An Assessment of Predicted and Measured Ionospheric Total Electron Content Using a Regional GPS Network

Transcription

1 An Aement of Predicted and Meaured Ionopheric Total Electron Content Uing a Regional GPS Network A. Komjathy and R.B. Langley Geodetic Reearch Laboratory, Department of Geodey and Geomatic Engineering Univerity of New Brunwick, P.O. Box, Fredericton, N.B. E3B A3 Canada BIOGRAPHIES Attila Komjathy received hi M.Sc.Eng. degree (199) from the Department of Geodey and Mining Surveying of the Univerity of Mikolc, Hungary, and ubequently worked a a reearch aitant for the Hungarian Academy of Science. In 1991, he enrolled a a Ph.D. tudent in geodey in the Department of Geodey and Geomatic Engineering of the Univerity of New Brunwick (UNB), Canada. Komjathy pecialize in extraterretrial poitioning technique for high preciion geodey and navigation. Hi current reearch focue on the Global Poitioning Sytem (GPS) and it application, concentrating on the tudy of the ionopheric effect on GPS ignal uing variou modelling technique. Richard Langley i a profeor in the Department of Geodey and Geomatic Engineering at the UNB, where he ha been teaching ince 191. He ha a B.Sc. in applied phyic from the Univerity of Waterloo and a Ph.D. in experimental pace cience from York Univerity, Toronto. After obtaining hi Ph.D., Dr. Langley pent two year with the Department of Earth and Planetary Science of the Maachuett Intitute of Technology where he carried out reearch involving lunar laer ranging and very long baeline interferometry. Dr. Langley ha worked extenively with GPS. He i a co-author of the bet-elling Guide to GPS Poitioning publihed by Canadian GPS Aociate and i a columnit for GPS World magazine. He ha helped develop and preent a number of eminar coure on GPS for both Canadian GPS Aociate and the Americanbaed Navtech Seminar Inc. Dr. Langley ha conulted extenively in the field of GPS with private companie and government agencie both in Canada and abroad. ABSTRACT The ignal from the atellite of the Navtar Global Poitioning Sytem (GPS) mut travel through the earth' ionophere on their way to GPS receiver on or near the earth' urface. To achieve the highet poible poitioning accuracie from GPS, one mut correct for the carrier phae advance and peudorange group delay impoed on the ignal by the ionophere. Wherea thee effect may be conidered a nuiance by mot GPS uer, they will provide the ionopheric community with an opportunity to ue GPS a a tool to better undertand the plama urrounding the earth. The diperive nature of the ionophere make it poible to meaure it total electron content (TEC) uing dualfrequency GPS obervation collected by permanent network of receiver. One uch network i that of the International GPS Service for Geodynamic (IGS). We have ued dual-frequency GPS peudorange and carrier phae obervation from ix European tation in thi network to derive regional TEC value. In thi reearch, we invetigated the effect of uing different elevation cutoff angle and ionopheric hell height on TEC etimate and atellite-receiver intrumental biae. We found that uing different elevation cutoff angle had an impact on TEC etimate at the TEC unit (TECU) level. We alo dicovered that uing different ionopheric hell height ha an effect on the ionopheric TEC etimate at about the TECU level depending on geographic location and time of the day. We found no ignificant change in the bia etimate uing different elevation cutoff angle. We compared our TEC etimate with TEC prediction obtained by uing the International Reference Ionophere 199 (IRI9) model. The reult of thi comparion are imilar to thoe of other tudie that were conducted uing data et at low olar activity time. After proceing the data from the 6 European tation collected over a 7 day period, we were able to follow highly varying ionopheric condition aociated with geomagnetic diturbance. INTRODUCTION One of the major error ource in GPS poitioning i ionopheric refraction which caue ignal propagation delay. The diturbing influence of the temporally and Preented at the ION National Technical Meeting, Santa Monica, CA, - January 1996

