Hydrological loading induced vertical displacements from GPS and GRACE
4.3 Datasets and their inconsistencies
filter), AG (anisotropic Gaussian filter), IGD (isotropic Gaussian filter combined with the de-striping filter),AGD(anisotropic Gaussian filter combined with the destriping filter). It should be noted that the combination of the destriping filter with the isotropic or anisotropic Gaussian filter is conducted by applying the destriping first.
4.3 Datasets and their inconsistencies 61
2003 2004 2005 2006 2007 2008 2009 2010 2011
-15 15 50 85 120 155 190 225 260 295 330 365 400 435
time/yr
mm
NYAL POLV GLSV ANKR SOFI UZHL JOZE PENC GRAZ WTZR ZIMM MORP POTS Original weekly time series Averaged monthly time series
Figure 4.2:ExemplaryGPSheight time series from Europe. In the figure, exceptNYAL,GPSheight time series of the rest stations are shifted for plotting purposes. Shaded areas are error bounds of each original weekly time series.
For the comparison over the Amazon area, theGPStime series fromSIRGAS GPSnetwork, which is processed byDGFI (German Geodetic Research Institute), are used. The whole SIRGAS GPS
network, see Fig.4.3, comprises 58IGSglobal stations and otherSIRGAS-CONregional stations, which makes up to around 300 stations in total up to now. The time series utilized in this comparison are the latest multi-year solutionSIR11P01 (Sánchez and Seitz,2011;Sánchez et al., 2013) and the detailed computation procedure is described in (Sánchez and Seitz, 2011). The final residual time series out of this multi-year solution are cleaned and detrended weekly solutions spanning from 2000 to 2011.29 when theIGS08 reference frame was introduced.
240˚
240˚
270˚
270˚
300˚
300˚
330˚
330˚
-60˚ -60˚
-30˚ -30˚
0˚ 0˚
30˚ SIRGAS 30˚
REFERENCE NETWORK
2014-08-07
IGS RNAAC SIRGAS Deutsches Geodaetisches Forschungsinstitut (DGFI)
IGS/IGS+ STATION SIRGAS REGIONAL STATION
/BOGA/ABCC/ABPD/ABPW
/QUEM/EPEC AACR/CRCP
AGCA
ALAR ALBE
ALEC
ALUM AMCO
AMHU ANDS
ANGO
APSA APTO
ARCA
AUCA
AUTF AZUE
AZUL
BABR
BAIL BAIR BARI
BATF BAVC
BCAR BECE
BEJA
BELE BERR
BLPZ/EMIB BNGA BOSC BQLA
CALI CASI CATR
CEEU/CEFT CESB CHEC
CHET CHIH
CHIS
CLEC CN30 COAT
COEC COL2
COTZ
CRAT CRCS
CRUZ
CSLO CUEC CULC
CUM3 DARI DAVI
DORA
EBYP ECEC
EESC ELEN
EREC ESMR
ESQU ETCG/RIDC
EXU0
FLOR FQNE GARA
GOGY GOJA GRE0 GTK0
GUAY
GVA1 GVRE
GYEC
GZEC HUEH
IACR IBAG
IBEC ICEP
IDGO
IGM1 IGN1
ILHA IMPZ ISCO
JBAL LIBE
LIMN
LJEC LREC
MA01
MABA MABB
MABS MAEC
MAGA
MAPA
MCL1/MGMC
MECO MEDE
MEXI
MGBH
MGIN MGRP
MGV1 MHEC
MOTE
MPL2 MRLS
MSCG
MSDR MTBA MTCN MTCO MTEC
MTSF MTSR
MTVB
MZAC MZAE MZAU
MZGA MZSR NARA
NEIL
NESA NEVA NICY
NJEC
OSOR
OURI PAAT PAIT PAMP
PAST
PBCG PBJP PDEC
PEAF
PEJO
PEPE PERA
PISR PITN PJEC
PMB1
POLI POPA
POPT
PRCV PREC
PRGU PRMA
PRNA PSTO
PTEC PUNT
QUIB
QVEC CXEC
RIOD/ONRJ RIOH
RJCG RNMO
RNNA
ROCD ROGM
ROJI
ROSA
RSAL RUBI SAGE SAYA
SCAQ SCCH SCEC
SCFL SCLA
SEAJ SEEC
SINC
SJRP
SJSP
SL01 SMAR SMRT
SNLR SNSN
SPAR
SPBO SPCA SPJA
SRLP SRNW
SRZN
SSA1 STEC
SVIC TAXI
TEG2
TERO TIKA
TINT
TNEC
