View Poster

LoHCo logo The LoHCo Project

1: Comparison of Ring-Diagram Local Helioseismology on GONG++, MDI and Mt. Wilson Data Sets

Richard S. Bogart1, Sarbani Basu2, John Bolding3, Thierry Corbard4, Deborah A. Haber5, Frank Hill3, Bradley W. Hindman5, Rachel Howe3, Rudolf Komm3, John W. Leibacher3, Edward J. Rhodes6, Perry J. Rose6, Jesper Schou1, Clifford G. Toner3, Juri Toomre5, the LoHCo Project

  1. Stanford University
  2. Yale University
  3. National Solar Observatory
  4. Observatoire de Nice
  5. JILA
  6. University of Southern California


Full deployment of the GONG+ enhanced observing network in October 2001 and implementation of ring-diagram helioseismology in the GONG++ analysis pipeline this year has enabled us to make a detailed intercomparison of results obtained through multiple paths, from observation through each of the analysis steps. Such comparisons can provide a certain degree of validation of the implementations of the analysis procedures, hints of systematic errors, and better characterization of the observations, possibly leading to improved calibrations. The Local Helioseismology Comparison (LoHCo) Project has been established to provide standards for intercomparison of results obtained with different local helioseismic analysis techniques applied to the available observational data sources.

Ring-diagram analysis was the first local helioseismic technique to be developed It is the one that is easiest to exploit both in terms of processing speed and ease of interpretation. Since the launch of SOHO in 1996, regular synoptic ring-diagram analyses have been performed on all MDI data of sufficient continuity, generally those taken during the annual two- or three-month Dynamics Programs. The most important results to come out of that analysis to date are the discovery and characterization of flow concentrations around active regions, and the structure of the near-surface global meridional circulation pattern, including a very surprising turnover at depth in the northern hemisphere only during the peak years of the solar activity cycle. Those analyses have been performed on an over-sampled grid of 15° diameter regions sampled 24 times per Carrington rotation. Here we report an attempts to duplicate that analysis using both GONG+ data for the same observing periods and the newly developed GONG++ analysis pipeline (designated A here) in addition to the MDI data and the traditional MDI pipeline (C )

Data Analysis

Reasonably continuous full-disc Doppler data from both the MDI instrument on SOHO and the GONG+ network are available throughout Carrington Rotation 1988 (2002/3/30 - 2002/4/26), with comparable duty cycles in most of the sampling intervals (Table 1). We selected this time period for a detailed comparison of the ring-diagram determinations of localized sub-surface flows and frequency shifts obtained from both sources in common observing intervals. Note that the comparison intervals of 1664-minute duration are timed by central meridian crossings as observed by SOHO; pipeline A used earth-based crossing times, which differ by about 10 minutes. (For some comparisons a few extra observation intervals at the beginning of CR 1989 were used.)

Table 1. Data coverage during the 24 comparison intervals in CR 1988 (yr 2002)
StartEndMDIGONG Carr.
36003.29 10:5303.30 14:36.913.881 18004.12 02:2704.13 06:10.951.733
34503.30 14:1103.31 17:54.943.954 16504.13 05:4504.14 09:28.963.802
33003.31 17:3004.01 21:13.968.897 15004.14 09:0204.15 12:45.801.626
31504.01 20:4804.03 00:31.958.796 13504.15 12:1904.16 16:02.669.686
30004.03 00:0604.04 03:49.974.974 12004.16 15:3604.17 19:19.856.995
28504.04 03:2404.05 07:07.948.996 10504.17 18:5204.18 22:35.973.933
27004.05 06:4204.06 10:25.906.825 9004.18 22:0904.20 01:52.955.777
25504.06 10:0004.07 13:43.952.781 7504.20 01:2504.21 05:08.976.984
24004.07 13:1804.08 17:01.977.876 6004.21 04:4204.22 08:25.904.889
22504.08 16:3504.09 20:18.960.962 4504.22 07:5804.23 11:41.976.767
21004.09 19:5304.10 23:36.932.725 3004.23 11:1404.24 14:57.966.776
19504.10 23:1004.12 02:53.911.869 1504.24 14:3004.25 18:13.898.497

The two analysis pipelines involve multiple independent steps, involving data selection and preprocessing; tracking; spectral cleaning, ring fitting, and inversion. Preprocessing is more or less tailored to the observational data, but from that point on data from either stream can, at least in principle, be run through a processing pipeline built from any combination of options in the two approaches.

