Figure 4 – uploaded by André Garcia Sá

See full PDF downloadDownload figure

Figure 4: Discretisation of the space by a tomographic grid where unknowns (wet refractivity or water vapour densities) are the values (N,,,) of each voxel. Dash lines are rays of propagation of GPS signals from satellites to ground stations. yr each GPS station within the voxel model, the integrated slants observations SLANTgps (in our case the SWDs or — Figure 4 Discretisation of the space by a tomographic grid where unknowns (wet refractivity or water vapour densities) are the values (N,,,) of each voxel. Dash lines are rays of propagation of GPS signals from satellites to ground stations. yr each GPS station within the voxel model, the integrated slants observations SLANTgps (in our case the SWDs or

Related Figures (20)

Figure 1: Observations of Cloud Top altitudes from a) GOME-2 at 23:56 UTC on 2010/03/05, and b) OMI satellite’s sensors at 04:07 UTC on 2010/03/06 (background from MODIS/Aqua). Prediction (NWP) of the Australian Bureau of Meteorology (Puri et al., 2013). GOME-2 and OMI UV-visible instruments A typical configuration of the atmosphere, with instability in the troposphere (e.g. triggered by the topography or due to low

Figure 2: 2D fields of [WV from a) ERA-interim NWP, b) ACCESS-A NWP, and c) GPS observations at 06:00 UTC on 2010/03/0€ (the black circles are the locations of GPS stations and the grey arrows show the horizontal IWV gradients).

Figure 3: a) GPS stations (red circles), Radiosonde sites (blue triangles), Radio-occultation profiles (white-purple lines), OMI cloud top altitude at 04:07 on 2010/03/06 (purple-blue-orange pattern; c.f. Fig. 1b) and the tomography grid (inner grid in thick black lines and outer grid in thin black lines); b) Red square represents altitudes above the sea level of the centres of each voxel of the tomography grid.

Table 1: List of Radio-occultations. x the four RO reported in Tab. 1 and shown in Fig. 3. All the profiles were acquired by the COSMIC satellite

Figure 5: Overview of the 5 models and retrievals achieved and apriori model, and the quality check for tomographic retriavals. This study uses 5 models with different inversion technique, i.e., SVD, Least-Square adjustment (LS) and SART. Retrievals

Table 2: Steps studied, divided into data type, apriori data inclusion, stacking observation and impact of uncertainty on solutions.

16a-, 17a+, 17a-) the same as step 4. 17b+, 17b-) used the same configuration as step 3, and steps (7a, 8a, 9a, 10a, lla, 12a, 13a, 14a+, 14a-, 15a+, I5a-, 16a An increased geometrical distribution can be a way to obtain a better stabilisation of tomography solution. More voxels

Table 3: Geometrical distribution and time processing of our sensitivity tests. To visualise the results and validate our methodological improvements, tomography retrievals are compared with externa

Table 4: Sensitivity of N,, tomography retrievals to apriori / time resolution, normalised RMS

Table 5: Sensitivity of p,,, tomography retrievals to data stacking without apriori data, normalised RMS is shown.

Figure 7: Illustration of additional pseudo-slant observations in GPS tomography. The dashed lines show slant observations. The blue and dashed-blue arrows represent respectively the direction of horizontal gradients and their opposites (for a defined distance of e.g. 20 km) that are used to generate pseudo-slants. This section investigates the interest of adding pseudo-slants observations (SLANTgps) in tomography retrievals. The

Table 7: Sensitivity of p,,, tomography retrievals with additional pseudo-slants, normalised RMS. 4.4 Results about the impact on the precision of slant retrievals To investigate the impact on the precision of slant retrievals, the uncertainties of parameters acting for obtaining SWDs and

“depends on the elevation of GPS satellite and data considered Table 8: Absolute uncertainty and relative error of parameters related to GPS tomography.

Table 9: Sensitivity of p,,, tomography retrievals to apriori / time resolution, normalised RMS. 9a, 8a and 7a. Thereby, the statistical results using more realistic uncertainty are presented in brackets (like in Tab. 8).

Figure 8: Comparison of py, retrieval from NWP models (ACCESS-A and ERA-Interim) and tomography models (BIRA, UB: WUELS; step 9b) with respect to Melbourne radiosonde (RS) over whole 5 days period. 1% better than ACCESS-A for the layer 8 — 13 km.

Fig. 9: Profiles of N, distribution determined on the basis of Radio Occultation technique (purple), ACCESS-A (blue) service and exponential pattern. Therefore, these profiles will not be discussed further. - — ~ _ * ~ = shows results from the 4" march at 04:39PM UTC (08:39AM local time). Distribution of N,, and py, were calculated usin; initial observations in 30 minutes of calculation interval, without stacking and taking into account the apriori data by 6 hours The results of the WUELS, BIRA and TUW models are similar, distinguishing two inversions in vertical profiles at height between 4 and 6 km (that we can define as N,, — anomaly). For the RO profile, there are also local minima at these heights however, the largest inversion occurs below 4 km. The RO — and tomography — based values of N,, and py, significantl differ from the ACCESS-A results in the lowest layer as later does not show any inversion below 4 km. All tomography based profiles show a similar inversion layer to the one represented by RO profile, however the location of that layer is lowe

Figure 4 – uploaded by André Garcia Sá

Related Figures (20)

Connect with 287M+ leading minds in your field