TOGU TOL2
TUCU TUMA
TUNA
UBA1 UBE1/MGUB
UCOR UGTO
UNPA UNRO UNSJ USLP
UYDU UYLP UYMO UYNI
UYPA
UYRO UYSO
UYTA VALL
VICO VIL2
VIVI
YEMA YOPA ZARZ
ABMF
ANTC AREQ/AREV
BOAV BOGT
BOMJ
BRAZ BRFT BRMU
BUE2 BUEN
CALL CANO CART
CATA CBSB
CEFE
CHAC
CHPI
CHTI
CONZ COPO
CORD
COYQ CRO1
CUCU
CUIB
FALK GLPS
GOLD
GOUG GUAT
HER2
ICAM
IMBT INEG
IPAZ
IQQE IQUI
ISPA
KOUG KOUR
LHCL LPGS MANA
MARA
MAS1 MDO1
MERI
MGUE MKEA
MTY2
NAS0
NAUS
NEIA
NKLG OAX2
OHI2 PALM/PALV PARC
PDEL PIE1
POAL POVE
PPTE QUI1
RECF
RIO2 RIOB RIOP
RWSN SAGA
SALU SAMA
SANT
SAVO
SCRZ SCUB
SSIA
STHL
SUTH TAMP
TGCV
THTI
TOPL
UFPR UNSA
URUS USNO
VALP
VBCA
VESL YCBA WIND
−20
−10 0 10
Latitude
−80 −70 −60 −50 −40 −30
Longitude
BELE BOAV
BOGA
POVE NAUS
CUIB AREQ
BOGT
BOMJ BUEN
CALI
CHPI CART
CUCU
GVAL IMPZ KOUR
MABA MAPA MARA
MCLA NEVA
PERA
PMB1
POPA
PPTE QUI1
RECF RIOB
RIOD ROJI
SAGA
SALU SAMA
SRNW SRZN
TOGU TOPL TUNA
UBER
UNSA VALL
VICO VIVI
BRAZ FLOR
−20
−10 0 10
Latitude
−80 −70 −60 −50 −40 −30
Longitude
Figure 4.3:Map of theSIRGAS GPSnetwork (courtesy:www.sirgas.org) and distribution of the selected 46
GPSsites from this regionalGPSnetwork.
To compare with GRACE in this area, GPS stations with time series which overlap with the
GRACE time frame more than three years are selected. As this study focuses on hydrological signals, the stations which are not located inside or around the Amazon basin and its nearby basins are ruled out. Eventually 46 stations out of 228 stations are used in the comparison, see Fig.4.3. A few time series are shown in Fig.4.4as examples. Since theSIRGAS GPSnetwork is under development, time series length fromSIRGASvaries among theGPSsites.
Relative toGPSheight time series from Europe, seasonal signals are much stronger in theSIR
-GASnetwork because significant water mass variations are happening in and around the Ama-zon area. For example,NAUS(located in Manaus, Brazil) displays a peak-to-peak 40 mm annual oscillation with an exception in 2009 when a severe flood happened (Chen et al.,2010). In ad-dition to seasonal behavior, stations likeBOGA(located in Bogota, Columbia) also show a clear non-linear trend signal.
4.3.2 GRACE products
T
HE GRACE GSM RL05a product from GFZ(Dahle et al.,2014) is used for the following rea-sons. Firstly, as reported byTesmer et al.(2011), the GRACE datasets fromGFZ, CSRandJPLdid not show significant systematic differences. Secondly, compared to theGRACE related
4.3 Datasets and their inconsistencies 63
2003 2004 2005 2006 2007 2008 2009 2010 2011
-25 25 75 140 180 215 265 300 330 360 400 430 460
time/yr
mm
BOGA POVE NAUS CUIB BOGT IMPZ MABA RIOB ROJI SAGA UBER VIVI
Original weekly time series Averaged monthly time series
Figure 4.4:Exemplary time series from SIRGAS. In the figure, exceptBOGA,GPS height time series of the rest stations are shifted for plotting purposes.
products from CSRandJPL, GFZprovides better metadata, e.g. the calibrated standard devia-tions of the Stokes coefficients, which are used in the regularization filtering.