In addition to the MDI and GONG+ data there is another independent data set of sufficient resolution, extent, and continuity to be suitable for synoptic ring-diagram analysis: the Doppler data from the magneto-optic filter on the 60-ft solar tower telescope at Mt. Wilson. These single-site data go back to the year 1988 and continue to the present. We have begun to integrate analysis of Mt. Wilson data into pipeline C and have produced some initial analysis results that appear promising. There are problems with the gap structure, however, that may require some modifications to the analysis program. In any case the data for the comparison period selected, CR 1988, are not yet available for processing at this time, so that comparison awaits a subsequent report.

Data Selection and Preprocessing

Apart from a small number of bad or incomplete images the MDI data are quite uniform and stable, so little preprocessing is required. In pipeline C the data are detrended by removing the mean of all images during the tracking period. This removes the large-scale Doppler signals due to mean observer motion (which changes little during the tracking period), solar rotation, and uncalibrated instrumental sensitivity gradients. Gaps and holes in the data are then zero-filled. Pipeline A applies a forward first-difference to sequential images, zero-filling the gaps.

GONG data must of course be merged as well as properly detrended. Pipeline A achieves this by scaling the individual site data to a common sensitivity level, first-differencing consecutive images, remapping individual site data to a common scale and location, and averaging together data for the same minute from multiple sites when more than one is available, typically 50 - 60% of the time during this month. In pipeline C the individual site data are scaled to a common sensitivity level, detrending by removing the effect of the known observer motion, and simply selecting data from one site when more than one is available, in this case the easternmost site.

Due to misunderstandings in time indexing (start vs. center of observing interval and use of TAI vs. GPS), the times selected for the processing intervals of MDI and GONG+ data differed by 60 seconds rather than near simultaneity when accounting for the 5-second light travel time difference.


Pipeline A remaps the images using a transverse cylindrical projection, which preserves distances along perpendiculars to the meridian, with a cubic spline interpolation (Corbard et al. 2002). Pipeline C uses Postel's projection, which preserves distances along great circles throught the center of the map, with cubic convolution interpolation. Of course both pipelines map to the same size and scale, 0°.125 arc heliographic per pixel with a map spatial extent of 16° (128 pixels), designed to preserve full-disc mode MDI resolution at disc center. All GONG data and MDI data away from disc center are consequently over-sampled. Also, both pipelines track at the same rate, a photospheric rotation rate appropriate to the latitude of the center of the tracked region.

For this comparison, we track at the locations used in the standard dense-pack sub-surface weather (SSW) analysis described in Haber et al. 2000. These are sampled in increments of 7°.5 in latitude and central-meridian longitude out to ±52°.5.

Spectral cleaning

Pipeline A applies a multi-taper to the spectrum; pipeline C uses a single spatial-temporal apodization. A pseudo-MTF correction of the azimuthal asymmetries in the ring spectra is applied for the SSW ring fitting in both pipelines.

Ring fitting

For synoptic analyses, both pipelines use a common 6-parameter fitting approach (Haber et al. 2000). For selected regions and averages, pipeline C also makes use of a much costlier 13-parameter fit, but one yielding better uncertainty estimates for inversions (Basu & Antia 2000).


RLS inversions are used for both pipelines for the SSW analyses. For pipeline C we use an OLA inversion for the fits from the averaged spectra.


The SSW analysis of both the MDI and GONG+ data processed through pipeline C show broad similarities and a few specific differences that may warrant further investigation (see Figure 1). As expected the results from the GONG data are somewhat noisier, as would be expected from the lower spatial resolution and generally somwehat lower duty cycle. Within the noise the only systematic differences appear at high latitudes. The flows at low latitudes and particularly the anomalous flows inferred in the neighborhood of active regions are generally similar, though the latter tend to be weaker in the GONG data.