2 patially varying ionization of the ionophere have great impact on atellite geodey, epecially on GPS. Dualfrequency obervation can be ued to eliminate almot all of the ionophere effect. To correct data from a ingle-frequency GPS receiver for the ionopheric effect, it i poible to ue empirical model. We are conducting an on-going tudy uing uch model. After Newby [199] invetigated the International Reference Ionophere (IRI6) model performance, we decided to include the new IRI9 model [Bilitza, 199] in our ionopheric reearch. We ued Faraday rotation data a ground-truth with which we compared the vertical ionopheric range error correction predicted by the GPS navigation meage [Klobuchar, 196] and IRI9 model. Some of our reult have been preented earlier [Komjathy et al., 199a]. Baed on the comparion between the Broadcat and IRI9 model, we concluded that both for day-time and night-time period the IRI9 model appeared to be more accurate than the Broadcat model. Thi concluion i pecific to low olar activity and mid-latitude condition baed on a limited et of data [Komjathy et al., 199b]. Since the availability of Faraday rotation data for ue a ground-truth i limited, we decided to ue dualfrequency peudorange and carrier phae GPS meaurement to infer ionopheric TEC. The literature of relevance to thi reearch i large. Early tudie ued ingle tation obervation to etimate the line-of-ight peudo-tec which i the um of the atellite-receiver intrumental biae and the actual lineof-ight TEC (e.g., Lanyi and Roth [19], Coco et al. [1991]). The neceity to produce global ionopheric map with more accurate TEC and bia etimate ha led the ionopheric community to ue multi-ite fitting technique. Several reearch group have tarted producing regional or global cale TEC map along with atellite-receiver intrumental biae depending on the type of ionopheric obervable ued. For generating ionopheric obervable, one can ue undifferenced dual-frequency peudorange, undifferenced dualfrequency carrier phae obervation or thee two combined. Mot reearch group ue the combined (phae-levelling) technique in which cae the integer ambiguity afflicted difference of the L1 and L (L1-L) carrier phae meaurement are adjuted by a contant value determined for each phae-connected arc of data uing precie peudorange meaurement. The L1-L ionopheric obervable ha a noie level 1- order of magnitude below the peudorange ionopheric obervable. The technique ha been decribed by Wilon and Mannucci [199], Runge et al. [199] and other. It i widely ued to etimate variou ionopheric model parameter a well a atellite-receiver intrumental biae (ee, e.g., Gao et al. [199] and Sardon et al. [199]). It i alo feaible to ue double-differenced L1- L carrier-phae obervation to etimate global or regional ionopheric model [Schaer et al. 199]. The advantage of thi technique i that by uing the doubledifferenced ionopheric obervable, one doe not have to etimate the atellite-receiver intrumental biae a they are differenced away. The price we have to pay i that we loe ome of the reolution of the ionopheric ignal. We have therefore choen to ue the former technique. ESTIMATION STRATEGY The ionopheric meaurement from a GPS receiver can be modelled with the commonly ued ingle-layer ionopheric model uing the obervation equation (1): [ I r( t k) = M( e r) a, r( t k) + a1, r( t k) dλ r + (1) a, r( t k ) dϕ r] + b r + b where Ir( t k) i the L1-L phae meaurement at epoch t k made by receiver r oberving atellite, Me ( r ) i the thin-hell elevation mapping function projecting the line-of-ight meaurement to the vertical (ee, e.g., Schaer et al. [199]), a, r,a 1, r,a, r are the parameter for patial linear approximation of TEC to be etimated auming a firt-order Gau-Markov tochatic proce [Gail et al. 1993], dλ r = λ r λ i the difference between a ubionopheric point (the interection of the ray path of a ignal propagating from the atellite to the receiver with a thin pherical hell) and the mean longitude of the un, dϕ r = ϕr ϕ r i the difference between the geomagnetic latitude of the ubionopheric point and the geomagnetic latitude of the tation, and b r, b refer to the receiver and atellite intrumental biae repectively. The parameter a, r,a 1, r,a, r in equation (1) are etimated uing a Kalman filter approach. The prediction and update equation for the tate etimation are decribed by Schwarz [197], Coter et al. [199] and van der Wal [199]. We allowed the model to follow a relatively high 1 TECU per minute change in the total electron content which reulted in the proce noie variance rate of change to be.. TECU / econd characterizing the uncertaintie of the dynamic ionopheric model. For the variance of the meaurement Preented at the ION National Technical Meeting, Santa Monica, CA, - January 1996

