Global Validation of GRACE Gravity Measurements by in-situ and modelled

Ocean Bottom Pressure

GSTM/SPP 1257Potsdam, 15.10.2007

C. Böning, A. Macrander, R. Timmermann, O. Boebel, J. Schröter

AWI Bremerhaven

PIES Deployment in Southern Ocean, RV Polarstern, 2006

Ground-truth validation of GRACE with OBP

GRACE GFZ RL04 GSM+GAD RMS variability [dbar of water equivalent]. GRACE: do2-50, 750 km Gauss filter. OBP ground truth sites

• GRACE measures mass variability on Earth

• large hydrological cycle over continents

• much smaller signal over oceans (0.01 … 0.05 dbar)

• Does GRACE capture real oceanic variability?

• Here: Global comparison with in-situ and modelled data of Ocean Bottom Pressure (OBP)

GRACE: monthly variability [RMS / dbar]

AWI POL/CNES/IFREMER

MOVE

RAPID

AWI

NOAA

POL

0.15

0.10

0.05

0GFZ RL04 GSM+GAD [RMS]

Ocean Bottom Pressure (OBP) p = ∫ g dz

Vertical integral of oceanic + atmospheric mass

• In-situ observations by instruments deployed at the sea floor

• Short term variability ≤ O(1 dbar)tides, planetary wavesmay cause aliasing in GRACE data, corrected by GAC/GAD model

• Monthly variability (0.01 to 0.05 dbar) water mass changes, geostrophic ocean currents

Ocean Bottom Pressure

OBP timeseries in AWI ACC array (PIES ANT 7-1) .

OBP [dbar] at 45°S 7°E: unfiltered data, 7 days LP

– 4677 dbar

– 4676 dbar

Validation of GRACE with OBP

Data used in this study:

I. GRACE

• different data centres: CSR, GFZ, JPL, GRGS, ITG

• different releasesRL01 to RL04

• different products: GSM monthly geoid + GAC/D monthly average of de-aliasing model

• spatial smoothing: degree/order 2 – 50,750 km Gauss filter or newly developed patch filtering

GRACE GFZ RL04 GSM+GAD RMS variability [dbar of water equivalent]. GRACE: do2-50, 750 km Gauss filter. OBP ground truth sites

GRACE: monthly variability [RMS / dbar]

AWI POL/CNES/IFREMER

MOVE

RAPID

AWI

NOAA

POL

0.15

0.10

0.05

0GFZ RL04 GSM+GAD [RMS]

Validation of GRACE with OBP

Data used in this study:

II. in-situ OBP data

Global OBP database at AWIcontains:

• dedicated GRACE-arrays: ACC (AWI), MOVE (IFM-GEOMAR/SIO)

• oceanographic OBP-sections: Drake Passage (POL/CNES), Kerguelen (POL/IFREMER), Framstrait (AWI), RAPID (NOC)

• tsunami warning system: DART (NOAA)

… further contributions appreciated

GRACE GFZ RL04 GSM+GAD RMS variability [dbar of water equivalent]. GRACE: do2-50, 750 km Gauss filter. OBP ground truth sites

GRACE: monthly variability [RMS / dbar]

AWI POL/CNES/IFREMER

MOVE

RAPID

AWI

NOAA

POL

0.15

0.10

0.05

0GFZ RL04 GSM+GAD [RMS]

• OBP variability of about ± 0.03 dbar• OBP anomalies strongly related to barotropic velocity anomalies• High pressure anomalies → anticyclonic currents• Low pressure anomalies → cyclonic currents

Data used in this study:

III. FESOM

• Finite Element Sea Ice Ocean Model

• hydrostatic primitive equation OGCM with sea ice coupling

• 1.5° horizontal resolution

• 26 z-levels

• atmospheric forcing: usually NCEP/NCAR daily reanalysis 1958-2005

• no restoring

Validation of GRACE with OBP

Framstrait at 79°N300 km from Greenland Ice Shield

• PIES data from AWI since 2003

• Monthly OBP variability O(0.05 dbar)

• GRACE captures real oceanic variability:

→ correlation improvements by recent releases, e.g. GFZ:RL03 GSM+GAC r = 0.55RL04 GSM+GAC r = 0.71RL04 GSM+GAD r = 0.76

→ best agreement of all GRACE products: GRGS (10day time axis; r = 0.80)

→ GAC, GAD de-aliasing models alone do not show observed variability, actual GRACE measurements (GSM) necessary

GRACE vs. in-situ OBP: Framstrait 79°N Arctic

Blue: In-situ data: Framstrait PIES F8 – 1-3 [A. Beszczynska-Möller, AWI]Other colours: GRACE do2-50, 750 km Gauss filter

GRACE vs. in-situ OBP: MOVE, Tropical Atlantic

Blue: In-situ data: MOVE PIES V404 1-5 [J. Karstensen, IFM-GEOMAR]Other colours: GRACE do2-50, 750 km Gauss filter

Tropical Atlantic at 16°N

• in-situ OBP variability small O(0.02 dbar)

• GRACE strongly overestimates variability

• unrealistic annual cycle in all GRACE products [GFZ, CSR,

GRGS, ITG, JPL]

→ see also: Poster of U. Neumann et al. this afternoon

Filtering GRACE data

Issues of Gaussian filtering methods due tosymmetrie of Gauss function:

•adds land signal to oceanic data

•ignores ocean circulation pattern

FESOM simulations indicate that OBP anomalies arecoherent over a certain area which corresponds tobottom topography

Spatial coherence of OBP

• OBP anomalies of 2002-2005 from 50-yr FESOM simulation

• 4-months high pass filter to subtract dominant seasonal cycle

• Cross-correlation of time series at one point with time series at all other points

• Cut-off at correlation <0.7 and radius 20º

• filter data by weighting with correlation coefficients

ANT 7

ANT 11

OBP cross-validation FESOM/PIES/GRACE

PIES: in-situ data 2002-2005FESOM: simulations 2002-2005, patch filteredGRACE: GFZ RL04, d/o 2-50, patch filtered

• increase in correlation of 0.1-0.2 • FESOM reproduces seasonal cycle

AWI ACC array: PIES ANT 7, ANT 11 (more to come)

NOAA

Comparison of Gauss and coherence-patch filtered data

GFZ RL04GSM+GAD

AWIPOL/CNES/IFREMER

AWI

NOAA

POL

AWIPOL/CNES/IFREMER

MOVE

RAPID

AWI

POL

Correlations GFZ RL04 (750 km Gauss)/in situ OBP

Correlations GFZ RL04 (patch filtered)/in situ

•Correlation ofGRACE and in situ data indicates animprovement due tothe new filteringmethod

•At many locationscorrelation increasesby 0.1-0.2

•Improvement atPOL Array in DrakePassage, butcorrelation stillnegative

RAPID

MOVE

GFZ RL04GSM+GAD

Conclusions

• Recent GRACE releases capture real oceanic OBP variability with r = 0.8 … 0.9 at some locations (all at high latitudes)

• Improvements by recent GAC, GAD de-aliasing models, but actual GRACE measurements (GSM fields) necessary to capture real variability

• Improvements RL03→RL04 and GAC→GAD

• Improvements by using the patch filtering method

• OBP database at AWI now available upon request Contact: Carmen.Boening@awi.de, Andreas.Macrander@awi.de