Turning to the mean flows averaged over a rotation (Figure 2), we see the same pattern of close agreement at low latitudes and increasing discrepancies at higher latitudes, and also at greater depths. In particular, the very intriguing counter cell in the meridional flow at depth in the northern hemisphere seen in the MDI data is not confimred by the GONG data, though there is indication of at least a slowing of the poleward rate. There appears to be a consistent offset in the mean zonal flows at all latitudes between the two data sources, even though the zonal structure matches quite well. This is equivalent to a discrepancy in tracking rates, which in turn is equivalent to first order to a plate scale discrepancy. The east-west pattern in the mean zonal flows is quite strange: MDI data show a previously-noted slowing toward the west limb, at least at the surface, which seems inexplicable except in terms of uncorrected image distortion. (It seems only to show up in the f-mode as evidenced by its disappearnce in the inverted results below about 5 Mm.) This effect is not present in the GONG data. On the contrary, there appears to be an acceleration near the east limb that is seen only at depth, i.e. presumably in the p-modes!

Results from processing of the two data streams through pipeline A are broadly similar, but there are a few noticeable differences (Figure 3) In particular, while the meridional flow structure inferred from the MDI data with both pipelines is similar, there are substantial differences in the high-northern-latitude structure inferred from the GONG data; however neither is in agreement with the MDI results. In general, there is somewhat less noise in the results from the C processing of the MDI data than the A processing of the same data. The fact that the derived zonal flows are in better agreement than the meridional flows at higher latitudes may be due to the differences in the mapping, since the A mapping is designed to minimize distortion in the east-west direction, while the C mapping is isotropic in its distortion pattern; but the differential distortions between the two mappings over such small regions are less than 10-3..

One very puzzling feature is that there is a marked south-east/north-west gradient on the projected solar disc in the inferred frequencies from the C pipeline applied to GONG data only (using both merges), as shown in Figure 4. We do not yet understand the source of this feature, although interpretation of positional metadata and distortion corrections in the mappings are certainly suspect. In any case, the inferred flows, resulting differential rather than absolute measurements in the ring power spectra, are insensitive to these mapping effects to first order.


  1. Synoptic maps of spatial ditribution of inferred flows during CR 1988. Results are shown for MDI data and GONG+ data both processed through pipeline C. The flow maps are overlaid on MDI synoptic magnetograms.
    1. Figure 1a - Near-surface flows (depth 0.90 Mm); GONG+ data above, MDI data below
    2. Figure 1b - Flows at depth 7.10 Mm.
    3. Figure 1c - Residual flows at depth 7.10 Mm, after removal of 7-day averages for disc locations.
  2. Inferred flows at various depths, averaged over a rotation. Results are shown for MDI data and GONG+ data both processed through pipeline C.
    1. Figure 2a - Meridional flows as a function of latitude
    2. Figure 2b - Zonal flows as a function of latitude
    3. Figure 2c - Meridional flows as a function of longitude along the equator
    4. Figure 2d - Zonal flows as a function of longitude along the equator
  3. Inferred flows as a function of depth and latitude, averaged over a rotation. Results are shown for MDI data and GONG+ data both processed through pipeline A.
    1. Figure 3a - Meridional flows, MDI data
    2. Figure 3b - Meridional flows, GONG+ data
    3. Figure 3c - Zonal flows, MDI data
    4. Figure 3d - Zonal flows, GONG+ data
  4. Spatial frequency gradients in pipeline C processing of GONG+ data (Rachel's last figure)


  1. Basu, S. & Antia, H.M. 2000, Solar Phys. 192, 469--480.
  2. Corbard, T., Toner, C., Hill, F., Hanna, K.D., Haber, D.A., Hindman, B.W. & Bogart, R.S. 2002, Local and Global Helioseismology: The Present and Future. ESA SP-517, 255--258.
  3. Haber, D.A., Hindman, B.W., Toomre, J. Bogart, R.S., Thompson, M.J. & Hill, F. 2000, Solar Phys. 192, 335--350.


This work is partially supported by grants from NASA and NSF.

RH and RK are partially supported by NASA Grant S-92698-F.
SB is partially supported by NASA grant NAG5-10912.

Rick Bogart - 10 Jun 2003, 3:14 pm PDT