRACE over 2002-2011and its impact on sea level

Oliver Baur1, Michel Kuhn2, Will Featherstone2

1Space Research Institute, Austrian Academy of Sciences, Graz2Western Australian Centre for Geodesy, Curtin University of Technology, Perth

Page 2: Continental mass change from GRACE over 2002-2011 and its ... · CSR, release 04 (temporal resolution ≈1 month) SLR-derived geocenter coordinates (taken from CSR) replaced by results

Outline

Methods

Results

Conclusions

Page 3: Continental mass change from GRACE over 2002-2011 and its ... · CSR, release 04 (temporal resolution ≈1 month) SLR-derived geocenter coordinates (taken from CSR) replaced by results

Methods

Mass-change trends from GRACE

Period of investigationGravity fields“Manipulation”

Degree-1 termsc20 coefficients

Post-processingDe-correlationSpatial averaging

Inference of mass variationSurface mass-densitiesLeakage consideration

Signal separationGIA model

Time-series fit

May 2002 to April 2011 (9 integer years)CSR, release 04 (temporal resolution ≈1 month)

SLR-derived geocenter coordinates (taken from CSR)replaced by results from SLR (taken from CSR)

according to Swenson and Wahr (2006)Gaussian smoothing with a radius of 500 km

according to Wahr et al. (1998)according to Baur et al. (2009)

according to Paulson et al. (2007)linear trend function together with four sinusoids

Page 4: Continental mass change from GRACE over 2002-2011 and its ... · CSR, release 04 (temporal resolution ≈1 month) SLR-derived geocenter coordinates (taken from CSR) replaced by results

Methods

Mass-change trends from GRACE

Period of investigationGravity fields“Manipulation”

Degree-1 termsc20 coefficients

Post-processingDe-correlationSpatial averaging

Inference of mass variationSurface mass-densitiesLeakage consideration

Signal separationGIA model

Time-series fit

May 2002 to April 2011 (9 integer years)CSR, release 04 (temporal resolution ≈1 month)

SLR-derived geocenter coordinates (taken from CSR)replaced by results from SLR (taken from CSR)

according to Swenson and Wahr (2006)Gaussian smoothing with a radius of 500 km

according to Wahr et al. (1998)according to Baur et al. (2009)

according to Paulson et al. (2007)linear trend function together with four sinusoids

Page 5: Continental mass change from GRACE over 2002-2011 and its ... · CSR, release 04 (temporal resolution ≈1 month) SLR-derived geocenter coordinates (taken from CSR) replaced by results

Methods

Mass-change trends from GRACE

Period of investigationGravity fields“Manipulation”

Degree-1 termsc20 coefficients

Post-processingDe-correlationSpatial averaging

Inference of mass variationSurface mass-densitiesLeakage consideration

Signal separationGIA model

Time-series fit

May 2002 to April 2011 (9 integer years)CSR, release 04 (temporal resolution ≈1 month)

SLR-derived geocenter coordinates (taken from CSR)replaced by results from SLR (taken from CSR)

according to Swenson and Wahr (2006)Gaussian smoothing with a radius of 500 km

according to Wahr et al. (1998)according to Baur et al. (2009)

according to Paulson et al. (2007)linear trend function together with four sinusoids

Page 6: Continental mass change from GRACE over 2002-2011 and its ... · CSR, release 04 (temporal resolution ≈1 month) SLR-derived geocenter coordinates (taken from CSR) replaced by results

Degree-1 terms (geocenter motion)

Methods

Page 7: Continental mass change from GRACE over 2002-2011 and its ... · CSR, release 04 (temporal resolution ≈1 month) SLR-derived geocenter coordinates (taken from CSR) replaced by results

Degree-1 terms (geocenter motion)

Methods

Geocenter motion correction implies mass loss in the Northern Hemisphere and mass gain in the Southern Hemisphere

Geocenter motion estimates have large error bounds

Questionable reliability of secular trends in SLR-derived geocentercoordinates: two-track processing scheme

Page 8: Continental mass change from GRACE over 2002-2011 and its ... · CSR, release 04 (temporal resolution ≈1 month) SLR-derived geocenter coordinates (taken from CSR) replaced by results

