> molb TjbjbNN &z$$%D22222228P4D&,R%T%T%T%T%T%T%,'RM*%2%M"22=&&M"M"M"P22R%M"FT62222R%M"M"$h22$W5!$$c&l&$*M"* $M"2tt3.5.2 Rapid Service/Prediction Centre
Processing Techniques
The algorithm used by the IERS Rapid Service/Prediction Centre for the determination of the quick-look Earth orientation parameters (EOP) is based on a weighted cubic spline with adjustable smoothing fit to contributed observational data (McCarthy and Luzum, 1991a). Contributed data are corrected for possible systematic differences. Biases and rates are determined with respect to the 97 C04 (before 14 June 2007) and 05 C04 (on and after 14 June 2007) systems of the IERS Earth Orientation Centre (EOC). Statistical weighting used in the spline is proportional to the inverse square of the estimated accuracy of the individual techniques. Minimal smoothing is applied, consistent with the estimated accuracy of the observational data.
Weights in the algorithm may be either a priori values estimated by the standard deviation of the residual of the techniques or values based on the internal precision reported by contributors. Estimated accuracies of data contributed to the IERS Rapid Service/Prediction Centre are given in Table 1. These estimates are based on the residuals of between the series and the combined RS/PC EOP solution for 2007.
Table 1: Estimated accuracies of the techniques in 2007. Units are milliseconds of arc for x, y, d(, d(, dX, and dY and milliseconds of time for UT1 UTC.
Operationally, the weighted spline uses as input the epoch of observation, the observed value, and the weight of each individual data point. The software computes the spline coefficients for every data point which are then used to interpolate the Earth orientation parameter time series so that x, y, UT1 UTC, d(, and d( values are computed at the epoch of zero hours UTC for each day. Since the celestial pole offset software is written in terms of d( and d(, the IAA VLBI dX and dY values are converted to d( and d( for the combination process. The LOD are derived from the UT1 UTC data. The analytical expression for the first derivative of the cubic spline passing through the UT1 UTC data is used to estimate the LOD at the epoch of the UT1UTC data.
The only data points that are excluded from the combination process are the points whose errors, as reported by the contributors, are greater than three times their average reported precision or those points that have a residual that is more than four times the associated a priori error estimate. Since all of the observations are reported with the effects of sub-daily variations removed, the input data are not corrected for these effects (see IERS Gazette No. 13, 30 January 1997).
Table 2: Mean and standard deviation of the differences between the Rapid Service/Prediction Centre solutions and 97/05 C04 EOP solutions for 2007. Polar motion x and y values are in milliseconds of arc and UT1UTC values are in units of milliseconds of time.
The uncertainties in the daily values listed in Bulletin A are derived from the quality of the spline fit in the neighborhood of the day in question. Table 2 shows the accuracies of Rapid Service/Prediction Centers combination solution for the running, the weekly, and the daily products compared to the 97/05 C04 series maintained by the IERS EOC at the Paris Observatory. The running solution is the combination solution over the past 365-day period. The statistics for the running solution at years end show the agreement between the Bulletin A running combination solution and the 97/05 C04 series for the entire year. The comparison of the 52 weekly solutions to the 97/05 C04 series gives the statistics of the residuals computed over the new combination results for the 7-days prior to the solution epoch. The statistics for the daily solution are the differences for the day of the solution epoch. EOP accuracies for the Bulletin A rapid weekly solution for the new combination for the day of the solution run and daily solution at the time of solution epoch are similar and therefore, not included in the table.
Figure 1 shows the residuals between the daily Bulletin A rapid solution and the 97/05 C04 and presents the data used in Table 2 for the determination of the Bulletin A daily solution statistics. This year Bulletin A had only small reductions in the mean difference and standard deviations. The small bias difference in the polar motion x component appears to be due to different corrections for the change in the International GNSS Service (IGS) series due to the switch from relative phase center to absolute phase center corrections. The two large residuals in the daily polar motion in the early part of the year are the result of an unexpected change in input data format from a contributor. The larger difference in UT1UTC is caused by differences in the way non-VLBI UT data sources are handled between the two centers. These UT1 differences are an area of ongoing investigation.
