Discussion of Error Contributions to Ro Tls Data

We investigated error sources regarding the retrieval of RO data (potential initialization biases, dry temperature assumption), the computation of climatologies (sampling error), and the related synthetic TLS computation procedure (weighting function versus radiative transfer model).

Potential Initialization Biases: In the CCRv2.3 retrieval observed bending angle profiles are combined with background bending angle profiles (ECMWF analyses) in a statistically optimal way to minimize residual biases in atmospheric profiles below 35 km. For each observed RO profile, collocated bending angle profiles are computed from the ECMWF analysis field and expanded at up to 120 km using MSISE-90 climatology. This allows the initialization of the hydrostatic integral at very high altitudes (120 km), where the upper-boundary initialization has no effect on the retrieved atmospheric parameters below 40 km. The resulting atmospheric profiles are background-dominated above the stratopause and observation-dominated below 40 km (see Schr0der et al. (2007) on the importance of the stratopause concerning RO initialization). From validation with independent data sets it was found that potential temperature biases are < 0.2 K in the global mean within 10 km-30 km. Latitudinally resolved analyses show observationally constrained biases of < 0.2 K to 0.5 K up to 35 km in most cases, and up to 30 km in any case, even if severely biased (about 10 K or more) a priori information is used in the high altitude initialization of the retrieval (Gobiet et al. 2007). CHAMP biases near 40 km can thus be expected not to exceed about 2 K. The effect of these levels of bias up to 40 km is small in MSU TLS data, due to the small weighting function contributions above 30 km. Perturbation tests are illustrated in Fig. 5 by means of perturbed minus unperturbed TLS temperatures. They showed that biases (or trends in biases over 5 years) linearly increasing from 0.1 K to 1 K or 0.2 K to 2 K from 30 km to 40 km lead to a bias in MSU TLS (or a trend in bias over 5 years) of < 0.006 K or < 0.015 K, respectively. Even an increase from 1 K at 30 km to a 5 K bias at 40 km (or a trend in bias of 5 K over 5 years) would lead to a spurious TLS change of < 0.04 K only.

Dry Temperature Assumption: RO dry temperature profiles are used throughout this study, i.e., the contribution of water vapor to refractivity is neglected. This dry air assumption always holds to < 0.1 K differences between actual physical and dry temperature at altitudes above 8 km in polar winter regimes and above 14 km in the tropics, respectively, with the dry temperature then becoming gradually colder than the actual one deeper into the troposphere as humidity increases (Foelsche et al. 2008b). We investigated this effect on the MSU TLS temperature based on checks with ECMWF analysis data by integrating over dry temperature as well as over physical temperature profiles. The difference in resulting synthetic TLS temperatures is shown in Fig. 6. We found a negligible mean difference (AT) over

2002-2006 of < 0.02 K in the tropics, of < 0.01 K globally and in the extratrop-ics, respectively, with a negligible trend of < 0.001 K over the 5 year period. The trend is indicated as ST5 yrs with the two sigma uncertainty of the trend in Figs. 6, 7, and 8.

Sampling Error: Sampling error characteristics are estimated for CHAMP RO climatologies based on ECMWF analyses (for details see Gobiet et al. 2005; Steiner et al. 2006; Pirscher et al. 2007; Foelsche et al. 2008b). The sampling error for these single-satellite climatologies results from uneven and sparse sampling in space and time. The typical average CHAMP dry temperature sampling error for monthly zonal means in 10°-latitude bands is < 0.3 K in the upper troposphere and lower stratosphere, with the local time component of sampling error being < 0.15 K (Pirscher et al. 2007). For the investigated CHAMP data set the average number of occultation events per month in the regions was 3364 globally, 832 for the tropics, 1044 for the NH extratropics, and 951 for the SH extratropics, respectively. Figure 7 shows the number of events over time as well as the corresponding average monthly sampling error for the TLS temperatures (TLS weighting function applied to the sampling error profile). The 2002-2006 standard deviation of the sampling error was estimated to 0.08 K globally, 0.07 K for the tropics, 0.18 K for the NH extratropics, and 0.17 K for the SH extratropics, respectively. Other than the 10 km-40 km bulk layer estimate of the sampling error (used in Steiner et al. 2007) the TLS specific sampling error interestingly shows a significant trend over 2002-2006 in the NH extratropics (-0.26 K±0.14 K, Fig. 7c) which also projects globally (-0.12 K±0.07 K, Fig. 7a) while tropical trends are small (-0.06 K±0.06 K, Fig. 7b). This transient change in the CHAMP sampling may possibly be due to orbit changes of the CHAMP satellite and to changes in the GPS constellation over the investigated time period, respectively. It may partly relate also to the major resolution upgrade in ECMWF analyses per February 2006 (see Sect. 2.2), however, since 2002-2005 sampling error trends (not shown) are consistently smaller than 2002-2006 ones (e.g., NH extratropics -0.22 K±0.17 K, global -0.07 K±0.08 K, tropics -0.02 K±0.07 K). The present estimates are thus possibly conservative estimates for the "true" sampling error trends. Alternative estimates based on NCEP (National Centers for Environmental Prediction) reanalyses, and a closer look into CHAMP-GPS sampling pattern changes over time, will help in future to understand the estimates in more detail.

TLS Computation Procedure: A further possible error source is the synthetic MSU TLS computation procedure. The difference in absolute TLS brightness temperature between using either the TLS static weighting functions on monthly mean profiles or radiative transfer modeling for the RTTOV channel MSU4 on individual profiles was found on average < 0.2 K (individual maxima of monthly values ~0.3 K) for all four regions used in this study. This is within the MSU4 and AMSU9 bias estimates of Christy et al. (2003). As shown in Fig. 8, the mean differences in TLS anomalies (using RTTOV minus using weighting functions) were found to be 0.002 K±0.02 K globally, 0.003 K±0.03 K in the tropics, and 0.001 K±0.05 K in the NH/SH extratropics. Overall they are < 0.1 K for all four regions, except for a few individual months in the extratropics (still < 0.2 K). The differences do not show any significant drift over 2002-2006 (i.e., < 0.02 K). These results are consistent with those of Santer et al. (2000) that using a radiative transfer model or a suitable global weighting function for computing TLS temperatures yields to a negligible difference in global and large-scale zonal means.

Was this article helpful?

0 0

Post a comment