Consistency of Data from Different Satellites

Previous studies addressed the consistency of data from different RO satellites by selecting RO profiles in close spatial and temporal vicinity (Hajj et al. 2004; Schreiner et al. 2007; Ho et al. 2009). Data from CHAMP and SAC-C (Satelite de Aplicaciones Científicas-C), e.g., showed a remarkable consistency of 0.1 K in the mean between 5 km and 15 km (Hajj et al. 2004).

In Foelsche et al. (2009) we adopted a different approach and looked at systematic differences between zonal mean climatologies from different satellites for the season SON (Sep-Oct-Nov) 2006. This is a very rigorous test of the consistency, since these differences contain the sampling errors of both climatologies (F3C satellites in final constellation never simultaneously sample the same region of the atmosphere). Nevertheless, the results were very encouraging with larger differences confined to the polar bins, where the sampling was still sparse in the early phase of the F3C mission (with comparatively low orbit altitudes).

In a next step we looked at "double-differences", where we subtracted the estimated sampling error field from each climatology. We found that seasonal temperature climatologies derived from different F3C satellites agree to within < 0.1 K almost everywhere in the considered domain between 8 km and 35 km altitude. Here we test if these results still hold for JJA 2007 (with F3C satellites in significantly more separated orbits). Figure 4 shows the systematic differences between climatologies derived from Flight Model 2 (FM2) and each of the other F3C satellites. The estimated sampling errors are again subtracted. We note that these double differences contain also the errors in the estimation of the sampling errors for both climatologies.

The systematic differences are again amazingly small (note the 0.1 K contour lines). Differences exceeding 0.1 K are confined to small parts of the domain; most

1 10

1 10

-85 -75-65 -55 -45 -35 -25 -15 -5 0 5 15 25 35 45 55 65 75 85 Latitude [deg]

-0.8-0.7-0.6-0.5-0.4-0.3-0.2-0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 [K]

0.8-0.7-0.6-0.5-0.4-0.3-0.2-0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 [K]

-0.8-0.7-0.6-0.5-0.4-0.3-0.2-0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 [K]

-0.8-0.7-0.6-0.5-0.4-0.3-0.2-0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 [K]

85 -75 -65 -55 -45 -35 -25 -15 -5 0 5 15 25 35 45 55 65 75 85 Latitude [deg]

0.8-0.7-0.6-0.5-0.4-0.3-0.2-0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 [K]

e 20

0.8-0.7-0.6-0.5-0.4-0.3-0.2-0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 [K]

-85 -75 -65 -55 -45 -35 -25 -15 -5 0 5 15 25 35 45 55 65 75 85 Latitude [deg]

-0.8-0.7-0.6-0.5-0.4-0.3-0.2-0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 [K]

0.8-0.7-0.6-0.5-0.4-0.3-0.2-0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 [K]

Fig. 4 Systematic differences between JJA 2007 zonal mean, seasonal mean dry temperature climatologies from Formosat-3/COSMIC FM2 and each of the other F3C satellites, with the estimated sampling errors subtracted

Fig. 4 Systematic differences between JJA 2007 zonal mean, seasonal mean dry temperature climatologies from Formosat-3/COSMIC FM2 and each of the other F3C satellites, with the estimated sampling errors subtracted

-85 -75-65 -55 -45 -35 -25 -15 -5 0 5 15 25 35 45 55 65 75 85 Latitude [deg]

e 20

85 -75 -65 -55 -45 -35 -25 -15 -5 0 5 15 25 35 45 55 65 75 85 Latitude [deg]

-85 -75 -65 -55 -45 -35 -25 -15 -5 0 5 15 25 35 45 55 65 75 85 Latitude [deg]

of them appear above 30 km altitude. Mean differences over the entire domain between 8 km and 35 km altitude are smaller than 0.03 K in any case. The values are: +0.028 K, -0.020 K, -0.014 K, -0.010 K, and -0.003 K, respectively. We note that this consistency is not a proof of absolute accuracy, since there is a possibility of common systematic errors. Nevertheless, we regard the results as very encouraging. These very good agreements show furthermore that the estimation of the sampling error is quite accurate. F3C sampling errors in the polar bins in JJA 2007 (not shown) are considerably smaller than in SON 2006 (see Foelsche et al. 2009) since the higher orbits allow for a better coverage of latitudes beyond 85°.