Fig. 1: Residuals between daily Bulletin A rapid solutions at each daily solution epoch for 2007 and the Earth orientation parameters available in 97/05 C04 series produced in April 2008.
Prediction Techniques
In 2007, the algorithm for polar motion predictions was changed to incorporate the least-squares, autoregressive (LS+AR) method created by W. Kosek and improved by T. Johnson (personal communication, 2006). This method solves for a linear, annual, semiannual, 1/3 annual, 1/4 annual, and Chandler periods fit to the previous 400 days of observed values for x and y. This deterministic model is subtracted from the polar motion values to create residuals, which are more stochastic in nature. The AR algorithm is then used to predict the stochastic process while a deterministic model consisting of the linear, annual, semiannual, and Chandler terms is used to predict the deterministic process. The polar motion prediction is the addition of the deterministic and stochastic predictions. The additional unused terms in the deterministic solution help to absorb errors in the deterministic model caused by the variable amplitude and phase of the deterministic components (T. Johnson, personal communication, 2006). For more information on the implementation of the LS+AR model, see Stamatakos et al. (2008).
The procedure for UT1UTC involves a simple technique of differencing (McCarthy and Luzum, 1991b). All known effects such as leap seconds, solid Earth zonal tides, and seasonal effects are first removed from the observed values of UT1UTC. Then, to determine a prediction of UT1UTC n days into the future, (UT2RTAI)n, the smoothed time value from n days in the past, <(UT2RTAI)n> is subtracted from the most recent value, (UT2RTAI)0
(UT2RTAI)n =2(UT2RTAI)0 <(UT2RTAI)n>.
The amount of smoothing used in this procedure depends on the length of the forecast. Short-term predictions with small values of n make use of less smoothing than long-term predictions. Once this value is obtained, it is possible to account for known effects in order to obtain the prediction of UT1UTC. This process is repeated for each days prediction.
The UT1UTC prediction out to a few days is strongly influenced by the observed daily Universal Time estimates derived at USNO from the motions of the GPS orbit planes reported by the IGS Rapid service. The IGS estimates for LOD are combined with the GPS-based UT estimates to constrain the UT1 rate of change for the most recent observation.
The UT1UTC prediction also makes use of a UT1-like data product derived from a combination of the operational NCEP and U.S. Navy NOGAPS AAM analysis and forecast data (UTAAM). AAM-based predictions are used to determine the UT1 predictions out to a prediction length of 5 days. For longer predictions, the LOD excitations are combined smoothly with the longer-term UT1 predictions described above. In October 2007, the length of AAM forecasts increased from 5 to 7.5 days. This change means that AAM forecasts are the basis of UT1 predictions out to 7 days. For more information on the use of the UT AAM data, see Stamatakos et al. (2008).
Errors of the estimates are derived from analyses of the past differences between observations and the published predictions. Formulas published in Bulletin A can be used to extend the tabular data. The predictions of d( and d( are based on the IERS Conventions (McCarthy, 1996; McCarthy and Petit, 2004). Table 3 shows the standard deviation of the differences between the Bulletin A daily solution predictions and the 97/05 C04 solution for 2007. Initial estimates indicated that the UT1UTC prediction performance would be improved by 42% at 10 days into the future by the addition of UTAAM to the combination and prediction process (Johnson et al., 2005). However, comparisons of the UT1UTC prediction performance from 2003 to those estimated in 2001 (before UTAAM was introduced) indicated a better than 50% improvement in prediction error at both 10 day and 20 days into the future.
