Processing Mixed-Mode GPS Networks for Deformation Monitoring Applications

Transcription

1 Volker Janssen/Chrs Rzos, Processng Mxed-Mode GPS Networks Fachbeträge Processng Mxed-Mode GPS Networks for Deformaton Montorng Applcatons Volker Janssen and Chrs Rzos Summary The Global Postonng System (GPS) can be utlsed n a wde range of deformaton montorng applcatons. Durng the past few years a methodology has been developed for processng data collected by GPS networks consstng of a mxed set of sngle-frequency and dual-frequency recevers. The strategy s to deploy a few permanent GPS statons wth dual-frequency, geodetc-grade recevers surroundng an nner network of low-cost, sngle-frequency GPS recevers. The dual-frequency GPS network s used to generate a fle of correctons, analogous to Wde Area DGPS correcton models for the dstance dependent bases. These correctons are then appled to the double-dfferenced phase observatons from the nner recevers to mprove baselne accuracy (prmarly through emprcal modellng of the resdual atmospherc bases that otherwse would be neglected). The performance of ths confguraton has been nvestgated by smulatng such a two-stage GPS network usng data collected n dfferent geographcal regons. Zusammenfassung Das Global Postonng System (GPS) kann für ene ganze Rehe von Deformatonsmessungen engesetzt werden. In den letzten paar Jahren wurde ene Methode entwckelt, um Daten von GPS-Netzen auszuwerten, de sch aus ener gemschten Anzahl von En- und Zwefrequenzempfängern zusammensetzen. De Stratege st herbe, en nneres Netz von kostengünstgen Enfrequenzempfängern mt engen, wengen GPS- Referenzstatonen von Zwefrequenzempfängern höchster Qualtät zu umgeben. Das äußere Zwefrequenznetz wrd benutzt, um ene Date von Korrekturen zu genereren (n Analoge zu Wde Area DGPS Korrekturmodellen für de dstanzabhänggen Fehler). Dese Korrekturen werden dann an de Doppeldfferenz-Phasenbeobachtungen der nneren Empfänger angebracht, um de Basslnengenaugket zu verbessern (hauptsächlch durch emprsche Modellerung der atmosphärschen Restenflüsse, de sonst vernachlässgt werden würden). De Lestungsfähgket deser Konfguraton wurde untersucht, ndem en solches gemschtes GPS-Netz mt Daten aus verschedenen geografschen Regonen smulert wurde. Introducton The Global Postonng System (GPS) can be utlsed n a wde range of deformaton montorng applcatons. The decreasng cost of GPS hardware, together wth the ncreased relablty of the technology, facltates such demandng applcatons as the montorng of actve volcanoes, tectonc fault lnes, landsldes, ground subsdence, brdges, dams, hgh-rse buldngs, etc. GPS deformaton measurements can be contnuous, automatc, conducted n all weather condtons, and provde three-dmensonal 28. Jg. 2/200 zfv 87

2 Fachbeträge Volker Janssen/Chrs Rzos, Processng Mxed-Mode GPS Networks postonng results. Hgher computng power also means that the complex mathematcs requred to process GPS baselnes can easly be handled n (near) real-tme. Deformaton montorng usng GPS s usually carred out by nstallng and operatng a local network of GPS recevers located on and around the deformng body. Numerous contnuous GPS networks of dfferent sze have been establshed on tall buldngs (Ogaja et al., 200), brdges (Wong et al., 200), dams (Behr et al., 998), volcanoes (Owen et al., 2000), and on tectonc faults (Tsuj et al., 995). However, these networks are comparatvely costly because they rely entrely on the use of dual-frequency nstrumentaton. In order to keep the cost of such a deformaton montorng system to a mnmum, sngle-frequency GPS recevers need to be used. On the other hand, some atmospherc bases (manly the onospherc delay) cannot be accounted for drectly f only one frequency s used. A sngle-frequency, carrer phase-trackng system s consdered approprate for small-scale contnuous GPS networks f the baselne lengths are not longer than 0 km. Ths rule-of-thumb mples that the dfferental onospherc and tropospherc delay between the two recevers s essentally zero, and therefore does not mpact on the baselne result. Durng solar sunspot cycle maxmum perods, however, onospherc dsturbances have ndeed corrupted