3 noie, we ued 1 TECU which decribe the uncertainty in the obervation. In our invetigation, we did not ue receiver with calibrated intrumental delay. We etimated the combined atellite-receiver intrumental delay for one tation of the network. The ionopheric reearch community tend to ue tation Madrid for that purpoe o that it i eaier to exchange reult among each other. In our network olution, we need to etimate additional biae for the other tation baed on the fact that the other receiver have different intrumental delay. Therefore, for each tation other than the tation Madrid an additional bia parameter wa etimated which i the difference between the receiver intrumental delay between a tation in the network and tation Madrid. Thi technique i decribed by Sardon et al. [199]. Since the reaon for the ionophere exitence i the interaction of ionizing radiation (principally from olar ultraviolet and x-ray emiion) [Langley, 199] with the earth atmophere and magnetic field, we choe a olar-geomagnetic reference frame baed on un-fixed longitude and geomagnetic latitude. Mannucci et al. [199] concluded that the ionophere varied much more lowly in a un-fixed reference frame compared to an earth-fixed one and reulted in more accurate ionopheric delay etimate when uing Kalman-filter updating. THE DATA SET With thi method, we analyed dual-frequency GPS data et from the European region coniting of 6 tation of the International GPS Service for Geodynamic (IGS) network collected by Turbo Rogue receiver. The tation are lited in Table 1 and are identified on the map in Figure 1. Station Madrid, Spain Grae, France Matera, Italy Bruel, Belgium Wettzell, Germany Onala, Sweden Geographic Geographic Geomagnetic lat. in deg. long. in deg lat. in deg Table 1. Lit of IGS tation ued for data analyi. The difference in geomagnetic latitude of tation Madrid, Grae, and Matera are le then degree, and 3.3 degree in cae of tation Bruel and Wettzell. Therefore, we can identify three ditinct latitude region in our tet network (1. Madrid, Grae, Matera;. Bruel, Wettzell; 3.Onala). 6 N N N N N 3 N Madrid Grae Onala Bruel Wettzell Matera 3 N 1 W 1 E E 3 E Figure 1. Location of IGS tation ued for data analyi. We proceed 7 day worth of data from all 6 tation panning the time period 1 to 1 October 199 (GPS week 3) during which a geomagnetic diturbance occurred [NGDC, 199]. The planetary equivalent amplitude of magnetic activity a p ugget that the magnetic diturbance tarted on 1 October, 199 and lated for about 6 day until 3 October, 199. The peak (a p = 111) occurred on 19 October, 199. We choe to proce data et for the whole of GPS week 3 encompaing both magnetically quiet and diturbed condition. RESULTS AND DISCUSSION We ued PhaEdit verion. automatic data editing program to detect bad point and cycle lip, repair cycle lip and adjut phae ambiguitie uing the undifferenced data. The program take advantage of the high preciion dual-frequency peudorange meaurement to adjut L1 and L phae by an integer number of cycle to agree with the peudorange meaurement [Freymueller, 199]. The Univerity of New Brunwick DIfferential POitioning Program (DIPOP) package wa extenively modified to etimate ionopheric parameter, and Preented at the ION National Technical Meeting, Santa Monica, CA, - January