For 2007, the prediction errors were, in general, better than those of 2006. The polar motion predictions errors returned to historical levels as the amplitude of the polar motion loops is much smaller than the amplitude of the polar motion in 2007. The prediction of polar motion has been improved by the switch to the LS+AR prediction method. The UT1UTC prediction shows a slight indication of improvement due to the switch from AAM forecast lengths being extended from 5 to 7.5 days. Further investigation to confirm this trend is needed.
Table 3: Root mean square of the differences between the EOP time series predictions produced by the daily Bulletin A rapid solutions and the 97/05 C04 combination solutions for 2007.
The predictions of celestial pole offsets (both dX/dY and d(/d( representations) are produced through the use of the KSV1996 model (IERS Conventions (1996)). In addition, a bias between the model and the last 20 days worth of celestial pole offset observations is computed. This bias is tapered so that as the prediction length is extended, the bias becomes increasingly small. Since celestial pole offsets
are based solely on VLBI data, if no new VLBI 24-hour session observations are available, a new rapid combination/prediction of these angles is not determined. Therefore, the predictions of celestial pole offset start before the solution epoch and the length of the prediction into the future can and does vary in the daily solution files. The differences between the daily Bulletin A predictions and the 97/05 C04 for 2007 are given in Table 4.
Table 4: Root mean square of the differences between the nutation prediction series produced by the daily Bulletin A rapid solutions and the 97/05 C04 solution for 2007.
Predictions of TTUT1 up to 2017 January 1, are given in Table 5. They are derived using a prediction algorithm similar to that employed in the Bulletin A predictions of UT1UTC. Up to twenty years of past observations of TTUT1 are used. Estimates of the expected one-sigma error for each of the predicted values are also given. These are based on analyses of the past performance of the model with respect to the observations.
Additional information on improvements to IERS Bulletin A and the significance for predictions of GPS orbits for real-time users is available (Luzum et al., 2001; Wooden et al., 2004; Stamatakos et al. 2008).
Table 5: Predicted values of TTUT1, 20082017. Note that UT1-TAI can be obtained from this table using the expression UT1TAI = 32.184s (TTUT1).
Centre Activities in 2007
During 2007 a number of changes occurred that affected the performance of IERS Bulletin A. On 14 June, the system of the Bulletin A was changed to match the system of the new 05 C04 solution of the IERS EOC. This change made the EOPs more consistent with the ITRF. The LS+AR polar motion prediction algorithm was implemented on 25 January. Electronic-VLBI (e-VLBI) became operational for certain aspects of the VLBI Intensive observations improving the quick-turnaround UT1 combination and short-term UT1 predictions. IGS Ultra data were added to the polar motion combination on 19 July, improving the quick-turnaround polar motion combination. The improvement can be seen in the statistics presented in Tables 2 and 3. These statistics show that there was a significant reduction in the scatter of the residuals after the inclusion of the IGS Ultras. This reduction is seen in both the daily combination and 1-day daily prediction values, as expected. The ILRS Series A was added to the operational procedures on 25 January, improving the robustness of the combined polar motion solution. On 4 October, the forecast length of the AAM data increased from 5 days to 7.5 days, improving the information available for near-term UT1 forecasts. Additional efforts included improving operational software, updating and monitoring currently used datasets, and investigating potential new data sets. Additional work to increase the robustness of an alternate site to mirror data storage for the combination processing was carried out.
New global solutions were received from GSFC, USNO, IAA, and IVS VLBI analysis centers. These new solutions were examined and new rates and biases were computed.