L baselne measurements over dstances less than 0 km, adversely affectng baselne repeatablty (Janssen et al., 200). It s therefore necessary to combne a sngle-frequency deformaton montorng network wth a small number of dual-frequency recevers n order to account for these atmospherc effects. Such a mxed-mode confguraton offers consderable flexblty and cost savngs for deformaton montorng applcatons, whch requre a dense spatal coverage of GPS statons, and where t s not possble, nor approprate, to establsh permanent GPS networks usng dual-frequency nstrumentaton. Hartnger/Brunner (2000) have used such a mxed-mode approach to montor landsldes n the Austran Alps. In ths study a network of three dual-frequency GPS recevers surroundng the deformaton zone s used to generate emprcal correcton terms (Rzos et al., 2000). These double-dfferenced correctons are then appled to the data from the sngle-frequency baselnes n the nner network to account for resdual atmospherc bases. In ths paper, the onospherc correcton model proposed for ths mxed-mode system and the data processng strategy are descrbed, and expermental results obtaned n dfferent geographcal regons are presented. 2 Ionospherc Correctons The onosphere s a band of the Earth s upper atmosphere located approxmately 50 km 000 km above the surface. The hgh spatal and temporal varablty of the onosphere has a major effect on GPS sgnals travellng from the satellte to the recever. Moreover, the condton of the onosphere s strongly related to the -year sunspot actvty cycle. The most recent solar maxmum occurred n , causng hgh onospherc actvty and havng a clear mpact on the results presented n ths paper. It s well known that the onosphere s most actve n a band extendng up to about 20 on ether sde of the geomagnetc equator. Ths s also one of the two regons where small-scale onospherc dsturbances (short-term sgnal varatons n ampltude and phase known as scntllatons) manly occur. The other beng the hgh-lattude regon close to the poles. In the equatoral regon scntllatons occur between approxmately one hour after sunset untl mdnght (Klobuchar, 996), and should have dsappeared by a.m. local tme (IPS, 2002). The occurrence of scntllatons also vares wth the seasons. Between Aprl and August they are less severe n the Amercan, Afrcan and Indan longtude regons, whle they are at a maxmum n the Pacfc regon. The stuaton s reversed from September to March (Seeber, 99). In md-lattudes scntllatons are rarely experenced, but Medum-Scale Travellng Ionospherc Dsturbances (MSTIDs) occur frequently, manly durng daytme n the wnter months, durng perods of hgh solar actvty, wth a maxmum around local noon (Wannnger, 999). Whle data from dual-frequency GPS recevers can account for the onospherc delay drectly by the approprate lnear combnaton of measurements made on both frequences, data from sngle-frequency recevers cannot be corrected n ths way. A smple onosphere model transmtted wthn the navgaton message can be used to account for about 50 % of the onospherc range delay on L (Klobuchar, 987). However, ths s not suffcent for deformaton montorng applcatons where t s desred to detect movements of a few centmetres or less. For deformaton montorng a mxed-mode GPS network approach can be adopted, where a fducal network of dual-frequency recevers surroundng the deformaton zone can be used to generate correcton terms. These can then be appled to the sngle-frequency observatons to account for the onospherc effects and hence mprove baselne accuracy. Fgure shows the deal confguraton of such a mxed-mode GPS network where the trangles denote fducal statons, whle the dots ndcate snglefrequency stes. The reference statons are to be stuated outsde the deformaton zone but wthn the local tectonc regon n order to avod unwanted dsplacements of the external network. Han/Rzos (996) and Han (997) have proposed a lnear combnaton model whch utlses a sngle onospherc layer model at a heght of 50 km. Ths approach can account for orbt bas and onospherc delay, as well as mtgate tropospherc delay, multpath and measurement nose across the network. Data from the GPS re- 88 zfv 2/ Jg.

