The image on the left displays height scatter after correcting for ionosphere derived from L1 and L2 pseudoranges.  The image in the center displays height scatter after applying the GPS broadcast ionosphere model corrections (ICD-GPS-200C) to the L1 pseudoranges.  The image on the right displays height scatter when no ionosphere correction is applied to the L1 pseudoranges.
 

Date	Dual Freq. Measured Ionosphere correction 95%	Single Freq. Broadcast Ionosphere correction 95%	Single Freq. No Ionosphere correction 95%
May 03	19.88 m	19.33 m	32.27 m
May 04	21.02 m	19.89 m	29.45 m
May 05	27.91 m *	21.94 m	32.08 m
May 06	20.73 m	21.69 m	31.32 m
May 07	20.59 m	22.45 m	31.43 m
May 08	21.25 m	24.20 m	32.17 m
May 09	19.89 m	21.96 m	31.88 m
May 10	20.50 m	22.86 m	31.08 m
May 11	20.78 m	21.07 m	31.54 m
May 12	21.36 m	19.04 m	30.33 m
May 13	-- data set incomplete --
May 14	19.46 m	17.68 m	29.83 m
May 15	20.36 m	19.56 m	32.81 m
May 16	20.71 m	20.34 m	33.83 m
May 17	19.50 m	20.99 m	29.49 m
May 18	21.02 m	20.11 m	32.95 m
May 19	23.35 m *	21.33 m	31.62 m
May 20	20.95 m	20.66 m	32.74 m
May 21	19.84 m	19.81 m	32.08 m
May 22	21.33 m	21.09 m	32.37 m
May 23	21.11 m	20.93 m	33.58 m
May 24	22.84 m	21.20 m	32.08 m
May 25	21.36 m	19.29 m	29.82 m
May 26	21.28 m	17.81 m	29.25 m
May 27	21.36 m	20.83 m	32.13 m
May 28	21.01 m	21.69 m	30.88 m
May 29	20.82 m	21.27 m	29.39 m
May 30	20.92 m	20.48 m	29.46 m
May 31	22.40 m	21.77 m	30.13 m
* Note: May 5 & May 19 data sets have L2 disturbance.

The table at left summarizes the accuracies using different ionosphere corrections. Click on any of the results for a full-size image of the underlying 24-hour data set. The numbers in the table represent the +/- limit in meters bounding 95% of the computed heights.

Now that SA has been removed, the impact of ionospheric and tropospheric error on height is clearly evident.  As seen in the figures, the dual frequency correction does reduce the height error somewhat, but substantial systematic error in height is present in all cases.  The patterns seen after the dual frequency correction are due to the troposphere error on the low elevation GPS satellites participating in the solution.

In the interest of completeness, it should be pointed out that dual frequency ionosphere correction represents a measurement of the ionosphere, and is subject to the multipath error on both the L1 and L2 pseudoranges.  The larger variation in the dual frequency results of May 5 and 19 are due to some disturbance (perhaps interference) in the L2 pseudoranges.

For additional information:

Dr. Richard Snay
National Geodetic Survey, NOAA
Richard Snay