Availability of Rapid Service
The data available from the IERS Rapid Service/ Prediction Centre consist mainly of the data used in the IERS Bulletin A. These data include: x, y, UT1 UTC, dX and dY from IAA VLBI; x, y, UT1 UTC, d( and d( from GSFC VLBI; x, y, UT1 UTC, d( and d( from USNO VLBI; x, y, UT1 UTC, dX and dY from IVS combination VLBI; UT1 UTC from Saint Petersburg University 1-day Intensives; UT1UTC from GSFC 1-day Intensives; UT1UTC from USNO 1-day Intensives; x, y from Institute of Applied Astronomy 1-day SLR; x, y from the Russian Mission Control Centre 1-day SLR; x, y, LOD from the International GNSS Service; UT from USNO GPS; UT from NRCanada (EMR) GPS; UT from NCEP AAM; UT from NAVY NOGAPS AAM; x, y, UT1UTC, d( and d( from the IERS Rapid Service/Prediction Centre; x, y, UT1 UTC, d( and d( from the IERS Earth Orientation Centre; and predictions of x, y, UT1 UTC from the IERS Rapid Service/Prediction Centre.
In addition to this published information, other data sets are available. These include: UT0UTC from University of Texas as Austin LLR, UT0UTC from JPL LLR; UT0UTC from CERGA LLR; UT0UTC from JPL VLBI; latitude and UT0UTC from Washington PZTs 1,3,7; latitude and UT0UTC from Richmond PZTs 2,6; LOD from ILRS 1-day SLR; x, y, UT1UTC from CSR LAGEOS 3-day SLR; x and y from CSR LAGEOS 5-day SLR; x and y from Delft 1-, 3- and 5-day SLR; and x, y, UT1UTC, d( and d( from IRIS VLBI.
The data described above are available from the Centre in a number of forms. You may request a weekly machine-readable version of the IERS Bulletin A containing the current ninety days worth of predictions via electronic mail from
ser7@maia.usno.navy.mil or through http://maia.usno.navy.mil/.
Internet users can also direct an anonymous FTP to
ftp://maia.usno.navy.mil/ser7
where the IERS Bulletin A and more complete databases can be accessed including the daily Bulletin solutions.
Centre Staff
The Rapid Service/Prediction Centre staff consisted of the following members:
William WoodenDirectorBrian LuzumProgram manager, research, and software maintenanceNick StamatakosOperational procedure manager, research, and software maintenanceGillian BrockettAssists in daily operations and support, research, and software maintenanceMerri Sue CarterAssists in daily operations and supportBeth StetzlerAssists in daily operations and support, research, and software maintenance
In the second half of 2007, Beth Stetzler joined the IERS Rapid Service and Prediction Center.
References
Johnson, T.J, 2002, Rapid Service/Prediction Centre, IERS Annual Report 2001, 4755.
Johnson, T.J, Luzum, B.J., and Ray, J.R., 2005, Improved near-term UT1R predictions using forecasts of atmospheric angular momentum, J. Geodynamics, 39(3), 209.
Luzum, B.J., Ray, J.R., Carter, M.S., and Josties, F.J., 2001, Recent Improvements to IERS Bulletin A Combination and Prediction, GPS Solutions, 4(3), 3440.
McCarthy, D.D. and Luzum, B.J., 1991a, Combination of Precise Observations of the Orientation of the Earth, Bulletin Geodesique, 65, 2227.
McCarthy, D.D. and Luzum, B.J., 1991b, Prediction of Earth Orientation, Bulletin Geodesique, 65, 1821.
McCarthy, D.D. (ed.), 1996, IERS Conventions (1996), IERS Technical Note No. 21, Paris Observatory, France.
McCarthy, D.D. and G. Petit (eds.), 2004, IERS Conventions (2003), IERS Technical Note No. 32, Verlag des Bundesamts fr Kartographie und Geodsie, Frankfurt, Germany.
Stamatakos, N., Luzum, B., Wooden, W., 2008, Recent Improvements in IERS Rapid Service/Prediction Center Products, accepted in Proc. Journes Systmes de Rfrence Spatio-Temporels, Paris, 1719 Sept. 2007.
Wooden, W.H., Johnson, T.J., Carter, M.S., and Myers, A.E., 2004, Near Real-time IERS Products, Proc. Journes Systmes de Rfrence Spatio-Temporels, St. Petersburg, 2225 Sept 2003,160163.