3 Volker Janssen/Chrs Rzos, Processng Mxed-Mode GPS Networks Fachbeträge α 2, α ( ) and α (2) X G u X G 0 mn where X G and X G are the user staton poston vector u and reference staton poston vector n the Gaussan plane coordnate system respectvely, whle α denotes the weght for reference staton. The lnear combnaton of the sngle-dfferenced observables can now be wrtten as: Fg. : Ideal network confguraton of a mxed-mode GPS deformaton montorng network ference staton network can be used to derve emprcal correctons to the double-dfferenced carrer phase data formed between the statons of the nner network. 2. Sngle-Dfferenced Model The sngle-dfferenced carrer phase observaton can be wrtten as (Han, 997): φ ρ+ dρ c dt+λ N - don + dtrop + d φ mp +ε φ where s the sngle-dfference operator (dfference between user and reference recever), φ s the carrer phase observaton n unts of metres, ρ s the dstance between recever staton and satellte, dρ s the effect of satellte ephemers error on a partcular recever-satellte range, c s the speed of lght, dt s the recever clock error wth respect to GPS tme, λ s the wavelength of the carrer phase, N s the nteger ambguty for a partcular satellte-recever phase measurement, d on s the onospherc delay, d trop s the tropospherc delay, d φ mp s the carrer phase multpath effect, and ε φ s the carrer phase observaton nose n the one-way observaton. If more than one reference staton s avalable, several sngle-dfferenced carrer phase observables can be obtaned between the montorng user recever(s) and the reference recevers. In order to account for the dstance dependent bases (orbt error, onospherc delay and tropospherc delay) n equaton (), weghts are ntroduced. These weghts are nversely proportonal to the baselne lengths and are used to average out the dfferent dstances between the user and the reference statons. Assumng a fducal network consstng of three reference statons, a set of parameters α can be determned such that the followng condtons are satsfed (Han/Rzos, 996; Wu, 994): () α φ α ρ+ α ρ d c α dt +λ α N α don, + α dtrop, φ + α dmp, +ε α φ Accordng to the second condton n equaton (2) the orbt bas term has been shown to be: α. (4) ρ d 0 Consderng a user recever u, the onospherc delay can be expressed as: don, don, α α2. don,2 d on, Wth ncreasng dstance between the GPS reference statons the resdual bas wll become larger due to errors n the onospherc delay nterpolaton. If the tropospherc delay can be nterpolated from the resdual tropospherc delay at the reference statons, the tropospherc delay can be represented as: α d d d on, on,u on, α d d d trop, trop,u trop, x y dtrop, dtrop, xu yu x2 y 2 dtrop,2 d trop,. () where (x, y ), (x 2, y 2 ) and (x u, y u ) are the coordnates of the reference and user recevers relatve to reference staton n the Gaussan plane coordnate system. How close ths term s to zero depends on the spatal correlaton of the tropospherc delay. The resdual tropospherc delay s largely nfluenced by the wet component of the troposphere, whch s hghly varable wth heght, tme and geographc locaton makng t very dffcult to model. It can be expected that the tropospherc delay wll be mtgated to some extent, but ts effectveness s an unknown functon of the dstance between the GPS recevers. (5) (6) 28. Jg. 2/200 zfv 89

4 Fachbeträge Volker Janssen/Chrs Rzos, Processng Mxed-Mode GPS Networks The carrer phase multpath effect can be expressed as: φ φ φ α dmp, dmp,u α dmp,. (7) The second term on the rght sde of equaton (7) s the weghted mean of the multpath values at the three reference recevers for ths satellte. Due to the random nature of multpath at dfferent recevers, ths weghted mean wll be reduced f all α (, 2, ) are postve and less than, although the weght α s not derved from ts standard devaton. On the other hand, the multpath at the user recever wll become a hgh-frequency bas, and wll therefore be assumed to be close to random nose (Zhang/Schwarz, 996). Hence t can be assumed that the multpath term has been sgnfcantly reduced and wll consequently be gnored n the functonal model. The standard devaton of the one-way carrer phase observaton can be approxmated as a functon of the elevaton angle. If all GPS network recevers are located wthn a regon of about 00 km radus, the elevaton of a satellte s approxmately the same. If σ j denotes the approxmated standard devaton of a one-way carrer phase observaton, the standard devaton of the lnear combnaton of sngle-dfferenced observatons can be represented as: σ +α +α +α σ α φ j 2. (8) The sngle-dfferenced functonal model can now be wrtten n a smplfed form: α φ α ρ c α dt +λ α N +ε α φ 2.2 Double-Dfferenced Model Consderng a user recever u, equaton (9) can be expressed as (Han, 997): φu, α φ, + α2 φ 2, ρ α ρ + α ρ u,, 2 2, u, c α dt +λ N α N, +α2 N2, +ε. α φ. (9) (0) The double-dfferenced observaton model can then be represented as: φ α φ + α φ ρ where denotes the double-dfference operator. Note that the recever clock bas term cancels by formng the between-satellte dfferences. Resdual vectors can be formed from the double-dfferenced observatons between reference statons & and 2&: V V u,, 2 2, u, α ρ +α ρ +λ N, 2 2, u, α N, +α2 N2, +ε φ N ρ,,,, φ N ρ 2, 2, 2, 2, (2). () The double-dfferenced observable can then be wrtten as: φ u, α V, + α2 V2, ρ + λ N + ε u, u, α φ. (4) After the ntalsaton of the reference statons the nteger ambgutes are known, and together wth the known coordnates, the correcton vectors V, and V 2, can be determned. The correcton term α V, +α 2 V2, can now be obtaned and appled to the user recever. If the baselne between two GPS statons j and k of the nner network s consdered, the lnear combnaton can be wrtten as: k,j k,j φk,j α V, + α2 V2, ρ + λ N + ε k,j k,j k,j α φ () (5) where α k,j s the dfference n the α value for statons k,j k,j j and k, and α V, +α2 V2, s the correcton term for the nner baselne between these user recevers. By formng the double-dfferenced observables between the nner sngle-frequency recevers, and usng the resdual vectors generated from the reference statons, the nner statons coordnates can be determned wthout the need to use any GPS reference staton observatons at all. In practce, holdng one reference ste fxed, the baselnes to the other two stes are processed and correcton terms are obtaned for both baselnes. These are then scaled (by means of the parameters α ) accordng to the poston of the nner statons nsde the reference net- 90 zfv 2/ Jg.