4 atellite-receiver intrumental biae uing a Kalman filter algorithm. We invetigated the effect of uing different elevation cutoff angle on the TEC etimate and atellite-receiver intrumental biae. Since the elevation dependence allow u to eparate TEC from atellite-receiver intrumental biae, it i crucial to know what i the minimum elevation cutoff angle we can ue without mimodelling the TEC due to the higher noie on obervation with lower elevation angle. Or alternatively, what i the maximum elevation cutoff angle that i till acceptable without mimodelling the atellitereceiver intrumental biae. We conidered three different elevation angle cutoff: 1,, and degree. Figure how the difference in TEC etimate of 1 and degree with repect to thoe computed uing a degree cutoff angle for Madrid, Bruel, and Onala. Thee difference are diplayed in eparate graph to illutrate the mall difference. We ued degree elevation cutoff angle a a reference for the other two ince thi olution wa expected to be the leat noiy a we can ee by looking at the r.m.. of L1-L ionopheric reidual in Table. Thee tation repreent the three different geomagnetic latitude region from the network of tation we invetigated (ee Table 1). It appear that the peak-to-peak variation of the difference formed by the 1 and degree elevation cutoff angle olution i larger than the one formed by the and degree olution which i what we would expect knowing that there are noiier obervation included when computing the olution with a 1 degree cutoff angle. Uing different elevation cutoff angle produce maximum difference in TEC etimate that can be characterized by a bia of.3 TECU and tandard deviation of. TECU. Thi maximum bia and aociated tandard deviation occured at tation Onala. Another parameter that affect the TEC etimation i the height of ionopheric hell which play an important role in computing the coordinate of the ubionopheric point. It i alo an input parameter of the M( e r ) mapping function - (ee equation (1)). The ingle-layer ionopheric model aume that the vertical TEC can be approximated by a thin pherical hell which i located at a pecified height above the urface of earth. Thi altitude i often aumed to correpond to the maximum electron denity of the ionophere. Furthermore, it i uually aumed that the ionopheric hell height ha no temporal or geographical variation and therefore it i et to a contant value regardle of the time or location of interet. In our invetigation we looked at fixed height of 3, 3, and km and alo included variable height computed by the IRI9 model uing F layer peak height. The TEC difference hown in Figure were obtained uing a 3 km hell height. In Figure 3, we have plotted the difference between correponding TEC etimate for the three repreentative tation uing different ionopheric hell height. We ued a degree elevation cutoff angle to generate the TEC etimate ued in thi figure. The difference tend to decreae a further north the tation i. Since the IRI9 F layer critical height take the geographical and temporal variation into account, there i a larger periodic variation of the difference involving IRI9 height. By looking at Figure 3, the larget difference between two different olution appear to be at the TECU level at the Madrid tation. We computed the mean of the daily atellite-receiver intrumental biae for all 7 day. We alo obtained a et of bia etimate computed by Deutche Forchungantalt f r Luft und Raumfahrt (DLR) Fernerkundungtation, Neutrelitz, Germany [DLR, 199]. After computing the mean of the correponding value obtained from DLR for all 7 day we computed the difference of the correponding biae. After that, we calculated the mean and tandard deviation of the difference computed by UNB (uing different elevation cutoff angle and different ionopheric hell height) and DLR reult (uing degree elevation cutoff angle with 3 km ionopheric hell height [Sardon et al., 199]). The mean and tandard deviation are lited in Table. The difference between UNB and DLR individual intrumental biae range between.1 and 1.1 n. It i clear that the mean of the difference in biae decreae a higher ionopheric hell height are choen. It appear that chooing different elevation cutoff angle ha no ignificant impact on the bia difference. Larget bia difference occur uing the F layer critical height computed by the IRI9 model. Thi i mainly becaue IRI9 predict the critical height to be between and 3 km depending on location and time of the day. The difference between the two intitution reult may be due to the fact that the UNB network conit of only 6 tation. A larger number of tation would provide more accurate intrumental bia value a well a TEC etimate. Our oftware ha been deigned to eaily accommodate more tation in future data analye. The magnetic diturbance on day 9 affected the diurnal variation of the total electron content. Thi can be een in Figure. A an example, we diplay the diurnal variation of our etimate of TEC for two day at tation Madrid: day when the magnetic field wa quiet and day 9 when the magnetic diturbance reached it peak. Preented at the ION National Technical Meeting, Santa Monica, CA, - January 1996