Methods

Mass-change trends from GRACE

Period of investigationGravity fields“Manipulation”

Degree-1 termsc20 coefficients

Post-processingDe-correlationSpatial averaging

Inference of mass variationSurface mass-densitiesLeakage consideration

Signal separationGIA model

Time-series fit

May 2002 to April 2011 (9 integer years)CSR, release 04 (temporal resolution ≈1 month)

SLR-derived geocenter coordinates (taken from CSR)replaced by results from SLR (taken from CSR)

according to Swenson and Wahr (2006)Gaussian smoothing with a radius of 500 km

according to Wahr et al. (1998)according to Baur et al. (2009)

according to Paulson et al. (2007)linear trend function together with four sinusoids

Page 9: Continental mass change from GRACE over 2002-2011 and its ... · CSR, release 04 (temporal resolution ≈1 month) SLR-derived geocenter coordinates (taken from CSR) replaced by results

Methods

Mass-change trends from GRACE

Period of investigationGravity fields“Manipulation”

Degree-1 termsc20 coefficients

Post-processingDe-correlationSpatial averaging

Inference of mass variationSurface mass-densitiesLeakage consideration

Signal separationGIA model

Time-series fit

May 2002 to April 2011 (9 integer years)CSR, release 04 (temporal resolution ≈1 month)

SLR-derived geocenter coordinates (taken from CSR)replaced by results from SLR (taken from CSR)

according to Swenson and Wahr (2006)Gaussian smoothing with a radius of 500 km

according to Wahr et al. (1998)according to Baur et al. (2009)

according to Paulson et al. (2007)linear trend function together with four sinusoids

Page 10: Continental mass change from GRACE over 2002-2011 and its ... · CSR, release 04 (temporal resolution ≈1 month) SLR-derived geocenter coordinates (taken from CSR) replaced by results

Methods

Mass-change trends from GRACE

Period of investigationGravity fields“Manipulation”

Degree-1 termsc20 coefficients

Post-processingDe-correlationSpatial averaging

Inference of mass variationSurface mass-densitiesLeakage consideration

Signal separationGIA model

Time-series fit

May 2002 to April 2011 (9 integer years)CSR, release 04 (temporal resolution ≈1 month)

SLR-derived geocenter coordinates (taken from CSR)replaced by results from SLR (taken from CSR)

according to Swenson and Wahr (2006)Gaussian smoothing with a radius of 500 km

according to Wahr et al. (1998)according to Baur et al. (2009)

according to Paulson et al. (2007)linear trend function together with four sinusoids

Page 11: Continental mass change from GRACE over 2002-2011 and its ... · CSR, release 04 (temporal resolution ≈1 month) SLR-derived geocenter coordinates (taken from CSR) replaced by results

Methods

Mass-change trends from GRACE

Period of investigationGravity fields“Manipulation”

Degree-1 termsc20 coefficients

Post-processingDe-correlationSpatial averaging

Inference of mass variationSurface mass-densitiesLeakage consideration

Signal separationGIA model

Time-series fit

May 2002 to April 2011 (9 integer years)CSR, release 04 (temporal resolution ≈1 month)

SLR-derived geocenter coordinates (taken from CSR)replaced by results from SLR (taken from CSR)

according to Swenson and Wahr (2006)Gaussian smoothing with a radius of 500 km

according to Wahr et al. (1998)according to Baur et al. (2009)

according to Paulson et al. (2007)linear trend function together with four sinusoids

Page 12: Continental mass change from GRACE over 2002-2011 and its ... · CSR, release 04 (temporal resolution ≈1 month) SLR-derived geocenter coordinates (taken from CSR) replaced by results

Leakage consideration

Methods

Page 13: Continental mass change from GRACE over 2002-2011 and its ... · CSR, release 04 (temporal resolution ≈1 month) SLR-derived geocenter coordinates (taken from CSR) replaced by results

Leakage consideration

Methods

Page 14: Continental mass change from GRACE over 2002-2011 and its ... · CSR, release 04 (temporal resolution ≈1 month) SLR-derived geocenter coordinates (taken from CSR) replaced by results

Methods

Mass-change trends from GRACE

Period of investigationGravity fields“Manipulation”