5 Volker Janssen/Chrs Rzos, Processng Mxed-Mode GPS Networks Fachbeträge work trangle to generate double-dfferenced correctons for the nner baselnes. The nature of these emprcally-derved double-dfferenced correcton terms has been nvestgated by Janssen et al. (200). A range of GPS data sets were processed ncorporatng a varety of baselne lengths, dfferent geographcal locatons and dfferent perods of sunspot actvty (and hence onospherc condtons). The standard devaton of the double-dfferenced correcton terms was found to be ncreasng lnearly wth ncreasng baselne length. The rate of ncrease was much more severe under solar maxmum condtons as opposed to perods of low solar actvty. Ths suggests that long baselnes between reference statons mght not be capable of generatng relable correctons under these condtons. However, the magntudes of these bases are not entrely a functon of dstance, hence t s dffcult to predct what should be the dmensons of the reference staton network that would fathfully model the dstance dependent bases. The geographc locaton of the network s certanly another contrbutng factor, as the onospherc effects for GPS stes n the equatoral regon are much larger compared to md-lattude stes. Sngle-Frequency Data Processng A sngle-frequency verson of the Baselne software package developed at UNSW s used to process the nner network ncorporatng the correcton terms, whch are determned usng a modfed verson of the Bernese software. For deformaton montorng applcatons multbaselne processng strateges should be used because all baselnes are then computed together, takng nto account the between-baselne correlatons whch arse from observng a GPS network smultaneously (Craymer/Beck, 992). For contnuous deformaton montorng a near real-tme, epoch-by-epoch soluton s desred n order to detect movements over a short perod of tme. For a number of deformaton montorng applcatons, such as local networks around actve volcanoes or dams, the deformng body tself wll obstruct part of the sky. If the usual base-staton/base-satellte approach s used n the GPS data processng, usually only the common satelltes are consdered. Ths results n the number of possble double-dfferences beng comparatvely low, hence a lot of potentally valuable nformaton can be lost. As descrbed n Janssen (200), the Baselne software utlses a procedure to optmse the number of double-dfferenced observatons used n the data processng. Ths method also consders satelltes that are only vsble from a small number of network statons. Hence, the number of ndependent double-dfferenced observables can be maxmsed n order to generate a more accurate and relable soluton. Ths data processng approach determnes the recever-to-satellte connectons for each ste of the network. A maxmum set of ndependent double-dfferenced combnatons s then computed usng vector space methods and the geometrc charactersatons of Boolean matrces, as suggested by Saalfeld (999). 4 Case Studes 4. Md-Lattude Regon Data from the Southern Calforna Integrated GPS Network (SCIGN) (SCIGN, 2002) were used to nvestgate the performance of the mxed-mode network confguraton n the md-lattude regon. Fgure 2 shows the locaton of the GPS stes, whch are all equpped wth dual-frequency recevers. The part of the network used n ths study conssts of an outer network of three stes (FXHS, FMTP, QHTP) surroundng an nner network of three stes (CSN, OAT2, CMP9). The outer stes were used as fducal GPS reference statons, ndcated by trangles n Fgure 2, whle the nner stes (ndcated by crcles) smulated sngle-frequency recever statons (by gnorng the observatons made on L2). The data were collected under solar maxmum condtons, usng an observaton rate of 0s, on three consecutve days from 8 0 August 2000 (DOY 22-22). The fducal baselnes FXHS-FMTP and FXHS-QHTP are used to model the onospherc condtons across the network. The L correcton terms for the nner baselnes CSN-OAT2 and CSN-CMP9 can be determned by formng the lnear combnaton accordng to equaton (5) on a satellte-by-satellte and epoch-by-epoch bass. As an example, Fgure shows the double-dfferenced correctons obtaned for day 22. It can easly be recognsed that onospherc actvty n md-lattudes s manly a daytme phenomenon. A comparson wth drectly generated correcton terms for the same baselnes (usng dual-frequency data) showed that the proposed pro- Fg. 2: SCIGN network statons used n ths study 28. Jg. 2/200 zfv 9