5 hell height in km mean of diff. in n Elevation angle cutoff in degree 1.d. of diff. in n r.m.. of UNB re. mean of diff. in n.d. of diff. in n r.m.. of UNB re. mean of diff. in n.d. of diff. in n r.m.. of UNB re. IRI Table. Summary of comparion of intrumental bia difference between UNB and DLR and r.m.. of UNB ionopheric reidual. We were intereted in finding out whether our algorithm wa able to follow the rapidly changing ionopheric condition on day 9. The two upper panel of Figure diplay three curve repreenting coefficient a, r,a 1, r,a, r (ee equation (1)) with the correponding error bar. Thee coefficient repreent a contant offet, local time lope and latitude lope of the modelled vertical TEC, repectively, in the vicinity of a tation. The two upper panel alo how vertical TEC at tation Madrid ha been computed by evaluating the vertical TEC model in equation (1) (ee expreion in bracket). The hape of the diurnal curve on day and 9 are ignificantly different. On day, the larget TEC value i around 1 TECU wherea on day 9, the larget TEC i around TECU. On the bottom panel of Figure, we plotted the UNB ionopheric reidual for all 6 tation for day 9. In Figure, we diplay all TEC diurnal curve for all tation for all 7 day. We have alo plotted the TEC value predicted by the IRI9 model. It i very intereting to ee that on day 9, at tation Madrid, Grae and Matera, the peak of TEC value increaed coniderably compared to peak for the previou day. On the other hand, for tation Bruel, Wettzell and Onala, the GPS-derived TEC etimate how peak with maller ize than the one on the previou day. On the bottom panel of Figure, we indicate the planetary equivalent amplitude of magnetic field variation [NGDC, 199]. We can ee that large TEC variation on day 9 were preceded by the magnetic field diturbance tarting on day 91. Day 9 eemed to be the mot variable among the 7 day under invetigation. The bottom panel of Figure ugget that there are no major difficultie for our algorithm to follow ionopheric variation induced by the geomagnetic field diturbance. Thi tatement may only hold for the current low olar activity time. The effect of major magnetic field variation on GPS-derived TEC etimate i yet to be invetigated. In Figure, we have alo diplayed the total electron content computed by the IRI9 model (darker, moother curve). Mean of difference S.d. of difference in TECU Madrid Grae Matera..3 Bruel. 1. Wettzell Onala.3 1. Table 3. Comparion of GPS-derived TEC etimate with IRI9 prediction. We computed the difference of IRI9 predicted and GPS-etimated TEC value. The mean of the difference and the tandard deviation for the 6 tation are ummarized in Table 3. When computing the mean of difference, we ued all 7 day of data. The value eem to be overly optimitic compared to the reult decribed by Komjathy [199b]. Thi may be due to the fact that our current invetigation repreent lower olar activity condition compared to thoe of the data in Komjathy [199b] which repreented olar activity condition from April 199 to January 199. Thi i alo upported by the reearch conducted by Newby [199]. In hi thei, Newby looked at high, medium and low olar activity condition. Newby howed that the IRI6 model performance baed on low olar activity condition howed good agreement with Faraday rotation data at the 1. TECU level. Preented at the ION National Technical Meeting, Santa Monica, CA, - January 1996