Degree-1 termsc20 coefficients

Post-processingDe-correlationSpatial averaging

Inference of mass variationSurface mass-densitiesLeakage consideration

Signal separationGIA model

Time-series fit

May 2002 to April 2011 (9 integer years)CSR, release 04 (temporal resolution ≈1 month)

SLR-derived geocenter coordinates (taken from CSR)replaced by results from SLR (taken from CSR)

according to Swenson and Wahr (2006)Gaussian smoothing with a radius of 500 km

according to Wahr et al. (1998)according to Baur et al. (2009)

according to Paulson et al. (2007)linear trend function together with four sinusoids

Page 15: Continental mass change from GRACE over 2002-2011 and its ... · CSR, release 04 (temporal resolution ≈1 month) SLR-derived geocenter coordinates (taken from CSR) replaced by results

GIA model

Methods

GIA models are subject to large error bounds (particularly over Antarctica)

Adopted uncertainty level: 30%

Page 16: Continental mass change from GRACE over 2002-2011 and its ... · CSR, release 04 (temporal resolution ≈1 month) SLR-derived geocenter coordinates (taken from CSR) replaced by results

Methods

Mass-change trends from GRACE

Period of investigationGravity fields“Manipulation”

Degree-1 termsc20 coefficients

Post-processingDe-correlationSpatial averaging

Inference of mass variationSurface mass-densitiesLeakage consideration

Signal separationGIA model

Time-series fit

May 2002 to April 2011 (9 integer years)CSR, release 04 (temporal resolution ≈1 month)

SLR-derived geocenter coordinates (taken from CSR)replaced by results from SLR (taken from CSR)

according to Swenson and Wahr (2006)Gaussian smoothing with a radius of 500 km

according to Wahr et al. (1998)according to Baur et al. (2009)

according to Paulson et al. (2007)linear trend function together with four sinusoids

Page 17: Continental mass change from GRACE over 2002-2011 and its ... · CSR, release 04 (temporal resolution ≈1 month) SLR-derived geocenter coordinates (taken from CSR) replaced by results

Time-series fit

Greenland Amazon basin

Methods

Trend function + annual and semi-annual signals + S2 and K2 tidal aliases

Adoption of linear vs. quadratic trend model of secondary importance

Page 18: Continental mass change from GRACE over 2002-2011 and its ... · CSR, release 04 (temporal resolution ≈1 month) SLR-derived geocenter coordinates (taken from CSR) replaced by results

Results – regional balance

Regions of interest

Region selection according to continental areas with dominant GRACE signal

Different delineations are compensated equally during the leakage-correction procedure

Separation between physically meaningful signals and GRACE errors/artefacts

Trade-off: continental regions with signal amplitudes stronger than the strongest signal magnitude over the world’s oceans

Artefacts may be interpreted as meaningful signals and vice versa

Continental signals outside the regions of interest balance close to zero

Page 19: Continental mass change from GRACE over 2002-2011 and its ... · CSR, release 04 (temporal resolution ≈1 month) SLR-derived geocenter coordinates (taken from CSR) replaced by results

Regions of interest

Results – regional balance

Page 20: Continental mass change from GRACE over 2002-2011 and its ... · CSR, release 04 (temporal resolution ≈1 month) SLR-derived geocenter coordinates (taken from CSR) replaced by results

Regiongeocenter

neglected corrected

1 Greenlanda -286±15 -323±26

2 West Antarctica -201±40 -169±46

3 Alaska -47±5 -56±7

4 Parana -27±6 -26±7

5 Euphrates/Tigris -25±5 -27±6

6 Brahmap./Ganges -24±4 -25±6

7 Volga -24±5 -28±6

8 Patagonia -12±2 -10±3

9 West Australia -9±2 -7±3

Regiongeocenter

neglected corrected

10 East Antarctica 51±10 65±13

11 Amazon 50±13 43±14

12 Zambezi/Okavango 38±7 40±8

13 Canada 51±110 25±111

14 Fennoscandia 18±23 12±23

15 Central Siberia 13±2 10±4

16 East Australia 12±3 14±3

17 Niger 9±2 8±2

18 Kamchatka 9±2 6±3

19 Godavari/Krishna 3±1 4±1

aIncluding Iceland, Svalbard and the Canadian Arctic archipelagoUncertainties are given at the 95% (2σ) confidence level