Before deriving displacements from theGRACEdata, theC20term is replaced in theGRACE GSM
data using the product fromCheng et al.(2011). The resultant monthlyGRACE GSMStokes co-efficients are then filtered using the deterministic filters and the regularization filter tabulated in Table4.2. TheDDKfiltered datasets are downloaded directly from theICGEMwebsite. Note that theC20term in theDDKfiltered datasets are replaced as well before deriving the displace-ments for comparison.
4.3.3 Inconsistencies between GPS and GRACE
A
Sdiscussed in the previous chapters,GPSandGRACEobserve two fundamentally different quantities and they both experience totally different data processing procedures. Several issues exist in reality affecting the agreement betweenGPSandGRACE. For example,van Dam et al.(2007) andTesmer et al.(2011) mentioned possible error sources in bothGPSandGRACEwhich could probably influence the consistencies. To be more precise, van Dam et al.(2007) pointed out possible error sources from the GPSpart, e.g. atmospheric mismodelling, bedrock thermal expansion, monument thermal expansion, phase center modeling and common orbital errors. Tesmer et al. (2011) also attributed disagreements in part to GRACE, e.g. externalC20
term used in GRACE, GRACE data filtering and atmospheric and oceanic dealiasing models.
Apart from these possible errors, however, two fundamental issues should be resolved before comparingGPSandGRACE: 1) the reference frame issue; 2) the atmospheric loading and non-tidal oceanic loading issue. These two issues are not dealt consistently in GPS and GRACE
products and they should be corrected in those solutions before comparison.
Reference frame issue Several reference frames exist in use in geodesy and they have been discussed in Section2.3. From there we know that the resultantGPSdata stay in theCFreference frame while GRACE GSM products lie in the CM frame. Besides, translation of the reference frame from one to another is essentially linked to the translation of the degree-1 terms. As
GRACE does not sense the geocenter motion, the degree-1 terms in the GRACE GSM gravity monthly solutions are set to zero. To keep GPS andGRACE consistent in the reference frame, several ways were adopted in the literature. However, in essence, the rule is to change from one into the other, that is to say, we either change GPSfrom theCFframe into theCM frame or the other way around.
Davis et al. (2004) firstly started to compare the displacements observed by GPS and derived from GRACE. They added the l = 1 contribution to the deformation inferred from GRACE, which was followed by Nahmani et al.(2012). Whilstvan Dam et al.(2007) corrected the ref-erence frame issues by removing the displacements due to the degree-1 effects from GPS (see Fig. 2.8) and this approach was also applied by Tesmer et al.(2011). It should be noted that the displacements computed byvan Dam et al.(2007) andTesmer et al.(2011) were simply an average of theX,Y,Zcomponents from a globalGPSnetwork. Apparently, this way is not suit-able for a regionalGPSnetwork study, e.g. theGPSnetwork in Europe and theSIRGASnetwork used in this chapter. Alternatively,Tregoning et al.(2009) directly restored the degree-1 Stokes coefficients from Munekane(2007). Fu et al.(2012) followedTregoning et al. (2009) but with the degree-1 Stokes coefficients obtained fromSwenson et al. (2008). In theory, all the above-mentioned approaches should be able to maintain the consistency in the reference frame issues.
However, different approaches using different datasets will certainly cause differences.
Here we conduct one experiment regarding the frame issue. Two geocenter motion datasets, which are provided in coordinates sensed by Satellite Laser Ranging (SLR) (Cheng et al.,2013) and in the degree-1 Stokes coefficients delivered bySwenson et al.(2008), are applied. One is to follow van Dam et al.(2007) but using the degree-1 displacements computed fromCheng et al.(2013) and the other way is to followFu et al.(2012). We find that the latter way provides slightly better but negligible consistencies than the former way in both two study areas. Thus, the degree-1 coefficients fromSwenson et al.(2008) are restored back to theGRACE GSM data