6 TEC Difference in TECU TEC Difference in TECU Total Electron Content (TEC) Difference Uing Different Elevation Cutoff Angle for Day to 9 at Madrid TEC(el=)-TEC(el=) 1 TEC(el=)-TEC(el=) Mean olar time at Madrid TEC Difference at Bruel TEC(el=1)-TEC(el=) 1 Mean olar time at Bruel TEC(el=1)-TEC(el=) TEC Difference TEC Difference TEC Difference in TECU TEC(el=)-TEC(el=) TEC Difference at Onala 1 Mean olar time at Onala TEC(el=1)-TEC(el=) Figure. Comparion of TEC etimate uing different elevation cutoff angle TEC Difference CONCLUSIONS We invetigated the effect of uing different elevation cutoff angle and ionopheric hell height on TEC etimate and atellite-receiver intrumental biae. We found that uing different elevation cutoff angle had an impact on TEC etimate at the TECU level. Alo, everal ionopheric hell height were looked at. We found that at the TECU level, the ionopheric etimate uing different height agree depending on geographic location and time of the day. We alo compared our bia etimate with reult from DLR. We found an agreement at the. n level. The difference of the biae compared by UNB and DLR indicate that uing different elevation cutoff angle appear not to have a ignificant influence on the biae. However, we found that uing higher ionopheric hell height decreaed the bia difference ignificantly. We alo compared our TEC etimate with TEC prediction obtained by uing the IRI9 model. We managed to follow highly varying ionopheric condition due to a magnetic diturbance. Due to the limited data et invetigated, it i alo important to point Preented at the ION National Technical Meeting, Santa Monica, CA, - January

7 out that the reult preented here are only pecific for mid-latitude tation at low olar activity condition. In the future, we hope to be able to provide TEC update for the IRI9 model by uing dual-frequency GPS meaurement after firt etablihing dependencie of TEC on latitude, longitude and local time. We will alo look at the effect of eaonal variation a well a magnetic and olar activity on thee difference. TEC difference Total Electron Content Difference Uing Different Shell Height (h) for Day to 9 at Madrid TEC(h=3)-TEC(h=IRI) TEC(h=3)-TEC(h=) 1 Mean olar time at Madrid a p index TEC(h=3)-TEC(h=3) a index 3. TEC Difference at Bruel TEC difference TEC(h=3)-TEC(h=IRI) TEC(h=3)-TEC(h=) TEC(h=3)-TEC(h=3) 1 3. Mean olar time at Bruel TEC Difference at Onala. TEC difference TEC(h=3)-TEC(h=IRI) TEC(h=3)-TEC(h=) TEC(h=3)-TEC(h=3) 1 Mean olar time at Onala Figure 3. Comparion of TEC etimate uing different ionopheric hell height. Preented at the ION National Technical Meeting, Santa Monica, CA, - January

8 . Total Electron Content near Madrid for Day Component, TECU/hour, TECU/1 km Latitude lope Contant offet TEC at Madrid Local time lope TEC at Madrid Mean olar time at Madrid Total Electron Content near Madrid for Day 9 Component, TECU/hour, TECU/1 km Contant offet Latitude lope TEC at Madrid Local time lope TEC at Madrid Mean olar time at Madrid Reidual Ionopheric Reidual for 6 Station on Day Mean olar time at Madrid Figure. An example of the diurnal variation of TEC for magnetically quiet and diturbed day, and ionopheric reidual. ACKNOWLEDGMENTS We wih to thank Jeffrey Freymueller for providing u with the latet verion of PhaEdit.. Alo, we greatly appreciate the aitance of Dieter Bilitza in providing u with an update of the IRI9 model. We would like to expre our incerely thank to DLR Fernerkundungtation, Neutrelitz, Germany and IGS to allow u to ue their data et to complete our reearch. Funding from the Univerity of New Brunwick and the Preented at the ION National Technical Meeting, Santa Monica, CA, - January 1996

9 TEC TEC 1 1 Mean olar time at Madrid TEC Mean olar time at Grae 1 TEC 1 1 Mean olar time at Matera TEC Mean olar time at Bruel TEC Mean olar time at Wettzell Planetary equivalent amplitude, a index Mean olar time at Onala Figure. Diurnal variation of TEC at 6 tation for 7 conecutive day uing a network of IGS tation (darker, moother curve repreent the IRI9 prediction) Preented at the ION National Technical Meeting, Santa Monica, CA, - January