Regions of interest – mass-change trends [Gt/yr]

Results – regional balance

Page 21: Continental mass change from GRACE over 2002-2011 and its ... · CSR, release 04 (temporal resolution ≈1 month) SLR-derived geocenter coordinates (taken from CSR) replaced by results

Zambezi/Okavango mass-change trend [Gt/yr]

This study 05/2002–04/2011: 40 ± 8

Llovel et al. (2010) 08/2002–07/2009: 21 ± 3

Re-computation 08/2002–07/2009: 27 ± 12

Subset period: 3 yearsSubset shift: 36 monthsSubset period: 5 yearsSubset shift: 6 months

Comparison with other studies …a serious pitfall

Amazon basin mass-change trend [Gt/yr]

This study 05/2002–04/2011: 43 ± 14

Llovel et al. (2010) 08/2002–07/2009: 78 ± 9

Re-computation 08/2002–07/2009: 71 ± 19

Amazon basin mass-change trend [Gt/yr]

This study 05/2002–04/2011: 43 ± 14

Llovel et al. (2010) 08/2002–07/2009: 78 ± 9

Comparison with other studies

Results – regional balance

Page 22: Continental mass change from GRACE over 2002-2011 and its ... · CSR, release 04 (temporal resolution ≈1 month) SLR-derived geocenter coordinates (taken from CSR) replaced by results

Sea-level change equivalent [mm/yr]

Translation of mass gain or mass loss over land areas into homogeneouswater changes over the world’s oceans

geocenter

neglected corrected

Sea-level rise 1.8±0.2 1.9±0.2

(Glaciated areasa 1.4±0.2 1.4±0.2)

Sea-level fall -0.7±0.4 -0.6±0.4

Balance 1.1±0.6 1.2±0.6aGreenland (incl. Iceland, Svalbard, Canadian Arctic archipelago), Antarctica, Alaska and PatagoniaUncertainties are given at the 95% (2σ) confidence level

Land-water mass accumulation compensates about 20% of the impact of ice-melt water influx to the oceans

Geocenter motion affects sea level rise and sea level fall estimates equally; the net effect is 0.1±0.1 mm/yr

Results – global balance

Page 23: Continental mass change from GRACE over 2002-2011 and its ... · CSR, release 04 (temporal resolution ≈1 month) SLR-derived geocenter coordinates (taken from CSR) replaced by results

Study Period Land-ice Land-water Total

Llovel et al. (2010) 08/2002-07/2009 - -0.2±0.1 -

Jacob et al.(2012) 01/2003-12/2010 1.5±0.3 - -

Riva et al. (2010) 02/2003-02/2009 1.0±0.2a -0.1±0.3 1.0±0.4

This study 05/2002-04/ 2011 1.4±0.2a -0.1±0.3 1.2±0.6

Comparison with other studies (rates are in [mm/yr])

aGreenland (incl. Iceland, Svalbard, Canadian Arctic archipelago), Antarctica, Alaska and PatagoniaUncertainties are given at the 95% (2σ) confidence level

Continental mass change

Study Period Total

Lombard et al. (2007) 2002-2006 1.2±0.5

Leuliette and Willis (2011) 01/2005-09/2010 1.1±0.6

Results – global balance

Ocean mass trend

Page 24: Continental mass change from GRACE over 2002-2011 and its ... · CSR, release 04 (temporal resolution ≈1 month) SLR-derived geocenter coordinates (taken from CSR) replaced by results

Conclusions

First study that selects the regions of interest exclusively on the basis of dominant GRACE signal (no a priori spatial information)

Isolation and quantification of individual contributions to the global budget

The integrated mass change within the selected regions is representative for the variation over the whole land area

It remains an open issue whether the overall sea-level change budget is closed

The computation of mass variation from GRACE has been becoming increasingly heterogeneous; “conventions” would improve consistency

Baur O., Kuhn M., Featherstone W. (2012) Continental mass change from GRACE over 2002-2011 and its impact on sea level, Journal of Geodesy, online first