6 Fachbeträge Volker Janssen/Chrs Rzos, Processng Mxed-Mode GPS Networks Fg. : Double-dfferenced L correctons for the nner baselnes (DOY 22) cedure does ndeed compute correct values for the nner baselnes. The Baselne software s then used to process the nner baselnes n sngle-frequency mode, wth and wthout usng the emprcally-derved onospherc correcton terms. It can readly be assumed that no ground deformaton has taken place durng the tme of observaton. Hence, the baselne repeatablty gves a good ndcaton of the accuracy that can be acheved wth the data processng strategy descrbed n ths paper. Fgure 4 shows the results obtaned for the nner baselnes usng the Baselne software wthout applyng onospherc correctons, whle Fgure 5 shows the results obtaned by applyng onospherc correctons on day 22. The followng days produce smlar plots. The graphs show the Eastng, Northng and Heght components over a 24-hour perod, each dot representng a sngle-epoch soluton. In both Fg. 4: Results for nner baselnes not usng onospherc correctons (DOY 22) Fg. 5: Results for nner baselnes applyng onospherc correctons (DOY 22) 92 zfv 2/ Jg.

7 Volker Janssen/Chrs Rzos, Processng Mxed-Mode GPS Networks Fachbeträge Day 22, Day 22, Day 222, Day 222, Day 22, Day 22, no corr. corr. no corr. corr. no corr. corr. Baselne CSN-OAT2 STD Eastng [m] STD Northng [m] STD Heght [m] Baselne CSN-CMP9 STD Eastng [m] STD Northng [m] STD Heght [m] Tab. : Standard devatons of the nner baselne components on days Baselne DOY East [%] North [%] Heght [%] CSN OAT ( km) CSN CMP (5 km) Average [%] Tab. 2: Average mprovement n the STD for both baselnes on days cases the Saastamonen model was used to account for the tropospherc bas, as recommended by Mendes (999). Table lsts the standard devatons (STD) of the results obtaned for the nner baselnes usng the two dfferent processng strateges (not applyng correctons versus applyng correctons) on three successve days. Comparng Fgures 4 and 5, and takng note of the nformaton gven n Table, t s evdent that the baselne results are mproved sgnfcantly by applyng the correcton terms. On average, the standard devaton of the baselne results has been reduced by almost 50 % n the horzontal and almost 40 % n the vertcal component (Table 2). Usng the proposed processng strategy, standard devatons of less than cm horzontally and.5 cm vertcally have been acheved for a sngleepoch baselne soluton (Table ). For baselnes nvolvng a sgnfcant dfference n staton alttude, as for example n GPS volcano deformaton montorng networks, the accuracy could be further mproved by estmatng an addtonal resdual relatve zenth delay parameter to account for the tropospherc bas. Among others, Abdn et al. (998) clam that global troposphere models alone are not suffcent n such cases, and the relatve tropospherc delay should be estmated. proposed network confguraton n the low-lattude regon. Fgure 6 shows the locaton of the GPS stes, whch are all equpped wth dual-frequency recevers. The network conssts of an outer network of three stes (HKKY, HKFN, HKSL) surroundng two nner stes (HKKT, HKLT). The outer stes were used as fducal GPS reference statons, ndcated by trangles n Fgure 6, whle the nner stes (ndcated by crcles) smulated sngle-frequency recever statons (by gnorng the observatons made on L2). Due to the absence of a thrd ste wthn the fducal trangle, the fducal ste HKKY was used to form the nner (sngle-frequency) baselnes. Note that HKLT s located just outsde the fducal trangle, whch should normally be