For SON 2006 we had found a small but systematic difference between CHAMP and F3C (Foelsche et al. 2009). In JJA 2007 (Fig. 5) the situation is very similar. The mean values over the entire domain between 8 km and 35 km altitude are -0.28 K and -0.23 K, respectively, with the largest values above 30 km.

The differences are slightly larger than those found by Ho et al. (2009) for co-located profiles. In the Ho et al. (2009) study, however, phase and orbit data for F3C and CHAMP have both been computed at UCAR, while our present study uses CHAMP phase and orbit data from GFZ, but F3C phase and orbit data from UCAR. We had therefore the suspicion that there might already be differences at phase and orbit data level. Since UCAR computes also CHAMP phase and orbit data we could test this by comparing CHAMP climatologies derived from UCAR and GFZ. The differences are very similar to those in Fig. 5, with a mean systematic difference of-0.29 K for JJA2007 (Fig. 6, left). Differences between F3C and CHAMP/UCAR are therefore much smaller (Fig. 6, right), with a mean value of +0.06 K over the entire domain. Largest (positive) differences are at higher altitudes. This seems reasonable if we assume that ECMWF is indeed somewhat too cold in the stratosphere: The F3C climatologies (which are less influenced by the background) should therefore be slightly warmer than CHAMP (which is closer to the background, since CHAMP data are noisier). The reasons for the small but systematic UCAR-GFZ differences are currently under investigation. Preliminary results indicate that they are stationary to a high degree and that they appear already at bending angle level

Fig. 5 Systematic differences between zonal mean, seasonal mean dry temperature climatologies from Formosat-3/COSMIC and CHAMP with the estimated sampling errors subtracted: SON 2006 (left) and JJA 2007 (right)

JJA 2007: (CHAMP-UCAR-Sampl Err) - (CHAMP-GFZ-Sampl Err)

JJA 2007: (CHAMP-UCAR-Sampl Err) - (CHAMP-GFZ-Sampl Err)

-0.8-0.7-0.6-0.5-0.4-0.3-0.2-0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 [K]

-85 -75 -65 -55 -45 -35 -25 -15 -5 0 5 15 25 35 45 55 65 75 85 Latitude [deg]

35 30

JJA 2007: (COSMIC-SampI Err) - (CHAMP-UCAR-SampI Err)

JJA 2007: (COSMIC-SampI Err) - (CHAMP-UCAR-SampI Err)

35 30

0.8-0.7-0.6-0.5-0.4-0.3-0.2-0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8[K]

-85 -75 -65 -55 -45 -35 -25 -15 -5 0 5 15 25 35 45 55 65 75 85 Latitude [deg]

-0.8-0.7-0.6-0.5-0.4-0.3-0.2-0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 [K]

0.8-0.7-0.6-0.5-0.4-0.3-0.2-0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8[K]

Fig. 6 Left: Systematic difference between JJA 2007 zonal mean, seasonal mean dry temperature climatologies from CHAMP based on UCAR and GFZ phase and orbit data, respectively. Right: Like the right panel of Fig. 5 but using UCAR phase and orbit data as basis for both Formosat-3/COSMIC and CHAMP climatologies. The estimated sampling errors are again subtracted

-85 -75 -65 -55 -45 -35 -25 -15 -5 0 5 15 25 35 45 55 65 75 85 Latitude [deg]

-85 -75 -65 -55 -45 -35 -25 -15 -5 0 5 15 25 35 45 55 65 75 85 Latitude [deg]

(GFZ bending angles being systematically larger than UCAR). Clearly these results highlight the importance of consistent data processing when attempting to build climate records from RO data.

0 0

Post a comment