Wooden, W.H., Johnson, T.J., Kammeyer, P.C., Carter, M.S., and Myers, A.E., 2005, Determination and Prediction of UT1 at the IERS Rapid Service/Prediction Center, Proc. Journees Systemes de Reference Spatio-Temporels, Paris, 2022 Sept 2004, 260264.
Brian Luzum, Nicholas Stamatakos, Gillian Brockett, Merri Sue Carter, Beth Stetzler, William Wooden
&<>H P
rt
lnpz|~BJy{wjwhOw6CJOJQJ^JhhOw6CJOJQJ^Jh5hOw6CJOJQJ^J jehOwCJOJQJ^J jyhOwCJOJQJ^JhOwCJOJQJ^JhOwCJOJQJ^Jh5hOw6OJQJ^JhOwOJQJ^JhOwhOw5OJQJ^JhOwCJOJQJ^J'&'=> !
rt01^gdOw^
^ ^`
STLM+,12"%"l#m##########$$$F)L)`*}}}o`o`o`ohOwH*OJQJ^JmHsHhOwOJQJ^JmHsHhOwH*OJQJ^Jhc'?hOw6OJQJ^JhOw5OJQJ^JhOw6CJOJQJ]^JhOwhOwCJOJQJ^Jh8hOwCJOJQJ^Jh?hOwOJQJ^Jh?hOwOJQJhOwOJQJ^JhhOwCJOJQJ^J$."/"##$$x%y%&&U)V)--//Z1\1H35555_7
gdOw
`*b*d*n*p*r*-1////Z1\11111115555661828888=+>]>>>>>>>?ż߱߱߱ߧߏwwodhuhOwOJQJhOwOJQJh*[9hOwOJQJh*[9hOwOJQJ^Jh0ChOwOJQJ^JhOw5OJQJ^JhOw
hOwCJhOw6OJQJ^Jh8hOw^Jh8hOwOJQJ^J jehOwOJQJ^JhOwOJQJ^J jyhOwOJQJ^JhOwOJQJ^J'_7`718288888>>????EE4H6HJJCJDJwJxJJJKKgdOw
????d@f@@@@@@@8A:ARsRRRRRRR?SźŮŮŮzpzpzphOwCJOJQJh#DhOwCJOJQJhOwCJOJQJ]^JhOw5CJOJQJ^JhOw6CJOJQJ^JhOwCJOJQJ^JhOw6OJQJ^JhOwOJQJ^JhOwCJOJQJ^JhEhOwOJQJ^JhOw5OJQJ^JhmAhOwOJQJ^J,MMMdMeMpMqMMgNOOOeP
QQRSS^gdOw^
Qkd{$$Ifl0,"LL
t64
l4a?SASvSxS}SSSSSSSSSSSSSSSSThOwhOwCJOJQJ^Jh#DhOwCJOJQJ^JhOwCJOJQJh#DhOw6CJOJQJh#DhOwCJOJQJSSSTT^gdOw$&P/R / =!"#$%}$$If!vh5L5L#vL:Vl
t065L}$$If!vh5L5L#vL:Vl
t065L}$$If!vh5L5L#vL:Vl
t065L}$$If!vh5L5L#vL:Vl
t065L}$$If!vh5L5L#vL:Vl
t065L}$$If!vh5L5L#vL:Vl
t065L8@8Normal_HmH sH tH J@J Heading 1$$@&^a$6CJ<@< Heading 2$@&>*CJ>@> Heading 3$@&
5>*CJN@N Heading 4$`@&^``
5>*CJ<@< Heading 5$@&5CJ>@> Heading 6$@&
56CJDA@DDefault Paragraph FontVi@VTable Normal :V44