avoded. However, ths does not have any effect on the processng n ths case, as HKKY-HKFN and HKKY-HKSL are used as fducal baselnes. The data were collected under solar maxmum condtons, usng an observaton rate of 0s, on three consecutve days from October 2000 (DOY ). Accordng to equaton (4) L correcton terms can be determned for the nner baselnes HKKY-HKKT and HKKY-HKLT. As an example, Fgure 7 shows the doubledfferenced correctons obtaned for day 285. A comparson wth drectly generated correcton terms for the same baselnes (usng dual-frequency data) showed that 4.2 Equatoral Regon Data from the Hong Kong GPS Actve Network (Chen et al., 200) were used to nvestgate the performance of the Fg. 6: Hong Kong GPS Actve Network statons used n ths study 28. Jg. 2/200 zfv 9

8 Fachbeträge Volker Janssen/Chrs Rzos, Processng Mxed-Mode GPS Networks Fg. 7: Double-dfferenced L correctons for the nner baselnes (DOY 285) the proposed procedure does ndeed compute correct values for the nner baselnes. It s evdent that onospherc actvty n the equatoral regon s much more severe than n md-lattudes, reachng values of a few cycles. As expected, t manly occurs after local sunset, although there s also a lot of daytme actvty. Ths may be explaned by ntensfed small-scale dsturbances n the onosphere durng a perod of ncreased solar actvty. Furthermore, t can be dentfed as the prmary durnal maxmum of the equatoral anomaly, also known as the fountan effect (hgh electron concentraton observed on ether sde of the geomagnetc equator at magnetc lattudes of around 0 20 ). Huang/ Cheng (99) state that the daly equatoral anomaly generally begns to develop at around 9 0 a.m. local tme, reachng ts prmary maxmum development at 2 p.m. local tme. In perods of solar maxmum cond- Fg. 8: Results for the nner baselnes not usng onospherc correctons (DOY 285) Fg. 9: Results for the nner baselnes applyng onospherc correctons (DOY 285) 94 zfv 2/ Jg.

9 Volker Janssen/Chrs Rzos, Processng Mxed-Mode GPS Networks Fachbeträge Day 285, Day 285, Day 286, Day 286, Day 287, Day 287, no corr. corr. no corr. corr. no corr. corr. Baselne HKKY-HKKT STD Eastng [m] STD Northng [m] STD Heght [m] Baselne HKKY-HKLT STD Eastng [m] STD Northng [m] STD Heght [m] Tab. : Standard devatons of the nner baselne components on days Baselne DOY East [%] North [%] Heght [%] HKKY HKKT (8 km) HKKY HKLT (7 km) Average [%] Tab. 4: Average mprovement n the STD for both baselnes on days tons, however, the anomaly s prone to peak after sunset, and gradents n the total electron content (TEC) are consderably larger at ths secondary durnal maxmum (Skone, 2000). Usng the UNSW-developed Baselne software the nner baselnes are processed n sngle-frequency mode, wth and wthout usng the emprcally-derved onospherc correcton terms. It can be assumed that no ground deformaton has taken place durng the tme of observaton. Fgure 8 shows the results obtaned for the nner baselnes usng the Baselne software wthout applyng onospherc correctons, whle Fgure 9 shows the results obtaned by applyng onospherc correctons on day 285. The followng days produce smlar plots. The graphs show the Eastng, Northng and Heght components over a 24-hour perod on three successve days, each dot representng a sngle-epoch soluton. As recommended by Mendes (999), the Saastamonen model was used to account for the tropospherc bas n both cases. Table lsts the standard devatons (STD) of the results obtaned for the nner baselnes usng the two dfferent processng methods (not applyng correctons versus applyng correctons) on three successve days. Although t s not obvous from Fgures 8 and 9, the baselne results are mproved by applyng the correcton terms, as ndcated n Table. On average, the standard devaton of the baselne results has been reduced by about 20 % n all three components (Table 4). The bggest mprovement (of approxmately 25 %) was acheved on day 287, the day wth comparatvely calm onospherc condtons. Ths ndcates that extreme onospherc