Spectral analysis of GPS-­‐based sta1on posi1oning 1me series from PPP solu1ons Chris)na Selle, Shailen Desai, Miquel Garcia Fernandez, Aurore Sibois Jet Propulsion Laboratory, California Ins)tute of Technology, Pasadena, CA, USA Contact: chris)[email protected] 2014 IGS Workshop Pasadena, CA, USA Abstract We used spectral analysis to inves)gate periodic signals in GPS sta)on posi)on )me series obtained using single receiver precise point posi)oning (PPP). In par)cular, we examined a feature of excess power with a 5.5 day period (66 cycles per year) in the JPL GPS PPP )me series [1] that was previously no)ced by Amiri-­‐Simkooei [2] and Ray et al. [3]. We found the strength of this signal to be sta)on-­‐dependent, with many sta)ons not exhibi)ng it at all. We studied the geographic distribu)on of sta)ons with a strong 5.5 day signature and found it to be somewhat uneven. Southern California had a par)cularly high concentra)on of sta)ons with a strong 5.5 day feature, while the feature was not found for sta)ons in East Asia, Oceania and Australia. We also reprocessed point posi)oning )me series for six sta)ons with a very strong 5.5 day feature using the same JPL orbit and clock products and GIPSY-­‐
OASIS soaware, but with a processing strategy that differs from that used to generate the JPL GPS PPP )me series [1]. The resul)ng spectra from our reanalysis show a significant reduc)on in the strength of the 5.5 day feature. Our results demonstrate that the power of the signal at 5.5 days is both sta)on-­‐dependent and may be related to the PPP processing approach used to generate the sta)on posi)on )me series, rather than the JPL orbit and clock products or GIPSY-­‐OASIS soaware. Spectral analysis 50
East
North
Vert
2
10
45
Used PPP )me series available at [1] Included sta)ons for which data are available on 80% of days in 1998-­‐2011 382 sta)ons included (16 outlier sta)ons were excluded based on spectra) Computed Lomb-­‐Scargle periodogram for each sta)on aaer removing linear fit Computed ra)o indica)ng the magnitude of the feature for each sta)on: 40
1
35
Power [cm2/cpy]
10
30
25
20
0
10
15
−1
10
& P: periodogram power at given frequency #%
∑ P (ω )
(
60≤ω ≤70
R = ∑ %
( ω: frequency in cycles/year E,N,V % ∑ P (ω ) + ∑ P (ω ) (
$ 50≤ω <60
70<ω ≤80
' E,N,V: East, North, Ver)cal components No apparent correla)on of R with )me frame used or receiver type Divided sta)ons into five groups based on distribu)on of R For each group, stacked and averaged individual sta)on periodograms The East and ver)cal components were mul)plied by 5 to prevent visual overlaps. •  The power of the 5.5 day feature varies from zero in the 20% of sta)ons with lowest R ra)o to very strong in the 20% of sta)ons with highest R ra)o 5
−2
0
1.4
1.6
1.8
2
2.2
2.4
2.6
2.8
10
3
2nd quintile (1.5670 < R < 1.6760)
JPL GPS time series
Reprocessed PPP time series
2
10
East
North
Vert
2
10
1
1
10
0
10
−1
East
North
Vert
1
−2
0
10
1
10
2
10
10
1
10
Frequency [cpy]
4th quintile (1.76495 < R < 1.8995)
5th quintile (R > 1.8995)
East
North
Vert
2
10
2
10
East
North
Vert
2
10
1
1
10
Power [cm2/cpy]
10
Power [cm2/cpy]
0
10
Frequency [cpy]
0
10
−1
0
10
−1
10
10
−2
East
North
Vert
0
10
10
−2
10
East
North
Vert
2
10
10
10
10
2
10
3rd quintile (1.6760 < R < 1.76495)
−1
2
1
10
Frequency [cpy]
10
Reprocessed 1me series •  Compared two )me series: •  JPL GPS PPP )me series [1] •  Reprocessed )me series using the same orbits, clock products, and GIPSY soaware, but different seongs •  Stacked periodograms for six sites with a large 5.5 day signal, )me frame 1998-­‐2011 •  CHUR, ROCK, THU2, TORP, USC1, WHC1 •  Stacked values were normalized, and the East and North components were mul)plied by 10 to separate them in the figure 0
10
Power [cm2/cpy]
• 
• 
• 
• 
10
Power [cm2/cpy]
• 
• 
• 
• 
• 
1st quintile (R < 1.567)
R ratio histogram
−2
0
10
1
2
10
10
10
0
1
10
10
Frequency [cpy]
2
10
Frequency [cpy]
1
10
10
Power [cm2/cpy]
Power [cm2/cpy]
Geographic distribu1on 0
10
−1
10
−2
0
10
•  High R ra)o (red) indicates strong 5.5 day feature •  5.5 day feature especially visible for some sites in the Los Angeles area •  Feature is weak or not present in IGS08 frame sites and sites along the Western Pacific Rim Southern California
−1
10
−2
10
10
log(R)
Frequency [cpy]
Frequency [cpy]
•  Further research is necessary to inves1gate the source of the 5.5 day feature in the JPL GPS 1me series and to determine which modifica1on in the PPP seNngs led to the decrease in prominence of the feature in the reprocessed 1me series. •  Results suggest that the 5.5 day signal does not originate with JPL orbit and clock products or GIPSY-­‐OASIS soQware. References 1.  JPL repro2011b )me series accessed from ap://sideshow.jpl.nasa.gov/pub/JPL_GPS_Timeseries/repro2011b/post/
−3
10
−3
0
10
1
10
2
10
10
0
10
1
10
1
2
10
0.9
0.8
0.7
0.6
0.5
0.4
0.3
2.  Amiri-­‐Simkooei, A. R. (2013), On the nature of GPS draconi)c year periodic paiern in mul)variate posi)on )me series, J. Geophys. Res. Solid Earth, 118, 2500–2511 3.  Ray, J., J. Griffiths, X. Collilieux, and P. Rebischung (2013), Subseasonal GNSS posi)oning errors, Geophys. Res. Lei., 40, 5854–5860 ©2014. California Ins)tute of Technology. Government sponsorship acknowledged.