la(k@(No ListTC@TBody Text Indent@^@`6CJ2B@2 Body TextCJLR@LBody Text Indent 2
^CJLS@"LBody Text Indent 3
^CJ0U@10 Hyperlink>*B*@YB@Document Map-D OJQJ4@R4Header
!4 @b4Footer
!@V@q@FollowedHyperlink>*B*6P@6Body Text 26HHBalloon TextCJOJQJ^JaJ<Z@<
Plain TextOJQJ^Jn@nmA
Table Grid7:V0_H'Dz!z!z!z!z zs.;'D|:&'=> !WXNO5
6
:;LM/0st""%%%%f'$)%)))}+~+O,P,,,--333333X7Y7@9A9):*:i:j:::::,;-;:;;;;;;;;;;;;6<7<H<<<<<<<)=*=+======>+??@@3ABBCCC(DH000000ʀ00ʀ00ʀ00ʀ0ʀ0ʀ00ʀ00ʀ0x0ʀ0ʀ00ʀ0x00 00ʀ0ʀ0ʀ0x0ʀ0000x0ʀ0 0ʀ0 0ʀ00ʀ0 0ʀ00ʀ0ʀ00ʀ0(0ʀ00ʀ000ʀ000ʀ0(0ʀ000ʀ0ʐ000ʀ00ʀ00ʀ00ʀ0000ʀ00000000Ҡ0ʠ0ʠ0ʠ0ʠ0ʠ0Ҡ0ʠ0ʠ0ʠ0ʠ0ʠ0Ҡ0ʠ0ʠ0ʠ0ʠ0ʠ0ʀ0ʀ0ʀ00000ʀ0ʀ0ʀ000ʀ0ʀ0H000--333333X7Y7@9A9):*:i:j:::::,;-;:;;;;;;;;;;;;6<7<H<<<<<<<)=*=+======>+??@@3ABCCC(D0ʀ0ʀ0ʀ0ʐ0ʀ0ʀ0ʀ0ʀ0ʀ0ʀ0ʀ0ʀ0ʀ0ʀ0ʀ0ʀ0ʀ0ʀ0ʀ0ʀ0ʀ0ʀ0ʀ`C07 `C07 `C07 U0ʠ07 07 U0ʠ0^7 0^7 U0ʠ07 07 U0ʠ08 08 0ʠ@H8 @H8 0ʠ0ʀ0ʀ0"0"0"0"0"0"0"0"0"0"0"0"0"0"06@`*?oL?ST.1359;_7KKnLMST/24678:<T0'D8@0(
B
S ?_Hlt6056815D:(D@g:(D[a&(.0tz249;2<Yc6,@,--i4k4p4r4@6H6L8P8{8888;;<<<<L=T=>>??@@@@@@AAAAAA3A=AAAAAAAAAjBrBsB{BBBBBmCuCvC~CCCCCCCDD%D(D.09;tuV
^
GHFS
x#~##H$%%f'i'(((())2 299*:A:::%D(D:::::::::::::::::::::4dwF
G3qH(?Xn,Sub-Bureau for Rapid Service and PredictionsLW2Brian Luzum
Oh+'0 ,8
Xdp
|'0Sub-Bureau for Rapid Service and PredictionsLW2NormalBrian Luzum70Microsoft Word 11.4.2@5@ZD@7@8=
a9
՜.+,D՜.+,\hp
'USNOzwF-Sub-Bureau for Rapid Service and PredictionsTitleL@'C_AdHocReviewCycleID_EmailSubject
_AuthorEmail_AuthorEmailDisplayName_PreviousAdHocReviewCycleID'nRIERS Annual Report 2005 wooden.william@usno.navy.milWooden, William H.
!"#$%&'()*+,-./0123456789:;<=?@ABCDEGHIJKLMNOPQRSTUVWXYZ[]^_`abcefghijknRoot Entry FWpData
>1TableF+WordDocument&zSummaryInformation(\DocumentSummaryInformation8dCompObjX FMicrosoft Word DocumentNB6WWord.Document.8