condtons, such as those experenced n close proxmty to the geomagnetc equator durng solar cycle maxmum perods, can reduce the effcency of the proposed method. Ths s most lkely due to short-term effects of the hghly varable onosphere that cannot be modelled adequately. When applyng the correcton terms, the standard devatons stll reach values of cm n the horzontal components and cm n the heght component values too large to permt relable detecton of ground deformaton at the desred accuracy level. Unfortunately, n spte of the shorter fducal baselne lengths, the promsng results obtaned n the md-lattudes could not be repeated for ths network, stuated n the equatoral regon. Ths emphaszes the sgnfcant effect of the onosphere on GPS deformaton montorng networks, especally at low-lattudes n perods of heghtened solar actvty. 5 Concludng Remarks A procedure to process a mxed-mode GPS network for deformaton montorng applcatons has been descrbed. Sngle-frequency GPS observatons have been mproved by generatng emprcal correctons obtaned from a fducal network of dual-frequency reference statons surroundng the nner sngle-frequency network. Ths method accounts for the onospherc bas that otherwse would have been neglected when usng sngle-frequency nstrumentaton only. Data from two GPS networks located n dfferent geographcal regons have been used to smulate such a network confguraton n order to nvestgate the mpact of the proposed processng strategy on baselne results. The generated correcton terms have hghlghted that the onosphere has a sgnfcant effect on GPS baselne results. Even for short baselnes ths effect should not be neglected f t s desred to detect deformatonal sgnals wth sngle-frequency nstrumentaton at a hgh accuracy. In the md-lattude regon the sngle-frequency baselne repeatablty has clearly been mproved by applyng 28. Jg. 2/200 zfv 95

10 Fachbeträge Volker Janssen/Chrs Rzos, Processng Mxed-Mode GPS Networks the emprcal correcton terms. The standard devaton of the baselne results has been reduced by almost 50 % n the horzontal and almost 40 % n the vertcal component. Standard devatons of less than cm horzontally and.5 cm vertcally have been acheved for a sngleepoch baselne soluton. In the equatoral regon an mprovement of approxmately 20 % n all three components has been acheved by applyng the correctons. However, the fndngs also ndcate that extreme onospherc condtons, such as those experenced n close proxmty to the geomagnetc equator durng solar cycle maxmum perods, can reduce the effcency of the proposed method. The standard devaton of the baselne components for a sngle-epoch baselne soluton could not be reduced below about cm horzontally and cm vertcally values too large to permt deformaton montorng at the desred accuracy level. Nevertheless, the approach of processng a mxed-mode GPS network descrbed n ths paper can be a cost-effectve and accurate tool for deformaton montorng sutable for a varety of applcatons. Acknowledgements SCIGN and ts sponsors, as well as A/Prof. Peter Morgan from the Unversty of Canberra, are thanked for provdng the data used n ths study. The frst author s supported by an Internatonal Postgraduate Research Scholarshp (IPRS) and fundng from the Australan Research Councl (ARC). The DVW s gratefully acknowledged for provdng travel support for the frst author to undertake hs Ph.D. studes n Sydney, Australa. References Abdn, H. Z., Melano, I., Suganda, O. K., Kusuma, M. A., Muhard, D., Yolanda, O., Setyadj, B., Sukhyar, R., Kahar, J. and Tanaka, T.: Montorng the Deformaton of Guntur Volcano Usng Repeated GPS Survey Method. Proc. th FIG Int. Congress, Commsson 5, Brghton, UK, 9 25 July, 5 69, 998. Behr, J. A., Hudnut, K. W. and Kng, N. E.: Montorng Structural Deformaton at Pacoma Dam, Calforna, Usng Contnuous GPS. Proc. ION GPS-98, Nashvlle, Tennessee, 5 8 September, 59 68, 998. Chen, W., Hu, C., Chen, Y., Dng, X. and Kwok, S. C. W.: Rapd Statc and Knematc Postonng wth Hong Kong GPS