10 Natural Science and Engineering Reearch Council of Canada (NSERC) i gratefully acknowledged. Without their help thi reearch could not have been completed. REFERENCES Bilitza, D. (ed.) (199). International Reference Ionophere 199. National Space Science Center/World Data Center A for Rocket and Satellite, Lanham, MD. Report Number NSSDC/WDC-A-R&S 9-. Coco, D.S., C.E. Coker, S.R. Dahlke, and J.R. Clynch (1991). Variability of GPS Satellite Differential Group Delay Biae. IEEE Tranaction on Aeropace Sytem, Vol. 7, No. 6, pp Coter, A.J., E.M. Gapochkin, and L.E. Thornton (199). Real-time ionopheric monitoring ytem uing GPS. Navigation: Journal of the Intitute of Navigation. Vol. 39, No., pp DLR Fernerkundungtation Neutrelitz (199). Ionopheric Databae, ftp://ftpnz.dlr.de, pub/navigation/bia, acceed December 199. Freymueller, J. (199). Peronal communication, Department of Geophyic, Stanford Univerity, Stanford, CA, January. Gail, W.B., B. Prag, D.S. Coco, and C. Coker (1993). A Statitical Characterization of Local Mid-latitude Total Electron Content. Journal of Geophyical Reearch, Vol. 9, No. A9, pp. 1,717-1,77. Gao, Y., P. Heroux, and J. Kouba (199). Etimation of GPS Receiver and Satellite L1/L Signal Delay Biae Uing Data from CACS. KIS9, Proceeding of the International Sympoium on Kinematic Sytem in Geodey, Geomatic, and Navigation, Banff, Alberta, 3 Augut - September, Department of Geomatic Engineering, The Univerity of Calgary, Calgary, Alberta, pp Klobuchar, J.A. (196). Deign and Characteritic of the GPS Ionopheric Time Delay Algorithm for Single- Frequency Uer. Proceeding of the PLANS-6 conference, La Vega, NV, -7 Nov., pp. -6. ( Komjathy, A., R.B. Langley, and F. Vejrazka (199a). A Comparion of Predicted and Meaured Ionopheric Range Error Correction. EOS Tranaction of the American Geophyical Union, Vol. 76, No. 17, Spring Meeting Supplement, S7. ( Komjathy, A., R.B. Langley, and F. Vejrazka (199b). Aement of Two Method to Provide Ionopheric Range Error Correction for Single-frequency GPS Uer. Preented at IUGG XXI General Aembly, Boulder, Colorado, - 1 July 199. Langley, R.B. (199). The Effect of the Ionophere and Tropophere on Satellite Poitioning Sytem. Proceeding of Sympoium on Refraction of Tranatmopheric Signal in Geodey, The Hague, The Netherland, 19- May, Netherland Geodetic Commiion, Publication on Geodey, Delft, The Netherland, No. 36, New Serie, pp. 97. Lanyi, G.E. and T. Roth (19). A Comparion of Mapped and Meaured Total Ionopheric Electron Content Uing Global Poitioning Sytem and Beacon Satellite Obervation, Radio Science, Vol. 3, No., pp Mannucci, A., B. Wilon, and D. Yuan (199). An Improved Ionopheric Correction Method for Wide- Area Augmentation Sytem. ION GPS-9, Proceeding of the th International Technical Meeting of the Satellite Diviion of The Intitute of Navigation, Palm Spring, CA, - 1 September, The Intitute of Navigation, Alexandria, Va., pp National Geophyical Data Center (199). Geomagnetic Databae, ftp:// IC_DATA/INDICES/KP_AP/199.v, acceed January Newby, S.P. (199). An Aement of Empirical Model for the Prediction of the Tranionopheric Propagation Delay of Radio Signal. M.Sc.E. thei, Department of Surveying Engineering Technical Report No., Univerity of New Brunwick, Fredericton, NB, Canada. Runge, T., U. Lindqwiter, A. Mannucci, M. Reye, B. Wilon, and D. Yuan (199). Generation of GPS Obervable for Global Ionopheric Mapping. EOS Tranaction of the American Geophyical Union, Vol. 76 No. 17, Spring Meeting Supplement, S7. Sardon, E., A. Riu and N. Zarraoa (199). Etimation of the Tranmitter and Receiver Differential Biae and the Ionopheric Total Electron Content from Global Poitioning Sytem Obervation. Radio Science, Vol. 9, No. 3, pp Schaer, S., G. Beutler, L. Mervart, and M. Rothacher (199). Global and Regional Ionophere Model Uing the GPS Double Difference Phae Obervable. Preented at the 199 IGS Workhop, Potdam, Germany, 1-17 May. Schwarz, K.P. (197). Kalman Filtering and Optimal Smoothing. Paper for the CISM Adjutment and Analyi Seminar, nd ed., The Canadian Intitute of Surveying and Mapping, Ottawa, Ontario, Canada, pp van der Wal, A. D. (199). Evaluation of Strategie for Etimating Reidual Neutral-atmophere Propagation Delay in High Preciion Global Poitioning Sytem Data Analyi. M.Sc.E thei, Department of Geodey and Geomatic Engineering Technical Report No 177, Univerity of New Brunwick, Canada. Wilon, B. and A. Mannucci (199). Extracting Ionopheric Meaurement from GPS in the Preence of Anti-poofing. ION GPS-9, Proceeding of the 7th International Technical Meeting of the Satellite Diviion of The Intitute of Navigation, Salt Lake City, Utah, -3 September, The Intitute of Navigation, Alexandria, Va., Vol., pp Preented at the ION National Technical Meeting, Santa Monica, CA, - January