Actve Network. Proc. ION GPS-200, Salt Lake Cty, Utah, 4 September, 46 52, 200. Craymer, M. R. and Beck, N.: Sesson Versus Sngle-Baselne GPS Processng. Proc. ION GPS-92, Albuquerque, New Mexco, 6 8 September, , 992. Han, S.: Carrer Phase-Based Long-Range GPS Knematc Postonng. PhD Dssertaton, UNISURV S-49, School of Geomatc Engneerng, The Unversty of New South Wales, Sydney, Australa, 997. Han, S. and Rzos, C.: GPS Network Desgn and Error Mtgaton for Real-Tme Contnuous Array Montorng Systems. Proc. ION GPS- 96, Kansas Cty, Mssour, 7 20 September, , 996. Hartnger, H. and Brunner, F. K.: Development of a Montorng System of Landslde Motons Usng GPS. Proc. 9 th FIG Int. Symp. on Deformaton Measurements, Olsztyn, Poland, Sep 999, 29 8, Huang, Y.-N. and Cheng, K.: Ionospherc Dsturbances at the Equatoral Anomaly Crest Regon durng the March 989 Magnatc Storms. Journal of Geophyscal Research 96(A8): , 99. IPS: Space Weather and Satellte Communcatons. Fact sheet, space_weather_satellte_communcatons.html, Janssen, V.: Optmzng the Number of Double-Dfferenced Observatons for GPS Networks n Support of Deformaton Montorng Applcatons. GPS Solutons 4(): 4 46, 200. Janssen, V., Roberts, C., Rzos, C. and Abdn, H. Z.: Experences wth a Mxed-Mode GPS-Based Volcano Montorng System at Mt. Papandayan, Indonesa. Geomatcs Research Australasa 74: 4 58, 200. Klobuchar, J. A.: Ionospherc Tme-Delay Algorthm for Sngle-Frequency GPS Users. IEEE Transactons on Aerospace and Electronc Systems AES-2(): 25, 987. Klobuchar, J. A.: Ionospherc Effects on GPS. In: Parknson, B. W. and Splker, J.J. (Eds.), Global Postonng System: Theory and Applcatons Volume I. Progress n Astronautcs and Aeronautcs, 6, Amercan Inst. of Aeronautcs and Astronautcs, Washngton, , 996. Mendes, V. B.: Modelng the Neutral-Atmosphere Propagaton Delay n Radometrc Space Technques. PhD Dssertaton, Dept. of Geodesy & Geomatcs Eng. Tech. Rept. No. 99, Unversty of New Brunswck, Fredercton, Canada, 999. Ogaja, C., Rzos, C., Wang, J. and Brownjohn, J.: A Dynamc GPS System for On-lne Structural Montorng. Int. Symp. on Knematc Systems n Geodesy, Geomatcs & Navgaton (KIS 200), Banff, Canada, 5 8 June, , 200. Owen, S., Segall, P., Lsowsk, M., Mklus, A., Murray, M., Bevs, M. and Foster, J.: January 0, 997 Eruptve Event on Klauea Volcano, Hawa, as Montored by Contnuous GPS. Geophyscal Research Letters 27(7): , Rzos, C., Han, S., Ge, L., Chen, H. Y., Hatanaka, Y. and Abe, K.: Low- Cost Densfcaton of Permanent GPS Networks for Natural Hazard Mtgaton: Frst Tests on GSI s GEONET Network. Earth, Planets & Space 52(0): , Saalfeld, A.: Generatng Bass Sets of Double Dfferences. Journal of Geodesy 7: , 999. SCIGN: Southern Calforna Integrated GPS Network (SCIGN) webste: Seeber, G.: Satellte Geodesy. De Gruyter, Berln, Germany, 99. Skone, S. H.: Wde Area Ionosphere Modelng at Low Lattudes Specfcatons and Lmtatons. Proc. ION GPS-2000, Salt Lake Cty, Utah, 9 22 September, , Tsuj, H., Hatanaka, Y., Sagya, T. and Hashmoto, M.: Cosesmc Crustal Deformaton from the 994 Hokkado-Toho-Ok Earthquake Montored by a Natonwde Contnuous GPS Array n Japan. Geophyscal Research Letters 22(): , 995. Wannnger, L.: The Performance of Vrtual Reference Statons n Actve Geodetc GPS-Networks under Solar Maxmum Condtons. Proc. ION GPS-99, Nashvlle, Tennessee, 4 7 September, , 999. Wong, K.-Y., Man, K.-L. and Chan, W.-Y.: Montorng Hong Kong s Brdges Real-Tme Knematc Spans the Gap. GPS World 2(7): 0 8, 200. Wu, J. T.: Weghted Dfferental GPS Method for Reducng Ephemers Error. Manuscrpta Geodaetca 20: 7, 994. Zhang, Q. J. and Schwarz, K. P.: Estmatng Double Dfference GPS Multpath under Knematc Condtons, Proc. IEEE Poston Locaton & Navgaton Symp. (PLANS 96), Atlanta, Georga, Aprl, , 996. Authors address Dpl.-Ing. Volker Janssen and Prof. Chrs Rzos School of Surveyng and Spatal Informaton Systems The Unversty of New South Wales Sydney NSW 2052 Australa v.janssen@student.unsw.edu.au, c.rzos@unsw.edu.au 96 zfv 2/ Jg.